Survival analysis of the association between antenatal care attendance and neonatal mortality in 57 low- and middle-income countries

An ena al Ca e and Neona al Ou come Su i al analysis o he associa ion be ween an ena al ca e a endance and neona al mo ali y in 57 low- and middle-income coun ies Da id T Doku 1,2 * and Subas Neupane 1 1 School o Heal h Sciences, Uni e si y o Tampe e, Tampe e Finland and 2 Depa men o Popula ion and Heal h, Uni e si y o Cape Coas , Cape Coas , Ghana *Co esponding au ho . School o Heal h Sciences, FI-33014, Uni e si y o Tampe e, Tampe e, Finland. E-mail: [email p o ec ed] Edi o ial decision 13 June 2017; Accep ed 16 June 2017 Abs ac Backg ound: Neona al mo ali y is unaccep ably high in mos low- and middle-income coun ies (LMICs). In hese coun ies, whe e access o eme gency obs e ic se ices is limi ed, an ena al ca e (ANC) u iliza ion o e s imp o ed ma e nal heal h and bi h ou - comes. Howe e , e idence o his is scan y and mixed. We explo ed he associa ion be- ween a endance o ANC and su i al o neona es in 57 LMICs. Me hods: Employing s anda dized p o ocols o ensu e compa ison ac oss coun ies, we used na ionally ep esen a i e c oss-sec ional da a om 57 LMICs (N¼464 728) o in es- iga e he associa ion be ween ANC isi s and neona al mo ali y. Cox p opo ional haz- a ds mul i a iable eg ession models and me a- eg ession analysis we e used o analyse pooled da a om he coun ies. Kaplan-Meie su i al cu es we e used o desc ibe he pa e ns o neona al su i al in each egion. Resul s: A e adjus ing o po en ial con ounding ac o s, we ound 55% lowe isk o neona al mo ali y [haza d a io (HR) 0.45, 95% con idence in e al (CI) 0.42–0.48] among women who me bo h WHO ecommenda ions o ANC ( i s isi wi hin he i s imes- e and a leas ou isi s du ing p egnancy) in pooled analysis. Fu he mo e, me a- analysis o coun y-le el isk shows 32% lowe isk o neona al mo ali y (HR 0.68, 95% CI 0.61–0.75) among hose who me a leas one WHO ecommenda ion. In addi ion, ANC a endance was associa ed wi h lowe neona al mo ali y in all he egions excep in he Middle Eas and No h A ica. Conclusions: ANC a endance is p o ec i e agains neona al mo ali y in he LMICs s udied, al hough di e ences exis ac oss coun ies and egions. Inc easing ANC isi s, along wi h o he known e ec i e in e en ions, can imp o e neona al su i al in hese coun ies. Key wo ds: Su i al analysis, neona al mo ali y, an ena al ca e, low- and middle-income coun ies V CThe Au ho 2017; Published by Ox o d Uni e si y P ess on behal o he In e na ional Epidemiological Associa ion 1668 This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion Non-Comme cial License (h p://c ea i ecommons.o g/licenses/by-nc/4.0/), which pe mi s non-comme cial e-use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. Fo comme cial e-use, please con ac jou nals.pe [email protected] In e na ional Jou nal o Epidemiology, 2017, 1668–1677 doi: 10.1093/ije/dyx125 Ad ance Access Publica ion Da e: 13 July 2017 O iginal a icle Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 In oduc ion Goal 4 o he wo ld’s jus -ended de elopmen al agenda, he millennium de elopmen goals (MDGs), ocused on e- duc ion in child mo ali y. 1 As a a ge , goal 4 aimed a educing unde -5 mo ali y by wo- hi ds be ween 1990 and 2015. The MDGs agenda has con ibu ed o a ema k- able educ ion in unde -5 mo ali y globally. 2 As a esul , he global numbe o dea hs o unde -5 child en has declined om 12.7 million in 1990 o 6 million in 2015, al- hough he numbe o bi hs has inc eased o e he pe iod. This decline is no only in e ms o absolu e numbe s bu also in mo ali y a es o child en in his age g oup. These ha e declined, o e he 25-yea pe iod by mo e han 50%, om 90 pe 1000 li e bi hs o 43 pe 1000 li e bi hs. 2 Al hough signi ican achie emen s we e seen in unde -5 mo ali y, neona al mo ali y, which is a main componen o he unde -5 mo ali y, has no seen a signi ican educ- ion o e he same pe iod. This has esul ed in an inc ease in he p opo ion o neona al mo ali y wi hin he unde -5 dea hs o e he wo and a hal decades. In 1990, neona al mo ali y was 37% o unde -5 mo ali y, and by 2012, he igu e ose o 44%: hus a de ici o 2.9 million neona al dea hs globally. 