Predictors and pathways of language and motor development in four prospective cohorts of young children in Ghana, Malawi, and Burkina Faso

P edic o s and pa hways o language and mo o de elopmen in ou p ospec i e coho s o young child en in Ghana, Malawi, and Bu kina Faso Elizabe h L. P ado, 1 Souheila Abbeddou, 1 Se h Adu-A a wuah, 2 Ma y A imond, 1 Pe Asho n, 3,4 Ulla Asho n, 3 Jaden Bendabenda, 3,5 Kenne h H. B own, 1,6 Sonja Y. Hess, 1 Emma Ko ekangas, 3 Anna La ey, 2 Kenne h Male a, 5 B ie a M. Oaks, 1 Eugenia Ocansey, 1,2 Ha ie Ok onipa, 1,2 Jean Bosco Ou ed aogo, 7 Anna Pulakka, 3,8 J e ^ ome W. Som e, 1,7 Ch is ine P. S ewa , 1 Robe C. S ewa , 9 S ephen A. Vos i, 10 Elizabe h Yakes Jimenez, 11 and Ka h yn G. Dewey 1 1 Depa men o Nu i ion, Uni e si y o Cali o nia Da is, Da is, CA, USA; 2 Depa men o Nu i ion and Food Science, Uni e si y o Ghana, Legon Acc a, Ghana; 3 Cen e o Child Heal h Resea ch, School o Medicine and Tampe e Uni e si y Hospi al, Uni e si y o Tampe e, Tampe e; 4 Depa men o Paedia ics, Tampe e Uni e si y Hospi al, Tampe e, Finland; 5 School o Public Heal h and Family Medicine, Uni e si y o Malawi College o Medicine, Blan y e, Malawi; 6 Bill & Melinda Ga es Founda ion, Sea le, WA, USA; 7 Ins i u de Reche che en Sciences de la San  e/DRO, Bobo-Dioulasso, Bu kina Faso; 8 Depa men o Public Heal h, Uni e si y o Tu ku and Tu ku Uni e si y Hospi al, Tu ku, Finland; 9 Di ison o Psychia y, Uni e si y o Edinbu gh, Edinbu gh, UK; 10 Depa men o Ag icul u al and Resou ce Economics, Uni e si y o Cali o nia Da is, Da is, CA; 11 Cen e o Educa ion Policy Resea ch, Uni e si y o New Mexico, Albuque que, NM, USA Backg ound: P e ious e iews ha e iden i ied 44 isk ac o s o poo ea ly child de elopmen (ECD) in low- and middle-income coun ies. Fu he unde s anding o hei ela i e in luence and pa hways is needed o in o m he design o in e en ions a ge ing ECD. Me hods: We conduc ed pa h analyses o ac o s associa ed wi h 18-mon h language and mo o de elopmen in ou p ospec i e coho s o child en who pa icipa ed in ials conduc ed as pa o he In e na ional Lipid-Based Nu ien Supplemen s (iLiNS) P ojec in Ghana (n=1,023), Malawi (n=675 and 1,385), and Bu kina Faso (n=1,122). In wo coho s, women we e en olled du ing p egnancy. In wo coho s, in an s we e en olled a 6 o 9 mon hs. In mul iple linea eg ession and s uc u al equa ion models (SEM), we examined 22 ou o 44 ac o s iden i ied in p e ious e iews, plus 12 addi ional ac o s expec ed o be associa ed wi h ECD. Resul s: Ou o 42 indica o s o he 34 ac o s examined, 6 we e associa ed wi h 18-mon h language and/o mo o de elopmen in 3 o 4 coho s: child linea and ponde al g ow h, a ie y o play ma e ials, ac i i ies wi h ca egi e s, die a y di e si y, and child hemoglobin/i on s a us. Fac o s ha we e no associa ed wi h child de elopmen we e indica o s o ma e nal Hb/i on s a us, ma e nal illness and in lamma ion du ing p egnancy, ma e nal pe cei ed s ess and dep ession, exclusi e b eas eeding du ing 6 mon hs pos pa um, and child dia hea, e e , mala ia, and acu e espi a o y in ec ions. Associa ions be ween socioeconomic s a us and language de elopmen we e consis- en ly media ed o a g ea e ex en by ca egi ing p ac ices han by ma e nal o child biomedical condi ions, while his pa e n o mo o de elopmen was no consis en ac oss coho s. Conclusions: Key elemen s o in e en ions o ensu e quali y ECD a e likely o be p omo ion o ca egi e ac i i ies wi h child en, a a ie y o play ma e ials, and a di e se die , and p e en ion o al e ing in linea and ponde al g ow h and imp o emen in child hemoglobin/i on s a us. Keywo ds: Language de elopmen ; mo o de elopmen ; isk ac o s; low- and middle-income coun ies; s imula ion; nu i ion; g ow h; lipid-based nu ien supplemen s; iLiNS P ojec . In oduc ion In high-income coun ies, ea ly li e expe iences, such as in au e ine g ow h es ic ion, p e e m bi h, and ea ly childhood educa ion ha e been shown o ha e long- e m cogni i e consequences (Fox, Le i , & Nelson, 2010). Child en in low and middle-income coun ies (LMICs) bea a g ea e bu den o ea ly li e isk ac o s o poo ea ly child de elopmen (ECD) (Walke e al., 2011). Fou sys- ema ic and quali a i e e iews ha e iden i ied 44 modi iable isk ac o s o poo ECD in LMICs (Jensen e al., 2015; Wachs & Rahman, 2013; Walke e al., 2007, 2011). Fu he unde s anding o he ela i e in luence o hese ac o s on child de elopmen and hei pa hways is needed o in o m he design o in e en ions. Dis al and p oximal isk ac o s, which cooccu in he con ex o en i onmen al ad e si y, may in lu- ence child de elopmen h ough mul iple pa hways. En i onmen al ac o s, such as limi ed household esou ces and poo quali y wa e and sani a ion, may a ec child de elopmen di ec ly, o h ough a ec ing ma e nal o child heal h o nu i ional s a us, o h ough in luencing ca egi ing p ac ices. Child heal h and nu i ional s a us may also in luence child de elopmen di ec ly o h ough child en’s physical ac i i y o h ough ca egi e beha io . This la e pa hway is based on he Con lic o in e es s a emen : See Acknowledgemen s o disclosu es. ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s 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. Jou nal o Child Psychology and Psychia y 58:11 (2017), pp 1264–1275 doi:10.1111/jcpp.12751 PFI_12mmX178mm.pd + eps o ma unc ional isola ion hypo hesis, which posi s ha child en who a e malnou ished, o en ill, o small o hei age a e less ac i e explo ing hei en i onmen , and/o ecei e less de elopmen al s imula ion om ca egi e s, bo h o which can a ec child de elop- men (B own & Polli , 1996). This pa hway also e lec s Same o ’s ansac ional model, in which he child’s e ec on he en i onmen is as impo an as he en i onmen ’s e ec on he child (Same o , 2009). En i onmen al ad e si y may also cause high le els o ch onic s ess, and he e o e ch onic expo- su e o ele a ed s ess ho mones, which can a ec b ain de elopmen (Lupien, McEwen, Gunna , & Heim, 2009). Eigh een p e ious s udies epo ed media ion o pa hway analyses o ac o s associa ed wi h child de elopmen in low- and middle-income coun ies ( o de ails, see supplemen a y ma e ial a ailable online only). Ou o hese 18 s udies, he maximum numbe o ac o s examined was se en, only one s udy included ma e nal in o ma ion ga he ed du ing p eg- nancy, and none examined he ela i e di ec and indi ec associa ions o mul iple en i onmen al, ma e nal, ca egi ing, and child ac o s wi h ECD. We examined 22 ou o he 44 ac o s iden i ied in p e ious e iews, plus 12 addi ional ac o s expec ed o be associa ed wi h ECD (Table S1). We analyzed 42 indica o s o hese 34 ac o s in ou coho s o child en (n=4,205) who pa icipa ed in he In e na- ional Lipid-Based Nu ien Supplemen s (iLiNS) P o- jec in Ghana, Malawi, and Bu kina Faso. The i s objec i e was o iden i y he ac o s associa ed wi h 18-mon h language and mo o de elopmen in each iLiNS coho . The second objec i e was o iden i y he pa hways h ough which hese ac o s ope a e. Ha - moniza ion o many o he a iables ac oss coho s allowed eplica ion o esul s in di e en con ex s. Based on se e al p e ious heo e ical pa h models (B own & Polli , 1996; Engle, Menon, Ga e , & Slack, 1997; Walke e al., 2007), we de eloped a concep ual pa h model o po en ial in luences on child de elopmen (Figu e 1). We es ed he ollowing pa hways, which co espond o he labels o he a ows in Figu e 1. Pa hway (1): Socioeconomic dispa i ies and o he en i onmen al e ec s on child de elopmen may be media ed by (1.1) ma e nal ac o s, (1.2) ca egi ing p ac ices, (1.3) child ac o s, o (1.4) may di ec ly a ec child de elopmen . Pa h- way (2): Ma e nal e ec s on child de elopmen may be media ed by (2.1) child ac o s, (2.2) ca egi ing p ac ices, o (2.3) may di ec ly a ec child de elop- men . Pa hway (3.1): E ec s o in an eeding p ac- ices on child de elopmen may be media ed by child nu i ional s a us. Pa hway (3.2): Ca egi ing p ac- ices may di ec ly a ec child de elopmen . Pa hway (4.1): Child en who a e anemic, small o hei age, o o en ill may elici di e en s imula ion om ca e- gi e s, which may media e e ec s on child de elop- men . Pa hway (4.2): Child en who a e anemic, small o hei age, o o en ill may be less ac i e, esul ing in less explo a ion o he en i onmen , which may media e e ec s on child de elopmen . Pa hway (4.3): Child en who a e s un ed o anemic may ha e highe basal co isol, which may a ec child de elopmen . Pa hway (4.4): Child illness and nu i ional s a us may di ec ly a ec child de elopmen . Me hods iLiNS P ojec T ial Designs In he iLiNS-DYAD-G ial in Ghana (n=1,320) and he iLiNS-DYAD-M ial in Malawi (n=869) p egnan women we e en olled be o e 20 weeks o ges a ion. In he iLiNS- DOSE ial in Malawi (n=1,932) and iLiNS-ZINC ial in Bu kina Faso (n=3,220) in an s we e en olled a age 6 and 9 mon hs, espec i ely. All pa icipan s we e assigned o ecei e a ious doses and o mula ions o lipid-based nu i- en supplemen s (LNS), o o con ol g oups un il