Bancala i, An onella; Be nal, Ped o; Celhay, Pablo A.; Ma inez, Sebas ian; Sánchez,
Ma ia Deni
Wo king Pape
An ounce o p e en ion o a pound o cu e: Basic heal h
ca e and e iciency in heal h sys ems
IDB Wo king Pape Se ies, No. IDB-WP-1231
P o ided in Coope a ion wi h:
In e -Ame ican De elopmen Bank (IDB), Washing on, DC
Sugges ed Ci a ion: Bancala i, An onella; Be nal, Ped o; Celhay, Pablo A.; Ma inez, Sebas ian;
Sánchez, Ma ia Deni (2024) : An ounce o p e en ion o a pound o cu e: Basic heal h ca e
and e iciency in heal h sys ems, IDB Wo king Pape Se ies, No. IDB-WP-1231, In e -Ame ican
De elopmen Bank (IDB), Washing on, DC,
h ps://doi.o g/10.18235/0005669
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/299428
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/3.0/igo/
A
n ounce o p e en ion o a pound o cu e:
Basic heal h ca e and e iciency in heal h
sys ems
A
n onella Bancala i
Ped o Be nal
Pablo Celhay
Sebas ian Ma inez
Ma ia Deni Sánchez
WORKING PAPER No IDB-WP-1231
In e -
A
me ican De elopmen Bank
Social P o ec ion and Heal h Di ision
Janua y 2024
A
n ounce o p e en ion o a pound o cu e:
Basic heal h ca e and e iciency in heal h
sys ems
A
n onella Bancala i
Ped o Be nal
Pablo Celhay
Sebas ian Ma inez
Ma ia Deni Sánchez
In e -
A
me ican De elopmen Bank
Social P o ec ion and Heal h Di ision
Janua y 2024
Ca aloging-in-Publica ion da a p o ided by he
In e -Ame ican De elopmen Bank
Felipe He e a Lib a y
An ounce o p e en ion o a pound o cu e: basic heal h ca e and e iciency in
heal h sys ems / An onella Bancala i, Ped o Be nal, Pablo Celhay, Sebas ián
Ma ínez, Ma ia Deni Sánchez.
p. cm. — (IDB Wo king Pape Se ies ; 1231)
Includes bibliog aphic e e ences.
1. Communi y heal h se ices-El Sal ado . 2. P e en i e heal h se ices-El
Sal ado . 3. P ima y heal h ca e-El Sal ado . I. Bancala i, An onella. II. Be nal,
Ped o. III. Celhay, Pablo. IV. Ma ínez, Sebas ián. V. Sánchez, Ma ia Deni.
VI. In e -Ame ican De elopmen Bank. Social P o ec ion and Heal h Di ision.
VII. Se ies.
IDB-WP-1231
h p://www.iadb.o g
Copy igh © 2024 In e -Ame ican De elopmen Bank ("IDB"). This wo k is subjec o a C ea i e
Commons license CC BY 3.0 IGO (h ps://c ea i ecommons.o g/licenses/by/3.0/igo/legalcode). The
e ms and condi ions indica ed in he URL link mus be me and he espec i e ecogni ion mus be
g an ed o he IDB.
Fu he o sec ion 8 o he abo e license, any media ion ela ing o dispu es a ising unde such license
shall be conduc ed in acco dance wi h he WIPO Media ion Rules. Any dispu e ela ed o he use o
he wo ks o he IDB ha canno be se led amicably shall be submi ed o a bi a ion pu suan o he
Uni ed Na ions Commission on In e na ional T ade Law (UNCITRAL) ules. The use o he IDB's name
o any pu pose o he han o a ibu ion, and he use o IDB's logo shall be subjec o a sepa a e
w i en license ag eemen be ween he IDB and he use and is no au ho ized as pa o his license.
No e ha he URL link includes e ms and condi ions ha a e an in eg al pa o his license.
The opinions exp essed in his wo k a e hose o he au ho s and do no necessa ily e lec he iews o
he In e -Ame ican De elopmen Bank, i s Boa d o Di ec o s, o he coun ies hey ep esen .
[email p o ec ed]
www.iadb.o g/SocialP o ec ion
An ounce o p e en ion o a pound o cu e:
Basic heal h ca e and e iciency in heal h sys ems *
An onella Bancala i, Ped o Be nal, Pablo Celhay,
Sebas ian Ma inez and Ma ia Deni Sánchez
Abs ac
We examine he e iciency gains in heal h sys ems gene a ed a e he na ional oll ou o
basic heal hca e in El Sal ado be ween 2010 and 2013. Using da a om o e 120 million
consul a ions and i e million hospi aliza ions, we demons a e ha he expansion o com-
muni y heal h eams, comp ising less-specialized heal h wo ke s, inc eases p e en i e ca e
and dec eases cu a i e ca e and p e en able hospi aliza ions. We also es ima e co e age
imp o emen s o p e iously una ended ch onic condi ions amenable o e ec i e p ima y
ca e. These esul s sugges ha decen aliza ion o asks o less-specialized heal h wo ke s
imp o es e iciency, main aining quali y o ca e.
JEL: I15, I18, H21, H51.
Keywo ds: communi y-based heal hca e, e iciency, co e age.
*Bancala i: Ins i u e o Fiscal S udies and IZA Ins i u e o Labo Economics (e-mail: an-
[email p o ec ed]); Be nal: In e -Ame ican De elopmen Bank (e-mail: [email p o ec ed]);
Celhay (co esponding au ho ): Escuela de Gobie no and Ins i u o de Economía, Pon i icia Uni e sidad Ca ólica
de Chile (e-mail: [email p o ec ed]); Ma inez: 3ie (e-mail: [email p o ec ed]); Sanchez: In e -Ame ican
De elopmen Bank (e-mail:[email p o ec ed]).
We hank he Minis y o Heal h in El Sal ado and he Salud Mesoame ica Ini ia i e o assis ance and unding o
his s udy. Bancala i g a e ully acknowledges inancial suppo om he RSE–Fulb igh Visi o Schola Fellowship,
and Celhay om ANID, FONDECYT Regula 1221461, ANID, and PIA/PUENTE AFB220003. We a e g a e ul o
Ma ía Fe nanda Ga cía Agudelo and Ma ias Muñoz o ou s anding esea ch assis ance, and pa icipan s a he
LACEA-LAMES Annual Mee ing o use ul commen s. All opinions in his pape a e hose o he au ho s and do no
necessa ily ep esen he iews o he Go e nmen o El Sal ado , o he In e -Ame ican De elopmen Bank, i s
Execu i e Di ec o s, o he go e nmen s hey ep esen .
1 In oduc ion
The subs an ial g ow h in heal hca e spending in ecen decades has b ough he e iciency o
heal h sys ems in o he spo ligh (Hall and Jones,2007;Ga be and Skinne ,2008;Ch is opou-
los and Ele he iou,2020). A no able sou ce o ine iciency is he unde u iliza ion o p ima y ca e,
leading o an o e eliance on hospi al se ices o illnesses ha could be mo e e ec i ely p e-
en ed o managed h ough p ima y ca e (Da ny and G ube ,2005;Ga be and Skinne ,2008;
Alexande e al.,2019;Pinchbeck,2019). In low- and middle-income coun ies (LMICs), hos-
pi al ca e accoun s o app oxima ely 60% o go e nmen heal hca e expendi u e (Pin o e al.,
2018). In de eloped coun ies, was e ul spending in heal h has become a signi ican sou ce o
ine iciency (OECD,2017). To his end, he eo ganiza ion and/o expansion o basic heal hca e
co e age as an al e na i e ha e ga ne ed signi ican in e es (Eina and Finkels ein,2023).
We s udy e iciency gains om a supply-side expansion o basic heal hca e h ough a na-
ionwide e o m in El Sal ado , which es ablished Communi y Heal h Teams (CHTs). These
eams, comp ising physicians, nu ses, and communi y heal h wo ke s, o e a ange o p e en-
i e heal h se ices, including ou pa ien consul a ions, home isi s, and communi y ou each
ac i i ies. While i is well unde s ood ha communi y-based heal hca e can imp o e heal h ou -
comes, less is known abou how i can imp o e e iciency in heal hca e p o ision in low-income
con ex s.
We i s documen he e ec s o he e o m on he a ailable inpu s o heal hca e deli e y (i.e.,
heal h uni s and heal hca e wo ke s) and he supply o p e en i e ca e se ices. To s udy e i-
ciency, we hen es ima e he e ec s o CHTs on he p oduc ion o wo ypes o se ices: (i) cu a-
i e consul a ions o condi ions ha can be e ec i ely managed, ea ed, o p e en ed h ough
p ima y ca e, e e ed o as amenable cu a i e consul a ions; and (ii) hospi aliza ions o con-
di ions ha can be e ec i ely managed, ea ed, o p e en ed h ough p ima y heal hca e and,
i no p ope ly add essed, could lead o unnecessa y hospi aliza ions o complica ions, e e ed
o as p e en able hospi aliza ions.
A key and no el aspec o ou s udy in ol es ca ego izing ou pa ien ca e p oduc ion in o p e-
en i e and cu a i e componen s. In ou con ex , mos ou pa ien ca e is p o ided in a p ima y
ca e se ing, wi h oughly 95% o p e en i e consul a ions and 80% o cu a i e consul a ions
happening a his le el. As CHTs complemen adi ional p ima y ca e uni se ices wi h home
isi s and communi y ou each, unde s anding how he p oduc ion o ou pa ien ca e se ices
changes wi h he p omo ion o CHTs p o ides a mo e comp ehensi e assessmen o he eco-
nomic alue o basic ca e co e age. Ou da ase enables his analysis, ma king ou s udy as
one o he pionee s in explo ing he e ogenei y in he p oduc ion o ou pa ien ca e based on he
ype o ca e p o ided.
Addi ionally, we can p ecisely di e en ia e be ween cu a i e ca e o communicable diseases
2
(CDs) and non-communicable diseases (NCDs) in ou pa ien isi s by u ilizing ICD-10 diag-
nosis codes. This p ecision is aluable as basic heal hca e has he po en ial o p e en CDs
and expedi e he imely diagnosis and ea men o NCDs, he eby educing he likelihood o
complica ions, cos ly hospi aliza ions, and ad e se heal h ou comes.
We employ an e en -s udy s a egy exploi ing he s agge ed oll-ou o he CHT sys em ac oss
municipali ies be ween 2010 and 2013.1We use a panel da ase o 254 municipali ies, combin-
ing a ious sou ces o de ailed, high-quali y adminis a i e eco ds and census da a. We con-
s uc his da ase agg ega ing o e 120 million indi idual consul a ion eco ds ha ook place
be ween 2009 and 2018, and almos i e million inpa ien hospi al eco ds spanning 2005–2018,
allowing us o measu e p e en i e and cu a i e consul a ions and hospi aliza ions pe 1,000
inhabi an s o each municipali y and yea . The du a ion o ou da a enables us o analyze
dynamics o a pe iod o up o 5 yea s ollowing he implemen a ion o he e o m.