3,4 Conce ed e o is needed o e e se hese ends in o de o imp o e o e all child su i al, pa - icula ly, in low- and middle-income coun ies (LMICs) whe e he bu den o neona al mo ali y is highes . 2–4 Neona al mo ali y also con ibu es o he inequali ies in heal h be ween high-income and low-income coun- ies. 5,6 Neona al mo ali y emains unaccep ably high in mos LMICs and he highes is in sub-Saha an A ica. 7,8 Acco ding o a s udy by Lawn e al., 4 he e is a lag o abou 100 yea s in neona al su i al be ween A ican coun ies on he one hand, and Eu ope and No h Ame ica on he o he . O e all, he slow p og ess made owa ds add essing neona al mo ali y has led o enewed commi men s o ackling he challenge. The sus ainable de elopmen goals (SDGs) a e one o such commi men s which open a new e a o ackle he un inished agenda o he MDGs as well as o he eme ging social, de elopmen al and global heal h challenges. The SDG 3 a ge 3.2 aims a educing neona al mo ali y o 12 pe 1000 li e bi hs by he yea 2030. 9 O he impo an enewed commi men s include he E e y Newbo n Ac ion Plan (ENAP) which seeks o end p e en - able s illbi h and newbo n dea h. 3 Da a om LMICs a e c i ical in o ming in e en ion and acking p og ess o- wa ds achie ing hese goals. An ena al ca e (ANC) o e s p egnan women an oppo - uni y o access p e en i e ca e. In LMICs whe e access o eme gency obs e ic se ices is limi ed, ANC p esen s a i- able op ion o p egnan women o be sc eened o po en- ial isks du ing p egnancy o deli e y. I also p o ides an oppo uni y o ea men and heal h educa ion including nu i ional ad ice and in e en ions abou isky beha iou such as smoking and alcohol cessa ion p og ammes. A ew s udies ha e in es iga ed he e ec o ANC u iliza ion on bi h ou comes, pa icula ly on neona al mo ali y. 10–13 Thei e idence is inconclusi e; whe eas some ound ANC u iliza ion, including he numbe o isi s, o be associa ed wi h educed isk o neona al mo ali y, 10,13 o he s ound ad e se o no ela ions be ween ANC u iliza ion and bi h ou comes. Hollowell e al., 11 conduc ing a sys ema ic e- iew o he associa ion be ween ANC and in an and neo- na al mo ali y in high-income coun ies, concluded ha he e was insu icien e idence ha ANC in e en ions educed neona al o in an mo ali y in ulne able popula- ions. Dowswell e al. 14 on he o he hand, in a ecen sys- ema ic e iew o al e na i e ANC packages o low- isk p egnan women, ound ha in LMICs in pa icula , a educed numbe o ANC isi s was associa ed wi h highe pe ina al mo ali y. Howe e , no such e idence was ound o hose in high-income coun ies no o o he child heal h ou comes. Fo women in LMICs, he Wo ld Heal h O ganiza ion (WHO) ecommends a leas ou ANC isi s o no mal p egnancy, and he i s isi is ecommended o be wi hin he i s 3 mon hs o concep ion. 15 Despi e hese ecommenda ions, s udies ha e epo ed low ANC Key Messages •This pape is he i s o in es iga e, using compa able da a, he associa ion be ween an ena al ca e isi s and neo- na al mo ali y in low- and middle-income coun ies. •In pooled analysis, 55% educed isk o neona al mo ali y was ound among women who me bo h WHO ecommen- da ions ( i s isi wi hin he i s imes e and a leas ou isi s du ing p egnancy). •Me a-analysis o coun y-le el isk shows ha mee ing a leas one an ena al ca e ecommenda ion educed he isk o neona al mo ali y by 32%. •Inc eased ANC isi s, along wi h o he known e ec i e in e en ions, will imp o e neona al su i al in low- and mid- dle-income coun ies. In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 1669 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 co e age among women in LMICs. 