age 18 mon hs, when child de elopmen was assessed (Adu- A a wuah e al., 2015; Asho n e al., 2015; Hess e al., 2015; Male a e al., 2015). The e ec s o he in e en ions on 18- mon h child de elopmen di e ed ac oss ials, wi h posi i e e ec s in Bu kina Faso (P ado, Abbeddou, Yakes Jimenez e al., 2016), bu no in Ghana (P ado, Adu-A a wuah e al., 2016) o Malawi (P ado, Male a e al., 2016; P ado, Phuka e al., 2016). E hical app o al o he s udy p ocedu es was ob ained om he Ins i u ional Re iew Boa d o he Uni e - si y o Cali o nia Da is o he E hics Commi ee a Tampe e Uni e si y, Finland as well as local ins i u ions in each coun y. All pa icipan s p o ided w i en in o med consen . Fo u he in o ma ion, see Appendix S1 o ex ended supplemen a y me hods. Pa icipan s In he pa h analyses epo ed he e, we included all child en who pa icipa ed in de elopmen al assessmen a age 18 mon hs, comp ising 1,023 child en in iLiNS-DYAD-G (Coho A), 675 in iLiNS-DYAD-M (Coho B), 1,385 in iLiNS- DOSE (Coho C), and a andom subsample o 1,122 child en in iLiNS-ZINC (Coho D). The easons o loss o ollow-up and he baseline cha ac e is ics o he de elopmen al samples compa ed o hose who en olled bu we e no assessed ha e been published p e iously (P ado, Abbeddou, Yakes Jimenez e al., 2016; P ado, Adu-A a wuah e al., 2016; P ado, Male a e al., 2016; P ado, Phuka e al., 2016). Fo he la e , Coho A Figu e 1 Concep ual model ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. doi:10.1111/jcpp.12751 Pa h analyses o child de elopmen in A ica 1265 was he only coho in which no signi ican di e ences we e ound; howe e , in coho s C and D, he absolu e di e ences we e small (0.1 SD/2 o 6 pe cen age poin s), hus, hey a e likely o li le p ac ical signi icance. In Coho B, en olled mo he s whose child en did no pa icipa e had highe BMI and household asse s, we e signi ican ly younge , and a g ea e p opo ion we e p imipa ous. P ocedu e De ailed epo s o he da a collec ion p ocedu es in each ial ha e been published elsewhe e; he e o e, we summa ize he p ocedu e o collec ion o a iables ha we e used in he analyses p esen ed he e. Table S2 p esen s u he de ails o he da a collec ion p ocedu es and a iable de ini ions. Fig- u e 2 shows he da a collec ion schedule o each a iable in each coho . Sociodemog aphic in o ma ion was ga he ed a en olmen . Ma e nal and/o child an h opome y was measu ed a mul- iple ime poin s. Capilla y o enous blood samples we e collec ed om mo he s and/o child en a mul iple ime poin s o he assessmen o (a) mala ia using a apid diagnos ic es , (b) bioma ke s o hemoglobin/i on s a us, including hemoglo- bin (Hb) concen a ion (g/dl), zinc p o opo phy in (ZPP) con- cen a ion (lmol/mol heme), and soluble ans e in ecep o (sT R, mg/l), and (c) bioma ke s o in lamma ion, including C- eac i e p o ein (CRP; mg/l) and alpha-1 glycop o ein (AGP; g/ l) concen a ions. In ZINC and DYAD-G, known HIV in ec ion was an exclusion c i e ion, bu HIV was no es ed, he e o e HIV s a us o women who en olled was unknown. In DOSE, HIV s a us was also unknown. In DYAD-M, women we e es ed o HIV a en ollmen . Ma e nal and/o child sali a samples in DYAD-M and DYAD-G we e collec ed a se e al ime poin s o he mea- su emen o co isol concen a ion (nmol/l) and ma e nal sel - epo ed s ess was measu ed in DYAD-M a mul iple ime poin s using he Pe cei ed S ess Scale (Oaks e al., 2016; C. P. S ewa e al., 2015). Mo he s we e in e iewed ega ding dep essi e symp oms a 6 mon hs pos pa um in DYAD-M using a locally alida ed adap a ion o he Sel - Repo ing Ques ionnai e (SRQ) and in DYAD-G wi h he Edinbu gh Pos -na al Dep ession Scale (EPDS) (R. C. S ewa e al., 2009). In DYAD-M, a 6 mon hs pos pa um, ma e nal cogni ion was assessed using digi span o wa d and back- wa d, e bal luency, men al o a ion, and unc ional heal h li e acy es s (P ado, Ullman, Muadz, Alcock, & Shanka , 2012). Child en we e isi ed weekly o mo bidi y su eillance. A hese isi s, ca egi e s we e asked whe he he child expe i- enced any illness symp oms, including e e , dia hea, omi - ing, cough, nasal discha ge, o espi a o y dis ess du ing he pas 7 days and/o da a collec o s measu ed he child’s au icula empe a u e. P e alence and/o incidence o dia - hea, e e , mala ia, and/o acu e espi a o y in ec ion we e calcula ed (Table S2). In DOSE and DYAD-M, physical ac i i y a age 18 mon hs was measu ed o e 1 week wi h he hip-wo n Ac iG aph GT3X+accele ome e (Pensacola, FL, USA) (Pulakka e al., 2015). In an eeding