Breastfeeding promotion and priority setting in health · PDF fileHEALTH POLICY AND PLANNING; 11(2): 156-168 Oxford University Press 1996 Breastfeeding promotion and priority setting

HEALTH POLICY AND PLANNING; 11(2): 156-168 Oxford University Press 1996

Breastfeeding promotion and priority setting in healthSUSAN HORTON,1 TINA SANGHVI,2 MARGARET PHILLIPS,3 JOHN FIEDLER,4 RAFAEL PEREZ-ESCAMILLA,3 CHESSA LUTTER,6 ADA RIVERA7 AND ANA MARIA SEGALL-CORREA8

1 University of Toronto Institute for Policy Analysis, Canada, 'International Science and TechnologyInstitute, Arlington, VA, USA Consultant, UK, 'International Science and Technology Institute,Sturgeon Bay, Wl, USA, department of Nutritional Sciences, University of Connecticut, USA"WELLSTART International, Washington DC, USA 7Social Security Institute, Ministry of Health,Honduras, and aUniversity of Campinas, Sao Paulo, Brazil

An increase in exclusive breastfeeding prevalence can substantially reduce mortality and morbidityamong infants. In this paper, estimates of the costs and impacts of three breastfeeding promotionprogrammes, implemented through maternity services in Brazil, Honduras and Mexico, are used todevelop cost-effectiveness measures and these are compared with other health interventions. Theresults show that breastfeeding promotion can be one of the most cost-effective health interventionsfor preventing cases of diarrhoea, preventing deaths from diarrhoea, and gaining disability-adjustedlife years (DALYs). The benefits are substantial over a broad range of programme types. Programmesstarting with the removal of formula and medications during delivery are likely to derive a high levelof impact per unit of net incremental cost. Cost-effectiveness is lower (but still attractive relative toother interventions) if hospitals already have rooming-in and no bottle-feeds; and the cost-effectivenessimproves as programmes become well-established. At an annual cost of about 30 to 40 US centsper birth, programmes starting with formula feeding in nurseries and maternity wards can reduce diar-rhoea cases for approximately $0.65 to $1.10 per case prevented, diarrhoea deaths for $100 to $200per death averted, and reduce the burden of disease for approximately $2 to $4 per DALY. Maternityservices that have already eliminated formula can, by investing from $2 to $3 per birth, prevent diar-rhoea cases and deaths for $3.50 to $6.75 per case, and $550 to $800 per death respectively, withDALYs gained at $12 to $19 each.

IntroductionThe importance of breastfeeding, particularly theprotective effects of exclusive breastfeeding inthe first 6 months of infancy against mortality,"and diarrhoea incidence and severity, is well-established.15 Yet, in almost all countries wheredata are available, actual feeding practices fall wellbelow internationally recommended standards, withthe duration of any breastfeeding very short in severalcountries, and exclusive breastfeeding practices rarebeyond the first few weeks.67 Studies on the deter-minants of infant feeding practices have shown thathealth facilities practices and health professionalscan have a significant effect.8'11 Among thebreastfeeding promotion interventions, one approachwhich has received considerable attention is the pro-motion of breastfeeding in health facilities, partic-ularly at the time of birth. This includes educationand support for mothers, and changing hospitalroutines to establish early breastfeeding contact,

rooming-in of babies with mothers, withdrawal ofroutine bottle feeding, and post-partum coun-selling.12 While there is some evidence that suchprogrammes have an impact on breastfeedingpractices,13"13 most studies have methodologicallimitations or have been research studies, conductedunder controlled conditions. Reliable estimates ofprogrammes under field conditions have not beenavailable and the costs of such programmes haverarely been measured.3-16

It is difficult to argue convincingly for investment inthe area of breastfeeding promotion without havingbetter information on the impacts and costs of suchprogrammes. Governments and donors have becomeacutely aware of the reality of resource scarcity andare beginning to demand that health budgets givepriority to interventions with a proven record ofreducing the burden of illness substantially and at lowcost.