13,16 The e is a need o e idence ega ding he bene i s o ANC u iliza ion on im- po an heal h ou comes, o suppo in e en ions which a e gea ed owa ds inc easing ANC u iliza ion in pa icu- la and imp o ing ma e nal, neona al and child heal h in gene al. Compa able na ional ep esen a i e da a om LMICs, in es iga ing his ela ionship, would shed ligh on his. A sys ema ic sea ch o he li e a u e e ealed ha no s udy using compa able da a has in es iga ed he associ- a ion be ween ANC u iliza ion and neona al mo ali y in many LMICs. Ou s udy aimed o in es iga e whe he ANC a endance imp o es neona al su i al in LMICs. We we e also in e es ed in looking a he egional di e - ences in neona al su i al. Me hods E hical app o al o he s udy was g an ed om he ele- an ins i u ions in he a ious coun ies and esponden s ga e w i en consen in all he coun ies. Pa icipan s ga e consen o he da a o be used o publica ion. Da a sou ce We used na ionally ep esen a i e c oss-sec ional da a om 57 LMICs, collec ed om 2005 o 2015, in he mos ecen Demog aphic and Heal h Su ey (DHS). The DHS uses a s anda dized ques ionnai e and me hodology o da a collec ion in o de o acili a e in e na ional compa i- son. De ails o he DHS a e published elsewhe e [h p:// www.dhsp og am.com/da a/da a-collec ion.c m]. In ou analysis we used ‘Bi h eco d’ iles published by he DHS. Pe mission o use he da a was g an ed by Measu e DHS. The da a we e publicly a ailable and no u he pe mission om he espec i e coun ies was equi ed o hei use. S udy popula ion Da a we e collec ed on he ou come o he mos ecen li e bi h wi hin 3 yea s p eceding he su ey o each woman o ep oduc i e age 15–49 yea s (N¼464 728). In Bangladesh, Egyp , Jo dan, Maldi es and Pakis an, he sample was collec ed among e e -ma ied women as opposed o all women (bo h e e -ma ied and o he women) in he o he coun ies included. The 57 coun ies we e g ouped in o six egions mod- elled on he WHO classi ica ion o egions. Ou o 57 coun ies, 33 (63%) we e in A ica, ou we e in Eas Asia & Paci ic (6.9%), six we e in Eu ope & Cen al Asia (2.0%), se en we e in La in Ame ica & Ca ibbean (9.5%), wo we e in Middle Eas & No h A ica (4.6%) and i e we e in Sou h Asia (13.5%). The ull lis o coun ies and su eys is shown in Table 1. Va iables The ou come a iable o his s udy was neona al dea h, de ined as dea h o a li e-bo n du ing he i s 28 comple e days o li e. The in o ma ion on mon h and yea o each bi h, child’s su i al s a us and cu en age o age a dea h, as applicable, was a ailable in he da a. Age a dea h was eco ded in days i he child died wi hin 1 mon h o bi h. The p ima y independen a iables we e he numbe o ANC isi s and iming o i s ANC isi . In his s udy, ANC e e s o heal h ca e se ice p o ided o mo he and e us du ing p egnancy o ensu e bes ou come o bo h. The esponses o he numbe o ANC isi s we e ca ego- ized in o ou : no isi s, one isi , wo o h ee isi s and ou o mo e isi s. The iming o i s ANC isi was ca e- go ized as: no isi , <4 mon hs, 4–5 mon hs, 6–7 mon hs and 8þmon hs. WHO ecommended a leas ou isi s, wi h he i s ANC isi as soon as possible in p egnancy, p e e ably in he i s imes e o heal hy women wi h no unde lying medical p oblem. 15 We he e o e c ea ed a new a iable based on he numbe and iming o i s ANC is- i s: ‘ hose who did no mee he WHO ecommenda ion’, ‘ hose who had he i s ANC isi wi hin he i s imes- e ’, ‘ hose who had a leas ou ANC isi s’ and ‘ hose who me bo h ecommenda ions’. The da a on backg ound cha ac e is ics o he mo he [age, place o esidence, body mass index (BMI), weal h quin ile, child en e e bo n, sex o child and educa ion] we e also a ailable and included in he analysis as co a i- a es in o de o in es iga e he independen associa ion be- ween ANC a endance and neona al mo ali y. The weal h quin ile is he composi e measu e o he house- hold’s cumula i e li ing s anda d based on owne ship o speci ied asse s spli in o quin iles: poo es , poo e , middle, iche and iches . 