p ac ices we e assessed a mul iple ime poin s h ough quali a i e 24-h and/o 7-day die a y ecall ques ionnai es (A imond e al., 2016). In DYAD-M, nu u ing and s imula ion we e measu ed a 6 mon hs pos pa um using he Home Obse a ion o he Measu emen o he En i onmen (HOME) In en o y (Caldwell & B adley, 2003). In all ou ials, de elopmen al s imula ion was measu ed a age 18 mon hs using he Family Ca e Indica o s (FCI) in e iew (Hamadani e al., 2010). The mo he was in e - iewed wi h ega d o he a ie y o play ma e ials and ac i i ies in which adul s engaged wi h he child in he pas 3 days. Figu e 2 Da a collec ion imeline in each coho . Abb e ia ions: CDI: communica i e de elopmen in en o y, IYCF-FPQ: in an and young child eeding— eeding p ac ices ques ionnai e, DMC: de elopmen al miles ones checklis , FCI: amily ca e indica o s, Hb: hemoglobin, HOME, home obse a ion o he measu emen o he en i onmen in en o y, KDI: Kili i de elopmen al in en o y, LNS: lipid-based nu ien supplemen s, PSS: pe cei ed s ess scale ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. 1266 Elizabe h L. P ado e al. J Child Psychol Psychia 2017; 58(11): 1264–75 De elopmen al assessmen in Bu kina Faso In ZINC, we assessed mo o and language de elopmen a age 18 mon hs using he De elopmen al Miles ones Checklis -II (DMC-II) (P ado e al., 2014). This ool e alua es mo o and language de elopmen h ough bo h in e iewing he ca egi e and obse ing he child. De elopmen al assessmen in Malawi and Ghana In he DOSE and DYAD ials, we assessed mo o de elopmen by he Kili i De elopmen al In en o y (KDI), which is a di ec assessmen o he child (Abubaka , Holding, an Baa , New- on, & an de Vij e , 2008). Language de elopmen was assessed using a 100-wo d ocabula y checklis by ma e nal in e iew based on he MacA hu -Ba es Communica i e De elopmen In en o y (Fenson e al., 2007). We de eloped he checklis in he local languages h ough in e iews wi h ca egi e s and pilo es ing (P ado, Adu-A a wuah e al., 2016; P ado, Male a e al., 2016). The KDI and ocabula y sco es showed high in e - a e ag eemen and mode a e o high es – e es eliabili y in all ials (Table S3). Analysis We calcula ed z-sco es on all con inuous a iables so ha he coe icien s can be in e p e ed as he change in 1 SD o he dependen a iable wi h e e y 1 SD change in he independen a iable. Fo u he de ails, see Appendix S1: Supplemen a y Me hods. S ep 1: Va iable selec ion. Fi s , we examined inde- penden associa ions be ween each p edic o and each de el- opmen al sco e and d opped any ha we e no associa ed a p<.05. Second, we examined bi a ia e associa ions be ween p edic o s o check o collinea i y. I wo a iables we e highly collinea ( >.6), we d opped he one ha was less s ongly associa ed wi h he de elopmen al sco e. Thi d, we examined six mul i a ia e models and d opped any a iables ha we e no associa ed a p<.05. The six mul i a ia e models we e: en i onmen al ac o s; ma e nal ac o s; ca egi ing ac o s; child mo bidi y and nu i ional s a us; child physical ac i i y; and a ing o he child’s beha io du ing assessmen . S ep 2: Pa h selec ion. We examined he associa ion be ween each pai o a iables on each pa hway in Figu e 1, o de e mine which a iables we e po en ial media o s. Unidi ec- ional a ows ep esen pa hways es ed. Bidi ec ional a ows ep esen associa ions ha we e no modeled in he pa h analysis, excep o check o collinea i y. I any a iables we e no associa ed a p<.05, we d opped ha pa hway. All analyses up o his poin we e conduc ed using SAS e sion 9.4 (SAS Ins i u e, Ca y, NC). Nex , o each independen a iable wi h po en ial media o s, we es ed he mul iple media ion model using S a a e sion 14.1 (S a aCo p, College S a ion, TX) bina y media ion p og am including all po en ial media o s oge he . Fo each po en ial media o o which he indi ec e ec was signi ican , we included he media ion pa hway in he o e all pa h model. Finally, we an he inal pa h model using he sem command in S a a wi h he mlm op ion o es ima e he model on he ull da ase using maximum likelihood es ima ion o missing alues. All models wi h language o mo o sco e as he dependen a iable included ou co a ia es: ial g oup, child sex and age a de elopmen al assessmen , and de elopmen al da a collec o . Fo language sco es, we also included whe he he child was exposed o mo e han one language. Since he in e en ion a ec ed de elopmen al sco es in ZINC only, we addi ionally examined he po en ial media o s o in e en ion e sus con ol g