Ref,Title,Journal,Date of publication,Data collection timeframe,Primary decade of follow up,Country of study population,Data source category,Include? EHRs?,Data source details,Main morbidity,Population description,Mean Age,Indicator if median age reported,Percent male,Other features of population,Purpose of analysis category,Purpose of analysis details,Study design,Min renal function measures for analysis,Analysis criteria,Sample size at study start [pre-exclusions],Sample size for main analysis,Number dropped out during follow up,Percent analysed main analysis,Percent dropped out main analysis,Sample size details,Kidney function marker,Formula used,Change in renal function measure,Method of derivation,"Progression definition term, if applicable","Progression definition, if applicable",Details,Kidney function as exposure or outcome?,"If kidney function change measure is exposure, what is outcome?",change in renal function primary or secondary aim?,Statistical tools used,Methods category,Methods details,Follow up time,Follow up years,Study conclusion,Handling sample represenativeness,Methods sample representativeness,Handling dropouts,Methods dropouts,Handling longitudinal missingness,Methods longitudinal missingness,Handling covariate missingness,Methods covariate missingness,Handling distibutional checks,Methods distibutional checks,Handling patient variability in kidney function,Methods patient  variability in kidney function,Handling heterogeneity,Methods heterogeneity,Handling confounding,Methods confounding
Joss N et al [1],Diabetic nephropathy: how effective is treatment in clinical practice?,QJM: An international journal of medicine,2002,1989-1999,1990-1999,UK,outpatient diabetic-renal ,INCLUDE; states EHRs used,EHRs for diabetic-renal clinic in Glasgow,Diabetic nephropathy,patients referred to diabetic-renal clinic with a diabetic nephropathy diagnosis between 1989-1999,58,1,49%,,risk factor identification / causal inference,identification of risk factors associated with progression; establishing whether clinics improve renal function ,retrospective cohort study,4,"4+ measures of creatinine over a minimum of 6 months, no death/RRT in first year",170,125,62,74%,50%,"170 pts identified with diabetic nephropathy;
MAIN analysis for FU >1 year: 125 (74%)
[excluded: 45 death/RRT in first year]
 [during FU to 3 years: 32 died/RRT & 30 LTFU]",estimated creatinine clearance ,Cockcroft and Gault,regression slope of ECC ,individual linear regression,none defined,none defined,,outcome,N/A,primary outcome,multivariate linear regression,linear regression,multivariate linear regression model over >1 year,Median 36 months if >1 year follow up,3,"Age, ACR, and cholesterol are independent predictors of a faster rate of progression of renal disease; 
referral to diabetic-renal clinic slows CKD progression",Partially acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,"Mentioned data completeness, but not implications",None,"Mentioned data completeness, but not implications",None,Mentioned or partially addressed,None,Not mentioned,None,Mentioned or partially addressed,Adjustment for covariates,Mentioned or partially addressed,Adjutstment for baseline covariates
Dean BB et al [2],Erythropoiesis-stimulating protein therapy and the decline of renal function: a retrospective analysis of patients with chronic kidney disease,Current medical research and opinion,2005,1998-2002,2000-2009,USA,outpatient,INCLUDE; states EHRs used,electronic medical records from the Veterans Administration; all outpatient facilities within the VA Greater Los Angeles Healthcare System (no inpatient data),CKD,pre-dialysis patients with CKD,72,,98%,veterans,risk factor identification / causal inference,to determine whether ESA treatment can reduce the rate of decline in renal function,retrospective cohort study,4,"at least 2 serum creatinine measures before and after ESP treatment, separated by a minimum of 6 months; ESP treatment to start within 90 days of ESP treatment; at least 8 ESP treatments required",not mentioned,122,not mentioned,not available,not available,,inverse creatinine,N/A,regression slope of inverse creatinine,individual linear regression,none defined,none defined,,outcome,N/A,primary outcome,multivariate linear regression,linear regression,"Adjustments for age, gender, hyperlipidemia, MI, Hgb at ESP-inititiation, ACE/Arb use; 
model weighted by inverse of the variance of difference between slopes (estimated as sum of variances of pre- and post- treatment slopes) in order to give more weight to patients with more creatinine measures",mean 1.6 years,1.6,Renal function declined at a slower rate after ESP therapy; none of the independent variables examined had a significant effect on rate of disease progression,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Patrially acknowledged implications of data completeness,None,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,None,Tackled,Adjustment for covariates,Tackled,Adjustment for covariates
Gallant JE et al [3],"Changes in renal funcion associated with tenofovir disoproxil fumarate treatment, compared with nucleoside reverse-transcriptase inhibitor treatment",HIV/AIDs,2005,2001-2004,2000-2009,USA,multiple care settings,INCLUDE; states EHRs used,"Patients enrolled at HIV clinic, but data collected from all healthcare settings; database updated every 6 months",HIV,HIV patients starting therapy between Jan 2001 and Dec 2003 with either TDF or NRTI as part of HAART regimen,38,1,72%,,risk factor identification / causal inference,to assess the effect of treatment on renal function in clinical practice,retrospective cohort study,3,"unclear; 
assume avg of 2 x creat pre-therapy plus max creat wihtin 1-year of therapy start (assume strong availability of data at baseline)",not mentioned,658,not mentioned,not available,not available,It is possible that 100% of population was analysed due to comprehensive database,estimated creatinine clearance (ECC) ,Cockcroft and Gault,percent change in ECC from baseline,N/A,decline in creatinine clearance,> 0% percent decline in creatinine clearance from baseline,compared avg of 2 baseline creatinine values with max creatinine within 1 year of therapy start ,outcome,N/A,primary outcome,multivariate linear regression,linear regression,"multiivariate linear regression for percent change in ECC; Also, Wilcoxon rank sum test for bivariate comparisons of absolute and percent change in ECC.",1 year,1,TDF treatment associated with a greater decline in renal function than NRTI treatment,Tackled methodologically,Detailed/ comprehensive database of EHRs used ,Partially acknowledged implications of losses to follow up,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,subgroup analyses,Tackled,Adjustment for covariates
Eriksen BO et al [4],The progression of chronic kidney disease: a 10-year population-based study of the effects of gender and age,Kidney international,2006,1994-2003,1990-1999,Norway,multiple care settings,INCLUDE; states EHRs used,"complete database of serum creatinine measurements in the community (including primary care, inpatient and outpatients); patient and renal survival follow up from hospital databases; well-defined geographical area, served by one provider of nephrology services (University Hospital of North Norway) and universal coverage",CKD,CKD stage 3 (eGFR 30-59 over >3 months),75,1,30%,all aged 20+,risk factor identification / causal inference,"to study the effects of age and gender on eGFR decline (and on initiation of RRT, and death)",retrospective cohort study,2,2 x eGFR 30-59 over 3+ months,3047,3047,1021,100%,34%,"38,241 patients had 1+ creatinine measure; 6,863 had 1+ eGFR 30-59; 3,074 had 2 eGFR 30-59 over 3+ monhts; 3,047 eligible after excluding patients initiating RRT;
Table 1 shows 1021 (33%) patients LTFU due to death (31%) or ESRD/RRT (2%)",eGFR,MDRD,regression slope of eGFR,linear mixed model,rapid progression,regression slope < -5 units per year,"[see methods details]; note: progression term and definition not used for analysis, purely descriptive in discussion of results",outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"well-designed multi-level model of change in eGFR, with baseline adjustment and slope interaction for age and sex, as well as random intercept and slope for patient  (also, nice interpretable model coefficients used); however, no adjustment for any other variables which may impact baseline eGFR and slope",median 3.7 years,3.7,Female gender was associated with slower decline in GFR and better patient and renal survival.,Fully acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Tackled methodologically,mixed modelling,Tackled methodologically or not an issue,Complete data was available for all covariates,Not mentioned,None,Tackled,mixed models,Mentioned or partially addressed,Interaction terms,Fully acknowledged,Adjustment for covariates
Jones C et al [5],An evaluation of a shared primary and secondary care nephrology service for managing patients with moderate to advanced CKD,American Journal of Kidney Disease,2006,1997-2006,2000-2009,UK,multiple care settings,INCLUDE; states EHRs used,Southampton Integrated Monitoring Nephrology (SIMON) Programme; Shared primary and nephrology care scheme (SCS); patients managed in primary care remotely assessed by nephrologists every 6-12 months as needed; SIMON database,CKD,New referrals of patients with CKD stages 3-5 to nephrology clinic or shared care scheme ,75,1,61%,mixed urban/rural area,risk factor identification / causal inference,"Secondary aim:
assess the change in GFR rate of decline for the 5 years before and after referral to SIMON",retrospective cohort study,6,at least 3 pre-referral and 3 post-referral serum creatinine measures,949,738,not mentioned,78%,not available,"949 new referrals; 738 had enough eGFR data to derive pre- and post-referral slopes; Of 211 without data, 158 had insufficient pre-referral data and 53 had insufficient post-referral data",eGFR,MDRD,regression slope of eGFR,individual linear regression,none defined,none defined,Individual regression slopes derived using linear regression (5 years pre-referral; 5 years post-referral),outcome,N/A,secondary outcome,simple statistical tests,mean difference paired t-test,"Individual linear regression to estimate pre-referral and post-referral eGFR slopes. Individual slopes were averaged to achieve a grouped slope. Slopes summarised overall and by subgroup (enrolled in SCS before 6 months in clinic, enrolled in SCS after 6 months in clinic, managed entirely in clinic)",not clear (max 10 years),not stated,"Both the SCS and clinic groups experienced a significant slowing in eGFR decline in the 5 years after referral, emphasizing that even moderate to advanced disease is not inevitably progressive, although an effect of regression to the mean cannot be discounted.",Partially acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Patrially acknowledged implications of data completeness,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,separate analyses by subgroup,Not mentioned,None
Chen SC et al [6],Slowing renal function decline in chronic kidney disease patients after nephrology referral,Nephrology (asian pacific),2008,2001-2006,2000-2009,Taiwan,outpatient renal,INCLUDE; states EHRs used,nephrology divisions of one medical centre and one regional hospital; labs and medication obtained from hospital informatics,CKD,"new referrals of patients with CKD stages 3-5 to nephrology divisions, eGFR < 60 at time of referral",65,,65%,,risk factor identification / causal inference,to study the decline in renal function and factors related to the change in renal function before and after referral,retrospective cohort study,6,">= 3 serum creatinine measures during the 1-year period before and after nephrology referral; 
excluding patients with AKI or initiating dialysis within 1 year post-referral (for reasons of data completeness & variable eGFR)",not mentioned,213,0,not available,not relevant (complete case analysis),"patients did not drop out of analysis since complete data required for analysis, but many patients would not have been analysed (no data)",eGFR,MDRD,regression slope of eGFR,individual linear regression,none defined,none defined,"Fot t-test: individual linear regression, pre- and post- referral;
Main analysis = GEE investigating eGFR vs time with interaction terms",outcome,N/A,primary outcome,generalised estimating equations,generalised estimating equations,"(1) t-test for difference in mean slope before and after referral;
(2) GEEs assessed the association between nehrology referral and eGFR, adjusted for time, age, sex, diabetes, hypertension, CKD stage; 
(3) GEE with interaction between time (slope) and other variables (showing impact of variables on slope of decline) -> not sure results correctly interpreted in article",not stated,not stated,Decline in renal function slows after nephrology referral,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Tackled,GEEs,Tackled,slope interaction terms,Tackled,Adjustment for covariates
O’Riordan A et al [7],Renal biopsy in liver transplant recipients,NDT,2009,1996-? ,2000-2009,UK,outpatient renal,INCLUDE; states EHRs used,renal and liver departments' prospectively compiled databases (1996 onwards); single-center,liver transplant recipients,liver transplant patients referred to renal services,53,,65%,,risk factor identification / causal inference,to compare renal outcomes and survival in liver transplant recipients later referred to renal clinics by whether they had a renal biopsy,retrospective cohort study,not mentioned,Patients followed up for (time to) single eGFR<15 (or RRT or death) but no info on how regularly followed up,2100,54,not mentioned,3%,not available,2100 adults had a liver transplant in the study period; 54 were referred for renal review; 23 had a biopsy and 31 had no biopsy; percent analysed based on liver transplant population which is seen as target population of interest (see discussion),eGFR,MDRD,binary progression to threshold eGFR,single eGFR measure at any time; time to eGFR<15,renal survival endpoint,eGFR<15,,outcome,N/A,primary outcome,KM / life-table,kaplan meier estimation + log-rank test,"Kaplan meier for time from LT to eGFR<15 (inc RRT), with censoring at death and end of follow up; log-rank test for difference in time to event comparing biopsy vs no biopsy; additional KM for time to death",3.2 years,3.2,Patient management and survival did not differ significantly for biopsied vs no biopsy. There were baseline differences between groups but no statistically significant difference in renal outcome.,Partially acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Tackled,Measures capturing AKI excluded,Mentioned or partially addressed,None,Not mentioned,None
Eriksen BO et al [8],Predictors of declining glomerular filtration rate in a population-based chronic kidney disease cohort,Nephron Clinical Practice,2010,1994-2003,1990-1999,Norway,multiple care settings,INCLUDE; states EHRs used,"database of all creatinine tests in routine care, including primary care and hospital inpatient and outpatient medical records",CKD,prevalent CKD stage 3 in population-based cohort,77,,43%,,risk factor identification / causal inference,to study predictors of rate of change in GFR,retrospective cohort study,2,2 x eGFR 30-59 over 3+ months,1398,1224,not mentioned,88%,not available,"Pop coverage = 58,086; Pop 1+ creat = 38,241; Pop 1+ creat indicating CKD stage 3-5 = 2,965; Pop 2+ creat indicating CKD stage 3 = 1,224; (23 excluded due to RRT and 174 excluded with no follow-up GFR after 3 months); In percent analysed, included those with missing FU GFR in denominator but some won't have CKD",eGFR,MDRD,regression slope of eGFR,linear mixed model,progressors (vs non-progressors),"slope < 0 
(vs slope >= 0)",slope of eGFR estimated by 2-level multivariate linear regression model,outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"multiple imputation for missing covariate data; two-level multivariate linear regression model, incorporating baseline data as covariates to estimate indiviudal changes in eGFR (4 models including different covariates); statistical tests for quadratic terms for time and age and interaction term for proteinuria; tabulation of characteristics of progressors vs non-progressors",4.0 years,4,"Male gender, diabetes, proteinuria, and higher MAP were independent predictors of faster decline in GFR. Results may indicate potential benefits for improved adherence to existing guidelines for treatment of older CKD patients.",Fully acknowledged implications of sample completeness,multiple imputation,Partially acknowledged implications of losses to follow up,None,Tackled methodologically,mixed modelling,Tackled methodologically or not an issue,multiple imputation,Not mentioned,None,Mentioned or partially addressed,random effects,Tackled,Interaction terms,Tackled,Adjustment for covariates
Cummings DM et al [9],Glycemic control patterns and kidney disease progression among primary care patients with diabetes mellitus,The journal of the american board of family medicine,2011,1998-2008,2000-2009,USA,primary care,INCLUDE; states EHRs used,an academic primary care practice,diabetes,"medical-record established diagnosis of type 2 diabetes, seen in the primary care practice during 1998-2002",54,,32%,75% african american,risk factor identification / causal inference,to examine relationship between fluctuations in glycaemic control (HbA1C) over time and changes in EGFR,retrospective cohort study,2,require 5 HbA1C measures; eGFR evaluated at baseline and most recent follow up visit,"2,155",791,not mentioned,37%,not available,"Exclusions based on strict criteria for exposure data (5 HbA1c measures); lack of eGFR data not mentioned but since only 2 eGFR measures required, assume used most recently available measure",eGFR,MDRD,absolute change in eGFR from baseline,absolute change from baseline at most recent eGFR,worsening in CKD stage,increase of one or more CKD stages from baseline to final measurement (~8 yrs),main analysis linear regression of change in eGFR from baseline; secondary analysis logistic regression for worsening CKD stage,outcome,N/A,primary outcome,multivariate linear regression,linear regression,"(1) linear regression for eGFR change from baseline; (2) logistic regression for whether increase from baseline in CKD stage
(adjusted for covariates using stepwise approach)
[we consider linear regression as the main analysis, and logistic regression as secondary]",mean 7.6 years,7.6,"age, SBP, baseline HbA1c, baseline eGFR and number of HbA1c values explained 27% of the variation in change in eGFR; measures regarding fluctuations in HbA1c explained very little additional variation",Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,Adjustment for covariates,Mentioned or partially addressed,Adjustment for covariates
Abdelhafiz et al [10],Natural history and predictors of faster glomerular filtration rate decline in a referred population of older patients with type 2 diabetes mellitus,Hospital Practice,2012,1993-2010,2000-2009,UK,outpatient diabetes,INCLUDE; states EHRs used,outpatient diabetes clinic in a district general hospital in the UK; laboratory test results during clinic follow up collected from laboratory; lab results downloaded from central lab database,type 2 diabetes,referred population of older patients (>= 65 years) with type 2 diabetes mellitus,70,,54%,,risk factor identification / causal inference,to study the natural history and predictors of faster GFR decline,cross-sectional study,not mentioned,attending clinic for >= 5 years; no minimum requirement for eGFR ,not mentioned,100,not mentioned,not available,not available,"frequency of eGFR tests during follow up was mean 39.1 (range 18-84); patients sampled based on having at least 5 years attendance, hence not fully allowing for drop-outs; percent of patients excluded on this basis not provided",eGFR,MDRD,regression slope of eGFR,individual linear regression,faster decline,eGFR slope decline > mean (1.5 units per year),regression slope expressed as annual rate of eGFR decline,outcome,N/A,primary outcome,GLMs,logistic regression,slope of eGFR decline per year dichotimised into above and below mean (1.5 units per year) to generate categorical variable for logistic regression; adjusted for covariates (only significicant variables from univariate logistic regressions added to multivariate model),mean 14 years,14,"Development of CVD at (LAST) clinic visit was only ""predictor"" of eGFR decline",Partially acknowledged implications of sample completeness,None,Mentioned losses to follow up but not implications,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Adjustment for covariates ,Mentioned or partially addressed,Adjustment for covariates
Boudville N et al [11],Factors associated with Chronic Kidney Disease Progression in Australian Nephrology Practices,Nephron Clinical Practice,2012,not stated,not stated,Australia,outpatient renal,INCLUDE; states EHRs used,"Nephrology practice electronic medical records, Audit4 software for specialists ",CKD,"CKD patients referred to nephrology practice, spread throughout Australia: prevalent and incdiencent adult CKD with at least two serum creatinine measurements at least 90 days apart; excluding RRT patients",69,,57%,,risk factor identification / causal inference,Exploring factors associated with rate of CKD progression,retrospective cohort study,2,"2 x eGFR, at least 90 days apart",not mentioned,1328,not mentioned,not available,not available,"number CKD pts treated in nephrology practice not provided;