17 S a is ical analysis We applied sample weigh o es ima e he dis ibu ion o independen and dependen a iables, o accoun o he clus e sampling s uc u e. The weigh ed dis ibu ion o he independen a iables and co a ia es was also p e- sen ed, wi h he numbe o neona al dea hs in each ca - ego y in he o al sample. The DHS conduc ed edi ing and impu a ion o missing p ocedu es be o e he da a we e eleased. We used Cox p opo ional haza d eg ession o es ima e he haza d a ios (HRs) wi h hei 95% con i- dence in e als (CIs) o neona al mo ali y. 1670 In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 Table 1. Dis ibu ion o he numbe o an ena al ca e (ANC) isi s and he iming o he s a o he i s ANC isi s by coun y and egion Coun y N a Numbe o ANC isi s Timing o i s ANC isi s (mon hs) No ca e 1 isi 2–3 isi s 4þ isi s <4 4–5 6–7 8þMedian mon hs A ica Benin (2011–12) 6640 11.6 2.4 23.1 58.3 48.3 27.5 6.7 4.7 3.8 Bu kina Faso (2010) 8320 4.3 4.2 58.4 33.1 40.6 39.9 14.0 1.1 4.3 Bu undi (2010) 4251 0.9 3.5 61.8 33.6 20.4 44.2 31.9 2.3 5.3 Came oon (2011) 6145 15.0 2.3 21.1 60.7 33.1 35.4 14.8 1.2 4.5 Chad (2014–15) 8984 34.3 3.4 28.8 31.3 28.9 26.7 8.0 0.8 4.2 Como os (2012) 1428 6.6 3.7 26.5 49.2 57.2 23.5 7.9 2.9 3.6 Congo (2011–12) 4560 7.2 0.8 14.2 77.2 43.6 39.5 9.4 0.3 4.1 Congo DR (2013–14) 9349 9.9 3.8 38.5 47.2 16.7 40.4 29.7 2.8 5.4 Co e d’I oi e (2011–12) 4055 7.2 9.8 38.9 43.3 29.0 34.4 25.6 3.2 5.0 E hiopia (2011) 5991 57.8 5.1 19.1 17.8 10.2 16.2 12.2 3.3 5.3 Gabon (2012) 2751 4.9 0.6 15.9 77.2 61.2 26.6 6.6 0.4 3.6 Ghana (2014) 3108 2.9 1.3 9.0 86.2 62.1 28.1 6.2 0.6 3.6 Guinea (2012) 3868 12.8 4.4 26.2 56.3 40.1 30.5 14.8 1.8 4.2 Kenya (2014) 10350 4.1 3.6 36.4 55.9 19.7 41.1 32.1 2.9 5.4 Leso ho (2014) 1900 4.7 1.6 20.1 73.3 40.6 33.7 18.2 2.6 4.4 Libe ia (2013) 3429 2.5 1.8 14.7 77.6 66.7 23.4 6.0 0.8 3.3 Madagasca (2008–09) 6484 9.6 4.0 38.3 47.2 25.5 42.4 19.7 1.8 4.8 Malawi (2010) 10668 1.6 2.8 51.3 44.2 12.4 48.0 35.8 2.0 5.6 Mali (2012–13) 5340 24.8 5.3 28.2 40.9 34.0 26.5 11.5 2.6 4.2 Mozambique (2011) 6381 9.1 4.8 35.7 49.5 12.7 46.3 29.0 2.5 5.5 Namibia (2013) 2127 3.3 1.1 10.6 62.4 41.5 38.8 14.3 1.5 4.3 Nige (2012) 6781 13.2 6.2 46.9 33.2 22.0 40.4 22.1 1.9 5.0 Nige ia (2013) 16178 33.7 1.9 10.7 51.0 17.6 29.8 16.7 1.5 5.0 Rwanda (2014–15) 4512 0.7 3.2 51.7 44.3 56.0 30.9 10.8 1.4 3.9 Sao Tome (2008–09) 965 0.7 4.6 15.7 71.8 49.7 30.9 12.7 1.4 3.9 Senegal (2014) 6635 2.8 5.9 43.4 47.1 57.8 27.8 9.4 1.3 3.6 Sie a Leone (2013) 5689 1.6 0.5 9.0 75.8 44.7 41.9 9.8 0.7 4.1 Swaziland (2006–07) 1535 2.8 1.0 15.3 78.6 23.8 48.1 23.5 1.4 5.1 Tanzania (2010) 4316 1.9 3.9 54.0 39.8 18.6 49.0 28.0 2.0 5.2 Togo (2013–14) 3699 7.1 3.3 32.9 56.4 27.0 41.9 21.4 2.5 4.9 Uganda (2011) 3981 4.2 4.1 43.0 47.2 20.7 42.7 29.1 3.1 5.2 Zambia (2013–14) 7044 1.4 2.0 41.6 54.1 24.4 54.3 18.4 1.1 4.8 Zimbabwe (2010–11) 3300 11.3 3.0 22.7 61.9 17.5 38.8 27.0 5.3 5.4 A ica pooled 180759 12.4 3.4 32.8 51.4 36.4 34.9 16.8 1.9 4.5 Eas Asia & Paci ic Cambodia (2014) 4107 3.9 2.9 16.7 75.6 79.2 12.7 3.7 0.4 2.5 Indonesia (2012) 9699 3.0 1.6 7.3 87.4 79.8 12.0 3.8 0.9 2.4 Philippines (2013) 3667 4.0 1.8 10.0 84.3 60.4 28.5 6.0 1.0 3.6 Timo -Les e (2009–10) 4835 12.7 3.5 28.5 54.7 43.8 32.0 10.1 0.9 4.0 Eas Asia pooled 22308 5.5 2.3 14.2 78.0 65.8 21.3 5.9 0.8 3.1 Eu ope & Cen al Asia Albania (2008–09) 792 1.9 3.9 24.9 69.0 79.9 14.4 2.2 1.5 3.0 A menia (2010) 826 0.7 0.2 3.4 93.8 81.3 17.1 0.9 0.0 3.3 Aze baijan (2006) 1195 19.4 7.5 20.7 49.2 56.6 13.4 6.8 2.8 3.4 Ky gyz Republic (2012) 2195 2.7 0.6 9.9 83.9 79.1 15.2 2.7 0.2 3.1 Moldo a (2005) 909 1.6 0.4 4.5 89.3 69.3 21.2 6.0 1.1 3.3 Uk aine (2007) 524 0.5 0.2 1.3 78.8 83.7 11.8 2.0 0.4 2.9 Eu ope pooled 6443 5.3 2.3 11.8 80.7 75.0 15.5 3.4 1.2 3.2 (con inued) In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 1671 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 We c ea ed su i al ime in days: he ime elapsed since bi h in he case o child en who we e s ill ali e, and be- ween bi h and dea h in he case o child en who had died. To calcula e he su i al ime in days, we impu ed 15 h day o he mon h o hose whose days o bi h we e missing. Adjus ed