oup in his coho . We epo p- alues unco ec ed and co ec ed o mul iple compa isons using he Benjamini–Hochbe g co ec ion (Benjamini & Hochbe g, 1995), which is ecommended o con olling he alse disco - e y a e in SEM (C ibbie, 2007). We applied he co ec ion sepa a ely o each model de ined by one ou come in one coho . Fo u he de ails, see Appendix S1: Supplemen al Me hods. Resul s Summa y s a is ics o all independen a iables a e p esen ed in Table S4. The a iable selec ion esul s a e shown in Tables S5–S8. The pa hway selec ion esul s a e shown in Tables S9–S12. Among en i onmen al a iables, language sco es we e posi i ely associa ed wi h highe household asse s (DYAD-G), households loca ed u he om he ma ke (ZINC), an imp o ed wa e sou ce (ZINC), o highe pa e nal educa ion (DYAD-M, DOSE). Mo o sco es we e associa ed wi h highe household asse s (ZINC), households loca ed close o he ma ke (DOSE), an imp o ed wa e sou ce (DOSE), o highe pa e nal educa ion (DOSE). O he ma e nal a iables, we obse ed highe language sco es among child en o mo he s who we e olde (DYAD-G), alle (DYAD-M, ZINC), o had highe cogni i e sco es (DYAD-M). Highe mo o sco es we e associa ed wi h ma e nal heigh (DOSE, ZINC), highe BMI (ZINC), lowe s ess (basal co i- sol) a 36 weeks ges a ion (DYAD-M), highe heal h li e acy (DYAD-M), o in lamma ion (highe CRP) a <20 weeks ges a ion (DYAD-G). This las associa ion was in he opposi e di ec ion han expec ed. Among he ca egi ing a iables, we obse ed highe language sco es among child en who we e ed mo e equen ly (ZINC), had g ea e die a y di e si y (DYAD-G, DOSE, ZINC), and ecei ed highe quali y s imula ion om he home en i on- men , as indica ed by 6-mon h HOME sco es (DYAD- M), 18-mon h a ie y o play ma e ials (all ou coho s), o 18-mon h ac i i ies wi h ca egi e s (DYAD-G, DOSE, ZINC). Highe mo o sco es we e associa ed wi h g ea e die a y di e si y (ZINC), and highe de elopmen al s imula ion, as indica ed by 18-mon h a ie y o play ma e ials (DYAD-M, DOSE, ZINC) and ac i i ies wi h ca egi e s (DYAD-G, DOSE, ZINC). O he child a iables, highe language sco es we e associa ed wi h ha ing olde siblings (DYAD-G, DYAD-M), LAZ in ea ly in ancy (bi h/6/9 mon hs; DYAD-G, DOSE, ZINC), g ea e linea g ow h in la e in ancy (bi h/6/9 o 18 mon hs; DYAD-G, DOSE, ZINC), WAZ a 9 mon hs (ZINC), o highe Hb/i on s a us, as indica ed by Hb a 6 mon hs (DYAD-G, DYAD-M), change in Hb om 6 o 18 mon hs (DYAD- G, DYAD-M), o ZPP a 9 mon hs (ZINC). Mo o sco es we e associa ed wi h LAZ in ea ly in ancy (bi h/6/9 mon hs; all ou coho s), change in LAZ in la e in ancy (bi h/6/9 o 18 mon hs; all ou coho s), ponde al g ow h s a us in ea ly in ancy ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. doi:10.1111/jcpp.12751 Pa h analyses o child de elopmen in A ica 1267 (bi h/6/9 mon hs; DYAD-G, DOSE, ZINC), ponde al g ow h in la e in ancy (bi h/6 o 18 mon hs; DYAD- G, DOSE), in lamma ion (lowe AGP) a 18 mon hs (DYAD-M), s ess (lowe basal co isol) a 18 mon hs (DYAD-M), physical ac i i y a 18 mon hs (DOSE), o Hb/i on s a us as indica ed by Hb a 9 mon hs (ZINC), change in Hb 9 o 18 mon hs (ZINC), ZPP a 6/9 mon hs (DOSE and ZINC), o change in ZPP om 6 o 18 mon hs (DOSE). Posi i e beha io du ing he KDI assessmen also s ongly p edic ed mo o sco es and exposu e o mul iple languages was posi i ely associa ed wi h language sco es. The inal models o language de elopmen accoun ed o 18% o he a iance in language sco es in DYAD-G, 14% in DYAD-M, 14% in DOSE, and 25% in ZINC. The inal models o mo o de elop- men accoun ed o 21% o he a iance in mo o sco es in DYAD-G, 36% in DYAD-M, 31% in DOSE, and 36% in ZINC. Figu es 3 and 4 p esen he coe icien s o he di ec and indi ec e ec s ha we e signi ican a p<.1 in he inal SEM. All coe icien s a e p esen ed in Tables S13 and S14. Rega ding pa hway (1), he ex en o which socioeconomic dispa i ies in language de elopmen we e media ed by o he ac o s a ied ac oss coho s, om 22% o 85% (Figu e S1). Associa ions be ween socioeconomic s a us (SES) and language sco es we e la gely media ed by ca egi ing p ac- ices (22%–76%) a he han by ma e nal o child ac o s (0%–20%). Fo language, ma e nal age was he only ma e nal ac o ha was a media o (DYAD-G), and LAZ in ea ly in ancy (bi h/6/ 9 mon hs) was he only child ac o ha was a media o (DYAD-G, DOSE, ZINC), while ou Figu e 3 Pa h diag am o language de elopmen in ou iLiNS coho s ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. 