Breastfeeding and priority setting 157

This paper offers a contribution to the growingliterature on the cost-effectiveness of health interven-tions. It presents evidence from a field study of thecost and effectiveness of three breastfeeding promo-tion programmes in Latin America, compares thesefindings with data from other health interventions anddraws conclusions about the relative attractiveness ofinvestment in health facilities-based breastfeedingpromotion. Details of varioils aspects of this study'smethodology and findings, including a paper on in-ternal efficiency issues, can be found in companionpapers 17-20

MethodologyCost-effectiveness estimates for each programmewere generated in several steps. The first step wasthe direct measurement of the impact on breastfeedingpractices and costs of promotion activities in threehospitals which have well-developed programmes,using three other control hospitals to establishbaseline breastfeeding and cost levels. To enable acomparison of the breastfeeding promotion activitieswith other health interventions, it was necessary totranslate the data on breastfeeding impact into moregeneralizable health units, i.e. mortality and morbid-ity reduction. There is a substantial body of datadocumenting the effect of breastfeeding practices onmorbidity and mortality risk. We drew selectively onthis data in order to translate differences inbreastfeeding behaviour measured in our study, intomortality and morbidity effects. The mortalityestimates were themselves then used as the basis forderiving estimates of disability-adjusted life years(DALYs) gained, an indicator promoted by the WorldBank for comparing health interventions.21"22 Themortality, morbidity and DALY impacts were(separately) combined with cost data to generate aset of cost-effectiveness measures. The following sec-tions provide details on each of these steps.

Estimating impact on breastfeeding practicesAs no baseline was available for measuring the im-provements in breastfeeding practices at each pro-gramme hospital, a cross-sectional design was used.Three pairs of hospitals were selected for comparison,one each in Brazil, Honduras and Mexico, based onsimilarity of client populations served and locationin the same city. One hospital in each pair served asthe control and the other, with a well-developed pro-gramme of breastfeeding promotion, served as the

experimental hospital. All are government-fundedhospitals serving low-income groups; the programmehospital in Honduras is financed and administered bythe Social Security Institute. The programme in Brazilis the longest established (since 1975); the programmein Honduras began in 1984 and the one in Mexicobegan in 1989.

Data collection in Santos (Brazil), San Pedro Sula(Honduras) and Mexico City (Mexico) was under-taken during April 1992 to March 1993. Between 200and 400 women in each hospital were interviewedprior to exit from the hospital and their exposure tobreastfeeding promotion was measured using 28variables.b The programme exposure variables in-clude: practices in the delivery room, formula feedingand rooming-in experiences in maternity wardsand nurseries, education and counselling, andbreastfeeding activities during pre- and post-natalcheck-ups. In addition to hospital exit interviews, thewomen were followed up in their homes at one monthin all countries and again at two (Brazil), three(Honduras) or four (Mexico) months to determinebreastfeeding practices. Age at second follow-upwas based on feeding problems and outcome in-dicators of greatest relevance for each country.For example, exclusive breastfeeding is the centralissue for Honduras and Brazil, and duration of anybreastfeeding for Mexico; and these parametersdemonstrate the greatest declines at the ages selectedfor second follow-up.

There were no significant differences in the womenlost to follow-up in Brazil or Honduras; however,more working women and lower income womendropped out than remained in the study in MexicoCity hospitals. The women in control and programmehospitals within each country had similar charac-teristics in terms of age, parity, years of education,birth weight and sex of newborn, work status,household possessions index and past breastfeedingexperience.

Infant feeding practices measured at the secondhousehold follow-up interview were used, based ona 24-hour recall in which the mother was asked toidentify all liquids and solids consumed by the infantfrom the time it awoke the previous day to the sametime on the day of the interview. Feeding practiceswere then categorized as exclusive, partial or nobreastfeeding. Exclusive is defined as not consum-ing anything but breastmilk - no water, tea or juice.However, the vast majority of partially breastfed

158 Susan Horton et al.

infants in all three countries were consuming othermilks in addition to water and teas.23 The propor-tions of infants exclusively, partially and not breastfedin the programme and control hospitals in eachcountry were used as a measure of programmeimpact.0