1328 (model 1,2); 
216 (model 3)",eGFR,MDRD,"rate of change in eGFR, not clearly defined",not detailed,none defined,none defined,measured between first entry of patient in Audit4 software and last creatinine measurement before data extraction,outcome,N/A,primary outcome,multivariate linear regression,linear regression,"univariate and multivariate linear regression comparing rate of change in eGFR with other variables; variables with p<0.2 or of theoretical importance included in multivariate model; ""subjects censored for death or ESKD""",mean 1.5 years,1.5,potential risk factors for CKD progression identified; current data quality may result in bias,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Adjustment for covariates,Mentioned or partially addressed,Adjustment for covariates
Dreyer G et al [12],Progression of chronic kidney disease in a multi-ethnic community cohort of patients with diabetes mellitus,Diabetic medicine,2013,2005-2010,2000-2009,UK,primary care,INCLUDE; states EHRs used,extraction of EHRs from practice computer databases of 134 GP practices in East London (covering 98% of practices in the area),diabetes,"diabetes (type 1 and 2); eGFR 15-60 over 3+ months; 
age 30-75; dialysis patients excluded",64,,56%,population with 51% ethnic minorities and severely deprived,risk factor identification / causal inference,to examine whether there are differences in rate of CKD progression by ethnicity,retrospective cohort study,3,2 x eGFR 15-60 over 3+ months; at least 3 years of annual follow up data,not mentioned,"3,855",583,not available,15%,"Flow chart provided with exclusions, but not possible to compute percentage analysed from this since age criteria included after data availability critieria",eGFR,MDRD,regression slope of eGFR,linear mixed model,none defined,none defined,"random effects (pt, GP); baseline covariate adjustment",outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"linear mixed model with random effects for patient and practice; interaction term for time*ethnicity to allow for different slopes by ethnic group (heterogeneity); adjustments for age, sex, deprivation, co-morbidities, proteinuria, eGFR, presciprtions, etc. Analysis repeated in proteinuria only. ",mean 4.3 years,4.3,Rate of CKD progression was significantly greater in South Asians than white patients,Partially acknowledged implications of sample completeness,None,Mentioned losses to follow up but not implications,None,Patrially acknowledged implications of data completeness,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Tackled,mixed models,Tackled,Interaction terms,Tackled,Adjustment for covariates
Herget-Rosenthal S et al [13],Progressive chronic kidney disease in primary care: modifiable risk factors predictive model,Preventive Medicine,2013,2003-2006,2000-2009,Germany,primary care,INCLUDE; states EHRs used,"longitudinal observational Chronic Kidney Disease in Primary Care in Germany (COPING) study, covering 10 GP practices in Germany",CKD,"Patients with or at high risk of CKD: 
CKD = eGFR<60;
high risk = hypertension for 5 years, diabetes for 10 years;
patients exclusively managed in primary care",65,,51%,,risk prediction,"3 objectives: 
(1) determine incidence/prevalence of CKD and progressive GFR decline;
(2) identify modifiable risk factors;
(3) develop model to predict progressive GFR decline
[we decided risk prediction was tha main analysis, based on focus of conclusions]",retrospective cohort study,2,not clear but looks like one eGFR in 2003 and one eGFR in 2006,not mentioned,803,0,not available,not relevant (complete case analysis),Patients only included if renal function data available; 100% of identified population analysed; missing covariate data imputed,eGFR,CKD-EPI,"rate of change in eGFR, not clearly defined","not stated (but looks like absolute change between 2 measures, divided by 3 years time elapsed) (annual eGFR decline)",progressive GFR decline,eGFR slope decline > 2.5 units per year,"Unclear method of evaluation of annual eGFR decline, but looks likely data only collected for 2003 and 2006, so likely to be slope estimated using only 2 eGFR measures (poor estimation likely to adversely affect prediction model)",outcome,N/A,primary outcome,GLMs,logistic regression,multivariate logistic regression for risk prediction model,3 years,3,The risk prediction model contained 7 readily obtainable variables and performed well predicting patients at high risk of progressive GFR decline in primary care,Partially acknowledged implications of sample completeness,Other imputation methods (to avoid exclusions),Not mentioned,None,Not mentioned,None,Tackled methodologically or not an issue,imputation with mean,Not mentioned,None,Not mentioned,None,Tackled,Adjustment for covariates,N/A,N/A
Malgor RD et al [14],A case-control study of intentional occlusion of accessory renal arteries during endovascular aortic aneurysm repair,Journal of Vascular Surgery,2013,1989-2009,2000-2009,USA,multiple care settings,INCLUDE in sub-analysis; not clear if EHRs used,No general details except study approved by IRB of the Mayo Clinic; Details specifically provided for BP measurements (hospital + primary care data sources),abdominal aortic aneurysm (AAA),abdominal aortic aneurysms (AAA) treated by endovascular abdominal aneurysm repair (EVAR),75,,92%,,risk factor identification / causal inference,to investigate association of presence of ARA and changes in eGFR (and other outcomes),retrospective cohort study,not mentioned,"at least 3 measures used to determine baseline eGFR, where possible",not mentioned,119,not mentioned,not available,not available,Numbers provided pre-exclusions for clinical criteria of study but not clear how many patients were excluded due to missing data.,eGFR,MDRD,absolute change in eGFR from baseline,absolute difference from baseline,none defined,none defined,"despite defining reanl function detioration in methods, no results on this or evidence of evaluation; focus on change in eGFR compared by ANOVA",outcome,N/A,primary outcome,ANCOVA,ANCOVA,eGFR compared among cases and controls using ANCOVA; time point for evaluation of changes unclear,3.1 years,3.1,"EVAR with ARA coverage carried no negative impact on renal function or blood pressure, even when applied to patients with more advanced stages of CKD",Mentioned care pathway and inclusion criteria but not sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,subgroup analyses,Not mentioned,None
Brosnan EM et al [15],Drug-induced reduction in estimated glomerular filtration rate in patients with ALK-positive non-small cell lung cancer treated with ALK inhibitor crizotinib,Cancer,2014,2009-2012,2010-2019,USA,not stated,INCLUDE; states EHRs used,electronic medical records in accordance with a University of Coloardo Instituional Review Board-approved protocol,lung cancer,stage IV ALK-positive non-small cell lung cancer treated with crizotinib,55,1,53%,,risk factor identification / causal inference,examining the effects of crizonitib therapy on eGFR in lung cancer patients,retrospective cohort study,not mentioned,not mentioned,38,38,not mentioned,100%,not available,Results state 38 patients with cancer were identified. 38 patients were analysed and it is assumed 100% had data for analysis (mixed model). Patients stopping treatment were mentioned. ,eGFR,CKD-EPI,regression slope of eGFR,piecewise linear mixed models,none defined,none defined,"piecewise linear mixed model with knots at 2, 4, 6, 8 and 12 weeks; post-estimation of mean changes between baseline and each knot and between each consecutive knot with p-values;",outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"piecewise linear mixed model with knots at 2, 4, 6, 8 and 12 weeks; post-estimation of mean changes between baseline and each knot and between each consecutive knot with p-values; additional subgroup analysis; main analysis does not adjust for confounding",12 weeks,0.25,"eGFR is reduced by treatment with crizotinib, but the majority of patients will recover their eGFR after the cessation of therapy (suggesting reduction in eGFR may be pharmalogical effect rather than nephrotoxic effect)",Not mentioned,None,"Mentioned care pathway follow up, but not losses to follow up",None,"Mentioned care pathway follow up, but not data completeness",None,Not mentioned,None,Not mentioned,None,Tackled,random effects,Mentioned or partially addressed,random effects,Mentioned or partially addressed,Adjustment for covariates
Chase HS et al [16],Presence of early CKD-related metabolic complications predict progression of stage 3 CKD: a case-controlled study,BMC Nephrology,2014,2006-2012,2000-2009,USA,primary care ,INCLUDE; states EHRs used,primary healthcare electronic records from the Columbia University Medical Centre of the New York Presbyterian Hospital.,CKD,"adults aged 21+ with CKD stage 3 (eGFR < 60 sustained over time, excluding persistent stage 4); cases: regression slope <-3; controls: regression slope -1 to 1",75,,36%,excluding ICD-9 codes for renal diseases known to cause renal function decline,risk prediction,to design a prediction model for CKD progression using laboratory values reflecting metabolic status,case control study,4,4+ creatinine measurements over at least 4 years,700,481,not mentioned,69%,not available,~700 patients identified with confirmed stage 3 CKD; 481 patients analysed meeting progressor or non-progressor definitions after regression analysis in patients with 4+ measures over 4+ years,eGFR,MDRD,regression slope of eGFR,individual linear regression,progressors,eGFR slope decline > 3 units per year,"binary progression outcome: progressors (slope <-3, and p<0.05 for slope of 0) vs non-progression (slope -1 to 1)",outcome,N/A,primary outcome,machine learning methods,Naïve Bayes classifier; logistic regression,"case control design; t-tests to compare continuous markers at baseline and FU, chi-squared for categorical variables comparing P vs NP; logistic regression and Naïve Bayes classifiers to predict P vs NP, where P and NP defined based on linear regression slopes in individuals",Avg. 6 years (SD 2.9),6,"Patients who enter stage 3 CKD and ultimately progress to stage 4 manifest a greater degree of metabolic complications than those who remain stable. Certain lab values are associated with CKD progression.  Naïve Bayes classifier predicts progression with 58% accuracy and predicts non-porgression with 22% accuracy. For logistic regression, 72% and 30%, respectively. ROC AUCs 0.73, 0.75 respectively.",Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Fully acknowledged,None,Mentioned or partially addressed,adjustment for covariates,not relevant,N/A
Chen H et al [17],Combined application of eGFR and albuminuria for the precise diagnosis of stage 2 and 3a CKD in the elderly,Journal of Nephrology (Italy),2014,2000-2012,2000-2009,China,hospital,INCLUDE in sub-analysis; not clear if EHRs used,inpatient and outpatient data from Beijing Friendship Hospital,CKD,elderly patients (aged >=65) diagnosed with stage 2 or 3a CKD,78,,73%,,risk factor identification / causal inference,to confirm whether eGFR leads to over-diagnosis of CKD and to examine the utility of eGFR combined with albuminuria for diagnosising CKD stage 3a in the elderly,retrospective cohort study,3,"inclusion required 3+ years of consecutive urinary albumin, urine volume, serum creatinine, and routine blood analysis (and stage 2 or 3a CKD)",not mentioned,365,0,not available,not relevant (complete case analysis),"patients did not drop out of analysis since complete data required for analysis, but many patients would not have been analysed (no data)",eGFR,CKD-EPI,"Rate of percentage change in eGFR, not clearly defined",not stated,"""rapid kidney function decline (RFKD)""",annual percentage decline >5%,no information on how annual percentage decline derived (3 eGFR measures over 3 consecutive years available),outcome,N/A,primary outcome,GLMs,logistic regression,"2 analyses: 
(1) comparison of means between baseline and year 3 (t-test), stratified by albuminuria category and baseline CKD stage; 
(2) logistic regression for rapid kidney function decline with multivariate adjustment",3 years,3,"albuminuria is a strong predictor of renal porogress in the elderly, and the overdiagnosis of stage 3a CKD may exist in patients without albuminuria",Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Stratified analysis; adjustment for covariates,Mentioned or partially addressed,Adjustment for covariates
Kose E et al [18],Effects on serum uric acid by difference of the renal protective effects with atorvastatin and rosuvastatin in chronic kidney disease patients,Biological and pharmaceutical bulletin (Japan),2014,2006-2011,2000-2009,Japan,outpatient,INCLUDE; states EHRs used,CKD outpatients in Yokosuka Kyousai Hospital from Nov-2006 to Oct-2011; information collected from medical records retrospectively; electronic medical records,CKD with hyperuricemia/hyperlipidemia,CKD stage 3  patients with hyperuricemia and hyperlipidemia complications and treated with ATV or ROS,66,,59%,excluding patients receiving a combination of drugs,risk factor identification / causal inference,to investigate the effects of atorvastatin (ATV) and rosuvastatin (ROS) on their renal protective effects in CKD patients,cross-sectional study,2,patients with eGFR data before treatment and three months after treatment,35,29,not mentioned,83%,not relevant (complete case analysis),"""Among 35 [eligible] patients of CKD stage 3, we surveyed the 29 patients with data before and after the administration of statins""",eGFR,not specified,absolute change in eGFR from baseline,mean difference paired t-test,none defined,none defined,normality checks performed,outcome,N/A,primary outcome,simple statistical tests,mean difference paired t-test,5% significance level,3 months,0.25,Evidence of renal protective effects in stage 3 CKD patients with hyperuricemia and hyperlipidemia complications with ATV administration. No significant change in ROS patients.,Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Tackled,Distributional checks,Mentioned or partially addressed,Paired t-test,Mentioned or partially addressed,None,Fully acknowledged,None
Nderitu P et al [19],Analgesia dose prescribing and estimated glomerular filtration rate decline: a general practice database linkage cohort study,BMJ Open,2014,2009-2010,2000-2009,UK,primary care,INCLUDE; states EHRs used,2 large general practices; a general practice database linkage cohort study,general population,patients aged 40+ with 2 eGFR 90 days apart,66,,45%,,risk factor identification / causal inference,to investigate the  effect of analgesia prescribing on eGFR decline,retrospective cohort study,2,aged 40+ AND 2 x eGFR separated by 90+ days between 1 Jan 2009 and 21 Dec 2010,"13,081","4,145",not mentioned,32%,not available,"13,081 aged 40+; 4,145 2 valid eGFR measures; infomative flow diagram of inclusions shown",eGFR,MDRD,rate of change in eGFR,absolute difference in 2 eGFR measures divided by time elapsed,eGFR decline,"change of >5 units per year between 2 measures (first, last)",only the first and last eGFR measures during the study period were used (follow up duration between 90 days to 2 years),outcome,N/A,primary outcome,GLMs,logistic regression,"logistic regression for fast/slow rate of decline, adjusted for baseline covariates",mean 270 days (9 months),0.75,"NSAID, aspirin and paracetamol prescribing over 2 years did not significantly affect eGFR decline with a reduced risk of eGFR decline in high-dose aspirin users with well-preserved renal function.",Partially acknowledged implications of sample completeness,None,Not mentioned,None,"Mentioned data completeness, but not implications",None,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,None,Mentioned or partially addressed,Adjustment for covariates; Stratification,Mentioned or partially addressed,Adjustment for covariates
Oetjens M et al [20],Utilization of an EMR-biorepository to identify the genetic predictors of calcineurin-inhibitor toxicity in heart transplant patients,Pacific Symposium of Biocomputing,2014,not stated,not stated,USA,outpatient,INCLUDE; states EHRs used,"BioVU, Vanderbilt University Medical Center's DNA repository linked to electronic medical records; eligibility requires an outpatient lab blood draw",heart transplant recipients,heart transplant recipients prescribed CNIs; excludes eGFR < 30 (since outcome threshold),53,1,66%,,risk factor identification / causal inference,to idenitfy genetic risk factors for CNI nephrotoxicity,retrospective cohort study,not mentioned,"min number of measures not mentioned but data completeness was generally high, with avg 5 measures per pat per year",127,115,not mentioned,91%,not available,127 heart transplant patients in database; 115 with DNA genotyped analysed; 25 died during follow up; not stated how many lost to follow up during intended follow up duration or what intended follow up duration was,eGFR,MDRD,binary progression to threshold eGFR,time to eGFR<30; monthly median eGFR < 30 at least 6 months post-transplant for at least 3 consecutive months,nephrotoxicity,monthly median eGFR <30 for 3 consecutive months,,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"Cox PH regression for time to eGFR<30, adjsuted for age, CKD, diabetes and baseline eGFR (other variables were tested for univariate association but not included in final model as p>0.05); secondary analysis of linear mixed model of post-transplant eGFR",8.8 years,8.8,"No SNP met the threshold for statistical significance, although suggestion of association for some SNPs to be investigated further",Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,mixed modelling,Not mentioned,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Fully acknowledged,Adjustment for covariates,Tackled,Adjustment for covariates
Annor FB et al [21],Psychosocial stress and changes in estimated glomerular filtration rate among adults with diabetes mellitus,Kidney research and clinical practice,2015,2005-2008,2000-2009,USA,primary care,INCLUDE; states EHRs used,2005 survey data was used to capture the main exposure (psychosocial stress) and some other baseline characteristics; eGFR and other laboratory data captured from linkage to 4-year clinical and pharmacy records; other baseline data captured from primary care medical records ,diabetes,"working adults aged 25-59, diagnosed with DM but without advanced microvascular or macrovascular complications",50,,41%,only african american (black) and caucasian ethnicity included; 54% of study population were black,risk factor identification / causal inference,to examine the association between psychosocial stress and eGFR; to invetigate other predictors of eGFR changes,retrospective cohort study,none stated,"Stated that ""for eGFR measures, 49% had a measure on all 4 waves (years of FU) and 91% had a measure on at least 2 waves""; but not clear if all patients included or some excluded; multiple imputation used for missing data",not mentioned,575,not mentioned,not available,not available,,eGFR,MDRD,regression slope of eGFR,growth model,decline in eGFR,eGFR slope decline > 0,"MI to impute missing data; 
structural equation model with eGFR measured at 4 time points (calendar year); eGFR random slope variable specified representing annual rate of eGFR change",outcome,N/A,primary outcome,multiple imputation; structural equation modelling; ,structural equation modelling,"structural equation model with longitudinal eGFR measures ""caused by"" latent constructs (random eGFR intercept, random eGFR slope, latent variable for psycosocial stress), with adjustment for confounders (age, sex, race, education, SES, smoking, med use, MAP, etc)",4 years,4,"no evidence of association between psychosocial stress variable and changes in eGFR; 
other predictors of eGFR changes were age, race, insulin use and mean arterial pressure",Partially acknowledged implications of sample completeness,multiple imputation,Not mentioned,None,Partially acknowledged implications of data completeness,multiple imputation,Partially acknowledged implications of data completeness,multiple imputation,Not mentioned,none,Tackled,random effects,Tackled,stratified analysis; random effects / latent variables,Tackled,
Chakera A et al [30],Prognostic value of endocapillary hypercellularity in IgA nephropathy patients with no immunosuppression,Journal of Nephrology (Italy),2015,not stated,not stated,UK,tertiary care,INCLUDE; states EHRs used,patients managed in a single tertiary unit (Oxford Kidney Unit); clinical and demographic data collected from electronic records,IgA nephropathy,patients with IgA nephropathy managed without (or with minimal) immuospressive therapy; excluding diabtes and henoch-schonlein purpura,40,1,76%,,risk factor identification / causal inference,to determine the impact of proliferative lesions on the natural history of IgAN,retrospective cohort study,not mentioned,not mentioned,237,147,not mentioned,not available,not available,"237 adults identified with IgA nephropathy; 147 analysed after exclusions. Could not compute percent analysed as population criteria and data completeness criteria mixed up.