We use he me hod p oposed by Bo usyak e al. (2024) o explo e dynamics o ea men e ec s
when policies a e olled ou a di e en pe iods in di e en a eas. B ie ly, he me hod de ines
g oups o municipali ies acco ding o he pe iod ha hey we e ea ed and es ima es coun e -
ac uals o each ea ed g oup using impu a ion p ocedu es and elying on no -ye ea ed and
ne e - ea ed uni s as con ols a each poin o ime.
Ou indings show ha CHTs imp o e e iciency in heal h sys ems h ough a ask-shi ing model,
whe ein ce ain asks a e app op ia ely delega ed o skilled ye less-specialized heal h wo ke s,
such as nu ses ins ead o doc o s.2CHTs no only expanded p e en i e ca e bu also educed
amenable cu a i e consul a ions o CDs, while enhancing managemen o NCDs, ul ima ely
dec easing p e en able hospi aliza ions.
Speci ically, we ind ha CHTs inc eased p e en i e consul a ions by 36.9% compa ed o ini ial
le els. The g ea e supply o p e en i e heal hca e se ices and ou each e o s we e su icien
o o e come demand-side ba ie s cha ac e is ic o LMICs (Dupas,2011).
The e ec o CHTs on cu a i e heal hca e is a p io i ambiguous. On one hand, inc eased p e-
en i e ca e could lead o ewe amenable cu a i e consul a ions, he eby sa ing esou ces, as
he la e end o be mo e expensi e (e iciency e ec ). On he o he hand, he e could be an
inc ease in cu a i e isi s, especially in he sho e m, i he e is a high demand o heal hca e
o una ended condi ions (co e age e ec ) (Hennessy,2008;Glaze and McGui e,2012). Ou
da a allows o disen angle bo h e ec s by di iding cu a i e ou pa ien consul a ions acco ding o
ype. We ind ha CHTs dec eased amenable cu a i e consul a ions o CDs by 8.5%, while in-
c easing co e age o amenable cu a i e consul a ions o NCDs by 17.9%, compa ed o ini ial
le els.
1Municipali ies a e he lowes ju isdic ional le el in El Sal ado . An a e age municipali y had 22,000 inhabi an s
by he 2007 Census.
2This is in line wi h he Wo ld Heal h O ganiza ion’s de ini ion o ‘ ask-shi ing’ (Campbell and Sco ,2011).
3
Such imp o emen s in case managemen wi hin p ima y ca e uni s led o a 10.8% educ ion in
p e en able hospi aliza ions ollowing he c ea ion o he CHTs. These educ ions we e p ima ily
d i en by a 13.6% d op in p e en able hospi aliza ions o diagnosis ela ed o CDs equi ing
inpa ien ea men , compa ed o ini ial le els.
We p o ide e idence ha he ollou o he CHT sys em ac oss municipali ies is unco ela ed
wi h po en ial con ounding ac o s ha would in alida e ou empi ical s a egy o iden i y ea -
men e ec s. Al hough he Minis y o Heal h p io i ized implemen ing CHTs in poo e munici-
pali ies, he ac ual s a o hei ac i i ies depended on he speed a which eams egis e ed he
popula ion, which could ha e been slowe in a ge ed poo e municipali ies. We demons a e
ha he esul s emain obus when con olling o ini ial municipal po e y by yea ixed e ec s
and also p esen e idence suppo ing he pa allel ends assump ion. Fu he mo e, we show
ha he esul s wi hs and se e al sensi i i y checks and he use o al e na i e wo-way ixed
e ec s es ima o s.
Finally, while we lack da a o s udy causal e ec s on heal h ou comes, we ind ha he expan-
sion o he CHT sys em is associa ed wi h a educ ion in he mo ali y a e o CDs ha can
be e ec i ely managed, ea ed, o p e en ed h ough ou pa ien ca e, e e ed o amenable
mo ali y.
We con ibu e o h ee main li e a u e s eams. Fi s , his pape con ibu es o s udies on he
e ec i eness o basic heal hca e p o ided h ough a model o communi y heal h. This model,
wi h i s longs anding his o y as a supply-side al e na i e o deli e ing p ima y ca e, has e-
cen ly seen a esu gence in LMICs (see, o example, Angwenyi e al. (2018), Kok e al. (2015),
Mo e al. (2023)). While much o he exis ing li e a u e ocuses on es ima ing heal h e ec s,
ou s udy aims o unde s and he mechanisms behind hese e ec s by examining e iciency in
heal hca e deli e y wi hin a na ional heal h sys em.
One o he pionee ing wo ks in his domain is by Goldman and G ossman (1988), who demon-
s a ed ha communi y heal h cen e s in he USA a e associa ed wi h a educ ion in in an mo -
ali y a es.3Since hen, a subs an ial body o li e a u e, ocusing p ima ily on ma e nal and child
heal h, has eme ged. Fo ins ance, Kose e al. (2022) in he USA, Das e al. (2013) e iewed
s udies in Asia and A ica, B azil’s Family Heal h P og am s udied by Macinko e al. (2006),
Aquino e al. (2009), Rocha and Soa es (2010), and He e a-Almanza and Rosales-Rueda
(2023) in Madagasca , o name a ew. Se e al o he s udies ha e examined how communi y-
based heal hca e imp o es ep oduc i e heal hca e. Fo ins ance, A ends-Kuenning (2001),
Ba ham (2012) and Joshi and Schul z (2013) in Bangladesh, Salehi-Is ahani e al. (2010) in
I an, and He e a-Almanza and Rosales-Rueda (2020) in Madagasca .
Ou s udy con ibu es o his li e a u e s eam by p o iding e idence ha one mechanism o
3Fu he insigh s in o he his o y and e olu ion o communi y heal h p og ams can be ound in Singh and Sachs
(2013) and Pe y e al. (2014).
4
he success o CHTs as an al e na i e in o ganizing p ima y heal hca e is a ask-shi ing model.
In his model, ce ain asks a e delega ed o skilled ye less-specialized heal h wo ke s wi hou
comp omising quali y. Ou e idence is ele an no only o LMICs, which ha e been imple-
men ing such p ima y ca e models, bu also o de eloped coun ies con empla ing s a egies
o add ess ope a ional was e in heal hca e spending, aiming o ensu e pa ien s ecei e simila
bene i s o ca e while u ilizing ewe expensi e esou ces (Ben ley e al.,2008;OECD,2017;
Sh ank e al.,2019).
Second, ou s udy complemen s he li e a u e s eam on heal h sys em e iciency. P e ious
wo k, p ima ily om ad anced economies, has shown ha expanding p ima y ca e imp o es
o e all heal h sys em e iciency by educing eme gency oom use and hospi aliza ions o a oid-
able NCDs (e.g., Da ny and G ube ,2005;Kols ad and Kowalski,2012;Mille ,2012;Dol on
and Pa hania,2016;Whi ake e al.,2016;Alexande e al.,2019;Pinchbeck,2019;Ding e al.,
2021;G ube e al.,2022). No ably, in B azil, Macinko e al. (2010) ound ha a majo expansion
o p ima y ca e dec eased unnecessa y hospi aliza ions, and Bhalo a e al. (2020) obse ed
ha u gen ca e cen e s educed hospi al ou pa ien p ocedu es and admissions, and ha his
is associa ed wi h imp o ed hospi al pe o mance, indica ed by a decline in inpa ien mo ali y.
While e iciency gains om a oiding eme gency hospi aliza ions a e well-sui ed o high-income
and uppe -middle-income coun ies, in lowe -income con ex s, he e iciency ma gins wi hin he
p oduc ion o p ima y ca e se ices emain la ge. Hospi als o eme gency ooms in LMICs
a e inaccessible o a po ion o he popula ion, as hei in as uc u e is o en loca ed in u ban
cen e s (Tho n on,2008;K eme and Glenne s e ,2011;Adh a yu and Nyshadham,2015).
In lowe -income se ings like ou s, bo h p e en i e and ou pa ien cu a i e ca e p edominan ly
occu a he p ima y ca e le el and sha e limi ed esou ces. Cu a i e ca e is ypically mo e
esou ce-in ensi e han p e en i e consul a ions, equi ing medica ion and signi ican medical
s a hou s. Mo eo e , cu a i e consul a ions a e usually unscheduled, dis up ing physicians’
daily schedules in heal h uni s (Hey and Pa el,1983;Cou bage and Rey,2006;Williams e al.,
2006;Nuschele and Roede ,2016;Wang,2018;Pe e ,2021). Heal h cen e s o en ope a e
unde igh capaci y cons ain s, making he oppo uni y cos s o using esou ces o o he wise
p e en able condi ions signi ican .
We con ibu e o his li e a u e by ocusing on e iciency gains in a low-income con ex , exam-
ining how CHTs can shi esou ces wi hin ou pa ien heal hca e ha mainly akes place a he
p ima y le el. A signi ican ad an age o ou da a is ha i enables us o analyze he u iliza ion
o ca e o condi ions ha a e amenable o e ec i e p ima y ca e. By sepa a ely ocusing on
ca e o CDs, which a e easily p e en able, and NCDs, which a e mo e esou ce-in ensi e, we
iden i y wo ypes o gains om he CHT sys em: one whe e mo e p e en i e ca e shi s e-
sou ces owa ds less expensi e cu a i e ca e (e iciency e ec ), and ano he ha ealloca es
esou ces om cu a i e ca e o CDs o cu a i e ca e o NCDs, he eby inc easing he capaci y
5
He e he se o 1[Kj =h]a e he lead and lag ea men indica o a iables acking he numbe
o yea s Kj = −Ejsince he yea o he CHTs c ea ion o he municipali y, Ej, a ≥0 and b ≥
0 a e he numbe s o included leads and lags o he e en indica o , espec i ely, and µj is he
e o e m. bis chosen such ha all possible lags in he sample a e co e ed. This speci ica ion
also includes yea ly p e- ends coe icien s, i.e. a= 3. Absen p e- ends, he coe icien s on
he lags a e in e p e ed as he dynamic pa h o causal e ec s: a h= 0, ..., b yea s a e he
c ea ion o CHTs.
τhcap u es ea men e ec dynamics wi h espec o leng h o exposu e o he ea men , i.e.
he c ea ion o CHTs. Fo each iming g oup ea ed a pe iod k, ne e - ea ed, no -ye - ea ed,
and al eady- ea ed se e as he con ol g oup.
I has been well documen ed ha adi ional wo-way ixed e ec s (TWFE) es ima o s, le e aging
s agge ed oll-ou , a e subjec o ‘nega i e weigh s’ because hey use al eady- ea ed uni s ac
as he con ol g oup, and ea men e ec s may a y o e ime (Goodman-Bacon,2021). To
add ess his conce n, we use he impu a ion es ima o p oposed by Bo usyak e al. (2024).
B ie ly, Bo usyak e al. (2024)’s me hod de ines g oups o municipali ies acco ding o he pe iod
ha hey we e ea ed and es ima es coun e ac ual ou comes o each ea ed g oup. Po en ial
con ol ou comes Yj (0) a e de i ed om municipali ies ha we e ne e ea ed (27% o o al
municipali ies), and hose ha we e ea ed la e on in each yea . The coun e ac uals a e
es ima ed using impu a ion p ocedu es a each poin o ime, which a e obus and e icien
unde he e oskedas ici y.