1268 Elizabe h L. P ado e al. J Child Psychol Psychia 2017; 58(11): 1264–75 ca egi ing ac o s we e media o s o SES (all ou coho s). Fo mo o de elopmen , socioeconomic dispa i ies we e ound in wo coho s and 21%– 86% o hese associa ions we e media ed by o he ac o s. In ZINC, hese associa ions be ween SES and mo o sco es we e media ed o a g ea e ex en by ca egi ing p ac ices (39%) han child ac o s (18%), while in DOSE hey we e media ed o a g ea e ex en by child ac o s (21%–56%) han ca egi ing p ac ices (0%–10%). Rega ding pa hway (2), he ex en o which asso- cia ions be ween ma e nal nu i ional s a us (heigh and BMI) and language o mo o de elopmen we e media ed e sus di ec e ec s also a ied by coho (0%–100%). These associa ions we e media ed by child an h opome ic s a us in ea ly in ancy (LAZ a 6/9 mon hs: 29%–86% and BMIZ a 9 mon hs: 10%–21%), and o a smalle deg ee by ca egi ing p ac ices (6%–20%). O he ma e nal ac o s ela ed o language de elopmen (age and cogni ion in DYAD-G and DYAD-M, espec i ely) we e media ed o a small ex en by ca egi ing (10%–20%), bu we e la gely di ec associa ions. O he ma e nal ac o s ela ed o mo o de elopmen (CRP in DYAD-G, heal h li e acy and la e ges a ion co isol in DYAD- M) we e di ec associa ions. Associa ions be ween in an eeding p ac ices and language o mo o de elopmen (pa hway 3) we e di ec associa ions and we e no ound o be medi- a ed by child nu i ional s a us. Resul s om he Figu e 4 Pa h diag am o mo o de elopmen in ou iLiNS coho s ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. doi:10.1111/jcpp.12751 Pa h analyses o child de elopmen in A ica 1269 ZINC coho suppo ed pa hway (4.1). Child en wi h poo e nu i ional s a us (linea g ow h, ponde al g ow h, Hb) we e p o ided wi h ewe play ma e ials and in e ac ed less wi h ca egi e s, media ing 21%– 24% o associa ions wi h language sco es and 14%– 30% o associa ions wi h mo o sco es. In DYAD-M, child en wi h lowe LAZ a bi h we e also p o ided wi h ewe play ma e ials, media ing 8% o he associa ion wi h mo o de elopmen . Resul s om he DOSE coho suppo ed pa hway (4.2). Child en whose linea and ponde al g ow h s a us (z-sco es), and i on s a us (ZPP) declined mo e apidly om 6 o 18 mon hs we e less ac i e a 18 mon hs, media ing 5%–19% o hese associa- ions wi h mo o de elopmen . We did no ind e idence ha child co isol media ed he associa ion be ween any o he child ac o s and de elopmen (pa hway 4.3). O e all, ca egi ing and child ac o s showed s onge associa ions wi h child mo o and language de elopmen , as compa ed o en i onmen al and ma e nal ac o s, e en when conside ing bo h indi- ec and di ec e ec s (Figu e 5). Ra ings o he child’s beha io du ing he KDI assessmen we e he s onges p edic o o he child’s mo o sco e, wi h coe icien s anging om .50 o .67 (Table S14; hese coe icien s a e no included in Figu e 5). Child beha io also media ed associa ions o ou o he ac o s wi h mo o de elopmen : pa e nal educa ion, unimp o ed wa e , linea g ow h (DOSE), and in lamma ion a age 18 mon hs (DYAD-M). Two child ac o s and wo ca egi ing ac o s medi- a ed he e ec o he iLiNS-ZINC in e en ion on de elopmen . Child en in he in e en ion g oup had g ea e linea g ow h om 9 o 18 mon hs, media - ing 13% o he e ec o he in e en ion on language and 23% o he e ec on mo o sco es. They also had g ea e inc ease in Hb om 9 o 18 mon hs, medi- a ing 16% o he e ec o he in e en ion on mo o de elopmen . Child en in he in e en ion g oup had a highe a ie y o play ma e ials, media ing 12% o he e ec o he in e en ion on language and 16% o he e ec on mo o sco es, as well as highe eeding equency, media ing 11% o he e ec on language sco es. Fo bo h domains, he di ec unmedia ed e ec o he in e en ion was s ong and signi ican (45%–64%). A e co ec ing p- alues o mul iple compa isons, only one coe icien o language and six coe icien s o mo o sco es inc eased om <.05 o >.05, in all cases emaining <.1. These coe icien s a e speci ied in Figu es 3 and 4. Discussion and Conclusion We conduc ed pa h analyses o 34 en i onmen al, ma e nal, ca egi ing, and child ac o s hypo hesized o be associa ed wi h 18-mon h language and mo o de elopmen in ou la ge p ospec i e coho s o child en in Ghana, Malawi, and Bu kina Faso. Ou o 42 indica o s o hese 34 ac o s examined, h ee indica o s eme ged as consis en p edic o s o bo h language and mo o sco es in h ee o ou coho s: linea g ow h h oughou in ancy and 18-mon h a ie y o play ma e ials and ac i i ies wi h ca e- gi e s. Fo language de elopmen , wo o he