[percent analysed not related to eGFR data; nothing about eGFR data mentioned]",eGFR,MDRD,"rate of change in eGFR, not clearly defined",not stated,"""rapid decline""",decline in eGFR >5 units per year,additional endpoint: time to ESRD (eGFR<15 or RRT),outcome,N/A,primary outcome,GLMs,logistic regression,"logistic regression for rapid decline, adjusted for covariates; (Cox regression for time to ESRD)",mean 7 years,7,endocapillary proliferation and tubular atrophy/ interstitial fibrosis are independent predictors of loss of renal function,Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,None,Fully acknowledged,None,Tackled,Adjustment for covariates
Cid Ruzafa J et al [22],Estimated glomerular filtration rate progression in UK primary care patients with type 2 diabetes and diabetic kidney disease: a retrospective cohort study,The international journal of clinical practice,2015,2006-2011,2000-2009,UK,primary care,INCLUDE; states EHRs used,Clinical Practice Research Datalink (CPRD) [data from 2006 selected due to QOF updates -> data availability],diabetic kidney disease,patients with type 2 diabetes diagnosis  and CKD diagnosis or renal function tests consistent with NICE/KDIGO-defined CKD (eGFR; albuminuria),68,,57%,,risk factor identification / causal inference,to examine factors associated with progression of diabetic kidney disease,retrospective cohort study,not mentioned,"Require data for age, sex, 12 months of computerised records, albuminuria reported in md/dL, confirmation of CKD; for transitions between CKD stages, at least 2 eGFR required (assumed same for mixed model but not stated)","60,867",15692,not mentioned,26%,not available,"60,867 patients with T2DM and presumed DKD; 15,692 patients analysed in longitudinal mixed model of eGFR (parsimonious model); Number of patients analysed depended on variables in model and data completeness; Only patients with ACR data included in statistical analysis.",eGFR,CKD-EPI,regression slope of eGFR,linear mixed model,none defined,none defined,,outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"mixed model in all patients and separate analyses by CKD stage; adjustment for baseline covariates, inlcuding some time-dependent covariates (time, age, albuminuria); inclusion of covariates in model required p<0.25 in univariate analysis",mean 3.7 years,3.7,"Increased albuminuria and older age were the most consistent predictors of subsequent eGFR decline (slightly inaccurate interpretation of results, only adjusted at baseline and no interactions)",Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Tackled,mixed models,Tackled,separate analyses by subgroup,Tackled,Adjustment for covariates
Diggle PJ et al [23],Real-time monitoring of progression towards renal failure in primary care patients,Biostatistics,2015,1997-2007+,2000-2009,UK,primary care,INCLUDE; states EHRs used,large sample of primary care patients; longitudinal cohort study run by Salford Royal Hospital Foundation Trust; SRFT dataset,at risk for CKD,unspecified risk factors for CKD,67,1,52%,,risk prediction,to develop a system for real-time monitoring of the progression of renal failure,retrospective cohort study,1,all patients had at least 1 creatinine measure; all patients were analysed,22910,22910,not mentioned,100%,not available,,eGFR,MDRD,predicted percent change in eGFR per unit time,linear mixed model,progression,Predicted decline >5% per year,,outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"modelling of eGFR over time, based on explanatory variables, random intercept and continuous time non-stationary stochastic process; derive predictive probabilities for meeting progression criteria for referral to secondary care; log-transformed eGFR; based on preliminary modelling, piecewise linear model for effect of age with change of slope at age 56",median 4.5 years,4.5,Probabilistic predictive inference linked to clinical criteria is a useful component of surveillance but should not dictate clinical decision-making,Mentioned sample completeness but not implications,None,Not mentioned,None,"Mentioned data completeness, but not implications",mixed modelling,Partially acknowledged implications of data completeness,None,Tackled,Consideration of alternative error distributions,Tackled,Random effects; modelling of stochastic process (main),Tackled,Adjustment for covariates,N/A,
Johnson F et al [31],The impact of acute kidney injury in diabetes mellitus,Nephrology (asian pacific),2015,2009-2012,2010-2019,UK,multiple care settings,INCLUDE; states EHRs used,"complete lab data covering the whole of a single NHS Health Board, linked to EHRs to collect clinical data (co-morbidity, recovery of renal function, mortality)",Diabetes,community-acquired and hosptital-acquired AKI in diabetic and non-diabetic patients,75,,52%,,risk factor identification / causal inference,"To investigate the impact of AKI on progressive renal dysfunction in diabetes, compared to non-diabetes",retrospective cohort study,not mentioned,Require KDIGO-defined AKI lab data (2xEGFR); Require post-AKI follow-up data for at least 6 months; definition of progression not clear - assume based on at least 2 values and may or may not not include baseline,493,200,not mentioned,41%,not available,"493 patients identified with diabetes/no-diabrtes and AKI; 200 eligible for long-term analysis of CKD progression requiring ""follow up data available for at least 6 months"" (assume just 1 measure needed any time 6 months post-AKI) [41% analysed overall, 69% of those alive at 3 months]",eGFR,not specified,"rate of change in eGFR, not clearly defined",not stated,progressive chronic kidney disease,fall in eGFR > 5 units per year each year,"Very confusing definition of progression, with no information on how derived - assumed absolute value between baseline and other measures post-baseline",outcome,N/A,secondary outcome,simple statistical tests,difference in proportions chi-squared test,"Comparison of proportion developing progressive CKD in diabetes vs non-diabetes, chi-squared test; separate analysis for community-acquired and hospital-acquired AKI","(not clear);
6 months - 3 years",1.75,"Diabetic patients experience less severe AKI which translates into better recovery and less progressive CKD 
(I dispute this conclusion, as may not account for confounding by testing)",Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,separate analyses by subgroup,Not mentioned,None
Kaga M et al [24],Risk of new-onset dyslipidemia after laparoscopic adrenalectomy in patient with primary aldosteronism,World Journal of Surgery,2015,1998-2013,2000-2009,Japan,hospital,INCLUDE in sub-analysis; not clear if EHRs used,Medical records from Chiba University Hospital; Medical records searched from Aug 1998 to March 2013,Primary aldosteronism (PA),Japanese patients who underwent unilateral laparoscopic adrenalectomy (surgery) for primary aldosteronism (PA),54,1,49%,,risk factor identification / causal inference,"Secondary aim: 
assess changes in eGFR 12 months after surgery",retrospective cohort study,2,pre-operative and 12 months post-operative eGFR,not mentioned,57,not mentioned,not available,not relevant (complete case analysis),Patients with incomplete data or follow up < 12 months were excluded from the study but no data provided pre-exclusions,eGFR,new formula developed in Japanese population,absolute change in eGFR from baseline,paired t-test,none defined,none defined,"paired t-test, comparing baseline eGFR and eGFR 12 months after surgery (only before/after means presented, not changes from baseline)",outcome,N/A,secondary outcome,simple statistical tests,mean difference paired t-test,"paired t-test, comparing baseline eGFR and eGFR 12 months after surgery; subgroups: with/without preoperative dyslipidemia; with/without posteroperative eGFR<60",1 year,1,A significant decrease in eGFR was identified 12 months postoperatively in most patients,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Mentioned or partially addressed,Paired t-test,Mentioned or partially addressed,subgroup analyses,Not mentioned,None
Lai CL et al [25],Effects of atorvastatin and rosuvastatin on renal function in patients with type 2 diabetes mellitus,American Journal of Cardiology,2015,2000-2010,2000-2009,Taiwan,multiple care settings,INCLUDE; states EHRs used,"Taiwan National Health Insurance Pay-for-Performance program for diabetes mellitus database (2006-2009), linked to National Health Insurance claims database (2000-2010); 2 databases covering different data sources ",type 2 diabetes,type 2 diabetic patients aged 40-100 newly prescribed statins; ethnic Chinese population,62,,48%,,risk factor identification / causal inference,To investigate the effects of atorvastatin and rosuvastatin on renal function ,retrospective cohort study,2,Excludes subjects without baseline serum creatinine and at least one follow up measure,76078,5569,not mentioned,7%,not available,"76,708 met pop criteria; 32,363 excluded due to missing baseline creatining; 38,146 excluded due to no follow up data; Min 1 follow up creat required for analysis; no info on how many and when patients dropped out",eGFR,MDRD,absolute change in eGFR from baseline,absolute change from baseline at most recent eGFR,none defined,none defined,Baseline eGFR compared to last available follow up eGFR.,outcome,N/A,primary outcome,multivariate linear regression,linear regression,"Multiple linear regression, adjusting for confounders",Avg approx 7.5 months,0.65,Neither treatment with atorvastatin nor rosuvastatin was associated with a significant change of renal function in type 2 diabetes patients,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Tackled,subgroup analyses,Mentioned or partially addressed,Adjustment for covariates
Perotte A et al [26],Risk prediction for chronic kidney disease progression using heterogeneous electronic health record data and time series analysis.,Journal of american medical informatics association,2015,up to 2012; approx 12 years prior follow up,2000-2009,USA,multiple care settings,INCLUDE; states EHRs used,"clinical data warehouse for New York Presbyterian (NYP) hosptial; patients identified in outpatient primary care clinic; ""heterogeneous sources of data""; longitudinal inpatient and outpatient ",CKD,Incident CKD identified in an outpatient primary care clinic; excluding HIV and transplant patients,67,,35%,"excluded patients entering dataset with biochemical CKD stage 3 already, HIV positive and transplant patients",risk prediction,Primary aim = risk prediction; Secondary aim = risk factor identification; Methodological proof of principle study that aims to use both discrete elements and narratives from EHR to develop a model that predicts CKD progression in individual patients,retrospective cohort study,2,All patients meeting definition for inclusion. (No other criteria mentioned.) Cohort entry required eGFR<60 confirmed over 3 monhts,2908,2908,not mentioned,100%,not available,"Looks like 100% of patients identified with incident CKD were analysed; for patients with missing data, imputation rules were used based on prior values if available (LOCF) or imputation of population mean; 
timeframe and data availability for evaluation of progression not stated (results state 342 progression events = 12% of cohort)",eGFR,CKD-EPI,binary progression to stage 4 CKD,"2 x eGFR < 30, over >= 3 months",CKD progression from stage III to IV,binary progression to stage 4 (eGFR < 30 over >=3 months),timeframe and data availability for evaluation of progression not stated (results state 342 progression events = 12% of cohort),outcome,N/A,primary outcome,"Latent variable methods; 
Cox PH regression",Kalman filter time series model; Cox PH regression,"5 Cox proportional hazards models are presented: 2 simple; 3 2-step models based on pre-CKD time series models extracting patient latent factors for disease state at time developed CKD stage 3. Models: (1) baseline eGFR and demog; (2) 19 baseline labs and demog; (3) patient latent variables represengting history of 19 lab tests; (4) patient latent variables representing history of text used in patient notes; (5) combined 4&5. 