haza d a ios o neona al dea h we e i s cal- cula ed o he o al sample, adjus ing o each o he main independen a iables (numbe o ANC isi s, iming o i s ANC isi and ecommended ANC isi s), he sociode- mog aphic a iables (ma e nal age, place o esidence, BMI, weal h quin ile, child en e e bo n, sex o child and ma e nal educa ion), yea o su ey and coun y. We also es ima ed he coun y- and egional-le el haza d a ios wi h hei 95% CIs o in es iga e he associa ion o neo- na al dea h wi h ANC a endance, adjus ed o all s udied soci-demog aphic a iables. These es ima es we e plo ed using he me a-analysis (me an) command in S a a. Indonesia, he Philippines, Uk aine and A menia we e excluded om he coun y-le el analysis due o ew o no cases o neona al mo ali y pe he ca ego ies o he inde- penden a iable ( ecommended ANC isi s) used. To examine he iming o dea h, we calcula ed he daily haz- a d a es o neona al dea h du ing he i s 28 comple ed days o li e, s a i ied by WHO ecommended ANC com- pliance. Su i al cu es o he in an du ing he i s mon h o li e, s a i ied by he egions o in es iga e he e- gional di e ences in neona al su i al a es, we e also plo - ed. All s a is ical analyses we e conduc ed using STATA/ SE 14.0. Resul s The dis ibu ion o main independen a iables, ANC a - endance (numbe o ANC isi s and iming o i s ANC isi s) o each s udied coun y and pooled alues o he egions a e p esen ed in Table 1. Adjus ed haza d a ios o he associa ions be ween neona al mo ali y, ANC a end- ance and o he ma e nal and demog aphic a iables a e shown in Table 2. The mul i a iable model shows ha hose who had wo o h ee ANC isi s (HR 0.70, 95% CI 0.64–0.77) and ou o mo e isi s (HR 0.54, 95% CI 0.50–0.59) had lowe isk o neona al dea h han hose who had no ca e. The isk o neona al dea h among hose who had i s ANC isi s be o e 4 mon hs was 41% lowe (HR 0.59, 95% CI 0.54–0.65) han hose who had no isi . When we combined he wo ecommenda ions in o one a iable, we ound ha he isk o neona al dea h was Table 1. Con inued Coun y N a Numbe o ANC isi s Timing o i s ANC isi s (mon hs) No ca e 1 isi 2–3 isi s 4þ isi s <4 4–5 6–7 8þMedian mon hs La in Ame ica & Ca ibbean Boli ia (2008) 4651 9.6 3.4 15.2 71.5 60.4 19.2 9.1 1.4 3.3 Colombia (2010) 8639 3.3 1.0 7.0 87.7 74.9 16.2 4.7 0.7 2.7 Dominican Republic (2013) 1945 0.5 0.3 2.4 95.4 81.1 14.6 3.3 0.3 2.6 Guyana (2009) 918 2.1 1.0 5.5 77.4 47.3 32.3 13.2 2.1 4.0 Hai i (2012) 3848 9.8 4.1 20.2 65.3 57.2 22.3 9.4 1.2 3.6 Hondu as (2011–12) 5825 3.3 1.7 6.1 88.9 76.3 15.3 4.2 0.9 2.8 Pe u (2012) 4923 1.5 0.7 3.5 94.2 73.6 17.6 6.3 0.9 2.9 La in Ame ica pooled 30750 4.6 1.8 9.0 84.6 67.3 19.6 7.2 1.1 3.1 Middle Eas Egyp (2014) 12579 8.7 0.6 6.6 82.8 76.3 11.1 2.8 0.9 2.6 Jo dan (2012) 4841 0.9 0.6 4.0 94.5 90.7 6.3 1.8 0.3 2.3 Middle Eas pooled 17420 6.9 0.7 6.2 86.2 83.5 8.7 2.3 0.6 2.5 Sou h Asia Bangladesh (2014) 4621 21.4 17.9 29.4 31.2 24.3 19.3 13.3 5.8 5.1 India (2005–06) 28655 23.0 6.0 33.5 37.0 43.0 22.7 8.3 2.3 3.8 Maldi es (2009) 1977 0.1 0.3 1.7 85.6 91.4 6.3 1.3 0.3 1.8 Nepal (2011) 2847 14.4 6.2 27.8 51.6 50.3 26.5 7.1 1.8 3.7 Pakis an (2012–13) 5765 22.6 13.3 26.8 37.1 42.6 15.4 13.1 6.2 3.7 Sou h Asia pooled 43866 21.3 8.2 31.0 39.4 50.3 18.0 8.6 3.3 3.6 All coun ies Pooled 301546 8.6 3.3 22.9 62.4 58.0 17.4 8.0 2.9 3.4 a Numbe o esponden s in each coun y and egion. 1672 In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 Table 2. Associa ion o neona al dea h wi h an ena al ca e (ANC) isi s and o he demog aphic cha ac e is ics a he mos e- cen bi h N a Weigh ed pe cen age Numbe o neona al dea hs Adjus ed HR b (95% CI) P- alue Numbe o ANC isi s No isi s 40939 13.4 1262 (3.1) 1.0 1 isi 11029 3.6 318 (2.9) 0.97 (0.83–1.12) 0.700 2–3 isi s 79927 26.2 1745 (2.2) 0.70 (0.64–0.77) <0.001 4þ isi s 172892 56.7 2764 (1.6) 0.54 (0.50–0.59) <0.001 Timing o i s ANC isi s (mon hs) No isi s 36637 12.3 1176 (3.2) 1.0 <4 mon hs 125865 42.4 2165 (1.7) 0.59 (0.54–0.65) <0.001 4–5 mon hs 87275 29.4 1714 (2.0) 0.62 (0.56–0.68) <0.001 6–7 mon hs 41662 14.0 721 (1.7) 0.58 (0.52–0.65) <0.001 8þmon hs 5399 1.8 118 (2.2) 0.55 (0.42–0.71) <0.001 WHO ecommenda ions o ANC isi s Had no me any ecommenda ions 110257 36.3 2694 (2.4) 1.0 Had i s ANC isi wi hin i s imes e 20918 6.9 573 (2.7) 0.74 (0.67–0.83) <0.001 Had a leas 4 ANC isi s 67945 22.3 1172 (1.7) 0.49 (0.46–0.53) <0.001 Has me bo h ecommenda ions 104946 34.5 1592 (1.5) 0.45 (0.42–0.48) <0.001 Ma e nal age g oup 15–19 32226 7.0 1233 (3.8) 1.0 20–24 120172 26.1 3770 (3.1) 0.68 (0.62–0.74) <0.001 25–29 134372 29.2 3480 (2.6) 0.46 (0.42–0.50) <0.001 30–34 91130 19.8 2403 (2.6) 0.39 (0.35–0.43) <0.001 35–39 55857 12.1 1675 (3.0) 0.40 (0.35–0.45) <0.001 40þ26877 5.8 956 (3.6) 0.45 (0.40–0.52) <0.001 A ea o esidence U ban 143975 31.3 3705 (2.6) 1.0 Ru al 316659 68.7 9812 (3.1) 1.02 (0.96–1.07) 0.463 BMI (kg/m 2 ) <18.5 42835 13.5 1521 (3.6) 1.11 (1.04–1.18) <0.001 18.5–24.9 198767 62.6 5780 (2.9) 1.0 25.0–28.9 53460 16.8 1496 (2.8) 1.10 (1.04–1.17) 0.001 30.0 22543 7.1 605 (2.7) 1.20 (1.09–1.31) <0.001 Weal h quin ile Poo es 109491 23.8 3435 (3.1) 1.0 Poo e 101460 22.0 3108 (3.1) 1.07 (1.01–1.14) 0.017 Middle 94152 20.4 2751 (2.9) 1.15 (1.08–1.23) <0.001 Riche 85573 18.6 2442 (2.9) 1.17 (1.09–1.25) <0.001 Riches 69960 15.2 1780 (2.5) 1.25 (1.15–1.36) <0.001 Child en e e bo n 1 child 76051 16.5 1690 (2.2) 1.0 2–3 190792 41.4 5372 (2.8) 1.30 (1.20–1.40) <0.001 4–5 103245 22.4 2977 (2.9) 1.58 (1.44–1.74) <0.001 6 90546 19.7 3476 (3.8) 2.21 (1.99–2.46) <0.001 Sex o child Male 234331 50.9 7778 (3.3) 1.0 Female 226304 49.1 5739 (2.5) 0.75 (0.72–0.78) <0.001 Ma e nal educa ion No educa ion 165692 36.0 5683 (3.4) 1.0 P ima y 142532 30.9 4165 (2.9) 0.90 (0.85–0.95) <0.001 Seconda y o mo e 152412 33.1 3668 (2.4) 0.94 (0.89–1.00) 0.058 a F equency dis ibu ion o each o he independen a iables in he o al sample. The o al o each o he a iables may no be same because o he missing in- o ma ion in ANC isi s a iables. b Adjus ed o all o he ac o s in able and o he coun y and yea o su ey a iables. In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 1673 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 26% lowe (HR 0.74, 95% CI 0.67–0.83) among women who had i s ANC isi wi hin he i s imes e o ges a- ion, 51% lowe (HR 0.49, 95% CI 0.46–0.53) among hose who had a leas ou ANC isi s and 55% lowe (HR 0.45, 95% CI 0.42–0.48) among hose who me bo h ecommenda ions compa ed wi h hose who did no mee any o he ecommenda ions. The isk o neona al mo ali y among hose who had a leas one o he wo WHO ecommended ANC isi com- pa ed wi h hose who did no mee any o he wo ecom- menda ions, s a i ied by coun y and egions and so ed om lowes o highes isk, a e p esen ed in Figu e 1. Coun y-le el analysis shows ha , in mos coun ies, ANC a endance is associa ed wi h lowe isk o neona al Figu e 1. Adjus ed haza d a ios (HRs) and hei 95% CI o he isk o neona al mo ali y among hose who me a leas one ANC ecommenda ion. The model was adjus ed o ma e nal age, a ea o esidence, BMI, weal h quin ile, child en e e bo n, sex o child and ma e nal educa ion. 1674 In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 mo ali y among hose who me a leas one WHO ecom- menda ion compa ed wi h hose who did no . Simila ly, pe egions, a lowe isk o neona al mo ali y was ound in all he egions among hose who me a leas one ANC ecommenda ion compa ed wi h hose who did no . The excep ion was in he Middle Eas & No h A ica egion (HR 0.97, 95% CI 0.65–1.45) whe e no di e ence was ound be ween hese g oups. The me a-analysis o he coun y-le el isk shows 32% lowe isk o neona al dea h among hose who me a leas one ecommenda ion (HR 0.68, 95% CI 0.61–0.75) han hose who did no . The haza d was highes in he i s week o li e. I dec eased sha ply a e he i s week (Figu e 2). In an s bo n o mo he s who did no mee he WHO ecommenda- ions had g ea e excess haza d o dying du ing he i s mon h o li e han hose bo n o mo he s who ul illed bo h o a leas one