When calcula ing g oup-speci ic a e age ea men e ec s by ime, we end up wi h many ea -
men e ec pa ame e s in a “ ully dynamic” speci ica ion. Fo ease o in e p e a ion, we ake he
mean o e all poin es ima es using a linea combina ion, as sugges ed by Cunningham (2021).
4.2 In e nal alidi y
A Cox haza d model e eals ha he iming o he c ea ion o CHTs was un ela ed o ini ial
demog aphic cha ac e is ics o he municipali y, he ini ial a ailabili y o inpu s o heal hca e
p oduc ion, as well as he ini ial le el o heal hca e se ices (Table B2, Column (2)). Al hough
he MoH implemen ed CHTs gi ing p io i y o poo e municipali ies (as discussed in Sec ion
2), he ac ual s a o CHTs ac i i ies (e.g. egis e ing amilies in he CHT sys em) was likely
slowe in poo e municipali ies. Thus, he ne e ec o po e y on he iming o he s a o CHTs’
ac i i ies in null.
In line wi h he MoH manda e, he ini ial pe cen age o he popula ion li ing in po e y was
highe in he ea men g oup han in he ne e - ea ed g oup o municipali ies (see Table B1 in
he appendix). This di e ence pe sis s e en when con olling o ini ial le els o o he municipal
cha ac e is ics, heal hca e inpu s, and ou comes (see Table B2, Column (1)). T ea ed dis ic s
12
also had a la ge sha e o u al popula ion and we e gene ally smalle in e ms o popula ion
size, esul ing in hem ha ing g ea e heal hca e inpu s pe 1,000 inhabi an s (see Table B1
in he appendix). Howe e , hese ini ial imbalances a e no p oblema ic, as he di e ences in
le els a e e ec i ely con olled o by municipali y ixed e ec s. Fu he mo e, as demons a ed
in Sec ion 5.5, he esul s emain obus o con olling o hese ini ial municipali y cha ac e is ics
in e ac ed wi h yea dummies.
Addi ionally, we show in Sec ion 5 ha he pa allel ends assump ion in ou comes hold. An
ad an age o he dynamic e en s udy is ha i allows o isually assess he pa e n o ea men
e ec s ela i e o he c ea ion o CHTs. In ou main esul s we p esen up o h ee-yea p e-
ends. While we can es o he signi icance o −1 o all he ea ed municipali ies, and o
−2 o 58% o he ea ed municipali ies, a limi a ion is ha we can only es o he signi icance
o −3 o 6% o he ea ed municipali ies when using he consul a ions da a. Municipali ies ha
implemen ed CHTs pos -2012 only ha e su icien da a om 2009 onwa ds o he p e- ends
analysis. To o e come his limi a ion, we addi ionally es o i e-pe iod p e- ends using da a
a he hal -yea le el. Fu he mo e, exploi ing he a ailabili y o hospi al eco ds om 2005 we
a e able o es o up o i e-pe iod p e- ends o all ea ed municipali ies. We ind no e idence
o p e- ends using hese al e na i es, as discussed in Sec ion 5.5.
5 The E ec s o Communi y Heal h Teams
5.1 Changes in Inpu s o Heal hca e
We s a by e alua ing how he e o m in El Sal ado a ec ed he a ailabili y o inpu s o he
p oduc ion o heal hca e in municipali ies. Using h ee ounds o da a, 2009, 2010 and 2015,
we es ima e a s a ic DiD model using Equa ion 1.
The e o m imp o ed access o p ima y ca e se ices by inc easing he numbe o p ima y ca e
uni s and human esou ces, in pa icula nu ses and suppo wo ke s. Table 1, Panel A, shows
ha on a e age p ima y ca e uni s inc eased by 0.06 uni s pe 1,000 inhabi an s in ea ed
municipali ies a e he c ea ion o CHTs, equi alen o a 4.1% inc ease om he 2009 mean
o ne e - ea ed municipali ies. Fu he mo e, he o al numbe o heal h s a in municipali ies
inc eased by 0.21 pe 1,000 inhabi an s on a e age (10.7%). This o e all inc ease is mos ly
d i en by an inc ease in he numbe o nu ses and suppo wo ke s (0.07 pe 1,000 inhabi an s,
18.4% and 17.5% espec i ely compa ed o he 2009 mean). Al hough also posi i e, he e ec s
on he numbe o doc o s and CHWs a e no p ecisely es ima ed.
Panel B shows how he e o m expanded p ima y ca e se ices p o ided by la ge mul i-disciplina y
eams, a he han elying on physicians alone. The composi ion o human esou ces changed,
wi h an inc ease by 1.0 and 2.0 pe cen age poin s (pp s) in he sha e o nu ses and suppo
13
wo ke s, espec i ely, and a dec ease by 2.0 pp s in he sha e o CHWs ou o he o al human
esou ces, on a e age.
The e o m expanded he inpu s used in he p oduc ion o heal hca e and changed he compo-
si ion o heal hca e wo ke s in municipali ies. The indings sugges ha CHTs we e based on a
model ha ocused on skilled bu less-specialized heal h wo ke s and suppo pe sonnel, which
could help lowe cos s while imp o ing heal h ou comes.
5.2 Expansion o P e en i e Ca e
As explained in Sec ion 2, CHTs kick-s a ed hei ac i i ies wi h ou each e o s when egis-
e ing indi iduals, coupled wi h p oac i e ollow-up in scheduling p e en i e appoin men s. As
such, we nex e alua e he e ec o CHTs on he numbe o p e en i e consul a ions.
Be o e del ing in o he analysis, we es o he p esence o p e- ends. Figu e 2, Panel A,
shows ha he p e- end coe icien s a e close o ze o and a e no s a is ically signi ican wi hin
con en ional le els. Table 2showing he linea combina ion o all he coe icien s es ima ed o
he yea s p io o he c ea ion o he CHTs con i ms ha he e ec in he p e- ea men yea s is
insigni ican (column (1)).
The c ea ion o CHTs d ama ically inc eased p e en i e heal hca e in municipali ies. The impac
becomes signi ican a e a one-yea pe iod, consis en wi h he imeline du ing which CHTs
conduc ed household isi s as pa o hei es ablishmen . These household isi s may ha e
unc ioned as a subs i u e o p e en i e consul a ions a heal h cen e s. The e ec jumps om
37 o 170 addi ional consul a ions pe 1,000 inhabi an s be ween + 1 and + 2, and i peaks in
+ 5 a 277 consul a ions pe 1,000 inhabi an s. The e ec emains high e en eigh yea s a e
he c ea ion o CHTs.
Table 3summa izes he dynamic e ec s in a single coe icien cap u ing he a e age ea men
e ec o e e y yea a e he c ea ion o he CHTs, based on Equa ion 2. Column (1) Panel A
shows ha , on a e age, he c ea ion o CHTs inc eased p e en i e consul a ions by 187.6 pe
1,000 inhabi an s. This e ec is equi alen o a 36.9% inc ease in p e en i e consul a ions wi h
espec o he p e- ea men mean, and i is signi ican a he 1% le el. We addi ionally es ima e
he s a ic DiD e ec s ollowing Equa ion 1. We ind ha he c ea ion o CHTs inc eased by
72.3 p e en i e consul a ions pe 1,000 inhabi an s du ing he pos - e o m pe iod (see Table
4). The lowe magni ude in he coe icien s o he s a ic es ima ion compa ed wi h hose om
he impu a ion es ima o is consis en wi h ‘nega i e weigh s’ in oduced in adi ional TWFE
models, as discussed in Sec ion 4.
14
5.3 E iciency and Co e age Gains in Cu a i e Ca e
We now in es iga e he e ec o he c ea ion o CHTs on he numbe o amenable cu a i e
consul a ions. As explained in Sec ion 3, hese include isi s o es o e heal h due o condi-
ions o which e ec i e managemen and ea men can be achie ed in a p ima y ca e se ing,
po en ially a oiding he need o specialized o e ia y ca e.
The absence o s a is ically signi ican p e- ends in Figu e 2, Panel B, and Figu e 3, Panels
A and B, bols e ou con idence o in e p e he impu a ion es ima ions as causal e ec s o he
a i al o CHTs on amenable cu a i e ca e. We con i m his in Table 2(column (2)), whe e he
a e age e ec in yea s p io o he c ea ion o CHTs is insigni ican .
We es ima e no signi ican e ec s on amenable cu a i e consul a ions (see Figu e 2, Panel B).
The a e age dynamic e ec o up o eigh yea s a e he c ea ion o CHTs shows a s a is ically
insigni ican dec ease in amenable cu a i e consul a ions (see Table 3, Panel A, column 2).
Figu e B1 in he appendix also shows an insigni ican e ec on o al cu a i e consul a ions (Panel
A).
We nex explo e whe he he null e ec on amenable cu a i e consul a ions is he esul o a
co e age e ec o se ing an e iciency e ec . To do his, we classi y amenable cu a i e consul-
a ions as due o ei he CDs o NCDs. Focusing on amenable cu a i e consul a ions due o CDs
allows us o in es iga e po en ial e iciency gains. The ini ial ou each e o s by CHTs, along
wi h hei p oac i e ollow-up in scheduling p e en i e appoin men s (as e idenced in Sec ion
5.2), migh ha e p e en ed he sp ead o in ec ions and o he CDs. CDs a e o en mo e easily
p e en able compa ed o NCDs, o e which heal hca e p o ide s ha e less con ol. NCDs yp-
ically equi e li es yle modi ica ions and long- e m managemen s a egies, aspec s ha la gely
all unde he pa ien ’s con ol.
Panel A in Figu e 3 e eal ha he c ea ion o CHTs dec eased cu a i e consul a ions o CDs.
The e ec is immedia e, a d op by 36 consul a ions pe 1,000 inhabi an s, consis en wi h he
ini ial ou each ac i i ies o CHWs helping p e en he need o cu a i e ca e o CDs in heal h
uni s. The nega i e e ec s s eng hen o e ime, ollowing wi h he inc ease in p e en i e con-
sul a ions in heal h uni s (as es ima ed in Sec ion 5.2). The magni ude o his nega i e e ec
inc eased o 79 consul a ions in he ou h yea and o 91 consul a ions a e eigh yea s (Fig-
u e 3, Panel A). On a e age, he c ea ion o CHTs dec eased amenable cu a i e consul a ions
due o CDs by 73.1 pe 1,000 inhabi an s, equi alen o a 8.5% d op wi h espec o he p e-
ea men mean (Table 3, column (2)). The e ec is signi ican a he 1% le el. This esul se es
as e idence o an e iciency e ec om g ea e p e en i e ca e.
Nex , ocusing on amenable cu a i e consul a ions due o NCDs allows us o in es iga e po-
en ial co e age gains om CHTs o h ee easons. Fi s , he e was a g ea e need o CDs,
e idenced by lowe co e age o cu a i e ca e o hese diseases p io o he c ea ion o he
15
CHTs (855 CDs s. 248 NCDs cu a i e consul a ions pe 1,000 inhabi an s). Second, ou each
ac i i ies by CHWs gene a ed e e als o heal h uni s o ea illnesses and ch onic condi ions.
Thi d, i is mo e esou ce-in ensi e o iden i y and ollow-up on ch onic condi ions, like diabe es
and as hma (Williams e al.,2006;Wang,2018), and hence lowe amenable cu a i e consul a-
ions due o p e en able CDs migh ha e eleased esou ces ha could be alloca ed o NCDs
ea men .