(Unsupervised) time series model = Kalman filter. (See paper for more details on natural language processing component).",not specified,not stated,A risk prediction model that takes longitudinal laboratory test results and longitudinal clinical documentation into consideration can statistically predict CKD progression from stage 3 to 4 more accurately than models that do not take all of these variables nor their longitudinal aspect into consideration.,Partially acknowledged implications of sample completeness,Other imputation methods (to avoid exclusions),Not mentioned,None,Partially acknowledged implications of data completeness,LOCF; imputation with mean (main),Partially acknowledged implications of data completeness,LOCF; imputation with mean (main),Not mentioned,None,Mentioned or partially addressed,random effects,Mentioned or partially addressed,Adjustment for covariates,N/A,N/A
Singh A et al [27],Incorporating temporal EHR data in predictive models for risk stratification of renal function deterioration,Journal of Biomedical Informatics,2015,not stated,not stated,USA,multiple care settings,INCLUDE; states EHRs used,EHR of Mount Sinai Medical Center ,multiple CKD risk factors,"mildly to moderately compromised kidney function (eGFR 45-90) as well as diagnoses of hypertension, diabetes or both",68,,40%,,risk prediction,compare and develop models to identify patients at high risk of loss of kidney function over the next year,retrospective cohort study,4,at least 2 years of medical history on record AND median eGFR in the first year of the database of 45-90; 2 measures prior to baseline and 2 measures post-baseline,not mentioned,6435,not mentioned,not available,not available,,eGFR,CKD-EPI,percent change in eGFR from baseline,percent drop in eGFR over 1 year (10% threshold; 20% threshold); percent change between median eGFR in 1-year before baseline and eGFR at 1 year,progression,2 progression thresholds: 10%; 20%,,outcome,N/A,primary outcome,GLMs,logistic regression,logistic regression; various different approaches for incorporating data captured during different time intervals; all predictors transformed to binary variables; patients split into training and holdout (80/20 split),not stated,not stated,Temporal eGFR data from EHRs can be used to predict loss of kidney function. Relative importance of different predictors varies over time. Multi-task learning based models can yield predictvie models with good performance.,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,LOCF,Partially acknowledged implications of data completeness,LOCF,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,Adjustment for covariates,N/A,
Vejakama P et al [28],Epidemiological study of chronic kidney disease progression: a large-scale population-based cohort study,Medicine,2015,1997-2011,2000-2009,Thailand,multiple care settings,INCLUDE; states EHRs used,"linkage of community health screening, hospitals and death registry data; includes out and in patients",CKD,"CKD diagnosed by KDIGO criteria, based on GFR and albuminuria",63,,36%,,risk factor identification / causal inference,to estimate CKD progression risk with comparisons by risk factors,retrospective cohort study,not mentioned,diagnosed with CKD and at least 3 months follow up (but assume based on end of study date rather than eGFR measure),not mentioned,32106,not mentioned,not available,not available,32106 patients identified with CKD based on eGFR and albuminuria; not stated how many excluded if less than 3 months of follow up,eGFR,CKD-EPI,percent change in eGFR from baseline,change in eGFR category with 25% or greater drop in eGFR from baseline,CKD progression,change of eGFR category with 25% or greater drop in eGFR from baseline,,outcome,N/A,primary outcome,competing risks survival models,competing risks survival models,"Analysis of time from CKD diagnosis to CKD progression or death, with censoring at end of follow up (Dec 2011); death treated as competing risk; time to CKD progression compared by diabetes status; GFR category as time-changing covariate; adjustments for baseline characteristics",4.5 years,4.5,"Diabetes, hypertension, CVD and albuminuria all increased the risk of CKD progression but BMI reduced the risk (not well explained)",Partially acknowledged implications of sample completeness,None,Tackled methodologically,competing risks survival models,"Mentioned care pathway follow up, but not data completeness",None,Tackled methodologically or not an issue,multiple imputation,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,stratified analysis,Tackled,Adjustment for covariates
Yun WS et al [29],Long-term follow up results of acute renal embolism after anticoagulation therapy,Annals of vascular surgery,2015,2006-2012,2000-2009,South Korea,not stated,INCLUDE in sub-analysis; not clear if EHRs used,,acute renal embolism (ARE),patients diagnosed with ARE by computed tomography (CT),61,,51%,,descriptive characterisation of changes in renal function,to investigate the clinical manifestations and long-term follow-up results of ARE,retrospective cohort study,2,not stated but reasonable to assume 2 measures (baseline + FU),47,31,not mentioned,66%,not available,47 patients meeting diagnosis criteria were enrolled; 31 patients had follow up creatinine; some LTFU data available but not specifically for patients analysed,serum creatinine,N/A,absolute change in serum creatinine,absolute change in serum creatinine,none defined,none defined,,outcome,N/A,secondary outcome,descriptive results only,descriptive result only; no statistical analysis,Change in serum creatinine level was summarised if follow up data were available,2.6 years,2.6,Renal function did not deteriorate during the measn follow up period. Mean change in serum creatinine level was 0.2 (-2.2 to 3.2) during the mean period of 31 months,Fully acknowledged implications of sample completeness,None,Mentioned losses to follow up but not implications,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,N/A,
Kim YG et al [32],Renal protective effect of DPP-4 inhibitors in type 2 diabetes mellitus patients: a cohort study,Journal of Diabetes Research,2016,2010-2015,2010-2019,South Korea,tertiary care,INCLUDE; states EHRs used,"clinical database of Ajou University Hospital, a tertiary hospital",Diabetes,Type 2 diabetes (ICD-10 E11) prescribed DPP-4i,59,,54%,,risk factor identification / causal inference,To determine the renoprotective effects of DPP-4i (oral antidiabetic drugs) ,retrospective cohort study,3,"urine albumin and creatinine at baseline, 1 year pre-meds and 1 year post-med initiation;
(additional analysis at 4 years)",not mentioned,414,not mentioned,not available,not relevant (complete case analysis),"414 analysed; no data pre-exclusions 
(78 patients in 4 year analysis after 4 years treatment)",eGFR,MDRD,absolute change in eGFR from baseline,paired t-test (of 2 x changes in eGFR),none defined,none defined,Changes in year before treatment compared to changes in first year of treatment (paired t-test for difference in CHANGES),outcome,N/A,secondary outcome,simple statistical tests,mean difference paired t-test,"eGFR changes compared before and after treatment with paired t-test at 1 year (main analysis) and at 4 years (secondary analysis); 4 year analysis stratified by level of proteinuria (normo, macro, micro)",2 years,2,"(1) eGFR was not changed 1 year after treatment compared with 1 year before treatment;
(2) [other analysis]: Administration of DPP-4i reduces urine albumin excretion and mitigates reduction of eGFR in T2DM patients",Not mentioned,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Mentioned or partially addressed,Paired t-test,Mentioned or partially addressed,Stratified analysis,Mentioned or partially addressed,None
Li XM et al [33],Clinicopathological characteristics and outcomes of light chain deposition disease: an analysis of 48 patients in a single Chinese center,Annals of Hematology,2016,2004-2015,2000-2009,China,hospital,INCLUDE in sub-analysis; not clear if EHRs used,"single center study; National Clinical Research Center of Kidney Diseases, Jinling hospital;
unlikely EHRs",light chain deposition disease (LCDD),"biopsy-proven light chain deposition disease (LCDD), a disorder which usually affects the kidneys",51,,60%,"Chinese, non-Caucasian",risk factor identification / causal inference,to identify risk factors for ESRD,retrospective cohort study,not mentioned,ESRD main outcome (may not require creat data); unclear what creat data required at baseline,48,44,not mentioned,92%,not available,"48 patients identified with LCDD; 44 patients ""with clinical follow up"" analysed but not clear criteria used for duration of follow up",serum creatinine,N/A,binary progression to threshold serum creatinine,time to ESRD; ESRD = SCr>6 mg/DL or RRT initiation,worsening renal function,>50% increase from baseline in SCr but less than 6 mg/dL,2 renal function outcomes: worsening renal function and progression to ESRD; analysis focus on time to ESRD so choose this as main renal outcome,outcome,N/A,secondary outcome,Cox PH regression,Cox PH regression,"Univariate and multivariate Cox regression models for time to ESRD; looks like 5 pre-specified variables included in models (includes biomarkers and treatment, no demographic variables); additional ROC analysis dichotomosing key predictor variables (SCr, urinary retinol-binding protein)",average 1.8 years,1.8,"In addition to serum creatinine, urinary retinol-binding protein was an independent risk factor for progression to ESRD",Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Not mentioned,None,Tackled,Stratified or separate analyses,Mentioned or partially addressed,Adjustment for covariates
Mirajkar N et al [34],The impact of bariatric surgery on estimated glomerular filtration rate in patients with type 2 diabetes: a retrospective cohort study,Surgery for Obesity and Related Diseases,2016,2005-2012,2000-2009,UK,outpatient diabetes,INCLUDE; states EHRs used,"electronic patient records; university hospital, patients undergoing surgery and attending diabetes clinic",type 2 diabetes,,53,,47%,,risk factor identification / causal inference,to investigate the impact of bariatric surgery in type 2 diabetes on eGFR,retrospective cohort study,2,patients included with baseline eGFR and another value during follow up,790,388,not mentioned,49%,not available,"565 patients met surgery criteria, of which 163 had baseline and FU eGFR and included in analysis; a further 225 patients were included as comparison group (which used data from a prospective study); assume complete data in the comparison group as no suggestion otherwise",eGFR,MDRD,absolute change in eGFR from baseline,calculated based on 2 measures,none defined,none defined,,outcome,N/A,primary outcome,simple statistical tests,simple non-parametric tests (Mann Whitney U),"Final eGFR analysed by linear regression, main exposure bariatric surgery, adjusting for age, baseline eGFR, baseline HbA1c, baseline BMI, HbA1c change, BMI change and follow up duration; subgroup analysis by baseline eGFR and type of surgical procedure",3.0 years,3,"Bariatric surgery results in significant improvements in eGFR in T2D patients, particularly those with eGFR<60, while routine care was associated with a decline in eGFR",Partially acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Mentioned or partially addressed,None,Fully acknowledged,None,Tackled,Stratified analysis,Tackled,Adjustment for covariates
Koraishy FM et al [35],"Rate of renal function decline, race and referral to nephrology in a large cohort of primary care patients",Family practice,2017,2008-2015,2010-2019,USA,primary care,INCLUDE; states EHRs used,electronic healthcare record data from primary care clinics,general population,all patients from primary care database with 2 x eGFR over 7-year period,60,,34%,black and white primary care adults only,risk factor identification / causal inference,To determine if rate of CKD progression was associated with referral and if race disparities in referral exist according to rate of CKD progression,cross-sectional study,2,at least 2 x eGFR at any time over the 7-year observation period,31569,2170,not mentioned,7%,not available,"29490 patients meeting population criteria (age 18+, race black or white); 2268 with 2 x eGFR; 2170 after exclusions for ESRD or missing demographic data",eGFR,MDRD and CKD-EPI (multiple labs used),rate of change in eGFR,"difference between consecutive measures only, divided by time elapsed",Fast CKD progression,"decline in eGFR >= 5 units per year, for any pairwise comparison of eGFR measures","Each pair of consecutive eGFR measures for a patient was used to compute the rate of change in eGFR per year; if a patient had any rate < - 5 units per year, they were considered fast progressors and otherwise slow progressors",exposure,nephrology referral,primary exposure,GLMs,logistic regression,"Logistic regression for nehrology referral, separate unadjusted models for each covariate and adjusted models including eGFR progression, race and all covariates; test for interaction between race and eGFR progression",not stated,not stated,"Fast CKD progression and black ethnicity were independently associated with increased odds of nephrology referral; no evidence of interaction between CKD progression and race. However, a large majority of fast progressors were not referred.",Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Tackled methodologically or not an issue,Complete data was available for all covariates,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,Interaction terms,Mentioned or partially addressed,Adjustment for covariates
Lv L et al [36],Persistent hematuria in patients with antineutrophil cytoplasmic antibody-associated vasculitis during clinical remission: chronic glomerular lesion or low-grade active renal vasculitis?,BMC Nephrology,2017,1996-2016,2000-2009,China,hospital,INCLUDE in sub-analysis; not clear if EHRs used,"Renal Division, Peking University First Hospital; retrospective analysis",Antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV),newly diagnosed AAV,59,,41%,,risk factor identification / causal inference,"to identify risk factors for renal outcome, and specifically the association of persistent hematuria",retrospective cohort study,not mentioned,assume 2 creat values required for analysis,219,208,not mentioned,95%,not available,11 patients excluded due to not meeting exposure criteria,eGFR,not specified,rate of change in eGFR,"final eGFR minus baseline eGFR, divided by time elapsed",end-point of renal survival,doubling in serum creatinine or halving of eGFR compared to baseline,renal outcome used for Cox reg analysis,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,multivariate Cox PH regression of renal outcome,median 3.1 years,3.1,Slope of decline in eGFR was significantly higher in patients with persistent hematuria and persistent hematuria was an independent risk factor for renal outcome in multivariate analysis,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Adjustment for covariates
Nishida Y et al [37],Comparative effect of calcium channel blockers on glomerular function in hypertensive patients with diabetes mellitus,Drugs in R&D,2017,2004-2012,2000-2009,Japan,hospital,INCLUDE; states EHRs used,"3 hospitals in Japan; inlcudes linked data for prescriptions, demographics, lab data, diagnoses",diabetes,diabetes and hypertension,66,,57%,,risk factor identification / causal inference,to study effect of drugs on longitudinal eGFR for up to 12 months after initation,retrospective cohort study,not mentioned,"5 timepoints studied, not clear how much data required for analysis",not available,1217,not mentioned,not available,not available,some numbers provided but can not differentiate between failure to meet population criteria and insufficient data for analysis,eGFR,Japanese Society of Nephrology formula,percent change in eGFR from baseline,"percent change between baseline eGFR and exposure period eGFR measures (every 3 months, max 12 months)",none defined,none defined,,outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"multivariate linear mixed model with compound symmetry, with observations grouped into time periods every 3 months up to 12 months, analysing effect of treatment group and treatment duration on eGFR; post-hoc multiple comparison test to compare least squares means among 5 treatment groups",12 months,1,Monotherapy with calcium channel blockers may have little influence on renal function and may be safely used in hypertensive patients with diabetes,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Tackled methodologically,mixed modelling,Partially acknowledged implications of data completeness,None,Not mentioned,None,Tackled,random effects,Tackled,Interaction terms,Tackled,Adjustment for covariates
Rincon-Choles H et al [38],Impact of uric acid levels on kidney disease progression,American journal of nephrology,2017,2005-2009,2000-2009,USA,outpatient,INCLUDE; states EHRs used,Cleveland Clinic CKD registry; a large EHR-based CKD registry,CKD,CKD stages 3-4,68,,58%,,risk factor identification / causal inference,"to assess the association between uric acid (UA), UA-lowering therapy (UALT) and renal outcomes",retrospective cohort study,3,2 x eGFR<60 over 90+ days plus at least one follow up eGFR; at least one outpatient UA within year prior to 2nd eGFR confirming CKD,29475,1676,not mentioned,6%,not available,38954 patients ahd eGFR 15-59; 29475 after exclusion for malignancy; 1947 after exclusion for missing UA; 1676 after exclusion for no follow-up eGFR,eGFR,CKD-EPI,Binary progression (eGFR changes/threshold combination),"(time to) 50% drop in eGFR or progression to ESRD; time since 2nd eGFR result confirming CKD (assume each follow up eGFR compared to 2nd eGFR, but not specified)",CKD progression,50% drop in eGFR or ESRD,,outcome,N/A,primary outcome,competing risks survival models,competing risks survival models,Cox models and competing risks models; multivariate adjsutment; death as competing risk; time-updated UALT; secondary analyses with linear mixed model of eGFR; mean value imputation for missing covariate data,median 2.8 years,2.8,"Hyperuricemia is associated with increased risk of progression to ESRD in CKD stages 3-4, but UALT does not ameliorate the risk, suggesting that the relationship is not causal",Partially acknowledged implications of sample completeness,None,Tackled methodologically,competing risks survival models,Not mentioned,None,Tackled methodologically or not an issue,imputation with mean,Not mentioned,None,Tackled,Outcome likely to identify real change,Mentioned or partially addressed,Adjustment for covariates,Tackled,Adjustment for covariates
Tsai CW et al [39],Serum Uric Acid and Progression of Kidney Disease: A Longitudinal Analysis and Mini-Review.,Plos one,2017,2003-2011,2000-2009,Taiwan,tertiary care ,INCLUDE; states EHRs used,electronic medical records from a tertiary medical centre in Taiwan,hyperuricemia,Chinese patients with hyperucricemia,63,,60%,excluded patients with dialysis/transplant within 30 days of baseline,risk factor identification / causal inference,investigating the association between uric acid levels and CKD progression,retrospective cohort study,3,>= 3 serum creatinine measures between 2003 to 2011; no RRT within 30 days of baseline,not mentioned,739,not mentioned,not available,not available,number of patients diagnosed hyperuricemia in medical centre EHR not provided; number analysed = 739; LTFU not mentioned,eGFR,MDRD,regression slope of eGFR,linear mixed model,progression to ESRD,eGFR<15 or dialysis,eGFR slope (linear mixed model),outcome,N/A,primary outcome,mixed modelling methods,linear mixed model, (1) adjusted linear mixed model investigating association between uric acid at baseline and longitudinal eGFR; (2) adjusted cox PH model for associatino between UA level and eGFR<15/dialysis/transplant,mean 4.25 years,4.25,"A higher uric acid level is associated with a significant rapid decline in eGFR, particularly in patients with proteinuria. ",Not mentioned,None,Not mentioned,None,Not mentioned,mixed modelling,Not mentioned,None,Not mentioned,None,Tackled,random effects,Tackled,Adjustment for covariates,Tackled,Adjustment for covariates
Yao X et al [40],Renal outcomes in anticoagulated patients with atrial fibrillation,Journal of the American College of Cardiology,2017,2010-2016,2010-2019,USA,not stated,INCLUDE; states EHRs used,"a large US administrative database linked to lab results; Optum-Labs Data Warehouse; private and Medicare patients of all ages and races throughout USA; ""diverse clinical practice settings"" but not clarified what they were",atrial fibrillation ,patients with non-valvular AF who started taking oral anticoagulants (new users),73,,55%,,risk factor identification / causal inference,to study the impact of oral anticoagulants on renal outcomes,retrospective cohort study,2,creatinine at baseline and follow up,not mentioned,9769,not mentioned,not available,not available,,eGFR,CKD-EPI,percent change in eGFR from baseline,percent change between baseline and each follow up eGFR measure,decline,30% decline in eGFR during follow up ,,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"Cox PH regression comparing different drugs for renal outcomes; main outcome was 30% decline in eGFR; other outcomes were creatinine doubling, AKI and kidney failure; inverse probability weighting was used to balance 60 baseline characteristics; model included date of treatment initiation to account for variable follow up bias",11 months,0.9,"non-vitamin K antagonist oral anticoagulatnts (NOACs), particularly dabigatran and rivaroxoban may be associated with lower risks of adverse renal outcomes than warfarin",Partially acknowledged implications of sample completeness,None,Tackled methodologically,competing risks survival models,Partially acknowledged implications of data completeness,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Propensity score methods
Beyer-Westendorf J et al [41],The CHA2DS2VASc score strongly correlates with glomerular filtration rate and predicts renal function decline over time in elderly patients with atrial fibrillation and chronic kidney disease,International Journal of Cardiology,2018,2008-2015,2010-2019,multiple European countries,primary care ,INCLUDE; states EHRs used,"2 primary care EHR cohorts:
(1) IMS-DA Germany;
(2) IMS-THIN UK",atrial fibrillation ,"patients with atrial fibrillation and CKD; patients with concurrent
diagnosis of AF and CKD stages 3-4 between 1 Jan 2008 to 31 Aug 2015 identified by ICD-10 codes; CKD checked by eGFR<60",79,,48%,,risk factor identification / causal inference,It is borderline whether this analysis represents risk factor identification and causal inference or risk prediction; however no new risk prediction model is generated and it seems more about causal inference or the utility of CHA risk prediction score for eGFR decline prediction,retrospective cohort study,1,at least one eGFR at or after baseline (diagnosis),not available,36779,not mentioned,53% German data; not available UK data,not available,"IMS-DA (Germany): 34,859 pts met diagnosis criteria; 19,192 of these had at least one eGFR at or after baseline; 18,539 qualified for analysis after excluding those previously diagnosed CKD stage 5, RRT or missing CHA2DS2VASc score and implausible eGFR values dropped;
IMS-THIN (UK): 20,780 pts met diagnosis criteria after exclusions performed by data provider (no pre-exclusion figrues provided); 18,240 qualified for analysis after implausible eGFR dropped",eGFR,CKD-EPI,regression slope of eGFR,linear mixed model,none defined,none defined,CKD-EPI used if eGFR not available from ERH but creatinine available,outcome,N/A,primary outcome,joint modelling,joint longitudinal survival model,"IMS-DA: straight-forward linear mixed model (no death data available); 
IMS:THIN: joint longitudinal survival model adjusting for drop out due to death;
BOTH models: fixed effects for polynomial time to allow non-linear trajectories over time plus random intercept and slope for eGFR to allow patient-specific baseline and trajectory;
eGFR modelled on log-scale for interpretability of results as percent change",median ~1.7 years,1.7,Increasing CHA2DS2VASc score is associated with lower baseline eGFR and predicts steeper decline in eGFR over time,Partially acknowledged implications of sample completeness,None,Tackled methodologically,Joint modelling of longitudinal changes and time to drop out,Tackled methodologically,mixed modelling,Not mentioned,None,Not mentioned,none,Tackled,random effects,Tackled,random effects,Tackled,Adjustment for covariates
Butt AA et al [42],"Effectiveness, treatment completion and safety of sofosbuvir/ledipasvir and paritaprevir/ritonavir/ombitasvir + dasabuvir in patients with chronic kidney disease: an ERCHVIES study",Alimentary pharmacology & therapeutics,2018,2014-2016,2010-2019,USA,multiple care settings,INCLUDE; states EHRs used,"the Electronically Retrieved Cohort of HCV Infected Veterans (ERCHIVES), a well-established national cohort of HCV-infected persons with age-sex-race-matched controls",hepatitis C virus,"patients with hepatitis C virus (HCV) and CKD, treated with HCV drugs [note: not all have eGFR<60]",62,1,97%,,risk factor identification / causal inference,"to determine efficacy, safety and tolerability of HCV treatment in CKD population (w.r.t. eGFR decline)",retrospective cohort study,3,"at least 2 eGFR 3 or more months apart at baseline, and eGFR estimations >= 12 weeks after therapy",37481,17624,7683,not available,44%,"Could not compute percent analysed as population criteria and data completeness criteria mixed up.