Estimating reductions in mortality and morbidityUsing risk ratios and numbers of cases and deathsfrom diarrhoea and ARI estimated by recent in-vestigations in the Latin America region, the observeddifferences in breastfeeding practices were translatedinto differences in diarrhoea morbidity rates andrespiratory and diarrhoeal mortality.d For diarrhoeaand ARI mortality relative risk by breastfeedingstatus, we used the conservative results of Victoraet al.4-24 which are also geographically relevant toour case.' In this case-control study, conducted intwo urban areas of southern Brazil, each of 170 in-fants who died from diarrhoeal disease after the ageof 7 days were compared with two neighbourhoodcontrols, and relative risks are estimated after con-trolling for age, occupation of the head of household,birth weight, type of housing, water supply andpreceding birth interval. The estimated relative risksof diarrhoeal morbidity are drawn from a similarstudy from southern Brazil.26

The proportionate reduction in mortality and mor-bidity risk was calculated as:

( * "^Control group nttrvtntioc grou group

where: TAR = total attributable risk = (1*EBF) +(RR^ * PBF) + (RR^ * NBF)

and: EBF = proportion of infants exclusivelybreastfedR R ^ = risk of mortality/morbidity inpartially breastfed group relative to ex-clusively breastfed infants

PBF = proportion of infants partiallybreastfedRR^j = risk of mortality /morbidity inthe non-breastfed group relative to ex-clusively breastfed infants

NBF = proportion of infants notbreastfed

Based on calculations of the proportion of deaths andmorbidity averted from differences in breastfeeding

practices and relative risks, the number of cases anddeaths averted was calculated by multiplying this dif-ference in morbidity and mortality risk by the baselinemorbidity and mortality rates for infants under 6months. These are taken from the literature asfollows:

• 3 episodes of diarrhoea per child per year forchildren under 6 months;27-28

• 13.5 diarrhoea deaths per year per 1000 childrenunder 6 months, based on an annual rate of 20.2diarrhoea deaths per 1000 children under 12months29 and an estimated two-thirds of theseoccurring in children under 6 months;30"32

• 10 ARI deaths per year per 1000 children under6 months, based on an annual rate of 15 ARIdeaths per 1000 children under 12 months with anestimated two-thirds of these occurring in childrenunder 6 months.

Use of these numbers of diarrhoea cases and deathsfrom diarrhoea and ARI assumes that there are nobenefits after the first 6 months - a conservativeassumption given that relative risks are in factslightly greater than 1 in the second half of infancy.Also, mortality from other infections (shown to be2.5 times greater in non-breastfed infants as com-pared with exclusively breastfed infants by Victoraet al.4) are not included. The results therefore cap-ture the bulk of the impact of the interventions onmortality and morbidity but not all, and our estimatesshould be considered the lower bound for actualvalues.

Estimating additional disability-adjusted life years(DALYs)The DALY indicator combines the impacts on livessaved with disability prevented, discounts and weightsthe years of life based on age, thus allowing com-parisons across varied health interventions, with dif-ferent health outcomes. It also permits comparisonsacross interventions with longer- versus shorter-termbenefits and benefits accruing to different age groups.For global estimates, the World Bank21 suggests thatthe death of an infant (girls and boys do not differsignificantly) is equivalent to the loss of approxi-mately 32.5 disability adjusted life years (DALYs).This is based on life expectancy of 80 years in a low-mortality population, a 3% discount rate for the valueof future years of life saved, and age weights whichrise steeply from 0 at birth to a peak at age 25, andtaper off with age. For this analysis, we multipliedthe number of diarrhoea and ARI deaths averted per

Breastfeeding and priority setting 159

year by 32.5 to provide a conservative estimate ofthe total DALYs gained from the intervention peryear. This does not include any effect on total DALYsof morbidity due to inadequate breastfeeding.

Estimating costs and cost-effectivenessThe costs of breastfeeding promotion activities at thecontrol hospitals were compared with those at pro-gramme hospitals to obtain an incremental cost whichrepresents the value of resources used to achieve themeasured impacts. Programme maintenance costswere the main focus of the study. Start up, and one-time activity costs are not included. Programmes forbreastfeeding promotion seldom involve large set-upor capital investments to get started. In the pro-grammes included here, start-up occurred over avariable period, and in Brazil and Honduras severalyears ago, making it difficult to accurately estimatethese costs. Also, sustainability is likely to be morea function of recurring, maintenance costs than one-time costs.