In he yea CHTs we e c ea ed, cu a i e consul a ions due o NCDs inc eased by 14 pe 1,000
inhabi an s, by 35 consul a ions in he ou h yea and o 101 consul a ions a e eigh yea s
(Figu e 3, Panel B). The c ea ion o CHTs inc eased cu a i e consul a ions due o NCDs, on
a e age, by 44.5 pe 1,000 inhabi an s (17.9%; Table 3, column (2)). This e ec is also signi i-
can a he 1% le el. This esul se es as e idence o a co e age e ec due o mo e esou ces
a ailable o manage p e iously una ended ch onic condi ions.
We addi ionally es ima e he s a ic e ec ollowing Equa ion 1. We show in Table 4 ha he
e ec on amenable cu a i e consul a ions is -25.0 pe 1,000 inhabi an s wi h a p- alue o 0.16,
and when spli by disease ype, he e ec is -31.4 pe 1,000 inhabi an s o CDs and 6.3 pe
1,000 inhabi an s, hough he la e is no signi ican . O e all, hese indings sugges ha he
absolu e gain in e iciency was g ea e in magni ude han he gain in co e age.
5.4 E iciency Gains in Hospi aliza ions
Did he expansion o communi y-based heal hca e ansla e in o e iciency gains in he sys em?
To answe his ques ion, we ocus on p e en able hospi aliza ions. As explained in Sec ion 3,
hese include condi ions ha can be e ec i ely managed, ea ed, o p e en ed in a p ima y ca e
o ou pa ien se ing, and hose ha i no app op ia ely add essed, could lead o unnecessa y
hospi aliza ions o complica ions.
Figu e 2Panel C con i ms he absence o signi ican p e- end es ima es. Column (3) in Table 2
also con i ms ha he a e age e ec s on p e en able hospi aliza ions we e s a is ically insigni -
ican be o e he c ea ion o CHTs. Because we ha e da a on hospi aliza ions since 2005, we
also p esen es ima es o up o i e-yea p e- ends in Sec ion 5.5. I is impo an o no e ha ,
while he nega i e poin es ima e o −1could be conce ning (mos ly o p e en able hospi al-
iza ions o CDs), i ge s close o ze o and e en posi i e when impu ing 5-yea p e- ends and
when using hal -yea da a (see Figu es C5 o C8 in he appendix). When omi ing his i s lead
o no maliza ions, al e na i e TWFE es ima o s clea ly show no p e- ends be ween −5and
−2and a signi ican d op a e (see Figu es C4 and C5 in he appendix).
The c ea ion o CHTs dec eased he numbe o p e en able hospi aliza ions. The e ec is an
immedia e d op by 0.72 hospi aliza ions pe 1,000 inhabi an s, which peaks h ee yea s la e a
-0.9 and again eigh yea s la e a -1.5 hospi aliza ions. Figu e B1 in he appendix shows ha
16
he c ea ion o CHTs had no e ec on o al hospi aliza ions (Panel C).
The a e age dynamic e ec a e he c ea ion o he CHTs is p esen ed in Table 3(Panel A, col-
umn 3). We ind ha CHTs dec eased p e en able hospi aliza ions by 0.8 pe 1,000 inhabi an s
–a d op equi alen o 10.8% wi h espec o he p e- ea men mean and s a is ically signi ican
a he 1% le el. Admissions we e educed due o p e en able condi ions as ex eme cases
we e a oided h ough be e case managemen h ough ou pa ien ca e.
We e alua e e iciency as done o cu a i e ca e. Figu e 3, Panels C and D, e eal ha a e
he in oduc ion o CHTs p e en able hospi aliza ions d opped o CDs and NCDs, hough he
e ec s on he la e a e only p ecisely es ima ed in and + 8. In he yea CHTs we e c ea ed,
p e en able hospi aliza ions o CDs d opped by 0.4 pe 1,000 inhabi an s and o NCDs hey
d opped by 0.3 pe 1,000 inhabi an s. The magni ude o he nega i e e ec on p e en able
hospi aliza ions due o CDs inc eased o 0.7 hospi aliza ions ou yea s la e and o 0.9 hospi-
aliza ions a e eigh yea s (Panel C).
The a e age dynamic e ec o p e en able hospi aliza ions by CDs is -0.6, equi alen o a d op
by 13.6% wi h espec o he p e- ea men mean. This e ec is s a is ically signi ican a he
1% le el. The a e age e ec on p e en able hospi aliza ions o NCDs is -0.3, hough i is no
s a is ically signi ican a con en ional le els (see Table 3, column 3). Consis en wi h he la ge
inc ease in p e en i e ca e and cu a i e ca e o amenable NCDs, mo e o hese cases seem
o ha e been esol ed h ough ou pa ien ca e a he han equi ing hospi aliza ion.
We addi ionally es ima e he s a ic e ec ollowing Equa ion 1, p esen ed in Table 4. The e ec
on p e en able hospi aliza ions is -0.8 pe 1,000 inhabi an s wi h a p- alue below 0.01. When
spli by disease ype, he e ec is -0.5 pe 1,000 inhabi an s o CDs and -0.3 pe 1,000 in-
habi an s (12.3% and 6.7% compa ed o he p e- ea men mean, espec i ely). All es ima ed
e ec s a e signi ican a con en ional le els.
As a placebo es , we addi ionally es ima e he dynamic e ec o he c ea ion o CHTs on hos-
pi aliza ions caused by ex e nal ac o s, such as inju y, poisoning, acciden s, assaul s and sel -
ha m. Communi y heal hca e should no a ec admissions by hese un o eseen condi ions ha
equi e specialized ca e. In line, we ind no s a is ically signi ican e ec on hospi aliza ions due
o ex e nal causes (see Figu e 2Panel D).
5.5 Robus ness Checks
Sensi i i y Checks. All ou es ima ed e ec s on p e en i e and cu a i e consul a ions and hos-
pi aliza ions a e obus o sensi i i y checks in which we se he yea o ea men as he one
in which a CHTs egis e ed 10%, 15% and 20% o he municipali y’s popula ion (see Table C1
in he appendix). Th oughou he di e en speci ica ions, he es ima es e ec s emain highly
17
signi ican . The magni ude o he e ec s on consul a ions, i any hing, inc eases sligh ly when
he c ea ion yea is se when a highe pe cen age o he popula ion was egis e ed, sugges ing
ha ou main es ima es a e conse a i e.
Robus ness o E ec s on Inpu s. To es o p e- ends be o e he in oduc ion o CHTs, we con-
duc a placebo es using da a om he yea s 2009 and 2010. We d op municipali ies ea ed in
2010 (T2010) and we es ima e a s a ic DiD wi h an indica o a iable ha equals o one in 2010
o municipali ies ea ed a e 2010. Be o e he c ea ion o CHTs, we ind no signi ican di e -
ence in inpu s o heal hca e p oduc ion, nei he in coun s no in sha es, ac oss la e - ea ed
(a e 2010) and ne e - ea ed municipali ies (see Table C2 in he appendix). We eplica e he
es ima ions in Table 1 o he sample o municipali ies included in he placebo es , and we ind
ha he esul s emain obus and e en sligh ly highe in magni ude (see Table C3 in he ap-
pendix). The la e able alle ia es conce ns ha a di e en composi ion o municipali ies d i es
he null e ec s in he placebo es , as we exclude 41% o he ea ed municipali ies in his es .
Al e na i e es ima o s. We compa e he esul s ob ained wi h he impu a ion es ima o o Bo usyak
e al. (2024) o he al e na i e es ima o s o De Chaisema in and d’Haul oeuille (2020) (DCHF),
Sun and Ab aham (2021) (SA), and Callaway and San ’Anna (2021) (CS) ha a e also obus o
ea men coho he e ogenei y (see Figu es C1,C2 and C3 in he appendix). The esul s ali-
da e he main indings based on he impu a ion es ima o , as he poin es ima es o he e ec s
o CHTs a e e y simila . These es ima ions alle ia e u he he conce ns ela ed o ‘nega-
i e weigh s’ in adi ional OLS es ima ions and bias in oduced by he di e en composi ion o
ea men coho s.
Addi ional p e- ends es s. Due o he a ailabili y o mo e ounds o hospi al discha ge eco ds
be o e he c ea ion o CHTs ( om 2005 onwa ds), we a e able o es o p e- ends in hese
ou comes o addi ional pe iods o all ea ed municipali ies. Figu es C4 and C5 in he appendix
show ha he e a e no signi ican p e- ends when es ima ing up o i e-yea p e- ends and
when using he al e na i e es ima o s o DCHF, SA and CS. Fo p e en able hospi aliza ions by
NCDs, he impu a ion es ima o yields signi ican posi i e coe icien s o −5, −3and −2. Ye ,
hese p e- end coe icien s a e in he in he opposi e di ec ion o he ea men e ec s and he
DCHF, SA and CS es ima o s yield p e- end coe icien s ha a e close o ze o and insigni ican
o his ou come.
To u he s udy he p e- ea men pa e ns, we also u ilize he a ailabili y o heal hca e u iliza ion
da a a he semi-annual le el. The ad an age o using semi-annual da a is ha i p o ides
mo e a ia ion in he ollou o he CHTs and allows us o es o longe p e- ends. When
analyzing semi-annual da a, we assume ha he ea men begins in he semes e be o e 5%
o he municipali y’s popula ion is egis e ed in he CHT sys em. This is because ou comes a e
now measu ed o e a sho e ime span, and he ini ial ac i i ies o he CHTs can al eady be
e lec ed in hese ou comes. Wi h his conse a i e app oach, 41.9% o ea ed dis ic s we e
18
ea ed in he i s hal o 2010, 43.0% in he second hal o 2010, 9.2% in he i s hal o 2011,
1.6% in he second hal o 2011, 2.7% in he i s hal o 2012, and 1.6% in he second hal
o 2012 (Figu e C6 in he appendix shows he haza d plo s o he e en ”C ea ion o CHTs”
compa ing hal -yea ly and yea ly da a). We can now es o up o wo-yea p e- ends in he
i s coho (41.9%) and o up o h ee-yea p e- ends in he second coho (43.0%), which
oge he cons i u e he majo i y o he ea men g oup. The nega i e side o using da a a he
semi-annual le el is ha he da a is mo e noisy, mos ly o a e e en s like hospi aliza ions.
Figu es C7 and C8 in he appendix eplica e Figu es 2and 3, espec i ely, using hal -yea da a.
We a e able o ule ou up o i e-yea p e- ends o all ou comes. We only ind a signi ican
posi i e coe icien in −1and −4 o cu a i e consul a ions due o CDs, bu his imbalance
is in he opposi e di ec ion o he ea men e ec . The absence o p e- ends is con i med
in Table C4 showing insigni ican a e age p e- end coe icien s o all ou comes when using
he semi-annual da a. Mo eo e , he a e age dynamic ea men e ec s emain in he same
di ec ion and, as expec ed, a e hal he magni ude when using semi-annual da a compa ed o
annual da a (see Table C5 in he appendix). No ably, in Figu e C7 we obse e a d op in + 1
o p e en i e consul a ions, exac ly in he pe iod when CHTs we e mo e in ensi ely isi ing
households o egis e hem in he CHTs sys em, c owding-ou p e en i e isi s in heal h uni s.