Note: number dropped out derived based on a table of number of patients with eGFR data at 3 months (12 weeks) [N=9941]",eGFR,CKD-EPI,absolute change in eGFR from baseline,absolute change from baseline calculated at any follow up time up to 12 weeks,"""progression""",> 10 unit decline in eGFR from baseline at any time during follow up,analysis stratified by treatment group and baseline CKD stage,outcome,N/A,primary outcome,simple statistical tests,difference in proportions chi-squared test,"Summary of proportion of patients who experienced a >10 unit eGFR decline at any time during follow up, stratified by treatment group and CKD stage",12 weeks,0.25,A decline in eGFR was observed in a substantial proportion of CKD patients.,Mentioned sample completeness but not implications,None,Mentioned losses to follow up but not implications,None,"Mentioned data completeness, but not implications",None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Stratified analysis;,Mentioned or partially addressed,None
Evans RDR et al [46],Clinical manifestations and long-term outcomes of IgG4-related kidney and retroperitoneal involvement in a United Kingdom IgG4-related disease cohort,Kidney International Reports,2018,2002-2018,2010-2019,UK,tertiary care,INCLUDE in sub-analysis; not clear if EHRs used,London-Oxford database of IgG4-RD patients and further review of medical records,IgG4-related disease with renal or retroperitoneal involvement,adults with a diagnosis of IgG4-RD between 2002 to 2018,58,,93%,,estimation of incidence/ prevalence,to describe the clinical characteristics and outcomes of renal and retroperitoneal involvement in a cohort of IgG4-RD patients from the UK,retrospective cohort study,2,"2 x eGFR, at presentation and during follow up ",28,24,0,86%,not relevant (complete case analysis),Complete data required for analysis; 4 patients excluded from outomce analysis due to incomplete followup,eGFR,MDRD,percent change in eGFR from baseline,percent change between baseline and latest eGFR,progressive renal impairment,>15% reduction from baseline,MDRD unadjusted for ethnicity,outcome,N/A,secondary outcome,descriptive results only,crude estimation / descriptive results only,"Prevalence of progressive renal impairment described overall and compared by use of immunosuppression, no p-value given but very small numbers",mean 5 years,5,"Renal function detiorated in 5 patients (20.8%), including 2 (8.3%) who reached ESRD;
Major organ manifestations of IgG4-RD should be identified early with prompt treatment to prevent progression to ESRD",Partially acknowledged implications of sample completeness,None,Mentioned losses to follow up but not implications,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Fully acknowledged,None,N/A,N/A
Lamacchia O et al [43],Normoalbuminuria kidney impairment in patients with T1DM: insights from annals initiative,Diabetology & Metabolic Syndrome,2018,2004-2011,2000-2009,Italy,outpatient diabetes,INCLUDE; states EHRs used,electronic medical records; patients attending diabetes clinics; Italian Association of Clinical Diabetologists database designed to monitor diabetes quality of care,diabetic kidney disease,type 1 diabetes and eGFR <60,63,,47%,covering one third of Italian diabetes clinics and deemed representative,risk factor identification / causal inference,investigating the role of albuminuria in kidney disease progression,retrospective cohort study,2,"at least one outpatient measurement of serum creatinine and albuminuria at baseline (any time, may vary by patient) and a re-evaluation of eGFR 4 years later",1395,582,not mentioned,42%,not available,1395 patients with T1DM extracted with at least 1 outpatient serum creatinine and albuminuria measure; subgroup of 582 with 4 years follow up analysed for CKD progression,eGFR,CKD-EPI,percent change in eGFR from baseline,percent eGFR change from baseline at 4 years,Loss of eGFR >30%,eGFR reduction >30% from baseline (at 4 years),Analysis focused on percent change between 2 measures over 4 years; supplementary analysis estimated absolute changes over time using individual linear regression,outcome,N/A,primary outcome,GLMs,logistic regression,"Logistic regression comparing loss of eGFR>30% at 4 years comparing albuminuria categories at baseline, with adjustment of selected covariates with p<0.05, including age and gender (complete case analysis but sample size not stated)",4 years,4,Patients with normoalbuminuria have a slower eGFR decline than those with albuminuria,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Outcome likely to identify real change,Mentioned or partially addressed,Adjustment for covariates,Mentioned or partially addressed,Adjustment for covariates
Park JM et al [56],Oncological and functional outcomes of laparoscopic radiofrequency ablation and partial nephrectomy for T1a renal masses: a retrospecive single-center 60 month follow up cohort study,Urological oncology,2018,2005-2014,2010-2019,South Korea,hospital,INCLUDE in sub-analysis; not clear if EHRs used,Chungnam National University Hospital; perioperative and postoperative data,renal cancer,all consecutive patients undergoing RFA or PN in 2005-2014,56,,74%,,risk factor identification / causal inference,to compare renal function outcomes of laparoscopic radiofrequency ablation (RFA) and partial nephrectomy (PN) for cT1a renal masses,retrospective cohort study,not mentioned,"patients included if ""min 2 years follow up"" but unclear what data required at baseline and follow up",not available,115,not mentioned,not available,not available,Reasons for exclusion stated but not clear how many excluded and for what reasons; inclusion requires min 2 years follow up; not clear how many patients dropped out after 2 years or why,eGFR,not specified,absolute change in eGFR from baseline,absolute difference between baseline eGFR and last follow up measure,none defined,none defined,,outcome,N/A,primary outcome,simple statistical tests,difference in means t-test, two sample t-test used comparing change in eGFR by treatment group,5.3 years,5.3,The treatment groups did not differ in terms of short-term and long-term changes in eGFR post-surgery,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None
VanWagner LB et al [44],Cardiovascular disease outcomes related to early stage renal impairment following liver transplantation,Transplantation,2018,2002-2012,2000-2009,USA,multiple care settings,INCLUDE; states EHRs used,local and national LT patient cohorts; includes outpatient and inpatient data; Vizient hospitalisation data linked to United Network for Organ Sharing,liver transplant patients,"adults undergoing liver transplant from Feb 2002 to Dec 2011, with follow up to 2012 ",56,,66%,note: population features based on local cohort analysis of 671 patients (used for trajectory analysis),identification of sub-populations AND risk factor identification / causal inference,(1) to characterise subgroups of eGFR trajectories; (2) to determine association between eGFR trajectory and CV complications,retrospective cohort study,3,3+ eGFR within 1 year of liver transplant,not mentioned,671,not mentioned,not available,not available,671 pts with 3+ eGFR within 1 year of LT,eGFR,MDRD,eGFR trajectory group,Latent mixture modelling,6 categories defined: normal-slow decrease; normal rapid decrease; mild stable; mild slow decrease; moderate stable; severe stable,(trajectory shape),,exposure (and outcome),composite CV event,primary exposure,latent variable methods,trajectory clustering using latent variables,"2-step modelling process:
(1) LCMM for eGFR trajectory group;
(2) logistic regression for CV risk, given eGFR trajectory group
[no joint modelling of drop out mentioned]",1 year,1,6 distinct eGFR trajectories were identified; greatest odds of 1-year CV complications were in the normal-rapid decrease group; more rapid declines in eGFR soon after liver transplant correlate with risk of adverse CV outcomes,Not mentioned,None,Partially acknowledged implications of losses to follow up,Sensitivity analysis in drop-out for RRT,Tackled methodologically,mixed modelling,Not mentioned,None,Not mentioned,None,Tackled,random effects,Tackled,latent class models,Tackled,Adjustment for covariates
Viazzi F et al [45],"Apparent treatment resistent hypertension, blood pressure control and the progression of chronic kidney disease in patients with type 2 diabetes",Kidney & Blood pressure research,2018,2004-2011,2000-2009,Italy,outpatient diabetes,INCLUDE; states EHRs used,EHRs; 90 diabetes centers with close patient follow up,type 2 diabetes,patients with type 2 diabetes and hypertension and stage 3 CKD (eGFR 30-690 at baseline),73,,50%,,risk factor identification / causal inference,to assess the role of apparent treatment resistent hypertension (aTRH) and time-updated BP control on progression of CKD,retrospective cohort study,6,"2 baseline eGFR measures + annual evaluation for 4 years; requiring complete data for BP, eGFR and albuminuria (in line with treatment protocols)",6999,2312,not mentioned,33%,not relevant (complete case analysis), 6999 met study pop criteria; exclusions for missing data; 2312 analysed for renal outcomes (Table 5) with complete data over 4 years,eGFR,CKD-EPI,percent change in eGFR from baseline,binary 30% reduction in eGFR from baseline at any time over 4 years (mixed models also performed separately),eGFR loss,30% eGFR reduction from baseline,,outcome,N/A,primary outcome,mixed modelling methods,logistic regression,"A little confusing but looks like main analysis is 30% change from baseline at any time, analysed by mixed logistic regression with multivariate adjustment (Table 5). Random effects for clinic; complete case analysis with multivariate adjustment. Additional linear mixed models performed for changes over time in eGFR.",4 years,4,In patients with stage 3 CKD the presence of aTRH entails a faster loss of eGFR. More effective prevention of aTRH should be implemented as this condition is associated with a burden of risk not modifiable by tight BP reduction,Mentioned sample completeness but not implications,None,"Mentioned care pathway follow up, but not losses to follow up",None,"Mentioned care pathway follow up, but not data completeness",None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Outcome likely to identify real change,Mentioned or partially addressed,Adjustment for covariates,Tackled,Adjustment for covariates
Horne L et al [47],Epidemiology and health outcomes associated with hyperkalemia in a primary care setting in England,BMC Nephrology,2019,2009-2013,2010-2019,UK,primary care,INCLUDE; states EHRs used,"Clinical Practice Research Datalink (CPRD), Hospital Episode Statistics (HES) and Office for National Statistics (ONS)",Hyperkalemia (HK),"adults with a HK event between 2009-2013, with at least 1 year follow up post event and with at least 1 visit to GP within the year",61,,48%,,estimation of incidence/ prevalence,"Note: evaluation of renal function secondary analysis; 
estimation of incidence of kidney function decline in patients with HK event, overall and by subgroup",retrospective cohort study,not mentioned,difficult to make assumption here; could be no measure since renal decline not focus; could be 1 measure (baseline),not mentioned,195178,not mentioned,not available,not available,"""195,178 with an index HK event analysed""; not clear how many had renal function data during follow up, if all made it into renal function analysis and how many LTFU or why",eGFR,MDRD,transition to CKD stage,"not stated (""decline from baseline CKD stage"")",declining kidney function,"""presence of diagnostic codes and/or eGFR showing a decline from baseline CKD stage""","assume any eGFR during follow up in a CKD stage lower than that at baseline considered ""declining kidney function"" but this was not clear",outcome,N/A,secondary outcome,Descriptive results only,Crude estimation / descriptive results only,"Crude incidence rates of ""declining kidney function"" summarised overall and by subgroup for patients with a HK event; no analysis except crude incidence rate and 95% CI",not stated,not stated,Higher rates of renal function decline were observed among those with more severe index HK,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Stratified analysis,N/A,N/A
Hsu TW et al [48],Comparison of the effects of dsnosumab and alendronate on cardiovascular and renal outcomes in osteoporotic patients,Journal of clinical medicine,2019,2005-2017,2010-2019,Taiwan,multiple care settings,INCLUDE; states EHRs used,"Chang Gung Research Database (CGRD) EHR dataset, containing diagnosis, prescription, laboratory tests in emergengy, inpatient and outpatient settings.",Osteoporosis,"adults age 20-89 at time of initiation of denosumab or alendronate between 2005-2017, excluding history of  RRT, MI/CHF/Stroke/CABG/PCI and cancer.",71,,17%,,risk factor identification / causal inference,To investigate whether denosumab may play a role in reducing CV events and renal function progression,retrospective cohort study,2,"at least 1 year admission records before treatment initiation, a baseline and follow up value for serum creatinine",not mentioned,5046,not mentioned,not available,not available,"Flow chart provided with exclusions, but not possible to compute percentage analysed from this since other population and analysis criteria applied after data availability critieria; As a guide, 57% of all patients treated with study drugs (before other exclusions) met eGFR exclusion critieria. Note, further exclusions due to propensity score matching.",eGFR,MDRD,percent change in eGFR from baseline,percent change between baseline and each eGFR measure,renal function decline,eGFR decline >= 30% from baseline,Text states eGFR was assessed every 6 months - it is not clear how many patients had missing data in each 6 month period over the maximum of 5 years follow up,outcome,N/A,secondary outcome,Cox PH regression,Cox PH regression,"Adjusted Cox PH regression for eGFR decline >= 30% from baseline, where eGFR was assessed every 6 months over max 5 year period. Reasons for censoring were death, loss to follow up, treatment change or latest date in dataset. Unclear how LTFU defined and data completeness for eGFR. Cohort propensity score matched at baseline (leading to smaller sample size). 7 x stratified analysis by subgroups. Adjustment for age, sex, baseline eGFR group, drug dose, co-morbidities, concomitant meds",max 5 years,5,"Denosumab may exert cardiovascular benefits but may have renal disadvantages in certain conditions and must be used with caution. Denosumab was associated with increased risk of renal function decline in males, those with renal sufficiency and AKI.",Mentioned care pathway and inclusion criteria but not sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Tackled,Stratified analysis,Tackled,Adjustment for covariates
Jalal K et al [49],Can billing codes accurately identify rapidly progressing stage 3 and stage 4 chronic kidney disease patients: a diagnostic test study,BMC Nephrology,2019,2007-2017,2010-2019,USA,outpatient,INCLUDE; states EHRs used,large insurer database containing outpatient creatinine measurements [not sure if also inpatient measures included],CKD,"CKD identified by 2x eGFR<60 or ICD coded CKD, using a large insurer database including outpatient serum creatinine measures",not stated; 77% aged over 65,,39%,,audit of care provision,evaluation of diagnostic accuracy of ICD-10 codes used to record CKD progression,retrospective cohort study,5,"eGFR-CKD: 2xeGFR<60 over 90+ days, no eGFR>60 
AND at least 3 years follow up AND at least 5 observations","28,762","10,927",not mentioned,38%,not available,"28,762 patients with eGFR-CKD (2xeGFR<60, over 90+ days, with no intervening measure >60) [14,657 with ICD-CKD];
10,927 met criteria for progression analysis",eGFR,CKD-EPI,regression slope of eGFR,linear mixed model,rapid progression,eGFR slope decline > 4 units/year,"longitudinal mixed model for patients with eGFR-CKD; 
fixed and random effects for time of measurement by quarter-year increment",outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,longitudinal mixed model to identify CKD progressors; agreement with ICD-progression compared in ROC analysis,at least 3 years,3,"ICD-10 codes display poor capacity to identify rapidly progressing CKD patients when compared to gold standard eGFR measures, demonstrating limitations of coding in CKD diagnosis",Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,mixed modelling,Not mentioned,None,Not mentioned,None,Tackled,Random effects,Mentioned or partially addressed,None,N/A,
Kim WJ et al [50],The role of a treat-to-target approach in the long-term renal outcomes of patients with gout,Journal of clinical medicine,2019,2007-2018,2010-2019,South Korea,not stated,INCLUDE; states EHRs used,"electronic medical record databases, no other details of care setting or database details",Gout,Patients with gout newly prescribed urate-lowering therapy (ULT) for at least 12 months,51,,97%,,risk factor identification / causal inference,To evaluate impact of treatment to target on changes in renal function ,retrospective cohort study,2,baseline eGFR and last on-treatment eGFR at least 12 months after treatment initiation,262,244,not mentioned,93%,not available,244 analysed; 18 no available data,eGFR,MDRD,absolute change in eGFR from baseline,absolute difference from baseline,none defined,none defined,"ANCOVA comparing change in eGFR across 3 treatment groups, adjusted for baseline eGFR groups",outcome,N/A,primary outcome,GLMs,logistic regression,"main: logistic regression for improvement in eGFR (last eGFR measure > baseline);

other: ANCOVA comparing change in eGFR across 3 treatment groups, adjusted for baseline eGFR groups (stage 1, stage 2, stage 3)",~2 years,2,"Improvement in eGFR for those treated-to-target compared to not meeting target, for patients with baseline CKD stage 3",Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,"Mentioned data completeness, but not implications",None,Not mentioned,None,Not mentioned,None,Tackled,ANCOVA,Tackled,Adjustment for covariates
Lai YJ [51],Effect of weight loss on the estimated glomerular filtration rates of obese patients at risk of chronic kidney disease: the RIGOR-TMU study,"Journal of Cachexia, Sarcopenia and Muscle",2019,2008-2016+,2010-2019,Taiwan,multiple care settings,INCLUDE; states EHRs used,"weight reduction intervention on GFR in obese patients with renal impairment TMU study (large long-term propensity matched cohort study for patients who registered at weight reduction centres); includes outpatient, inpatient, lab results, drug prescriptions",Obesity,"Obese patients with CKD, registered at weight reduction centres, aged 20-70",37,,40%,,risk factor identification / causal inference,To investigate the effects of bariatric surgery on eGFR,retrospective cohort study,3,"Excludes patients with misssing ""CKD risk"" (2 x eGFR (or evidence of proteinuria) over 3+ months); propensity score matched only",4332,1620,not mentioned,not available,not available,"Data completeness criteria applied prior to PS matching critiera; Prior to matching, 4059 of 8391 (48.4%) excluded due to missing ""CKD risk""",eGFR,Taiwan CKD-EPI,percent change in eGFR from baseline,calculated percent change for each eGFR measure during 1-year follow up period,CKD progression,eGFR decline from baseline >= 25%,CKD progression: eGFR decline >= 25% at any time during a 1-year follow up period,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"Cox regression for time to eGFR decline >= 25%, comparing bariatric surgery as primary exposure; additional subgroup analyses in proteinuric, diabetic and hypertensive patients.",max 1 year,1,Bariatric surgery was associated with eGFR preservation in all obese patients and particularly those with moderate-to-high CKD risks,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Propensity score methods