The costing methodology is based on classical ap-proaches developed by Mills et al.33 and Phillips etal.34 and extensively used in health programme plan-ning and analysis.35 Each hospital programme wasfirst carefully elaborated in terms of activities under-taken for breastfeeding promotion. The nature andlevel of resources associated with each activity werethen determined. This included disaggregating thetime of multi-purpose staff and attributing propor-tions of staff time, space, materials and equipmentto breastfeeding services delivered. Unit costs wereobtained from expenditure records, market prices orreplacement costs. Donated goods were valued atmarket prices.

From the profile of total costs developed, the dif-ference in costs between breastfeeding activities atprogramme and control hospitals - incremental cost- was obtained. Savings due to reduced formula andother supplies were then subtracted from incremen-tal costs to obtain net incremental costs/ For eachprogramme hospital, the net incremental cost ofbreastfeeding activities for one year (1992) wasestimated. A depreciation rate of 3% was used tocalculate the annual costs for capital goods. Costs andsavings are expressed in 1992 prices for Hondurasand Mexico and 1993 prices for Brazil.36"38 For theBrazil programme, Fiedler38 provfdes estimates fora range of scenarios. We use the estimates that bestreflect typical practices at a service hospital (the pro-gramme Brazil hospital is a teaching hospital).

Postnatal clinics are estimated to dedicate 50% oftheir resources to breastfeeding activities and nursesalaries have been substituted for salaries of physiciantrainees (medical residents) in the postnatal clinics.For Honduras, it was assumed that the programmecould be targeted to non-working women at the samecost per birth as the current programme which coversall women, working and non-working.

A set of cost-effectiveness indicators were derivedby dividing annual net incremental costs by theestimated annual number of (i) cases of diarrhoeaaverted; (ii) diarrhoea deaths averted; and (iii)DALYs gained (including both diarrhoeal and ARIdeaths averted). Since the objective of this analysisis to determine relative priorities among health in-terventions, the data on costs and effectiveness arecombined to provide as close a valid comparison aspossible to cost-effectiveness estimates for other in-terventions. The World Development Report21 andJamison and Moseley22 provide the most recentestimates for other health interventions which are us-ed here for comparisons. The principles used fordeveloping estimates are also used here for estimatingthe cost-effectiveness of breastfeeding strategies. Forcomparisons with diarrhoeal control interventions weestimate the cost of breastfeeding promotion per diar-rhoea case averted and per diarrhoea mortalityprevented; for comparisons across all health interven-tions, cost per DALY gained is used.

In the cases of Honduras and Brazil, cross-sectionalcomparisons of the programme hospital net incre-mental costs do not adequately capture the cost-effectiveness of breastfeeding strategies. Forexample, the full benefit of savings generated atprogramme hospitals from reductions in formula(and other modifications such as rooming-in) thatoccurred during the course of the development of thebreastfeeding programme is not included because inboth countries control hospitals already limit formulafeeding to very few cases (and have institutedrooming-in). To illustrate this point, and to developa more accurate estimate of cost-effectiveness whengoing from formula feeding to no/limited formulafeeding, we estimated cost-effectiveness based onchanges at the programme hospital over time. Underthis scenario, according to recall by staff who werepresent at the time, breastfeeding activities weresimilar (during pre-1975, prior to the breastfeedingprogramme) to those currently practiced at the con-trol hospital, in all respects except that approximately50% of all infants were fed formula.

160 Susan Horton et al.

In all, 6 cost-effectiveness estimates were developedfor each indicator as follows:

• Brazil (restricted formula and low education com-pared with restricted formula and high education):this is based on comparing current practices at thecontrol hospital with current practices at the pro-gramme hospital;

• Brazil (50% formula and low education comparedwith restricted formula and high education): thisis based on comparing pre-1975 practices with cur-rent practices, both at the programme hospital;

• Honduras (restricted formula and low educationcompared with restricted formula and moderateeducation): this is based on comparing currentpractices at the control hospital with the pro-gramme hospital;

• Mexico (mostly formula and low education com-pared with restricted formula and moderate educa-tion): this is based on comparing current practicesat the control hospital with the programmehospital.