Al e na i e de ini ion o p e en able hospi aliza ions. We conduc obus ness checks by employ-
ing al e na i e classi ica ions o p e en able hospi aliza ions. Figu e C9 in he appendix shows
ha he es ima ed e ec s on hospi aliza ions emain obus (and a e sligh ly la ge in magni ude
du ing he i s yea s) when using only K uk e al. (2018)’s lis o condi ions amenable o e ec i e
p ima y ca e, he same classi ica ion used o amenable cu a i e consul a ions, as well as when
using only Rod iguez Ab ego (2012)’s lis o ambula o y-ca e sensi i e condi ions (ACSC).
Adding con ols. We add ess conce ns ega ding ini ial municipali y cha ac e is ics in luencing
he ends in ou ou comes o in e es . Conside ing ha he MoH p io i ized poo e municipal-
i ies, and ha we obse e his in he da a when compa ing ea men and con ol municipali-
ies, we con ol o ini ial po e y in e ac ed wi h yea dummies. We measu e ini ial po e y as
he sha e o he popula ion in a municipali y ha all below he po e y line, which was mea-
su ed in he 2007 census. Figu e C10 in he appendix shows ha he esul s emain obus .
No ably, some poin -wise con idence in e als become na owe , pa icula ly o he e ec s on
p e en able hospi aliza ions due o NCDs. Ou esul s also emain obus o con olling o ini ial
popula ion and he sha e o u al popula ion in e ac ed wi h yea dummies.
O he heal h shocks. We add ess conce ns abou o he heal h policies and heal h shocks, such
as epidemics, clima e change, gangs and es ic ions o mobili y, a ec ing di e en ly he egions
whe e CHTs we e deployed. Fi s ly, no o he heal h e o m was in oduced in he a ge ed
egions. Secondly, he e is no speci ic geog aphical di e ence ac oss ea ed and non- ea ed
municipali ies. Figu e 1shows ha CHTs we e deployed h oughou he na ional e i o y wi hou
19
a speci ic spa ial pa e n. Thi dly, Figu e C1, Panel (D), alle ia es conce ns abou o he heal h
shocks due o iolence being co ela ed wi h CHTs c ea ion, as he e a e no signi ican e ec s on
hospi aliza ions due o ex e nal causes. Finally, i o he heal h shocks in ea ed municipali ies,
co ela ed wi h he iming o he c ea ion o CHTs, we e d i ing he esul s, we would expec o
see e ec s in o al cu a i e consul a ions and o al hospi aliza ions. Howe e , Figu e C1 in he
appendix demons a es ha his is no he case.
5.6 Amenable Mo ali y
Finally, we an icipa e ha CHTs ha e imp o ed heal h ou comes due o he inc ease in p e-
en i e ca e and he expanded co e age in cu a i e ca e. To explo e his, we u ilize da a on
mo ali y a es a ailable om 2011 o 2018 and es ima e a s a ic DiD model in acco dance wi h
Equa ion 1.
We use da a on cause o dea h o compu e mo ali y a es o condi ions ha a e amenable o
e ec i e managemen and ea men ha can be achie ed in a p ima y ca e se ing, po en ially
a oiding he need o specialized o e ia y ca e, ollowing K uk e al. (2018)’s classi ica ion.
We u he spli amenable mo ali y a es by CDs and NCDs a he municipali y le el.
Consis en wi h ou p e ious esul s, Table D1 in he appendix shows ha mo ali y caused by
CDs amenable o heal hca e dec eased by 10.7 dea hs pe 1,000 inhabi an s a e he c ea ion
o CHTs (Panel A). The poin es ima e is equi alen o a d op o 24.7% wi h espec o he
2011 mean o ne e - ea ed municipali ies and s a is ically signi ican a he 5% le el. The es-
ima ed associa ion wi h mo ali y caused by NCDs amenable o heal hca e and wi h mo ali y
no amenable o heal hca e is also nega i e, bu no s a is ically signi ican . These la e esul s
a e encou aging as CHTs a e expec ed o dec ease mo ali y caused by diseases amenable o
e ec i e p ima y heal hca e, and ha a e easy o p e en .
As only 6% o he ea ed municipali ies implemen ed CHTs a e 2011, we a e unable o ake ad-
an age o he s agge ed ea men oll-ou and es o p e- ends in mo ali y a es. Hence, we
in e p e he esul ing coe icien s wi h cau ion. As addi ional e idence, we compa e amenable
mo ali y a es be ween la e - ea ed (a e 2011) and ne e - ea ed municipali ies in 2011, d op-
ping T2010 and T2011 municipali ies. In 2011, be o e being ea ed, we ind no signi ican di -
e ence in mo ali y a es ac oss la e - ea ed and ne e - ea ed municipali ies (Panel B). Addi-
ionally, we es ima e a DiD s a ic model d opping T2010 and T2011 municipali ies in he sample,
and we ind ha he esul s emain obus . Mo ali y caused by CDs amenable o heal hca e
dec eased by 12 dea hs pe 1,000 inhabi an s a e he c ea ion o CHTs, while he e is no sig-
ni ican e ec on mo ali y a es caused by NCDs and diseases and complica ions ha a e no
amenable o p ima y ca e (Panel C).
20
5.7 Cos -e ec i eness o he Re o m
In his sec ion, we discuss he cos -e ec i eness o in oducing CHTs, o which we unde ake
some back-o - he-en elope calcula ions. We ocus on mone izing p e en able hospi aliza ion
gains because he o e all e ec on amenable cu a i e ca e is ze o, as co e age and e iciency
gains o se each o he . Fo his, we use da a on hospi aliza ion cos s om he MoH o El
Sal ado and epo s om he CHTs implemen a ion.
We i s calcula e how much an a e age municipali y sa ed om he educ ion in p e en able
hospi aliza ions pe yea . Using he coe icien o he e ec on hese hospi aliza ions o -0.8 pe
1,000 inhabi an s and pe yea pe municipali y (Table 3, Panel A, column 3), and conside ing
he cos pe hospi aliza ion o USD 772.70, we es ima e a sa ing pe yea and municipali y
equi alen o USD 615,841.90.7
Nex we iden i y how cos ly a e CHTs o an a e age municipali y pe yea . The cos o unning a
CHT pe yea is USD 45,654.47. Using he pos - ea men popula ion o ecas s and guidelines
desc ibed in Sec ion 2, he median numbe o CHTs in a ea ed municipali y wi h an en i ely
u al popula ion is 3.1, and in a municipali y wi h an en i ely u ban popula ion is 1.0. Hence,
he o al cos o un CHTs anges be ween USD 45,654.5 and USD 141,528.9 pe yea pe
municipali y.
This calcula ion sugges s ha he in oduc ion o CHTs in Sal ado was highly cos -e ec i e.
Pe USD 1 in es ed in CHTs, El Sal ado sa ed oughly be ween USD 4.4 and USD 13.5 in
expendi u es o p e en able hospi aliza ion. This calcula ion is a lowe bound, as we a e no
mone izing gains in he heal h s a us o he popula ion (as sugges ed by Table D1 in he ap-
pendix).
6 Conclusions
Ou s udy in es iga es he e iciency gains esul ing om a supply-side expansion o p ima y
ca e h ough a na ionwide e o m in El Sal ado , which in oduced Communi y Heal h Teams
(CHTs). These mul idisciplina y eams, comp ising physicians, nu ses, and communi y heal h
wo ke s, o e a a ie y o p e en i e heal h se ices, including ou pa ien consul a ions, home
isi s, and communi y ou each.
While he bene i s o communi y-based heal hca e on heal h ou comes a e acknowledged, less
is known abou i s impac on e iciency in heal hca e p o ision, pa icula ly in low-income con-
ex s. Ou empi ical s a egy le e ages he s agge ed ollou o he CHT sys em ac oss munic-
ipali ies be ween 2010 and 2013, cons uc ing a comp ehensi e da ase co e ing 254 munici-
7The es ima e o cos pe hospi aliza ion in El Sal ado is ob ained om (Minis y o Heal h o El Sal ado ,2015).
21
Whi ake , W., L. Anselmi, S. R. K is ensen, Y.-S. Lau, S. Bailey, P. Bowe , K. Checkland,
R. El ey, K. Ro hwell, J. S okes, e al. (2016). Associa ions be ween ex ending access o
p ima y ca e and eme gency depa men isi s: a di e ence-in-di e ences analysis. PLoS
medicine 13(9), e1002113.
Williams, A., A. Lloyd, L. Wa son, and K. Rabe (2006). Cos o scheduled and unsched-
uled as hma managemen in se en Eu opean Union coun ies. Eu opean Respi a o y Re-
iew 15(98), 4–9.
28
Figu e 1: Spa ial Dis ibu ion o CHTs’ c ea ion in El Sal ado
Ne e ea ed
2010
2011
2012
2013
No es: This map shows he da e in which CHTs we e c ea ed, p oxied by he yea in which municipali ies
egis e ed a leas 5% o i s popula ion.
29
Table 1: Inpu s o Heal hca e P oduc ion
(1) (2) (3) (4) (5) (6)
P ima y uni s Human esou ces
To al Doc o s Nu ses CHWs Suppo
Panel A: Coun
CHTs c ea ion 0.06 0.21 0.03 0.07 0.05 0.07
(0.01) (0.08) (0.02) (0.02) (0.03) (0.02)
[0.00] [0.01] [0.20] [0.00] [0.13] [0.00]
P e- ea men mean 1.47 1.96 0.31 0.38 0.54 0.40
Panel B: Sha e
CHTs c ea ion 0.00 0.01 -0.02 0.02
(0.01) (0.00) (0.01) (0.00)
[0.72] [0.00] [0.01] [0.00]
P e- ea men mean 0.14 0.19 0.29 0.19
Muni-yea 729 729 729 729 729 729
Municipali y 250 250 250 250 250 250
No es: Es ima ed coe icien s om an linea eg essions o he dependen a iable on a bina y ea men
indica o ha akes alues equal o one o ea ed municipali ies, a e he c ea ion o CHTs (i.e. egis-
e ed a leas 5% o i s popula ion), and ze o o he wise, ollowing Equa ion 1. Dependen a iables by
panel: (A) absolu e numbe s, and (B) sha e ou o o al human esou ces. Dependen a iables in Panel
(A) a e de ined pe 1,000 inhabi an s. We include municipali y and yea ixed e ec s in all es ima ions.
Only h ee ime pe iods a e included in he da a: 2009, 2010 and 2015. Ou come alues a e missing o
ou municipali ies included in he main analysis. S anda d e o s clus e ed by municipali y in pa en he-
ses and p- alues in b acke s.