Leither MD et al [52],The impact of outpatient acute kidney injury on mortality and chronic kidney disease: a retrospective cohort study,NDT,2019,not stated,not stated,USA,multiple care settings,INCLUDE; states EHRs used,"Fairview Health Services, large health system in Minnesota; EHRs covering inpatients and outpatients (6 hospitals and medical centers, 40 primary care clinics, 55 specialty clinics), with linkage to death data",general population,"all adult patients receiving primary care in the health system, with 2 primary care visits separated by at least 18 months",46,,43%,,risk factor identification / causal inference,to investigate whether outpatient AKI is associated with all-cause mortality and renal events,retrospective cohort study,1,"no baseline eGFR data was required for analysis; patients were categorised according to AKI status, with one value being no creatinine data. 49% had no baseline creatinine data.","384,869","196,209",not mentioned,51%,not available,"Patients with no creatinine data could not be analysed for renal outcome but were analysed for mortality outcome; Unclear how many patients LTFU and how this may affect evaluation of renal outcome; Patients were censored at ""last data point in HER"" (but unclear if requires creatinine data)",eGFR,MDRD,Binary progression (eGFR changes/threshold combination),50% change from baseline and eGFR <30 (baseline = last value during exposure period; current eGFR based on at least 2 measurements during follow up period),renal event,at least 50% decline from baseline (last value during 18-month exposure period) to eGFR < 30 (on at least 2 measurements during follow up period),,outcome,N/A,secondary outcome,Cox PH regression,Cox PH regression,"Cox PH regression for time to renal event, with adjustment for covariates selected a-priori (comorbidities, vitals, labs, hospitalisation events, previous AKI events); additional adustment for (binary) missingness of each covariate; main exposure was outpatient AKI group (no AKI, AKI with recovery, AKI without recovery, AKI with repeat Cr and no Cr during exposure period)",average 5.3 years,5.3,"Outpatient AKI is more prevalent than inpatient AKI and is a risk factor for all-cause mortality and renal events, even amonghtose who recover kidney function",Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Tackled methodologically or not an issue,Adjustment for missingess,Not mentioned,None,Tackled,Outcome likely to identify real change,Mentioned or partially addressed,Adjustment for covariates,Tackled,Adjustment for covariates
Liu D et al [53],Serum immunoglobin G provides early risk prediction in immunoglobin A nephropathy,International Immunopharmacology,2019,2009-2014,2010-2019,China,hospital,INCLUDE in sub-analysis; not clear if EHRs used,Second Xiangyu Hospital,IgA nephropathy,,32,,44%,,risk factor identification / causal inference,investigating the role of serum IgG in progression of IgA,retrospective cohort study,not mentioned,"follow up time of at least 2 years, baseline serum IgG data, number of creat measures required unclear",620,455,not mentioned,73%,not available,"620 patients eligible before missing data exclusions; 455 patients with baseline IgG and 2 years follow up; follow up criteria, censoring criteria and how many patients LTFU are unclear",eGFR,MDRD,Binary progression (eGFR changes/threshold combination),Time to ESRD or an irreversible 50% eGFR reduction; Criteria for ESRD not provided; calculation of reduction of eGFR not provided,not named,ESRD or an irreversible 50% eGFR reduction,,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"multivariate Cox PH regression, primary exposure IgG, different models with adjustments for traditional risk factors and significant risk factors",median 3.7 years,3.7,A decreased serum IgG level is independently associated with a poor renal outcome in IgAN patients,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Outcome likely to identify real change,Tackled,Adjustment for covariates,Tackled,Adjustment for covariates
Morales-Alvarez MC et al [54],Renal function decline in latinos with type 2 diabetes,Kidney International Reports,2019,2002-2015,2000-2009,USA,outpatient diabetes,INCLUDE; states EHRs used,Joslin Diabetes Center; outpatient visits data,type 2 diabetes,patients with type 2 diabetes and Hispanic background,57,,43%,Hispanic only,risk factor identification / causal inference,undestanding risk factors of progression,retrospective cohort study,2,decline derived using first and last eGFR; more than one medical visit; more than one creatinine; time interval not specified,916,594,not mentioned,65%,not available,"920 patients ""met inclusion cirtieria""; 298 excluded due to missing visits or creatinine; 4 excluded due to ESRD at baseline; 24 patients not accounted for in exclusion reasons",eGFR,not specified,rate of change in eGFR,"difference between first and last eGFR, divided by time elapsed (reported in units per year)",rapid progression,loss of >3 units per year in eGFR,,outcome,N/A,primary outcome,simple statistical tests,difference in means t-test,"descriptive comparison of patient characteristics comparing progressors and non-progressors; t-tests and chi-squared tests for comparisons; (other analysis included Cox PH regression for time to development of CKD, but cohort was not restricted to GFR>60)",not stated,not stated,"Renal function decline in Latinos is associated with expected but modifiable variables, such as uncontrolled diabetes, uncontrolled hypertension and being overweight",Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,Adjustment for covariates
O’Neill RA et al [55],Evaluation of long-term intravitreal antivascular endothelial growth factor injections on renal function in patients with and without diabetic kidney disease,BMC Nephrology,2019,2012-2018,2010-2019,UK,multiple care settings,INCLUDE; states EHRs used,"electronic healthcare records; Belfast Health and Social Care Trust; EHRs from the Northern Ireland Electronic Care Record (ECR) [contains data from multiple settings: GP, hospital, clinics]",diabetes,treatment for diabetic macular oedema,64,,59%,,risk factor identification / causal inference,to study effect of anti-VEGF injections on renal function in patients treated for diabetic macular oedmea (DMO),retrospective cohort study,not mentioned,"renal function measures required prior to first injection and after last injection; exclusions for ""insufficient"" number of renal function measures, not clarified in detail",not available,85,not mentioned,not available,not available,90 patients met population criteria; 5 excluded due to AKI; analysis requires creatinine after last injection but no mention of any losses to follow up,eGFR,CKD-EPI,regression slope of eGFR,individual linear regression,none defined,none defined,looks like individual linear regression models were used to compute slope of change in eGFR over time,outcome,N/A,primary outcome,multivariate linear regression,linear regression,"simple and multivariate regression model for estimating the association between change in eGFR slope and number of injections (assume 2-step process used: (1) estimate change in eGFR; (2) estimate association), adjustments for diabetes, CVD, hypertension and PPIs but not demographics",2.6 years,2.6,No association between increasing number of injections and rate of eGFR decline; long-term treatment does not significantly impact change in eGFR and/or ACE,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Tackled,Distributional checks,Not mentioned,None,Mentioned or partially addressed,descriptive comparison by important subgroup,Mentioned or partially addressed,Adjustment for covariates
Posch F et al [57],"Longitudinal kidney function trajectories predict major bleeding, hospitalisation and death in patients with atrial fibrillation and chronic kidney disease",International Journal of Cardiology,2019,2009-2015,2010-2019,UK,primary care,INCLUDE; states EHRs used,"primary care electronic database, IMS-THIN; 616 GP practices",atrial fibrillation ,newly diagnosed concurrent atrial fibrillation and CKD stage 3/4,80,,45%,elderly cohort,risk factor identification / causal inference,to idnetify the role of kidney function decline to predict individual patient adverse outcomes (some discussion around potential for individual risk prediction but not achieved in this specific research; instead causal inference statements made about effects of eGFR changes on outcomes),retrospective cohort study,1,"exclusions for ""missing eGFR values"" and ""follow-up <6 months""; details of criteria not clarified; min number of measures identified from statistics on range of number of eGFR measures per patient",20780,18240,not mentioned,88%,not available,"20780 patients met morbidity criteria; 2540 patients excluded due to missing data; death was only reason provided for LTFU, no other losses to follow up (or poor/reduced follow up) discussed; rates of LTFU prior to end of study date not provided with data capture simply capped at last entry in database",eGFR,not specified,rate of change in eGFR,annualised absolute change in eGFR after baseline,none defined,none defined,,exposure,"(time to) stroke, major bleeding, hospitalisation, all-cause death",primary exposure,joint modelling,joint longitudinal survival model,"Joint model specified as follows: (1) linear mixed model for eGFR with random intercept and slope, (2) Weibull proprtional hazards model for time to 4 clinical outcomes, (3) ""1st derivative"" specification of association parameter alpha; (4) un-structured variance-covariance matrix;
(other simpler models included Cox regression, Fine & Gray competing risk regression, Semi-Markov Multistate Models); [joint models did not adjust for baseline variables]",median 1.9 years (between first and last GFR); longer for outcomes follow up,1.9,Declining kidney function is a critical determinant of unfavourable outcomes in patients with AF and CKD. Longitudinal kidney function trajectories may enable a much more individualised prediction of adverse outcomes.,Fully acknowledged implications of sample completeness,Detailed/ comprehensive database of EHRs used ,Tackled methodologically,Joint modelling of longitudinal changes and time to drop out,Tackled methodologically,mixed modelling,Not relevant (no covariate analysis),None,Not mentioned,None,Tackled,Random effects,Mentioned or partially addressed,random effects,Mentioned or partially addressed,None
Posch F et al [58],Exposure to vitamin k antagonists and kidney function decline in patients with atrial fibrillation and chronic kidney disease,Research and practice in thrombosis and haemostasis,2019,2009-2015,2010-2019,Germany,primary care,INCLUDE; states EHRs used,primary care electronic database; IMS Disease Analyzer Germany (IMS-DA); data routinely collected from ~1300 primary care physicians in Germany,atrial fibrillation ,newly diagnosed concomitant atrial fibrillation and CKD stage 3/4,78,1,52%,,risk factor identification / causal inference,to evaluate the impact of vitamin k exposure on kidney function,retrospective cohort study,1,Require an eGFR measure at or after baseline; require baseline CHA2DS2VASc score,37476,14432,not mentioned,39%,not available,"37,476 met AF/CKD criteria, of which 14,432 met inclusion criteria (mostly based on data completeness)",eGFR,not specified,regression slope of eGFR (absolute scale and percent scale),linear mixed model (eGFR scale and log-eGFR),not named,30% decline in eGFR during follow up at least once,,outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"linear mixed model with random intercept and slope and linear, quadratic and cubic terms for time, unadjusted and adjusted for age, CHA2DS2VASc score, and baseline eGFR; interaction terms to allow different exposure effects on slope; both absolute eGFR scale and log-eGFR scale used providing linear and percent interpretations of regression slopes; sensitivity analysis with propensity score adjustment; further sensitivity analysis for effect modification of impact of treatment on slope by CHA-score; complete case analysis to avoid added complexity of MI",median 1.4 years,1.4,"VKA use is associated with accelarated eGFR decline, independent of age, CHA-score and baseline eGFR",Fully acknowledged implications of sample completeness,None,Fully acknowledged implications of losses to follow up,None,Tackled methodologically,mixed modelling,Fully acknowledged implications of data completeness,None,Not mentioned,None,Tackled,Random effects,Tackled,Interaction terms,Tackled,Adjustment for covariates
Spanopoulos D et al [59],Temporal variation of renal function in people with type 2 diabetes mellitus: a retrospective UK clinical practice research datalink cohort study,"Diabetes, Obesity and Metabolism",2019,2009-2016,2010-2019,UK,primary care,INCLUDE; states EHRs used,"CPRD including primary care data for 10 million people covering 10% of UK practices and population representation (age, sex, eth, BMI) with linkage to HES to improve ethnicity coding",type 2 diabetes,type 2 diabetes diagnosis between 2009-2011,62,,56%,,descriptive characterisation of changes in renal function,to characterise longitudinal variability of eGFR (to support appropriate dosing of glucose-lowering medications that depends on renal function),retrospective cohort study,6,require 1 measure of creatinine after T2DM diagnosis and at least one measure in 5 yearly intervals post first creatinine; at least 12 months registered at practice,46813,7766,not mentioned,17%,not available,46813 had newly diagnosed T2DM; 7766 met inclusion criteria (based on creatinine data); drop out not possible since sampling based on 5 years FU (survival bias),eGFR,CKD-EPI,transition to CKD stages,raw change between eGFR measures,none defined,none defined,,outcome,N/A,primary outcome,descriptive results only,descriptive result only; no statistical analysis,"Descriptive analysis of number of times a patients changes CKD category, compared each year over a 5 year period, with comparison by basleine CKD stage; additional analysis comparing baseline and year 5 transition probabilities",5 years,5,"There was considerable variation between GFR categories and individuals, highlighting issues for prescribing glucose-lowering therapies requiring dose adjsutment based on GFR. Regular monitoring is important.",Partially acknowledged implications of sample completeness,None,Fully acknowledged implications of losses to follow up,None,Partially acknowledged implications of data completeness,None,Partially acknowledged implications of data completeness,Data linkage to improve data completeness,Not mentioned,None,Mentioned or partially addressed,None,Mentioned or partially addressed,stratified analysis,N/A,
Wang Y et al [60],Implications of a family history of diabetes and rapid eGFR decline in patients with type 2 diabetes and biopsy-proven diabetic kidney disease,frontiers in Endocrinology,2019,2007-2017,2010-2019,China,hospital,INCLUDE; states EHRs used,EHRs for West China Hospital,diabetic kidney disease,type 2 diabetes and biopsy-proven diabetic kidney disease (DKD),52,,76%,,risk factor identification / causal inference,to identify risk factors for rapid decline in eGFR,retrospective cohort study,3,at least 3 creatinine measures post-biopsy; all creat performed in same hospital; follow-up min 1 year,not mentioned,128,not mentioned,not available,not available,no info on patients excluded from analysis and why; min 1 year follow up required but no data or reasons on LTFU,eGFR,CKD-EPI,regression slope of eGFR,individual linear regression,relatively rapid eGFR decline,eGFR slope worse than median slope (-8.1 units per year),,outcome,N/A,primary outcome,GLMs,logistic regression,Multivariate logistic regression to identify risk factors for a rapid decline in eGFR; not explicitly stated but looks like only variables with statistically significan univariate association were included in multivariate model,2 years,2,"A family history of diabetes is an independent risk factor for rapid decline in eGFR, suggesting that early diagnosis and treatment is important for these patients",Partially acknowledged implications of sample completeness,None,Not mentioned,None,"Mentioned data completeness, but not implications",None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Tackled,Stratified analysis,Tackled,Adjustment for covariates
Yoo H et al [61],Effects of sarpogrelate on microvascular complications with type 2 diabetes,International Journal of Clinical Pharmacy,2019,2010-2015,2010-2019,South Korea,tertiary hospital,INCLUDE; states EHRs used,a 1108-bed tertiary university hospital; electronic medical records; clinical data warehouse,type 2 diabetes,type 2 diabetes treated with antidiabetics,55,,59%,,risk factor identification / causal inference,"to investigate the effects of sarpogrelate with metformin-based antidiabetic therapy on microvascular complications including nephropathy, neuropathy and retinopathy",retrospective cohort study,not mentioned,"patients excluded ""without lab values for serum creatinine"" (no further details); propensity score matched only",5334,478,not mentioned,not available,not available,"6643 patients with DM treated with antidiabetics; 2952 excluded due to no creatinine labs; 1309 exclude due to co-morbidity exclusions; only 478 included after propensity score matching. Could not compute percent analysed, as data compelteness and population critiera mixed up.",eGFR,MDRD,percent change in eGFR from baseline,percent change between baseline and follow up measures,CKD progression,"30% reduction from baseline, 30-50% and 50%",,outcome,N/A,secondary outcome,KM / life-table,kaplan meier estimation + log-rank test,Propensity score matched cohort comparing sarpogrelate vs no sarpogrelate; kaplan meier and log-rank test for 30% reduction in eGFR outcome,5.7 years,5.7,No significant difference in time to 30% reduction in eGFR by treat groups,Fully acknowledged implications of sample completeness,None,Not mentioned,None,"Mentioned data completeness, but not implications",None,Fully acknowledged implications of data completeness,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,propensity score matching,Tackled,Propensity score methods
Zhao J et al [62],Predicting outcomes of chronic kidney disease from EMR data based on random forest regression,Mathematical Biosciences,2019,2009-2017,2010-2019,USA,primary care,INCLUDE; states EHRs used,"dataset of EHRs for 120,495 patients aged 20-80; region receiving primary care from Sanford Health; Sanford EMR database",general population,all patients in primary care database with 3+ eGFR measures over 7 years,57,,44%,99.9% white,risk prediction,to develop and validate a prediction model of eGFR to be used in clinical practice,cross-sectional study,3,"at least 3 eGFR values between 2011 to 2017; to predict an individual's eGFR in 2015, a value in 2015 is required plus two further values from 2011-2014; similar criteria apply to 2016 and 2017",120495,61740,not mentioned,51%,not available,120495 patients aged 20-80 receiving primary care from Sanford Health (unclear if requires at least 1 eGFR over 10 years); 61740 patients with at least 3 eGFR measures over 7 years,eGFR,CKD-EPI,eGFR prediction at multiple time points,"eGFR value in years 2015, 2016, 2017",none defined,none defined,,outcome,N/A,primary outcome,machine learning methods,random forest regression,"Random forest regression to predict 1-year, 2-year and 3-year eGFRs from historical eGFR records between 2011-2014, with adjustment for covariates (baseline eGFR, age, sex, ethnicity, BMI, hypertension, diabetes, obesity); sci-kit learn package used for all models; estimated eGFR values classified in CKD stages;",not stated,not stated,"A model using real-world electronic medical records data can accurately predict future kidney functions and provide clinical decision support; R-squared goodness-of-fit was high (~0.95), but decreased over the 3 years; Previous eGFR was most important predictor/feature, followed by BMI, then age. Other features were much less impactful (race, eth, hyp, diabetes, obesity)",Fully acknowledged implications of sample completeness,None,Not mentioned,None,Partially acknowledged implications of data completeness,Imputation with median,Not mentioned,None,Tackled,Distributional checks,Mentioned or partially addressed,None,Mentioned or partially addressed,Adjustment for covariates,N/A,