ecommenda ion. The isk was lowes among hose who me bo h ecommenda ions. The Eu ope and Cen al Asia egion expe ienced be e su i al, whe eas he Sou h Asia egion had he wo s su i al (Figu e 3). The egional di e ences in he neona al su i al we e s a is ically signi ican (P<0.001). Discussion Using a unique and compa able da a se (N¼464 728) col- lec ed om 2005 o 2015, we conduc ed su i al analysis on he in luence o ANC a endance on neona al dea h in 57 LMICs. ANC a endance is p o ec i e agains neona al mo ali y in mos o he coun ies s udied. A e adjus ing o po en ial con ounding ac o s in pooled analysis, we ound 55% lowe isk o neona al mo ali y among women who me bo h WHO ecommenda ions o ANC. Fu he mo e, ANC a endance was ound o be p o ec i e agains neona al mo ali y in all he egions excep in he Middle Eas and No h A ican egion whe e no di e ence was ound. Addi ionally, me a-analysis shows a 32% educed isk o neona al mo ali y among women who me a leas one WHO ecommenda ion in he LMICs s udied. As expec ed, we ound a g ea e haza d du ing he ea ly neona al pe iod, simila o ha ound by an ea lie s udy which explo ed o he isk ac o s o neona al mo ali y in some LMICs. 18 None heless, child en bo n o mo he s who did no mee any o he WHO ecommenda ions, o me jus one o hem, had excess isk h oughou he neo- na al pe iod compa ed wi h hose who me bo h ecommenda ions. ANC isi s a e po en ially bene icial o bo h he baby and he mo he . ANC is an oppo uni y o p omo ing heal hy beha iou among mo he s as well as p omo ion o pa en ing skills, which a e pa icula ly impo an o new mo he s. In 2016, he WHO in oduced new guidelines o ANC, which ecommend a minimum o eigh ANC con ac s du - ing p egnancy and he i s con ac is ecommended o be wi hin he i s 12 weeks o ges a ion. 19 The goal o his new guideline is o ully use he oppo uni y o p o iding ANC o sa e li es. Ou indings sugges ha his new ec- ommenda ion is impo an , imely and can con ibu e o ending some p e en able neona al dea hs in LMICs. Fu he mo e, in LMICs whe e HIV in ec ion is high, he ANC isi o e s mo he s he oppo uni y o be sc eened o he i us, and hose who es posi i e can be ea ed o p e en mo he o child ansmission. This migh also con- ibu e o neona al su i al. Women who a end ANC migh be ad ised o seek ou skilled deli e y ca e o ins i u- ional deli e y, as well as pos na al ca e (PNC). I has been shown ha women who a end ANC a e mo e likely o ha e ins i u ional ca e du ing labou . 20 An associa ion be- ween ANC a endance and skilled deli e y ca e has also been epo ed, 21 al hough wi h some excep ions. 22 Bo h Figu e 2. Daily haza d o neona al mo ali y o in an s du ing he i s mon h o li e, s a i ied by ecommended ANC isi o mo he . Figu e 3. Daily su i al cu es om neona al mo ali y o in an du ing he i s mon h o li e s a i ied by he egion. In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 1675 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017 ins i u ional deli e y and skilled deli e y ca e can con ib- u e o imp o ing neona al ou come. This could explain ou inding o an associa ion be ween ANC a endance and neona al su i al. In addi ion, ou inding o highe child mo ali y in he i s ew days o li e emphasizes he im- po ance o PNC alongside ANC. In his espec , he qual- i y o ANC is essen ial. I is unknown whe he women who me he ecommended ANC isi s also ecei ed hese ec- ommended se ices o whe he he se ice p o ide s ha e he adequa e knowledge, skills, supplies and equipmen o deli e high quali y ca e. This s udy unde sco es he c ucial impo ance o ANC o child su i al in LMICs. Apa om he ole o ANC a - endance in p edic ing neona al mo ali y, we also explo ed he ole o o he sociodemog aphic ac o s, including ma e nal educa ion and BMI, in he isk o neo- na al dea h in LMICs. These indings a e simila o hose epo ed in ea lie s udies. 