30
Figu e 2: Consul a ions and Hospi aliza ions
A. P e en i e consul a ions B. Amenable cu a i e consul a ions
-100
0
100
200
300
400
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-100
-50
0
50
100
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
C. P e en able hospi aliza ions D. Hospi aliza ions due o ex e nal causes
-3
-2
-1
0
1
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-10
0
10
20
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
No es: Coe icien s om he ully dynamic speci ica ion ollowing Equa ion 2and es ima ed using he
impu a ion es ima o de eloped by Bo usyak e al. (2024). The y-axis shows he a e age ea men e -
ec s and he x-axis he yea ela i e o he c ea ion o he CHTs. Dependen a iables by panel: (A)
P e en i e consul a ions: o al consul a ions o p e en i e ca e; (B) Amenable cu a i e consul a ions:
o al cu a i e consul a ions due o condi ions amenable o e ec i e p ima y heal hca e; (C) P e en able
hospi aliza ions: o al hospi al discha ges due o condi ions ha can be e ec i ely managed, ea ed,
o p e en ed in a p ima y ca e o ou pa ien se ing, and ha i no app op ia ely add essed, could lead
o unnecessa y hospi aliza ions o complica ions; and (D) Hospi aliza ions due o ex e nal causes: o al
hospi al discha ges due o acciden s and ci cums ances as he cause o en i onmen al e en s and ci -
cums ances as he cause o inju y, poisoning and o he ad e se e ec s. All ou comes a e measu ed pe
1,000 inhabi an s. Con idence in e als a he 95% le el.
31
Figu e 3: Amenable Cu a i e Consul a ions and P e en able Hospi aliza ions, by Disease Type
A. Amenable cu a i e consul a ions, CDs B. Amenable cu a i e consul a ions, NCDs
-200
-100
0
100
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-50
0
50
100
150
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
C. P e en able hospi aliza ions, CDs D. P e en able hospi aliza ions, NCDs
-1.5
-1
-.5
0
.5
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-1
-.5
0
.5
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
No es: Coe icien s om he ully dynamic speci ica ion ollowing Equa ion 2and es ima ed using he
impu a ion es ima o de eloped by Bo usyak e al. (2024). The y-axis shows he a e age ea men
e ec s and he x-axis he yea ela i e o he c ea ion o he CHTs. Panels (A) and (B) co espond
o Amenable cu a i e consul a ions, o al cu a i e consul a ions due o condi ions amenable o e ec i e
p ima y heal hca e, spli by communicable (CDs) and non-communicable diseases (NCDs), espec i ely.
Panels (C) and (D) co espond o P e en able hospi aliza ions, o al hospi al discha ges due o condi ions
ha can be e ec i ely managed, ea ed, o p e en ed in a p ima y ca e o ou pa ien se ing, and ha i
no app op ia ely add essed, could lead o unnecessa y hospi aliza ions o complica ions, spli by CDs
and NCDs, espec i ely. All ou comes a e measu ed pe 1,000 inhabi an s. Con idence in e als a he
95% le el.
32
Table 2: P e- ea men E ec s on Consul a ions and Hospi aliza ions
(1) (2) (3)
P e en i e
consul a ions
Amenable cu a i e
consul a ions
P e en able
hospi aliza ions
Panel A: To al
CHTs c ea ion 7.17 18.80 0.08
(23.90) (34.88) (0.24)
[0.76] [0.59] [0.76]
P e- ea men mean 508.03 1103.59 7.41
Panel B. Communicable diseases
CHTs c ea ion 27.25 -0.13
(19.15) (0.15)
[0.15] [0.39]
P e- ea men mean 855.16 4.03
Panel C. Non-communicable diseases
CHTs c ea ion -8.46 0.21
(18.17) (0.14)
[0.64] [0.14]
P e- ea men mean 248.43 3.38
Muni-yea 2540 2540 3556
Municipali y 254 254 254
No es: Es ima es co espond o a linea combina ion o he p e- end coe icien s es ima es in Figu es 2
and 3 o each co esponding ou come ollowing Equa ion 2and using Bo usyak e al. (2024)’s me hod-
ology. Dependen a iables by column: (1) P e en i e consul a ions: o al consul a ions o p e en i e
ca e; (2) Amenable cu a i e consul a ions: o al cu a i e consul a ions o condi ions amenable o e ec-
i e p ima y heal hca e; and (3) P e en able hospi aliza ions: o al hospi al discha ges due o condi ions
ha can be e ec i ely managed, ea ed, o p e en ed in a p ima y ca e o ou pa ien se ing, and ha i
no app op ia ely add essed, could lead o unnecessa y hospi aliza ions o complica ions. These h ee
ou comes a e spli by communicable diseases in Panel B and non-communicable diseases in Panel C.
All ou comes a e measu ed pe 1,000 inhabi an s.
33
Table 3: E ec s on Consul a ions and Hospi aliza ions
(1) (2) (3)
P e en i e
consul a ions
Amenable cu a i e
consul a ions
P e en able
hospi aliza ions
Panel A: To al
CHTs c ea ion 187.56 -28.60 -0.80
(24.75) (22.90) (0.31)
[0.00] [0.21] [0.01]
P e- ea men mean 508.03 1103.59 7.41
Panel B. Communicable diseases
CHTs c ea ion -73.06 -0.55
(21.28) (0.18)
[0.00] [0.00]
P e- ea men mean 855.16 4.03
Panel C. Non-communicable diseases
CHTs c ea ion 44.46 -0.25
(8.56) (0.17)
[0.00] [0.14]
P e- ea men mean 248.43 3.38
Muni-yea 2540 2540 3556
Municipali y 254 254 254
No es: Es ima es co espond o a linea combina ion o he a e age ea men e ec s es ima es in Fig-
u es 2and 3 o each co esponding ou come ollowing Equa ion 2and using Bo usyak e al. (2024)’s
impu a ion es ima o . Dependen a iables as p esen ed in Table 2, all measu ed pe 1,000 inhabi an s.
S anda d e o s clus e ed by municipali y in pa en heses and p- alues in b acke s.
34
Table 4: S a ic DiD - Consul a ions and Hospi aliza ions
(1) (2) (3)
P e en i e
consul a ions
Amenable cu a i e
consul a ions
P e en able
hospi aliza ions
Panel A: To al
CHTs c ea ion 72.27 -25.02 -0.75
(21.88) (17.90) (0.28)
[0.00] [0.16] [0.01]
P e- ea men mean 543.68 778.06 7.88
Panel B. Communicable diseases
CHTs c ea ion -31.36 -0.49
(14.64) (0.17)
[0.03] [0.00]
P e- ea men mean 551.08 3.99
Panel C. Non-communicable diseases
CHTs c ea ion 6.34 -0.26
(7.27) (0.15)
[0.38] [0.09]
P e- ea men mean 226.98 3.90
Obse a ions 2540 2540 3556
Municipali ies 254 254 254
No es: Same no es as Table 3. Coe icien s co espond o es ima es o he e ec o “CHTs c ea ion”
using equa ion (1). “CHTs c ea ion” is an indica o a iable ha equals o one om he i s yea in which
CHTs s a ope a ions in a municipali y.
35
APPENDIX
This online appendix p o ides addi ional in o ma ion on he da a, me hods, and obus ness
checks.
1
A Addi ional ma e ial o Sec ion 3, Da a
Table A1: ICD-10 Codes o Cu a i e Consul a ions and Hospi aliza ions
Amenable o Heal hca e Ambula o y
Ca e
ICD-10 Desc ip ion Communica-
ble
Non-
communicable
Sensi i e
Condi ions
A00 Chole a x x
A01 Typhoid and pa a yphoid e e s x x
A02 O he salmonella in ec ions x x
A03 Shigellosis x x
A04 O he bac e ial in es inal in ec ions x x
A05 O he bac e ial oodbo ne in oxica ions, no elsewhe e clas-
si ied
x x
A06 Amoebiasis x x
A07 O he p o ozoal in es inal diseases x x
A08 Vi al and o he speci ied in es inal in ec ions xx
A09 O he gas oen e i is and coli is o in ec ious and unspeci ied
o igin
x x
A15 Respi a o y ube culosis, bac e iologically and his ologically
con i med
x x
A16 Respi a o y ube culosis, no con i med bac e iologically o
his ologically
x x
A17 Tube culosis o ne ous sys em x x
A18 Tube culosis o o he o gans x x
A19 Milia y ube culosis x
A20 Plague x
A21 Tula aemia x
A22 An h ax x
A23 B ucellosis x
A24 Glande s and melioidosis x
A25 Ra -bi e e e s x
A26 E ysipeloid x
A27 Lep ospi osis x
A28 O he zoono ic bac e ial diseases, no elsewhe e classi ied x
A30 Lep osy [Hansen disease] x
A31 In ec ion due o o he mycobac e ia x
A32 Lis e iosis x
A33 Te anus neona o um x x
A34 Obs e ical e anus x
A35 O he e anus x x
A36 Diph he ia x x
A37 Whooping cough x x
A38 Sca le e e x
A39 Meningococcal in ec ion x
A40 S ep ococcal sepsis x
A41 O he sepsis x
A42 Ac inomycosis x
A43 Noca diosis x
A44 Ba onellosis x
A46 E ysipelas x x
A48 O he bac e ial diseases, no elsewhe e classi ied x
Con inued on nex page
2
Table A1 – Con inued om p e ious page
ICD-10 Desc ip ion AHC Commu-
nicable
AHC Non-
communicable ACSC
Q22 Congeni al mal o ma ions o pulmona y and icuspid al es x
Q23 Congeni al mal o ma ions o ao ic and mi al al es x
Q24 O he congeni al mal o ma ions o hea x
Q25 Congeni al mal o ma ions o g ea a e ies x
Q26 Congeni al mal o ma ions o g ea eins x
Q27 O he congeni al mal o ma ions o pe iphe al ascula sys em x
Q28 O he congeni al mal o ma ions o ci cula o y sys em x
Z72 P oblems ela ed o li es yle x
Z73 P oblems ela ed o li e-managemen di icul y x
Z74 P oblems ela ed o ca e-p o ide dependency x
Z75 P oblems ela ed o medical acili ies and o he heal h ca e x
Z76 Pe sons encoun e ing heal h se ices in o he ci cums ances x
Z80 Family his o y o malignan neoplasm x
Z81 Family his o y o men al and beha iou al diso de s x
Z82 Family his o y o ce ain disabili ies and ch onic diseases
leading o disablemen
x
Z83 Family his o y o o he speci ic diso de s x
Z84 Family his o y o o he condi ions x
Z85 Pe sonal his o y o malignan neoplasm x
Z86 Pe sonal his o y o ce ain o he diseases x
Z87 Pe sonal his o y o o he diseases and condi ions x
Z88 Pe sonal his o y o alle gy o d ugs, medicamen s and biolog-
ical subs ances
x
Z89 Acqui ed absence o limb x
Z90 Acqui ed absence o o gans, no elsewhe e classi ied x
Z91 Pe sonal his o y o isk- ac o s, no elsewhe e classi ied x
Z92 Pe sonal his o y o medical ea men x
Z93 A i icial opening s a us x
Z94 T ansplan ed o gan and issue s a us x
Z95 P esence o ca diac and ascula implan s and g a s x
Z96 P esence o o he unc ional implan s x
Z97 P esence o o he de ices x
Z98 O he pos su gical s a es x
Z99 Dependence on enabling machines and de ices, no else-
whe e classi ied
x
* Pa ially coded as Ambula o y Ca e Sensi i e Condi ions
9
Table A2: Desc ip i e S a is ics: Mos common ICD-10 codes
Amenable cu a i e consul a ions P e en able hospi aliza ions
(1) (2) (3) (4)
ICD-10 Desc ip ion P e- ea men mean % P e- ea men mean %
Communicable diseases
A04 O he bac e ial in es inal in ec ions 0.14 1.83
A06 Amoebiasis 19.64 1.92 0.24 3.31
A08 Vi al and o he speci ied in es inal in ec ions 0.20 2.68
A09 O he gas oen e i is and coli is o in ec ious and unspeci ied o igin 48.80 4.76 1.92 25.90
B35 De ma ophy osis 22.67 2.21
B82 Unspeci ied in es inal pa asi ism 43.47 4.24
J00 Acu e nasopha yngi is [common cold] 233.78 22.81 0.08 1.15
J02 Acu e pha yngi is 133.40 13.01
J06 Acu e uppe espi a o y in ec ions o mul iple and unspeci ied si es 143.68 14.02 0.10 1.37
J15 Bac e ial pneumonia, no elsewhe e classi ied 0.18 2.41
J18 Pneumonia, o ganism unspeci ied 14.57 1.42 0.07 0.94
J20 Acu e b onchi is 24.75 2.41 0.33 4.43
J21 Acu e b onchioli is 0.32 4.36
J30 Vasomo o and alle gic hini is 18.69 1.82
Non-communicable diseases
E11 Type 2 diabe es melli us 24.62 2.40 0.90 12.20
E14 Unspeci ied diabe es melli us 18.02 1.76 0.15 2.01
G40 Epilepsy 11.55 1.13 0.30 4.05
I10 Essen ial (p ima y) hype ension 121.14 11.82 0.49 6.66
I11 Hype ensi e hea disease 0.06 0.87
I15 Seconda y hype ension 4.33 0.42
I64 S oke, no speci ied as haemo hage o in a c ion 0.09 1.24
I67 O he ce eb o ascula diseases 0.12 1.61
J40 B onchi is, no speci ied as acu e o ch onic 5.58 0.54
J44 O he ch onic obs uc i e pulmona y disease 3.60 0.35 0.36 4.85
J45 As hma 18.32 1.79 0.63 8.52
J46 S a us as hma icus 0.07 0.91
K40 Inguinal he nia 2.96 0.29
K80 Choleli hiasis 2.90 0.28
No e: Columns (1) and (3) epo he municipali y a e age by he en mos common ICD-10 codes in he
p e- ea men pe iod ac oss municipali ies. Columns (1) and (3) show he mean o each condi ion and
(2) and (4) show he mean as a pe cen age o he o e all mean o each ou come.