Cabrera CS et al [67],Impact of CKD Progression on Cardiovascular Disease Risk in a Contemporary UK Cohort of Individuals With Diabetes,KI reports,2020,2005-2015,2010-2019,UK,primary care,INCLUDE; states EHRs used,CPRD GOLD,CKD,Incident CKD patients in population of type 2 diabetes,71,,46%,,risk factor identification / causal inference,to evaluate the association between CKD progression and risk of CVD,retrospective cohort study,3,Min 2 x eGFR required for baseline eGFR assessment; Min 1 x eGFR post-baseline; Exclude if >2 year gap to next eGFR at any time,53174,30222,not available,57%,not available,"From flow chart, 56,504 CKD patients with prior diabetes diagnosis meeting population critieria, but 3,330 met CVD exclusion criteria -> approx 53,174 target pop but could be a bit less;
Note: patients excluded if no eGFR post-baseline or >2 year gap between follow-up measures;
Patients with CVD events in 2-year baseline period post-index date were excluded (survival bias?);
No data on drop-outs after inclusion criteria met",eGFR,CKD-EPI,regression slope of eGFR,individual linear regression (time-updated),"""Updated eGFr slope <-3""? ",time-updated indiviudal linear regressions over overlapping 2-year periods,,exposure,"CV outcomes: HF, MI, stroke + composite (MACE plus)",primary exposure,Cox PH regression,Cox PH regression,"Cox PH regression with time-updated mean eGFR and time-updated eGFR slope, adjusted for confounders; time-updated eGFR slopes were calculateed throughout follow-up in overlapping 2-year periods",4.3 years,4.3,Faster eGFR decline was asscoiated with increased CV risk,Mentioned sample completeness but not implications,None,Not mentioned,None,Mentioned data completeness but not implications,None,Partially acknowledged implications of data completeness,Adjustment for missingness,Not mentioned,None,Mentioned or partially addressed,None,Tackled,Adjustment for covariates,Tackled,Adjustment for baseline confounders
Cleary F et al [68],Feasibility of evaluation of the natural history of kidney disease in the general population using electronic healthcare records,Clincal Kidney Journal,2020,2008-2016,2010-2019,UK,primary care,INCLUDE; states EHRs used,primary care EHRs; NCKDA database,general population,Patients with eGFR tests in primary care,not stated,,48%,adults with at least 3 eGFR test results in primary care data,descriptive characterisation of changes in renal function,to estimate slope of decline in population subgroups and evaluate sensitivity of estimation to measurement methods,retrospective cohort study,3,At least 3 valid eGFR results required for analysis; Sensitivity analysis based on data availability restrictions,6513000,1597629,39091,25%,2.40%,Drop-out assessed as no eGFR test in last 3 years despite 3x prior eGFR,eGFR,MDRD,regression slope of eGFR,individual linear regression,none defined,none defined,,outcome,N/A,primary outcome,linear regression,linear regression,"individual linear regression of eGFR, aggregated in all adults and separately in subgroups",5.7 years,5.7,Evaluation of long-term changes in renal function will be possible in high risk patients if creatinine calibration problems are overcome,Fully acknowledged implications of sample completeness,None,Fully acknowledged implications of losses to follow up,None,Fully acknowledged implications of data completeness,None,Not relevant,None,Not mentioned,None,Fully acknowledged,None,Tackled,subgroup analyses,N/A,N/A
Faraj KS et al [69],The effect of urinary diversion on long-term kidney function after cystectomy,Urologic oncology,2020,2007-2018+,2010-2019,USA,outpatient,INCLUDE; states EHRs used,EHR of Mayo Clinic,urinary system disorders,patients underoing cystectomy with urinary diversion (surgery) with any indication,72,,79%,,risk factor identification / causal inference,to compare long-term kidney function after cystectomy by type of urinary diversion ,retrospective cohort study,not stated,at least 1 x follow up creatinine; but unclear if require pre-op creatinines,573,563,not available,98%,not available,573 patients identified for EHR search; 5 excluded for dialysis and 5 excluded due to no follow up creatinine; 563 patients in analysis; no data on dropouts,eGFR,CKD-EPI,regression slope of eGFR,linear mixed model,none defined,none defined,,outcome,N/A,primary outcome,mixed modelling methods,linear mixed model,"linear mixed model with inverse probability of treatment weighting (IPTW), adjusted for covariates; sensitivity analyses in CKD-EPI vs MRDR and eGFR>/<40 and in benign indications",median 3.9 years,3.9,Kidney deterioration may be lower over time for neobladder (NB) than ideal-conduit (IC) urinary diversion.,Tackled methodologically,Detailed/ comprehensive database of EHRs used,Mentioned losses to follow up but not implications,None,Tackled methodologically,Mixed modelling,Not mentioned,None,Not mentioned,None,Tackled,random effects,Tackled,subgroup analyses,Tackled,IPTW
Inaguma D et al [70],Increasing tendency of urine protein is a risk factor for rapid eGFR decline in patients with CKD: A machine learning-based prediction model by using a big database,PLOS ONE,2020,2004-2019,2010-2019,Japan,outpatient,INCLUDE; states EHRs used,"big hospital database; electronic medical records system; healthcare setting unclear; ""outpatients suffering from various diseases involving CKD""",CKD,CKD = eGFR < 60 and/or urine protein >1+ (dipstick method) over 90+ days,69,,58%,,risk prediction,to predict rapid decline in kindney function,case control study,not stated,"unclear; patients exhibiting (and instances of) eGFR decline were identified first, and then non-declining patients/instances were matched to declines using some unclear matching criteria",19894 (pre-PS matching),9911,not relevant (complete case analysis),not available,not relevant (complete case analysis)," 19,894 in database met CKD criteria; 9,911 met analysis criteria based on having rapid decline or being matched to a patient with rapid decline; Unclear what data availability was required to identify patients with decline / no decline",eGFR,not specified,"Rate of percentage change in eGFR, not clearly defined","decline of 30% or more in eGFR within a period of 2 years, unclear how derived",rapid eGFR decline,"decline of 30% or more in eGFR within a period of 2 years, unclear how derived","""to avoid temporal spikes"", averaged eGFR measures over a period of 90 days (not clear how incoporated in derivation)",outcome,N/A,primary outcome,machine learning algorithms,Logistic regression; Random forest regression,2 machine learning algorithms: logistic regression and random forest regression. Case control design where decliners and non-decliners identified first and look back at prior characteristics over various time-periods,not stated,not stated,Model AUCs of 0.71 and 0.73; Urine protein is a prominent risk factor for decline,Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Tackled methodologically,LOCF,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,Features in machine learning classification,N/A,N/A
Lee JS et al [63],Recovery of renal function in patients with lupus nephritis and reduced renal function: the beneficial effect of hydroxychloroquine,Lupus,2020,1995-2018,2000-2009,South Korea,tertiary hospital,INCLUDE; states EHRs used,electronic medical records; tertiary referral hospital,lupus nephritis,lupus nephritis and eGFR<60,38,,28%,,risk factor identification / causal inference,to investigate renal function recovery and associated factors,retrospective cohort study,2,one baseline eGFR and one eGFR at 6 months,90,90,0,100%,not relevant (complete case analysis),90 patients with diagnosed LN and eGFR<60; percentages provided for outcomes suggest all analysed at 6 months; further data that 87 patients had data for one year strengthens this assumption,eGFR,MDRD,binary progression to threshold eGFR,"threshold eGFR value at 6 months; eGFR >60 at 6 months post-baseline, where baseline eGFR<60",Renal function recovery,eGFR > 60 six months post-baseline,,outcome,N/A,primary outcome,GLMs,logistic regression,logistic regression analysis with stepwise elimination to identify factors associated with renal function (variables included if p<0.15 on univariate analysis but require p<0.05 to remain in multivariate analysis); both univariate and multivariate analysis results provided,6 months,0.5,"Renal function recovery occurred in 51% of patients, and factors associated with renal function recovery were use of hydroxychloroquine, shorter duration of LN and higher grade tubular atrophy",Fully acknowledged implications of sample completeness,Detailed/ comprehensive database of EHRs used ,Tackled methodologically,Complete follow up ,Not mentioned,None,Fully acknowledged implications of data completeness,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Adjustment for covariates,Mentioned or partially addressed,Adjustment for covariates
Nakamura A et al [64],Impact of sodium-glucose cotransporter 2 inhibitors on renal function in participants with type 2 diabetes and chronic kidney disease with normoalbuminuria,Diabetology & Metabolic Syndrome,2020,not stated,not stated,Japan,outpatient diabetes,INCLUDE in sub-analysis; not clear if EHRs used,outpatient center at Kurihara clinic,type 2 diabetes,type 2 diabetes and CKD,66,,54%,,risk factor identification / causal inference,"effects of SGLT2 inhibitors on renal funciton in type 2 diabetes and CKD, compared by albuminuria status",retrospective cohort study,2,eGFR at baseline and 2 years,not available,87,not mentioned,not available,not relevant (complete case analysis),553 patients treated for 2 years; 87 analysed; reasons for exclusion include eGFR>60 and incomplete data but no breakdown of percent excluded for lack of data rather than failing to meet population criteria,eGFR,not specified,percent change in eGFR from baseline,percent change in eGFR between 2 measures (baseline and at 2 years post-treatment),none defined,none defined,,outcome,N/A,primary outcome,ANOVA,ANOVA,"One-way ANOVA for comaprison of change in eGFR over 2 years, comparing albuminuria at baseline groups",2 years,2,"More favourable treatment effects were observed in the normoalbuminuria group, suggesting reno-protective effects depend on degree of albuminuria",Not mentioned,None,Not mentioned,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,None,Not mentioned,None
Nichols GA et al [71],Kidney disease progression and all-cause mortality across estimated glomerular filtration rate and albuminuria categories among patients with vs. Without type 2 diabetes,BMC Nephrology,2020,2006-2016,2010-2019,USA,multiple care settings,INCLUDE; states EHRs used,"EHR of Kaiser Permanente Northwest (KPNW); integrated delivery system serving approx 550,000 patients, covers insured patients; representative of service area",Diabetes,,61,,52%,,risk factor identification / causal inference,to compare CKD progression in diabetes vs no diabetes,retrospective cohort study,2,"Required 1x baseline eGFR duing 2006-2012 (and urine ACR or dipstick proteinuria within 6 months of baseline eGFR), plus >=1 x eGFR during follow up
 [Note: initial eGFR decline analyses did not require ACR/proteinuria data]",78590,approx 36727,not available,47%,not available,"39,295 patients met type 2 diabetes definition prior to 2013 and were 1:1 matched to non-diabetics = 78,590; 9613 excluded with no eGFR in 2006-2012; 3854 excluded due to no follow up creatinine; initial analysis cohort = 65132; Number of losses to follow up not mentioned; approx 28405 patients had no ACR/proteinuria data (calculated from Table 1, not explicitly stated) -> 36,727 in main analysis requiring ACR data",eGFR,CKD-EPI,Transition between CKD stages,2 x eGFR in KDIGO-defined CKD stages over 90-365 days,progression of kidney dysfunction to next CKD stage,2 x eGFR in KDIGO-defined CKD stages over 90-365 days,"Progression to next CKD stage summarised separately by initial CKD stage (G1/G2->G3a/3b, G3a/3b->G4,G4->ESKD); Those without 2xeGFR in next CKD stage deemed to have remained at baseline stage",outcome,N/A,primary outcome,KM / lifetable,life-table analysis,"Various analyses of CKD progression in 1:1 matched cohort (age, sex, year of baseline creatinine) of diabetics and non-diabetics: (1) Comparison of annual rate of eGFR decline, adjusted for age, sex, number of eGFR measures in diabetes vs non-diabetes; (2) life table analysis for probability of tranistion between CKD stages at 1,3 and 5 years; (3) Progression to ESKD (but not eligible outcome due to inclusion of RRT);
Analysis (2) considered MAIN analysis although could be argued otherwise",5 years,5,Presence of albuminuria was associated with accelerated eGFR decline independent of diabetes status,Fully acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Partially acknowledged implications of data completeness,None,Not relevant (no covariate analysis),None,Not mentioned,None,Fully acknowledged,Outcome likely to identify real change,Tackled,Stratified analyses,Fully acknowledged,None
Peng YL et al [72],Comparison of uric acid reduction and renal outcomes of febuxostat vs allopurinol in patients with chronic kidney disease,Scientific reports,2020,2010-2015,2010-2019,Taiwan,multiple care settings,INCLUDE; states EHRs used,"electronic healthcare records from a large healthcare delivery system in Taiwan; Chang Gung Research Database (CGRD) from a group of hospitals in Taiwan; detailed diagnositic, prescription and lab test results from emergency department, inpatient and outpatient settings",CKD,Adult CKD patients newly treated with urate-lowering therapy (ULT),67,,76%,,risk factor identification / causal inference,to investigate the effect of ULT on renal function,retrospective cohort study,not stated,"Excluding: lack of any medical records 365 days before and after index date, lack of SUA and creatinine lab results, and no CKD diagnosis during study period;
Require CKD ICD-9 code within 1-year prior to index date; at least 2 serum creatinine at baseline but unclear if follow-up restrictions",5977 (pre-PS matching),1050,not available,not available,not available,"From flow chart, 5977 new drug users with CKD; 5,628 new drug users fulfilled study inclusion criteria; 1,050 in 1:1 propensity-score matched cohort. Number ""LTFU"" (and therefore censored) not mentioned; avg follow up years was 1.4 years, despite intended 2.5 years follow up period. Since PS matching applied after exclusions due to eGFR data completeness, cannot compute percent of target population analysed after application of data completeness rules. 95 of 5977 excluded prior to PS matching (98% of original cohort met analysis criteria). ",eGFR,MDRD,regression slope of eGFR,linear mixed model,eGFR decline,Raw percent change from baseline (where baseline is avg of at least 2 measures) >= 30%,"2 x eligible renal outcomes: (1) change in eGFR from baseline, (2) incident eGFR decline >= 30% from baseline; adjusted analysis in binary outcome",outcome,N/A,secondary outcome,Cox PH regression,linear mixed model,Main analysis: Linear mixed model with adjustment; Other analyses: Adjusted Cox regression;  t-tests for change from baseline.,1.5 years,1.5,There were no significant differences in mean eGFR change over time or risk of eGFR decline >= 30% from baseline between treatment groups,Partially acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Partially acknowledged implications of data completeness,Mixed modelling,Not mentioned,None,Not mentioned,None,Tackled,random effects,Mentioned or partially addressed,subgroup analyses,Tackled,Adjustment for baseline confounders
Rej S et al [73],Association of Lithium Use and a Higher Serum Concentration of Lithium With the Risk of Declining Renal Function in Older Adults: A Population-Based Cohort Study,Early Career Psychiatrists,2020,2007-2015,2010-2019,Canada,multiple care settings,INCLUDE; states EHRs used,"linked healthcare databases in Ontario, Canada; residents of Ontario, Canada; inpatient and outpatient data; outpatient serum creatinine data",general population,general population cohort; older adults,71,,41%,residents of Ontario province aged >= 66 with 2 prescriptions of either lithium or valporate,risk factor identification / causal inference,"to compare risk of renal function decline in lithium users vs valporate, and to investigate any interaction with baseline lithium levels ",retrospective cohort study,1,at least 1 x creatinine in year prior to index date (when prescription filled); AND at least 1 lithium value in year prior to index date (lithium users only),12544 (pre-PS),6226,not available,not available,not available,"12,544 drug-users prior to creaitnine/lithium data exclusions; 10,114 post data exclusions; 3113 lithium users matched 1:1 to 3113 valporate users resulting in 6226 patients for analysis; no data on LTFU;
Data completeness not available post-PS matching; 2430 of 28778 excluded due to data completeness pre-PS matching (92% met critieria prior to PS matching)",eGFR,CKD-EPI,percent change in eGFR from baseline,"raw percent change in eGFR from baseline, for any measure during follow up","""clinically important decline""",">= 30% loss in eGFR from baseline, at any time during follow up",,outcome,N/A,primary outcome,Cox PH regression; Propensity score matching,Cox PH regression,1:1 propensity score matched analysis cohort; Cox PH regression for time to >= 30% loss in eGFR comparing lithium and valporate users; censoring for end of data availability (Sep 30 2015)) or death; subgroup analyses by baseline lithium levels and by CKD status; interaction explored by lithium levels also,median 3.1 years,3.1,Lithium was associated with a 14% increased risk of clinically important renal function decline compared to valporate users,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Propensity score matched cohort
Sise ME et al [65],Direct-acting antiviral therapy slows kidney function decline in patients with Hepatitis C virus infection and chronic kidney disease,Kidney International,2020,2013-2017,2010-2019,USA,hospital,INCLUDE; states EHRs used,"Patients seen in the Partners Healthcare system, Research Patient Data Repository (registry of EHRs); ",hepatitis C virus (HCV) infection,HCV infected patients receiveing direct-acting antiviral (DAA) therapies from 2013-2017,56,,64%,,risk factor identification / causal inference,to determine the effect of DAA therapy on rate of decline of eGFR,retrospective cohort study,4,at least 2 creatinine measures in the 3 years before therapy start and at least 2 creatinine measures in up to 3 years after,1953,1178,not mentioned,60%,not available,"2319 patients prescribed DAAs; 1953 after exclusions for not meeting population criteria (no treatment, organ transplant, RRT); 1590 with baseline creatinine; 1178 included in primary analysis of slopes requiring 4 creatinine measures",eGFR,CKD-EPI,regression slope of eGFR,GEEs,none defined,none defined,,outcome,N/A,primary outcome,generalised estimating equations,generalised estimating equations,"Generalised estimating equations (GEEs) estimating pre-therapy and post-therapy slope with a sandwhich estimator of variance to account for correlations within subjects, includes interaction between therapy and slope to estimate difference in slopes pre- and post- therapy, adjusted for age, diabetes and baseline eGFR; gender/race removed from model since ""not significant confounders""; analysis stratified by baseline CKD status (eGFR<60)",1.6 years (post-therapy),1.6,DAA therapy for HCV infection may slow CKD progression,Partially acknowledged implications of sample completeness,None,Not mentioned,None,Partially acknowledged implications of data completeness,GEEs,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,GEEs,Tackled,stratified analysis,Tackled,Adjustment for covariates
