Naive Bayes classifiers for verbal autopsies: comparison to physician-based classification for 21,000 child and adult deaths

184 35 1779 3 5993 document 1779 File1-WHO_DeathCauseCategories.doc application/msword d7969cc7d10a074562417efb1b9142eb34cadc4046425ad9e0b6fb2fe1419b24 SHA256 51200 2016-03-10 16:03:13 https://datacompass.lshtm.ac.uk/id/eprint/184/1/File1-WHO_DeathCauseCategories.doc 184 1 1 application/msword Dataset en public cc_by

File1-WHO_DeathCauseCategories.doc

data indefinite Mapping of WHO cause of death categories 1780 3 5995 document 1780 File2-Rcode.zip application/x-zip c2b6ab59832b1307d127a750d5371e13d4c2dfcb51a064f37050e2646e87bb30 SHA256 103908 2016-03-10 16:03:16 https://datacompass.lshtm.ac.uk/id/eprint/184/2/File2-Rcode.zip 184 2 2 application/x-zip Software en public cc_by

File2-Rcode.zip

study_instrument indefinite R code used to produce the results in this study 1781 3 5997 document 1781 File3-InterVA-4_TariffMethods.doc application/msword e0ef68016d8b2ea715009561aa385dd2fa254eecb1dc250487352a15379d6a30 SHA256 26112 2016-03-10 16:03:18 https://datacompass.lshtm.ac.uk/id/eprint/184/3/File3-InterVA-4_TariffMethods.doc 184 3 3 application/msword Text en public cc_by

File3-InterVA-4_TariffMethods.doc

documentation indefinite Description of InterVA-4 and Open-source Tariff Method 1782 3 5999 document 1782 File4-MDS_Cause_of_death.doc application/msword 703f9242c9c51f1068f2b8ff143e06163ae282938783e4e080ffd14db25551a1 SHA256 70656 2016-03-10 16:03:19 https://datacompass.lshtm.ac.uk/id/eprint/184/4/File4-MDS_Cause_of_death.doc 184 4 4 application/msword Dataset en public cc_by

File4-MDS_Cause_of_death.doc

data indefinite Sensitivity and specificity of assignment by cause of death in Indian Million Death Study 1783 3 6001 document 1783 File5-Agincourt_Cause_of_death.doc application/msword 716377f4b2d4e5650b4ae3e213129fb1da4216464812bebe41ac4d7557603d7a SHA256 72704 2016-03-10 16:03:21 https://datacompass.lshtm.ac.uk/id/eprint/184/5/File5-Agincourt_Cause_of_death.doc 184 5 5 application/msword Dataset en public cc_by

File5-Agincourt_Cause_of_death.doc

documentation indefinite Sensitivity and specificity of assignment by cause of death, on Agincourt data 1784 3 6003 document 1784 File6-Matlab_Cause_of_death.doc application/msword 92b7945440fbfcd4714cdc6e4aa8c3ee6bcb5880e634d98bcdc4779760e1691f SHA256 70656 2016-03-10 16:03:23 https://datacompass.lshtm.ac.uk/id/eprint/184/6/File6-Matlab_Cause_of_death.doc 184 6 6 application/msword Dataset en public cc_by

File6-Matlab_Cause_of_death.doc

data indefinite Sensitivity and specificity of assignment by cause of death, on Matlab data 1785 1 6019 document 1785 indexcodes.txt text/plain 4d823756aae06f37d78f6e1522e11c5d1ff7c3754556a84101cd1033b7f93e89 SHA256 382 2016-03-10 16:13:37 https://datacompass.lshtm.ac.uk/id/eprint/184/7/indexcodes.txt 184 7 7 text/plain other en public

indexcodes.txt

http://eprints.org/relation/isVersionOf https://datacompass.lshtm.ac.uk/id/document/1779 http://eprints.org/relation/isVolatileVersionOf https://datacompass.lshtm.ac.uk/id/document/1779 http://eprints.org/relation/isIndexCodesVersionOf https://datacompass.lshtm.ac.uk/id/document/1779 indefinite Generate index codes conversion from Dataset to indexcodes 1786 1 6020 document 1786 indexcodes.txt text/plain b6b6904ffdf7fda13c6f016dd245246666640754c3b1dc85fe383e9764dcb0ee SHA256 1756 2016-03-10 16:13:37 https://datacompass.lshtm.ac.uk/id/eprint/184/8/indexcodes.txt 184 8 8 text/plain other en public

indexcodes.txt

http://eprints.org/relation/isVersionOf https://datacompass.lshtm.ac.uk/id/document/1781 http://eprints.org/relation/isVolatileVersionOf https://datacompass.lshtm.ac.uk/id/document/1781 http://eprints.org/relation/isIndexCodesVersionOf https://datacompass.lshtm.ac.uk/id/document/1781 indefinite Generate index codes conversion from Text to indexcodes 1787 1 6021 document 1787 indexcodes.txt text/plain 519bea42936c6c436ef830dcbe1d60ec0d83c4170646d1619ad0eca632121c36 SHA256 333 2016-03-10 16:13:37 https://datacompass.lshtm.ac.uk/id/eprint/184/9/indexcodes.txt 184 9 9 text/plain other en public