18,23 A numbe o s udies ha e also epo ed some o hese sociodemog aphic ac o s as p edic o s o ANC. 13,16 Ou indings, o independen associa ion be ween ANC a endance and neona al mo ali y in ela ion o hese ea - lie s udies, emphasize he complexi y o he isk o neo- na al dea h, which should in o m in e en ions in LMICs. Ea lie s udies ha e iden i ied a numbe o ba ie s o ANC u iliza ions, including cos and p o ide ba ie s, in de eloping coun ies. 24,25 Al hough ba ie s migh be coun y-speci ic, he e idence o subs an ial educ ion in neona al mo ali y isk by ANC a endance, ound in he p esen s udy, means ha conscious e o s a add essing hese ba ie s a local, na ional and egional le els a e ne- cessa y o imp o e neona al heal h in LMICs. We ound egional di e ences: ANC a endance was ound o be p o- ec i e agains neona al mo ali y in all he egions excep in he Middle Eas and No h A ican egion. Gi en ha da a om only wo coun ies in his egion we e a ailable and included in his s udy, he in e p e a ion o he e- gional di e ences should be done wi h cau ion. In he DHS, an ena al heal h ca e u iliza ion was sel - epo ed and may be biased by social desi abili y wi hin he socie y whe e he women li ed. Also, ANC a endance was measu ed o bi hs wi hin he pas yea s p eceding he s udy; hence i may be a ec ed by ecall bias. Wha cons i u es ANC in each coun y may a y–ANC migh no necessa ily mean ca e by skilled heal h pe sonnel, and hus he esul should be in e p e ed wi hin his con ex . Causal in e ence is limi ed due o he c oss-sec ional na u e o he da a. Fu he mo e, da a on bi hs and dea hs om which he neona al mo ali y was es ima ed we e epo ed e ospec - i ely by mo he s and could be subjec o ecall bias. Howe e p e ious s udies, which alida ed such measu es in e ospec i e and longi udinal su eys, ound hem o be accu a e. 26 Ini ial assessmen o he heal h da a in he DHS-I sugges s ha hey a e accu a e es ima es. 27 We did no in es iga e he quali y o an ena al ca e in his s udy, as he ocus was mainly on he associa ion be ween ANC a endance and neona al mo ali y. Fu u e s udies, which will explo e he quali y o ANC deli e y in LMICs and how i ela es o neona al mo ali y as well as he aspec s o ANC ha a e bene icial o neona al su i al, will con- ibu e u he o he p esen indings. Such s udies should also explo e he link be ween ANC and PNC and hei e - ec s on neona al su i al in hese coun ies. Conclusions We conclude ha ANC is p o ec i e agains neona al mo - ali y in he LMICs s udied, al hough di e ences exis ac oss he coun ies and by egion. Ou s udy p o ides a comp ehensi e o e iew on he associa ion be ween ANC isi s and neona al mo ali y in nea ly hal o he wo ld’s LMICs. This s udy con ibu es o he li e a u e on he sub- jec and cla i ies he impo ance o ANC isi s on a heal h ou come o global impo ance. To add ess he huge bu den o neona al mo ali y in hese coun ies, i is impo an o inc ease ANC co e age and a endance. The da a a e publicly a ailable a [h p://dhsp og am. com/Da a/]. Pe mission o use he da a is equi ed om Measu e DHS. Acknowledgemen s We would like o hank M Jani Rai anen o he School o Heal h Sciences, Uni e si y o Tampe e, Finland, and M Geo ge Adjei o he Kin ampo Heal h Resea ch Cen e, Ghana, o hei in ol emen in he alida ion o he da a. Au ho s con ibu ions S.N. and D.T.D. concep ualized he s udy and de eloped he analy - ical s a egy. D.T.D. and S.N. conduc ed he s a is ical analysis and d a ed he manusc ip . Bo h au ho s did he c i ical e ision o he manusc ip and app o ed he inal e sion. Con lic o in e es : None decla ed. Re e ences 1. Uni ed Na ions. O icial Lis o MDG Indica o s. 2008. h p:// mdgs.un.o g/unsd/mdg/Hos .aspx?Con en ¼indica o s/o icial- lis .h m (29 July 2016, da e las accessed). 2. Uni ed Na ions. Millennium De elopmen Goals Repo 2015. h p://www.un.o g/millenniumgoals/2015_MDG_Repo /pd / MDG%202015%20 e %20%28July%201%29.pd (28 July 2016, da e las accessed). 1676 In e na ional Jou nal o Epidemiology, 2017, Vol. 46, No. 5 Downloaded om h ps://academic.oup.com/ije/a icle-abs ac /46/5/1668/3960260 by Tampe e uni e si y lib a y use on 30 No embe 2017