birth median applicable richer survival category male wealth private married adjusting female all direction deliveries survived 114 doctor refernce order auxiliary urban women marital status s2 after public discharge midwife wantedness heatmap day hours formerly 999 twins 115 103 variables section vaginal 107 education child then multiple birthweight models other nurse age singleton short reference grams higher attendant richest regression centercenter 000 too births later completed poorer index length wanted facility 112 died sd poorest iqr residence sex mean across table not years 500 determinant currently middle sector level secondary none linear rural showing country staff stay 499 more logistic before cesarean primary never quintile effects 111 triplets