Weldegiorgis M et al [66],Socioeconomic disadvantage and the risk of advanced chronic kidney disease: results from a cohort study with 1.4 million participants,Nephrol Dialysis Transplant,2020,2000-2014+,2000-2009,UK,primary care,INCLUDE; states EHRs used,English CPRD (2000-2014) linked to HES and mortality data;,general population,"Valid BMI 2000-2014, age 20-87, BMI 15-60; excluding CKD stage 4+ (coded or eGFR<30)",48.6,,42%,Required BMI data for inclusion (not relevant for analysis; selection bias),risk factor identification / causal inference,to investigate effect of socioeconomic deprivation on CKD progression/outcomes,retrospective cohort study,0,"Main analysis did not require eGFR data at baseline and outcome could potentially be obtained by CKD codes without requiring eGFR data;
at least 2 years of ""data"" before baseline date and at least 3 years of follow up thereafter (FU end at death / leaving practice / 30-3-2014 [ref 14])",not available ,1397573,not available,not available,not available,"4.4 million patients in UK CPRD database before excluding 25% ineligible practices not in England;
1,405,016 = Valid BMI record between Jan 2000 - March 2014 (based on cohort from a previous study looking at BMI associations) = main study population; 1,397,573 = MAIN analysis, excluding those with CKD stage 4+ at baseline or in first 3 years follow up; Secondary analysis adjusted for eGFR requiring eGFR data = 1,031,377",eGFR,CKD-EPI,(time to) progression to stage 4 CKD,"Algorithm of CKD codes + labs (eGFR<30 over 90+ days, no intermmediate eGFR >30)",incident CKD stages 4-5,binary progression to stage 4-5 CKD (codes+labs),,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"Cox PH regression; 4 models: (1) unadjusted; (2=MAIN MODEL) adj age+sex; (3) adj age, sex, smoking, BMI, diabetes, BP, CVD; (4) mod-3 + baseline eGFR; Subgroup analyses also performed",7.5 years,7.5,Socioeconomic deprivation is independently associated with an increased hazard of CKD stages 4-5 and ESRD.,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Not mentioned,None,Mentioned or partially addressed,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Adjustment for baseline confounders
Jackevicius CA et al [74],Bleeding Risk of Direct Oral Anticoagulants in Patients With Heart Failure And Atrial Fibrillation,Circulation: Cardiovascular Quality and Outcomes,2021,2010-2018,2010-2019,USA,multiple care settings,INCLUDE; states EHRs used,linked veterans administration databases; inpatient and outpatient data,heart failure / atrial fibrillation,patients with heart failure newly prescribed warfarin or DOACS for atrial fibrillation,72,,99%,,risk factor identification / causal inference,"to compare bleeding outcomes in use of DOACs and warfarin, investigating any interaction with renal function changes",retrospective cohort study,not stated,"Analysis crtieria for eGFR unclear; 4,338 patients excluded in flow chart for ""no follow up"" but not defined.",53796,49458,not available,92%,not available,"53,796 met inclusion criteria before follow up criteria applied; 49458 included after follow up criteria applied; missing data at baseline was imputed so no patients excluded on that basis",eGFR,MDRD,"Rate of percentage change in eGFR, not clearly defined",percent change in eGFR at any time during follow up (not clear if from baseline or between adjacent measures),transient or persistent renal function decline,20% reduction in eGFR at any time during follow up,full derivation not described,exposure,bleeding events,secondary exposure,Cox PH regression,Cox PH regression,"Secondary analysis of changes in renal function:
IPTW-weighted Cox PH regression for bleeding events, compared by whether renal function decline occurred, separately by treatment group;
(other analyses included comparison of change in eGFR from baseline by treatment group)",approx 1.4 years,1.4,DOACs were associated with lower risk of bleeding; a 20% decline in renal function was independently associated with increased bleeding in DOAC-treated patients and increased monitoring for dose-adjustments are required,Tackled methodologically,Other imputation methods to avoid exclusions,Partially acknowledged implications of losses to follow up,Sensitivity analysis adjusting for competing risk of death,Not mentioned,None,Tackled methodologically,imputation with mean,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,Stratified or separate/ subgroup analysis,Tackled,IPTW
Neuen BL et al [75],Changes in GFR and Albuminuria in Routine Clinical Practice and the Risk of Kidney Disease Progression,American Journal of Kidney Disease,2021,2000-2015,2000-2009,UK,primary care,INCLUDE; states EHRs used,"CPRD, 700 primary care practices in UK, representing 8% of entire UK population; inpatient eGFR/UACR were not included",general population,"indiviudals with spot UACR and eGFR assessments approx 3 years apart between 2000-2015, aged 20-79",65,,57%,,risk factor identification / causal inference,to investigate whether combined eGFR/ACR predict advanced kidney disease better than either measure alone,retrospective cohort study,2,Require UACR and eGFR assessments approx 3 years apart between 2000-2015,8157969,91319,not available,1%,not available,"This was a general population cohort with analysis based purely on repeat eGFR/UACR availability; From flow chart, 8,157,969 patients were identified in linked database; 91,319 had paired UACR and eGFR tests over approx three-year exposure window",eGFR,CKD-EPI,percent change in eGFR between measures,Change in eGFR was calculated over an approximate 3-year exposure window using the eGFR value closest to the paired UACR value,advanced CKD,2 x eGFR<30 over 90+ days with no intermediate eGFR>30,"Here, changes in renal function were exposures and outcomes. Since the study looked at progression to CKD stage 4 as the outcome for long-term renal function changes, this was used as the main measure summarised here",exposure (and outcome),advanced CKD,primary exposure,Cox PH regression,Cox PH regression,"Multivariate Cox PH regression, using 3 models; Model 1 = basic adjustment, Model 2 = full adjustment, Model 3 = Model 2 variables with multiple imputation, using predictive mean matching for continuous variables and multinomial logistic regression for categorical variables; Model 3 multiple imputation was ""sensitivity analysis... to assess bias due to missing data""; assessment of effects of changes in UACR and eGFR separately and combined, comparing discrimination statistics between models; also subgroup analyses and ""heterogeneity assessed by"" exposure*subgroup interaction terms",median 2.9 years,2.9,The combination of changes in UACR and eGFR predicted kidney outcomes better than either alone; changes in eGFR were more important than changes in UACR in prediction of progression,Partially acknowledged implications of sample completeness,Multiple imputation to avoid exclusions,Not mentioned,None,Not mentioned,None,Tackled methodologically,Multiple imputation,Tackled,Distributional checks,Tackled,Outcome likely to identify real change,Tackled,Interaction terms,Tackled,Adjustment for baseline confounders
Niu SF et al [76],Early Chronic Kidney Disease Care Programme delays kidney function deterioration in patients with stage I-IIIa chronic kidney disease: an observational cohort study in Taiwan,BMJ Open,2021,2012-2017+,2010-2019,Taiwan,hospital,INCLUDE; states EHRs used,"Taipei Medical University Research Database; 3 hospitals; CKD patients identified from hospital database; hospital data used, unclear if just outpatient data or also any inpatient data collected",CKD,CKD stages 1-3a,67,,66%,,risk factor identification / causal inference,to investigate the impact of enrollment in a CKD Care Programme on CKD progression,retrospective cohort study,1,at least 2 medical visits between 2012-2017; looks like at least 1 x eGFR,not available ,3114,not available,not available,not available,"307,762 patients had 2 x medical visits, of which 159,774 adults had CKD stages <= 3a; 3114 patients included after propensity score matching; No mention of number of patients dropping out over course of follow up. Cannot compute percent analysed as only patients with at least 2 medical visits shown in flow chart",eGFR,MDRD,(time to) progression to stage 3b CKD,single eGFR < 45,Progression to CKD stage 3b,single eGFR < 45,,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"1:2 propensity score matched cohort using age, sex, eGFR and CKD stage (intervention: control); multivariate Cox PH regression for progresson to CKD stage 3b. Also subgroup analysis by CKD stage (3a, <3a)",median 3.0 years,3,The Early CKD Programme was an independent protective factor against progression of early CKD,Partially acknowledged implications of sample completeness,None,Partially acknowledged implications of losses to follow up,None,Mentioned care pathway follow up but not data completeness,None,Not mentioned,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Adjustment for baseline confounders
Robinson DE et al [77],Safety of Oral Bisphosphonates in Moderate-to-Severe Chronic Kidney Disease: A Binational Cohort Analysis,Journal of Bone and Mineral Research,2021,1997-2016,2000-2009,multiple European countries,multiple care settings,INCLUDE; states EHRs used,"primary care EHRs from 2 cohorts: CPRD GOLD (1997-2016, UK) and SIDIAP (2007-2015, Catalonia); linked to hospital records; SIDIAP also linked to ESRD registry data; eGFR from primary care data but transplant/dialysis from linked data",CKD,"CKD stages 3b-5 (2xeGFR<45), age 40+",80,,27%,"Mean age, percent male shown in PS matched cohort",risk factor identification / causal inference,to investigate the safety of oral biophsophonates (primary outcome CKD stage worsening) in moderate to severe CKD,retrospective cohort study,3,2 x eGFR<45 + follow-up eGFR,not available ,19324,not available,not available,not available,Patients excluded if missing IMD or no follow-up eGFR. Flow chart provided. Data on specific exclusions in CPRD but not Spain data so cannot compute percent analysed overall. Some analyses in full cohort and some in PS matched reduced cohort. No data on losses to follow up over time. Matched analysis considered primary and sample size shown.,eGFR,CKD-EPI,(time to) transition between CKD stages,"Increase in CKD stage from 3b to 4/5 or 4 to 5, based on eGFR CKD stages and coded dialysis/transplant",CKD stage worsening,"Increase in CKD stage from 3b to 4/5 or 4 to 5, based on eGFR CKD stages and coded dialysis/transplant",,outcome,N/A,primary outcome,competing risks survival models,competing risks survival models,Adjusted Cox PH regression in unmatched cohort; Fine and Gray models in matched cohort to account for any difference in death rate due to PS matching (primary analysis); outcome increase in CKD stage; propensity score matched cohort using logistic regression based on 45 covariates; missing covariate data imputed using MI (assuming MAR). Results combined from both cohorts using meta-analysis. Interactions tested and subsequent stratified analyses.,approx 3.7 years,3.7,Biophosphonate use was associated with a modest increased risk of CKD progression in both cohorts,Partially acknowledged implications of sample completeness,Multiple imputation to avoid exclusions,Partially acknowledged implications of losses to follow up,competing risks survival models,Not mentioned,None,Partially acknowledged implications of data completeness,Multiple imputation,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,Interaction terms,Tackled,Adjustment for baseline confounders
Tangri N et al [78],Metabolic acidosis is associated with increased risk of adverse kidney outcomes and mortality in patients with non-dialysis dependent chronic kidney disease: an observational cohort study,BMC Nephrology,2021,2007-2017,2010-2019,USA,multiple care settings,INCLUDE; states EHRs used,large US community based cohort; Optum's de-identified integrated electronic health records; Optum database holds longitudinal clinical data on 81 million insured and uninsured patients from several large healthcare provider organisations in all 50 US states and Puerto Rico; Data inlucded inpatient and outpatient EHRs and admin systems,CKD,non-dialysis-dependent CKD stages 3-5,73,,47%,at least 2 serum bicarbonate results,risk factor identification / causal inference,"to investigate the association between metabolic acidosis with CKD progression and mortality, to idenitfy new modifiable risk factors in CKD",retrospective cohort study,3,"min 1 year of EHR activity; 
3 x eGFR <60; 3 x serum bicarbonare 12-29;
2 x serum bicarbonate measures, 28-365 days apart, both in same category (normal = 22-29; metabolic acidosis: 12-22); baseline eGFR 10-60",not available ,"32,007",not available,not available,not available,"Flow chart provided but does not show primary analysis population (with complete ACR data). 81 million in database; 319,126 extracted from database with 3 x eGFR<60, 3 x serum bicarbonate, 1 x serum bicarbonate 12-29; 51,558 with serum bicarbonate category definitions, 2 years FU data, baseline eGFR 10-60; 32007 with ACR data eligible for primary analysis.
Not possible to compute percent of target population availale for analysis before and after data completeness exclusions, due to being mixed up in flow chart;
Min 2 years EHR activity required for analysis, up to 10 years, median 3.9 years",eGFR,CKD-EPI,percent change in eGFR from baseline,">=40% decline in eGFR from baseline at any time, based on 90-day averages at baseline and follow-up (this is the eGFR component of composite outcome)",RRT40: >=40% decline in eGFR from baseline or RRT,"percent change in baseline >=40% from baseline at any time; baseline = avg eGFR over 90 days prior to index date; ""event"" eGFR = point at which average eGFR over last 90 days results in a >=40% decline",time to event analysis,outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"Primary analysis was composite of 40% eGFR decline, RRT and death (DD40); secondary analyses in components, including kidney specific outcome RRT40 (40% decline of RRT). Cox PH regression for full duration analysis; logistic regression for additional 2-year analysis; Cox models censored at last ineraction with health system or end of FU (March 2017); Kidney-specific outcome (RRT40) additionally censored for death; subgroup analyses by age and by ACR missing data status; sensitivity analyses: single eGFR for decline outcome; non-oversampled analysis cohort",median 3.9 years,3.9,Presence of metabolic acidosis is an independent risk factor for the composite outcome of CKD progression and death,Mentioned sample completeness but not implications,None,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,subgroup analyses,Tackled,Adjustment for baseline confounders
Vesga JI et al [79],Chronic kidney disease progression and transition probabilities in a large preventive cohort in colombia,International Journal of Nephrology,2021,2009-2018,2010-2019,Colombia,multiple care settings,INCLUDE; states EHRs used,Data on patients enrolled in a CKD prevention program; primary and secondary care,CKD,CKD stages 3-4 diagnosed by a nephrologist,70,,48%,,estimation of incidence/ prevalence,to estimate the probability of transitions between CKD stages over time,retrospective cohort study,2,"at least 2 x eGFR <60 for inclusion in main analysis;
 at least 4 x eGFR to estimate CKD progression rate (linear model);","3,048",2752,1783,90%,65%,"3048 patients identified; 2752 met eGFR inclusion criteria; 969 completed study;
",eGFR,CKD-EPI,Transition between CKD stages,"percent moved to each CKD stage category over 6-month periods, looks like patients included for every 6-month period but no adjustment for pt correlation",transition between CKD stages,probabilities of transition to CKD stages over 6-month periods,,outcome,N/A,secondary outcome,descriptive results only,crude estimation,"Main analysis of eGFR: to estimate probabilities of transition between CKD stages over 6 month periods; looks like patients are included in percentage calculations every time they start from a certain CKD stage, throughout the study. If patients are included multiple times, there does not appear to be adjustment for patient correlation. Analysis repeated by subgroup.","""minimum of 5 years"" stated but actual follow up not clear",not stated,Patients generally tend to remain in their current CKD stage over 6-month periods; Probability of transition to the next CKD stage is higher in diabetic and glomerulonephritis CKD patients,Mentioned sample completeness but not implications,None,Mentioned losses to follow up but not implications,None,Not mentioned,None,Not relevant (no covariate analysis),None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,subgroup analyses,N/A,N/A
Yanagawa et al [80],Retrospective study of factors associated with progression and remission/ regression of diabetic kidney disease-hypomagnesemia was associated with progression and elevated serum alanine aminotransferase levels were associated with remission or regression,Diabetology international,2021,2003-2019,2010-2019,Japan,hospital,INCLUDE; states EHRs used,"electronic medical records of institution (Nerima General Hospital); unclear what data used, likely outpatient only but not clear",diabetic kidney disease,patients with type 2 diabetes receiving treatment with oral antidiabetic drugs other than SGLT2 inhibitors,62,,66%,,risk factor identification / causal inference,to investigate the association between clinical factors and progression and remission//regression of diabetic kidney disease,retrospective cohort study,not stated,not stated,690,681,complete case analysis,99%,complete case analysis,No flow chart; analysis based on patients with complete follow up data; 690 patients extracted with diabetes ; 681 patients with complete data for analysis (follow up duration varies between patients depending on last follow up),eGFR,Japanese formula,(time to) transition between CKD stages,"transition between:
low risk or
moderately increased risk or 
high risk categories 
to the very high risk category between first day of prescription and last consultation (risk categories based on combined proetinuria and eGFR)",diabetic kidney disease progression,"transition between:
low risk or
moderately increased risk or 
high risk categories 
to the very high risk category between first day of prescription and last consultation (risk categories based on combined proetinuria and eGFR)","RISK CATEGORIES:
low risk:
prot (-) eGFR>=60.
moderate:
prot (-) eGFR 45-59.9;
prot (+-) eGFR >=60;
high:
prot (-) eGFR 30-44.9;
prot (+-) eGFR 45-59.9;
prot (+) eGFR >=60;
very high:
prot (-) eGFR <30;
prot (+-) eGFR <45;
prot (+) eGFR <60.
[+- = trace]",outcome,N/A,primary outcome,Cox PH regression,Cox PH regression,"Progression of DKD defined as transition between low, moderate and high risk categories to very high risk categories, where categories combine proteinuria and eGFR data. Cox PH regression for progression. Additional analyses of eGFR decline per year (not defined) using logistic regression. Patients in the very high risk category were generally included in analysis, with 2 different models used. Model 1: very high -> very high = ""non-progression""; Model 2: exclude very high -> very high from analysis but include very high -> other categories",mean 6.2 years,6.2,"Advancing age, serum Mg, and serum HbA1c were associated with progression of DKD",Tackled methodologically,Detailed/ comprehensive database of EHRs used,Not mentioned,None,Not mentioned,None,Partially acknowledged implications of data completeness,None,Not mentioned,None,Tackled,Outcome likely to identify real change,Tackled,Adjustment for covariates,Tackled,Adjustment for baseline confounders