consis- en p edic o s in h ee o ou coho s we e die a y di e si y and indica o s o child Hb/i on s a us (Hb in DYAD-G and DYAD-M; ZPP in ZINC). Fo mo o de elopmen , one o he consis en p edic o in h ee coho s was ponde al g ow h s a us in ea ly in ancy (BMIZ o WLZ a bi h/6/9 mon hs). O e all, ca e- gi ing and child ac o s showed s onge and mo e consis en associa ions wi h child mo o and lan- guage de elopmen han en i onmen al and ma e - nal ac o s, e en when conside ing bo h indi ec and di ec e ec s. While speci ic pa hways di e ed ac oss coho s, in gene al di ec e ec s we e s on- ge and mo e p e alen han indi ec e ec s. A s eng h o he s udy was he a ailabili y o da a om ou coho s o child en in h ee A ican coun- ies, wi h many a iables ha monized ac oss coho s, enabling he eplica ion o esul s. O he s eng hs we e he la ge numbe o child en in each coho and he la ge numbe o ac o s examined (34 ac o s, as compa ed o a maximum o 7 ac o s in p e ious s udies). Ano he s eng h was ha he coho s we e en olled ei he du ing p egnancy o a 6 o 9 mon hs pos pa um and ollowed p ospec i ely h ough 18 mon hs pos pa um, allowing hese isk ac o s o be examined h oughou mos o he i s 1,000 days o b ain de elopmen . A weakness was ha we examined associa ions and he e o e we e no able o es ablish causali y. Ano he limi a ion was ha we we e no able o explo e ce ain isk ac o s o ime poin s in hese coho s due o lack o da a, o example, de elop- men al s imula ion in ea ly in ancy was measu ed only in one coho , o because hese coho s did no include child en exposed o ce ain isk ac o s. Ou s udy is also suscep ible o alse posi i e indings due o mul iple compa isons, howe e , we ha e mi iga ed his isk by epo ing p- alues co ec ed o mul iple compa isons and by de i ing ou con- clusions om indings ha we e eplica ed ac oss mul iple coho s. In addi ion, hese samples may no be ep esen a i e o hei popula ions due o he supplemen s hey ecei ed. Howe e , we would expec ha any bias in oduced by he supplemen- a ion would end o dec ease a he han inc ease he magni ude o he associa ions and he e o e ou gene al conclusions ega ding consis en signi ican p edic o s would no change. Fo example, in Malawi and Ghana, we ound ha ma e nal LNS a enua ed he associa ion o linea g ow h wi h mo o and language sco es (P ado, Abbeddou, Adu-A a wuah e al., 2016). Ou indings in o m he design o in e en ions o pu sue Uni ed Na ions’ Sus ainable De elopmen ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. 1270 Elizabe h L. P ado e al. J Child Psychol Psychia 2017; 58(11): 1264–75 Goal 4.2 o ‘ensu e ha all gi ls and boys ha e access o quali y ea ly childhood de elopmen so ha hey a e eady o p ima y educa ion.’ The consis en esul s om hese ou coho s sugges ha key aspec s o in e en ions o p omo e quali y ECD a e likely o be p omo ion o ca egi e ac i i ies wi h child en, p o ision o a a ie y o play ma e i- als, and p o ision o a di e se die , as well as Figu e 5 Coe icien s o di ec and indi ec e ec s on language and mo o de elopmen in he inal models. No e. Coe icien size ep esen s he change in SD o he language o mo o sco e wi h 1 SD change in he independen a iable, adjus ed o all o he a iables in he model ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. doi:10.1111/jcpp.12751 Pa h analyses o child de elopmen in A ica 1271 p e en ion o al e ing in linea and ponde al g ow h and imp o emen in Hb/i on s a us. Since ou analysis canno de e mine causali y, andomized ials should be conduc ed e alua ing la ge-scale p og ams ha ocus on hese key elemen s. Cos analyses should also be included o iden i y he mos cos -e ec i e s a egies o p omo ing ECD. Among ac o s ha we e only measu ed in one o wo coho s, ac o s ha we e associa ed wi h language de elopmen we e ma e nal cogni ion and equency o child eeding, while ac o s ha we e associa ed wi h mo o de elopmen we e ma e nal heal h li e - acy, ma e nal and child basal co isol, 18-mon h in lamma ion, and 18-mon h physical ac i i y, sug- ges ing ha hese a e impo an ac o s o u he esea ch. A me a-analysis o 21 in e en ions aimed a enhancing de elopmen al s imula ion and 18 in e - en ions ha p o ided supplemen al nu i ion in child en unde age 2 yea s in LMICs ound ha s imula ion had a medium e ec size o .42 and supplemen al nu i ion had a small e ec size o .09 on child de elopmen (Aboud & Yousa zai, 2015). Toge he wi h ou indings, his highligh s he impo ance o a ge ing de elopmen al s imu- la ion om bi h o age 2 yea