indexcodes.txt

http://eprints.org/relation/isVersionOf https://datacompass.lshtm.ac.uk/id/document/1782 http://eprints.org/relation/isVolatileVersionOf https://datacompass.lshtm.ac.uk/id/document/1782 http://eprints.org/relation/isIndexCodesVersionOf https://datacompass.lshtm.ac.uk/id/document/1782 indefinite Generate index codes conversion from Dataset to indexcodes 1788 1 6022 document 1788 indexcodes.txt text/plain 9915f4bc1a99fe93ace66dff9f24fef63b2ff861e23a47202cf90f96738a0b68 SHA256 342 2016-03-10 16:13:38 https://datacompass.lshtm.ac.uk/id/eprint/184/10/indexcodes.txt 184 10 10 text/plain other en public

indexcodes.txt

http://eprints.org/relation/isVersionOf https://datacompass.lshtm.ac.uk/id/document/1783 http://eprints.org/relation/isVolatileVersionOf https://datacompass.lshtm.ac.uk/id/document/1783 http://eprints.org/relation/isIndexCodesVersionOf https://datacompass.lshtm.ac.uk/id/document/1783 indefinite Generate index codes conversion from Dataset to indexcodes 1789 1 6023 document 1789 indexcodes.txt text/plain f70dbb047e7b464fc6a63252e8a32392414898a8b069a11a8864b22289ccc4c0 SHA256 343 2016-03-10 16:13:38 https://datacompass.lshtm.ac.uk/id/eprint/184/11/indexcodes.txt 184 11 11 text/plain other en public

indexcodes.txt

http://eprints.org/relation/isVersionOf https://datacompass.lshtm.ac.uk/id/document/1784 http://eprints.org/relation/isVolatileVersionOf https://datacompass.lshtm.ac.uk/id/document/1784 http://eprints.org/relation/isIndexCodesVersionOf https://datacompass.lshtm.ac.uk/id/document/1784 indefinite Generate index codes conversion from Dataset to indexcodes archive 3 disk0/00/00/01/84 2016-03-10 16:11:19 2022-02-09 18:32:00 2016-03-10 16:11:19 data_collection show Miasnikof Pierre FALSE Giannakeas Vasily FALSE Gomes Mireille FALSE Aleksandrowicz Lukasz TRUE 28ad009a8239bcdc63a08778ec50cb40 Shestopaloff Alexander Y. FALSE Alam Dewan FALSE Tollman Stephen FALSE Samarikhalaj Akram FALSE Jha Prabhat FALSE St. Michael’s Hospital University of Toronto University of the Witwatersrand Ryerson University London School of Hygiene & Tropical Medicine Naive Bayes classifiers for verbal autopsies: comparison to physician-based classification for 21,000 child and adult deaths EPPH Cause of death Computer-coded verbal autopsy InterVA Tariff Naive Bayes classifier Physician certified verbal autopsy Verbal autopsy BACKGROUND: Verbal autopsies (VA) are increasingly used in low- and middle-income countries where most causes of death (COD) occur at home without medical attention, and home deaths differ substantially from hospital deaths. Hence, there is no plausible "standard" against which VAs for home deaths may be validated. Previous studies have shown contradictory performance of automated methods compared to physician-based classification of CODs. We sought to compare the performance of the classic naive Bayes classifier (NBC) versus existing automated classifiers, using physician-based classification as the reference. METHODS: We compared the performance of NBC, an open-source Tariff Method (OTM), and InterVA-4 on three datasets covering about 21,000 child and adult deaths: the ongoing Million Death Study in India, and health and demographic surveillance sites in Agincourt, South Africa and Matlab, Bangladesh. We applied several training and testing splits of the data to quantify the sensitivity and specificity compared to physician coding for individual CODs and to test the cause-specific mortality fractions at the population level. RESULTS: The NBC achieved comparable sensitivity (median 0.51, range 0.48-0.58) to OTM (median 0.50, range 0.41-0.51), with InterVA-4 having lower sensitivity (median 0.43, range 0.36-0.47) in all three datasets, across all CODs. Consistency of CODs was comparable for NBC and InterVA-4 but lower for OTM. NBC and OTM achieved better performance when using a local rather than a non-local training dataset. At the population level, NBC scored the highest cause-specific mortality fraction accuracy across the datasets (median 0.88, range 0.87-0.93), followed by InterVA-4 (median 0.66, range 0.62-0.73) and OTM (median 0.57, range 0.42-0.58). CONCLUSIONS: NBC outperforms current similar COD classifiers at the population level. Nevertheless, no current automated classifier adequately replicates physician classification for individual CODs. There is a need for further research on automated classifiers using local training and test data in diverse settings prior to recommending any replacement of physician-based classification of verbal autopsies. 2015-11-25 published Figshare Dataset St. Michael’s Hospital University of Toronto University of the Witwatersrand Ryerson University London School of Hygiene & Tropical Medicine CompilationSynthesis public Dataset Quantitative en Data record - Figshare https://doi.org/10.6084/m9.figshare.c.3629489_D1.v1 dataresource Naive Bayes classifiers for verbal autopsies: comparison to physician-based classification for 21,000 child and adult deaths http://researchonline.lshtm.ac.uk/id/eprint/2534637 2015-11-25