10
Table A3: Desc ip i e S a is ics: Consul a ions and Hospi aliza ions by Type
All T ea ed
(1) (2) (3) (4)
P e- ea men Pos - ea men P e- ea men Pos - ea men
Panel A: Consul a ions
To al consul a ions 2398.66 2621.09 2582.37 2862.98
P e en i e consul a ions 465.77 830.28 508.03 917.97
% o o al consul a ions 20.75 32.31 21.41 32.96
Cu a i e consul a ions 1932.88 1790.80 2074.34 1945.02
% o o al consul a ions 79.25 67.69 78.59 67.04
Cu a i e amenable consul a ions 1024.71 844.16 1103.59 916.14
% cu a i e consul a ions 53.23 47.29 53.50 47.30
Cu a i e amenable consul a ions - CDs 791.60 551.78 855.16 598.05
% amenable cu a i e consul a ions 77.32 65.11 77.73 65.24
Cu a i e amenable consul a ions -
NCDs
233.11 292.38 248.43 318.09
% amenable cu a i e consul a ions 22.68 34.89 22.27 34.76
Panel A: Hospi aliza ions
To al hospi aliza ions 54.35 67.40 54.89 67.98
P e en able hospi aliza ions 7.32 8.64 7.41 8.58
% o o al hospi aliza ions 13.14 12.44 13.17 12.29
P e en able hospi aliza ions - CDs 3.89 4.12 4.03 4.14
% p e en able hospi aliza ions 52.96 48.02 54.26 48.55
P e en able hospi aliza ions - NCDs 3.43 4.52 3.38 4.44
% p e en able hospi aliza ions 47.04 51.98 45.74 51.45
No e: This able epo s he mean absolu e numbe s and pe cen age ac oss municipali ies o each
ype o ca e and condi ion. P e- ea men pe iod o cu a i e and p e en i e consul a ions is 2009,
and 2005-2009 o hospi aliza ions. Pos ea men pe iod is 2010-2018.
11
B Addi ional ma e ial o Sec ion 5, The E ec s o CHTs
Table B1: Balance in ini ial cha ac e is ics, by ‘pu e’ con ol and ea men g oups
Con ol T ea men Di e ence
(1) (2) (3)
Municipal cha ac e is ics:
To al popula ion 33542.54 18203.39 -15339.16**
[49880.75] [33273.57] (6504.22)
% Ru al popula ion 0.49 0.65 0.16***
[0.28] [0.21] (0.04)
% Pop in po e y 0.38 0.48 0.09***
[0.09] [0.10] (0.01)
Inpu s pe 1,000 inhabi an s:
P ima y uni s 0.11 0.20 0.09***
[0.11] [0.23] (0.02)
To al HR 1.44 2.18 0.73***
[0.84] [1.26] (0.14)
Doc o s 0.18 0.34 0.17***
[0.15] [0.39] (0.03)
Nu ses 0.29 0.42 0.13***
[0.21] [0.35] (0.04)
CHWs 0.41 0.61 0.20***
[0.27] [0.37] (0.04)
Admin 0.28 0.45 0.16***
[0.19] [0.41] (0.04)
No e. Municipal cha ac e is ics a e om he 2007 census, and inpu s o p ima y heal hca e p oduc ion a e om 2009.
Columns 1 and 2 epo sample mean wi h s anda d de ia ion in b acke s o he con ol and o he ea men g oup,
espec i ely. Column 3 epo s he di e ence be ween he ‘pu e’ con ol g oup (ne e ea ed) and he ‘pu e’ ea men g oup
( ea ed a some poin ), es ima ed using OLS, wi h obus s anda d e o s epo ed in pa en heses. S a is ical signi icance
deno ed by *** p<0.01, ** p<0.05, * p<0.1.
12
Table B2: T ea men s a us and iming o s a CHTs
S a us Timing
OLS Cox haza d model
(1) (2) (3) (4)
Municipal cha ac e is ics:
To al popula ion 0.00 0.00 0.00 0.00
(0.00) (0.00) (0.00) (0.00)
% Ru al popula ion 0.05 0.00 0.27 0.30
(0.18) (0.19) (0.49) (0.55)
% Pop in po e y 0.99*** 1.04*** 1.21 1.32
(0.34) (0.35) (1.06) (1.09)
Ou comes
P e en i e consul a ions 0.00*** 0.00** 0.00 0.00
(0.00) (0.00) (0.00) (0.00)
Cu a i e consul a ions CDs -0.00 -0.00 0.00 0.00
(0.00) (0.00) (0.00) (0.00)
Cu a i e consul a ions NCDs 0.00 0.00 -0.00 -0.00
(0.00) (0.00) (0.00) (0.00)
Hospi aliza ions CDs 0.01 0.01 -0.05 -0.05
(0.02) (0.02) (0.06) (0.06)
Hospi aliza ions NCDs -0.03 -0.03 0.00 0.02
(0.03) (0.03) (0.07) (0.08)
Inpu s pe 1,000 inhabi an s:
P ima y uni s -0.21 0.20
(0.25) (0.72)
To al HR 0.07 0.01
(0.11) (0.29)
Doc o s 0.10 0.27
(0.15) (0.51)
Nu ses -0.19 0.11
(0.14) (0.46)
CHWs 0.14 -0.32
(0.14) (0.39)
Admin -0.07 -0.34
(0.20) (0.59)
Obse a ions 254 250 186 184
No e. Municipal cha ac e is ics a e om he 2007 census, inpu s o p ima y heal hca e p oduc ion a e om 2009, and
ou comes a e he a e age o hal -yea obse a ions in 2009 o consul a ions and 2005-2009 o hospi aliza ions. Column
(1) shows coe icien s o an OLS eg ession o being ea ed o e ini ial cha ac e is ics. Column (2) shows coe icien s o
a Cox eg ession o iming un il he s a o CHTs. S anda d e o s a e epo ed in pa en heses. S a is ical signi icance
deno ed by *** p<0.01, ** p<0.05, * p<0.1.
13
Figu e B1: E ec s on Cu a i e Consul a ions and Hospi aliza ions
A. To al cu a i e consul a ions
-100
0
100
200
300
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
B. To al hospi aliza ions
-4
-2
0
2
4
6
A e age causal e ec
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
No es: Same no es as Figu e 2. Dependen a iables by panel: (A) Cu a i e consul a ions: o al consul-
a ions o cu a i e ca e; (B) To al hospi aliza ions: o al hospi al discha ges.
14
C Addi ional ma e ial o Sec ion 5.5, Addi ional es s
Table C1: Sensi i i y Analysis by Yea o CHTS c ea ion
(1) (2) (3)
P e en i e consul a ions Amenable cu a i e consul a ions P e en able hospi aliza ions
Panel A: To al
CHTs c ea ion 5% 187.56 -28.60 -0.80
(24.75) (22.90) (0.31)
[0.00] [0.21] [0.01]
P e- ea men mean 5% 508.03 1103.59 7.41
Municipali ies ea ed 5% 186 186 186
CHTs c ea ion 10% 199.96 -28.11 -0.65
(24.78) (22.77) (0.30)
[0.00] [0.22] [0.03]
P e- ea men mean 10% 512.95 1108.11 7.35
Municipali ies ea ed 10% 180 180 180
CHTs c ea ion 15% 216.14 -31.78 -0.58
(24.60) (22.54) (0.29)
[0.00] [0.16] [0.05]
P e- ea men mean 15% 520.02 1120.24 7.26
Municipali ies ea ed 15% 172 172 172
CHTs c ea ion 20% 230.59 -31.95 -0.55
(24.56) (22.93) (0.28)
[0.00] [0.16] [0.05]
P e- ea men mean 20% 527.82 1137.37 7.33
Municipali ies ea ed 20% 163 163 163
Panel B. Communicable diseases
CHTs c ea ion 5% -73.06 -0.55
(21.28) (0.18)
[0.00] [0.00]
P e- ea men mean 5% 855.16 4.03
Municipali ies ea ed 5% 186 186
CHTs c ea ion 10% -74.58 -0.47
(21.32) (0.17)
[0.00] [0.01]
P e- ea men mean 10% 859.70 4.02
Municipali ies ea ed 10% 180 180
CHTs c ea ion 15% -78.40 -0.47
(21.24) (0.17)
[0.00] [0.01]
P e- ea men mean 15% 870.22 3.98
Municipali ies ea ed 15% 172 172
CHTs c ea ion 20% -80.56 -0.46
(21.66) (0.16)
[0.00] [0.01]
P e- ea men mean 20% 883.95 4.03
Municipali ies ea ed 20% 163 163
Panel C. Non-communicable diseases
CHTs c ea ion 5% 44.46 -0.25
(8.56) (0.17)
[0.00] [0.14]
P e- ea men mean 5% 248.43 3.38
Municipali ies ea ed 5% 186 186
CHTs c ea ion 10% 46.47 -0.18
(8.41) (0.17)
[0.00] [0.27]
P e- ea men mean 10% 248.41 3.33
Municipali ies ea ed 10% 180 180
CHTs c ea ion 15% 46.63 -0.12
(8.36) (0.16)
[0.00] [0.46]
P e- ea men mean 15% 250.03 3.28
Municipali ies ea ed 15% 172 172
CHTs c ea ion 20% 48.61 -0.10
(8.38) (0.15)
[0.00] [0.53]
P e- ea men mean 20% 253.43 3.30
Municipali ies ea ed 20% 163 163
No es: Same no es as Table 3. In each speci ica ion, he yea o he c ea ion o CHTs a ies depending
on he pe cen age o he municipali y’s popula ion ha heal h eams en olled.