s, e en be o e child en s a p eschool. Many LMICs ha e begun o in es in p og ams o p omo e adequa e nu i- ion du ing he i s 1,000 days, howe e , ew la ge-scale p og ams a ge ing de elopmen al s im- ula ion cu en ly exis . O he 21 s imula ion in e en ions e iewed, mos had small sample sizes, wi h only h ee s udies wi h >350 child en. P og ams o enhance de elopmen al s imula ion need o be de eloped and e alua ed ha can be implemen ed on a la ge scale. Despi e ou inding ha die a y di e si y, child g ow h, and Hb/i on s a us we e consis en ly associa ed wi h child de elopmen in hese coho s, he inconsis en e ec o LNS on child de elopmen in ou s udies and he ela i ely small e ec o supplemen a ion ac oss he 18 s udies e iewed (Aboud & Yousa zai, 2015) demons a es a need o de elop new nu i- ion in e en ions ha a e mo e e ec i e o enhance child de elopmen . The compa able mag- ni ude o he coe icien s o bo h ca ego ies o ac o s highligh s he impo ance o p og ams in eg a ing bo h nu i ion and de elopmen al s im- ula ion in ea ly childhood. The ea ly eme gence o socioeconomic dispa i ies in child de elopmen in hese coho s also unde - sco es he impo ance o in e en ions a ge ing his ea ly pe iod. Ou indings a e consis en wi h se e al s udies in LMICs showing ha de elopmen al sco es o child en in low-SES households di e ged om hose o child en in high-SES households e en be o e age 2 yea s (Fe nald, Ka ige , Hid obo, & Ge le , 2012; Hamadani e al., 2014). While he esul s o mo o de elopmen we e no consis en ac oss ou coho s, socioeconomic dispa i ies in language de elopmen we e media ed o a g ea e ex en by ca egi ing p ac ices han by ma e nal o child ac o s. This is consis en wi h h ee p e ious s udies in LMICs showing ha indica o s o de el- opmen al s imula ion in he home media ed a g ea e pe cen age o he associa ion o cogni i e de elop- men wi h SES han media ed by linea g ow h (Fe nald e al., 2012; Hamadani e al., 2014; Rubio-Codina, A anasio, & G an ham-McG ego , 2016). Fac o s ha we e no associa ed wi h language o mo o de elopmen in hese coho s we e indica o s o ma e nal educa ion, low ma e nal Hb/i on s a us, ma e nal illness and in lamma ion, ma e nal sel - epo ed s ess and dep ession, lack o exclusi e b eas eeding, and child dia hea, e e , mala ia, and acu e espi a o y in ec ions. One possible expla- na ion is ha ea ly language and mo o de elopmen may be esilien o hese isk ac o s in hese coho s. Ano he possibili y is ha inaccu acy in measu e- men makes i di icul o de ec hese associa ions. Fo example, al hough he alidi y o he ma e nal dep ession ques ionnai es was igo ously e alua ed in Malawi (R. C. S ewa , Uma , Tomenson, & C eed, 2013), he Pe cei ed S ess Scale was no alida ed in he local con ex and he e o e may no ha e accu- a ely measu ed ma e nal pe cep ions o s ess. Addi- ionally, while posi i e associa ions ha e been shown be ween child de elopmen and b eas eeding e sus no b eas eeding in p e ious s udies (Dee, Li, Lee, & G umme -S awn, 2007), in ou samples b eas eed- ing was almos uni e sal and he e o e we we e no able o examine his con as . In conclusion, in e en ions o p omo e quali y ECD in LMICs should begin in ea ly li e. In e en- ions a ge ing ca egi ing p ac ices, such as encou - aging ca egi e ac i i ies wi h child en, p o ision o a a ie y o play ma e ials and a di e se die , a e likely o ha e posi i e e ec s on ea ly language and mo o de elopmen and educe inequali ies. Randomized ials o la ge-scale p og ams a ge ing hese aspec s o ca egi ing should be conduc ed, including e alu- a ing cos -e ec i eness. P e en ion o al e ing in linea and ponde al g ow h and imp o emen in Hb/ i on s a us a e also key a ge s o in e en ion. Fu he wo k is equi ed o design nu i ion in e - en ions o enhance hei e ec i eness o imp o e child de elopmen . Suppo ing in o ma ion Addi ional Suppo ing In o ma ion may be ound in he online e sion o his a icle: Appendix S1: Supplemen al Me hods. Table S1. Fac o s associa ed wi h child de elopmen in low- and middle-income coun ies iden i ied in ou p e ious e iews and hose included in ou analysis. Table S2. Desc ip ion o a iables. Table S3. In e sco e ag eemen and es – e es eliabili y o he de elopmen al assessmen s in Malawi and Ghana. ©2017 The Au ho s. Jou nal o Child Psychology and Psychia y published by John Wiley &Sons L d on behal o Associa ion o Child and Adolescen Men al Heal h. 1272 Elizabe h L. P ado e al. J Child Psychol Psychia 2017; 58(11): 1264–75