15
Table C2: Placebo o Inpu s o P ima y Heal hca e P oduc ion using La e -T ea ed
(1) (2) (3) (4) (5) (6)
P ima y uni s Human esou ces
To al Doc o s Nu ses CHWs Suppo
Panel A: Coun
CHTs c ea ion 0.00 -0.01 -0.00 -0.00 -0.00 -0.00
(0.00) (0.02) (0.00) (0.00) (0.00) (0.00)
[0.95] [0.75] [0.43] [0.75] [0.58] [0.97]
P e- ea men mean 1.561 1.764 0.233 0.323 0.539 0.340
Panel B: Sha e
CHTs c ea ion -0.00 0.00 -0.00 0.00
(0.00) (0.00) (0.00) (0.00)
[0.319] [0.353] [0.319] [0.319]
P e- ea men mean 0.13 0.18 0.31 0.18
Muni-yea 329 329 329 329 329 329
Municipali y 165 165 165 165 165 165
No es: Same no es as Table 1. Coe icien s co espond o es ima es o he e ec o ‘CHTs c ea ion’ using
equa ion (1). ‘CHTs c ea ion’ is an indica o a iable ha equals o one in 2010 o municipali ies ea ed
a e 2010 as a placebo es . Analysis excludes he coho o municipali ies ea ed in 2010 (T2010).
Sample includes he yea s 2009 and 2010.
Table C3: Inpu s o P ima y Heal hca e P oduc ion, Excluding T2010
(1) (2) (3) (4) (5) (6)
P ima y uni s Human esou ces
To al Doc o s Nu ses CHWs Suppo
Panel A: Coun
CHTs c ea ion 0.06 0.21 0.03 0.07 0.05 0.07
(0.01) (0.08) (0.02) (0.02) (0.03) (0.02)
[0.00] [0.01] [0.20] [0.00] [0.13] [0.00]
P e- ea men mean 1.47 1.96 0.31 0.38 0.54 0.40
Panel B: Sha e
CHTs c ea ion 0.00 0.01 -0.02 0.02
(0.01) (0.00) (0.01) (0.00)
[0.72] [0.00] [0.01] [0.00]
P e- ea men mean 0.14 0.19 0.29 0.19
Muni-yea 719 719 719 719 719 719
Municipali y 240 240 240 240 240 240
No es: Same No es as Table 1. Analysis excludes he coho o municipali ies ea ed in 2010 (T2010).
Sample includes he yea s 2009, 2010 and 2015.
16
Figu e C1: Consul a ions and Hospi aliza ions, Al e na i e TWFE Es ima o s
A. P e en i e consul a ions
-100
0
100
200
300
400
T ea men e ec s
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
Bo usyak e al. de Chaisema in-D'Haul oeuille
Callaway-San 'Anna Sun-Ab aham
C. P e en able hospi aliza ions
-2
-1
0
1
T ea men e ec s
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
Bo usyak e al. de Chaisema in-D'Haul oeuille
Callaway-San 'Anna Sun-Ab aham
No es: Same no es as Figu e 2. In addi ion o he impu a ion es ima o o Bo usyak e al. (2024), we
use h ee obus es ima o s: De Chaisema in and d’Haul oeuille (2020), Sun and Ab aham (2021), and
Callaway and San ’Anna (2021).
17
Figu e C2: Amenable Cu a i e Consul a ions by Disease Type, Al e na i e TWFE Es ima o s
A. Amenable cu a i e consul a ions, CDs
-300
-200
-100
0
100
T ea men e ec s
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
Bo usyak e al. de Chaisema in-D'Haul oeuille
Callaway-San 'Anna Sun-Ab aham
B. Amenable cu a i e consul a ions, NCDs
-50
0
50
100
150
T ea men e ec s
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
Bo usyak e al. de Chaisema in-D'Haul oeuille
Callaway-San 'Anna Sun-Ab aham
No es: Same no es as Figu e 3. In addi ion o he impu a ion es ima o o Bo usyak e al. (2024), we
use h ee obus es ima o s: De Chaisema in and d’Haul oeuille (2020), Sun and Ab aham (2021), and
Callaway and San ’Anna (2021).
18
Table C4: P e- ea men E ec s on Consul a ions and Hospi aliza ions, Hal -Yea Da a
(1) (2) (3)
P e en i e
consul a ions
Amenable cu a i e
consul a ions
P e en able
hospi aliza ions
Panel A: To al
CHTs c ea ion 0.37 11.97 0.04
(10.69) (14.63) (0.12)
[0.97] [0.41] [0.72]
P e- ea men mean 254.01 551.79 3.70
Panel B. Communicable diseases
CHTs c ea ion 14.30 -0.03
(9.36) (0.08)
[0.13] [0.70]
P e- ea men mean 427.58 2.01
Panel C. Non-communicable diseases
CHTs c ea ion -2.33 0.07
(10.25) (0.07)
[0.82] [0.28]
P e- ea men mean 124.22 1.69
Muni-yea 5080 5080 7112
Municipali y 254 254 254
No es: Same No es as Table 2. Da a a he hal -yea le el.
25
Table C5: Consul a ions and Hospi aliza ions, To al and by Type, Hal -Yea Da a
(1) (2) (3)
P e en i e
consul a ions
Amenable cu a i e
consul a ions
P e en able
hospi aliza ions
Panel A: To al
CHTs c ea ion 60.47 -15.31 -0.38
(8.68) (9.82) (0.16)
[0.00] [0.12] [0.02]
P e- ea men mean 254.01 551.79 3.70
Panel B. Communicable diseases
CHTs c ea ion -28.73 -0.28
(8.80) (0.09)
[0.00] [0.00]
P e- ea men mean 427.58 2.01
Panel C. Non-communicable diseases
CHTs c ea ion 13.42 -0.09
(3.14) (0.09)
[0.00] [0.31]
P e- ea men mean 124.22 1.69
Muni-yea 5080 5080 7112
Municipali y 254 254 254
No es: Same No es as Table 3. Da a a he hal -yea le el.
26
Figu e C9: E ec s on Amenable Hospi aliza ions and ACSC Hospi aiza ions
A. Amenable hospi aliza ions, CDs B. Amenable hospi aliza ions, NCDs
-2
-1
0
1
A e age causal e ec
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-1
-.5
0
.5
1
1.5
A e age causal e ec
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
C. ACSC hospi aliza ions, CDs D. ACSC hospi aliza ions, NCDs
-1
-.5
0
.5
A e age causal e ec
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-3
-2
-1
0
1
2
A e age causal e ec
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
No es: Same no es as Figu e 3. Dependen a iables a e: in Panels (A) and (B) he o al hospi al dis-
cha ges due o condi ions amenable o e ec i e p ima y ca e ollowing K uk e al. (2018)’s classi ica ion
( he same one used o amenable cu a i e ca e), spli by communicable (CDs) and non-communicable
diseases (NCDs), espec i ely; in Panels (C) and (D) he o al hospi al discha ges due o ambula o y ca e
sensi i e condi ions, ollowing Rod iguez Ab ego (2012)’s classi ica ion, spli by communicable (CDs)
and non-communicable diseases (NCDs), espec i ely.
27
Figu e C10: Consul a ions and Hospi aliza ions, Po e y x Yea Dummies
A. P e en i e consul a ions C. P e en able hospi aliza ions
-100
0
100
200
300
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-4
-3
-2
-1
0
1
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
A. Amenable cu a i e consul a ions, CDs B. Amenable cu a i e consul a ions, NCDs
-200
-100
0
100
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-50
0
50
100
150
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
C. P e en able hospi aliza ions, CDs D. P e en able hospi aliza ions, NCDs
-2
-1.5
-1
-.5
0
.5
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
-2
-1.5
-1
-.5
0
.5
A e age causal e ec
-3 -2 -1 0 1 2 3 4 5 6 7 8
Yea s since he c ea ion o CHTs
No es: Same no es as Figu es 2and 3. All speci ica ions con ol o he ini ial sha e o poo popula ion
( om he 2007 census) in e ac ed wi h yea dummies.
28
D Addi ional ma e ial o Sec ion 5.6, Amenable Mo ali y
Table D1: E ec s on Mo ali y Ra es
(1) (2) (3)
Amenable MR Non-amenable MR
Communicable Non-communicable To al
Panel A: S a ic e ec
CHTs c ea ion -10.68 -16.77 -5.35
(5.07) (19.88) (22.99)
[0.04] [0.40] [0.82]
Municipali y-yea 2032 2032 2032
Municipali y 254 254 254
Panel B: Placebo using la e ea ed
T ea ed municipali ies 1.16 3.17 23.33
(9.01) (23.72) (33.49)
[0.90] [0.89] [0.49]
Municipali ies 79 79 79
Panel C: S a ic e ec using ea ly ea ed
CHTs c ea ion -11.73 -14.98 -4.06
(5.29) (20.59) (24.31)
[0.03] [0.47] [0.87]
2011 con ol mean 42.33 123.37 331.03
Municipali y-yea 632 632 632
Municipali y 79 79 79
No e: This able epo s he esul s o he e ec es ima ed om an linea eg essions o he dependen
a iable on a bina y ea men indica o ha akes alues equal o one o ea ed municipali ies, a e he
c ea ion o CHTs (i.e. en olled a leas 5% o i s popula ion), and ze o o he wise, ollowing Equa ion 1.
Dependen a iables by column: (1) Communicable: amenable mo ali y a e caused by communicable
diseases; (2) Non-communicable: amenable mo ali y a e caused by non-communicable diseases; (3)
No AMR: mo ali y a e by diseases no amenable o heal hca e. Amenable mo ali y a e dea hs a oid-
able h ough access o quali y heal hca e, which we classi y ollowing he de ini ion by K uk e al. (2018).
All ou comes a e measu ed pe 1,000 inhabi an s. Panel da a o mo ali y a es is a ailable yea ly be-
ween 2011 and 2018. Panels A and C p esen coe icien s o a s a ic di e ence-in-di e ence es ima ion
ollowing Equa ion 1. Panel B p esen s coe icien s o a c oss-sec ional OLS es ima ion o being ea ed
la e (a e 2011) as opposed o ne e ea ed using 2011 da a. Panels B and C d op om he sample o
analysis municipali ies ha we e ea ed be o e o in 2011 (T2010 and T2011). We include municipali y
and yea ixed e ec s in all es ima ions in Panel A and C. S anda d e o s clus e ed by municipali y in
pa en hesis and p- alues in b acke s.
29