ResultsDifferences in the nature, quality and intensity of pro-motional activities in the control and programmehospitals, as reported at exit by mothers who gavebirth in the hospitals, are shown in Table 1. Allinfants (except those with severely ill mothers orhaving serious complications) were roomed in withtheir mothers in five of the six hospitals. The dif-ferences in breastfeeding promotions varied acrossand within countries. In the cases of Brazil andHonduras, it was essentially education and supportthat varied since even the control hospitals had well-established policies of rooming-in and very limitedbottle-feeding. In the case of Mexico, routines suchas rooming-in and no bottle-feeding are vastly dif-ferent in the programme and control hospitals.

Table 2 shows the estimated additional cost per headof investment in breastfeeding promotion activitiesfor the three programmes. It also shows the net costswhen savings due to differences in bottle-feeding,rooming-in and drug routines are taken into account.This is shown for two scenarios in Brazil: 1) whencompared with current norms and routines in con-trol hospitals, and 2) when compared with the normsand routines that existed historically in the pro-gramme hospital.

The incremental annual costs of the four programmeswere similar, in the range of US $2.61 to US $2.73per birth. However, the costs are relative to differentstarting points. For example, in Mexico, the in-cremental cost reflects a start-up phase in which thetotal coverage with services is still low; in the casesof Honduras and Brazil the comparison is being madewith a more advanced baseline (control hospitalswhich already have several breastfeeding activitiesunderway). The levels of savings generated fromchanges in routines are also different because thelevel of formula use at baseline varies from 100%in Mexico to 50% in Brazil (historical) to almost noformula in Honduras and (current) Brazil. Net incre-mental costs of breastfeeding when moving from asituation of no rooming-in and high formula use(Mexico, pre-programme Brazil) are lower (US $0.28to US $0.37 per birth). Net incremental costs aresomewhat higher when moving from a situation ofrooming-in with formula feeding already removed,where the main difference is in the quality and cover-age of maternal education and counselling (US $2.01to US $2.81 for Brazil and Honduras respectively).

The differences in breastfeeding practices in the threeprogramme and control hospitals are presented inTable 3. Each of the three country studies found thatany breastfeeding and exclusive breastfeeding werehigher in the programme hospitals as compared withcontrol hospitals.

In a separate analysis, Lutter et al.20 compared dif-ferences in the duration of exclusive breastfeedingafter controlling for differences in maternalcharacteristics in each pair of hospitals shown inTable 3, and found significant increases in Brazil andHonduras. In Brazil, median duration of exclusivebreastfeeding increased from 22 days in the controlto 75 days in the programme hospital. The differencefor Honduras was 14 days in the control and 40 daysin the programme hospitals Differences in Mexicowere not significant for exclusive breastfeeding(median durations being 6 and 7 days for control ver-sus programme hospitals, respectively, after controll-ing for potential confounders). However, in Mexico,significant differences in probability of any breast-feeding were found after controlling for confoundingvariables.19 This suggests that the breastfeeding pro-gramme was more effective in reducing the propor-tion of non-breastfeeders in Mexico, and the changefrom partial to exclusive breastfeeding was lessdramatic. Although some a priori differences existedin each pair of (control and programme) hospital

Breastfeeding and priority setting 161

Table 1. Differences in breastfeeding promotion activities in programme and control hospitals (% of mothers who answered yes dur-ing exit interviews)

Coverageindicator

No formulafeeding

BF in deliveryroom

Help BF firsttime

Shown how toexpress milk

Info, on milkadequacy

Info, on how toincrease supply

Info, on whento start liquids

BrazilProgramme

n=236

99.6

65.3

72.0

68.2

49.2

61.0

32.6

Controln=206

90.3

2.2**

33.7***

5.4***

3.9***

5.3***

2.9***

HondurasProgramme

n=501

98.0

39.3

30.3

23.4

17.8

45.9

20.9

Controln=488

98.0

39.3

9.4***

24.0

97***

23.3***

24.9

MexicoProgramme

n=333

88.2

3.9

31.7

25.3

10.3

20.6

10.6

Controln=247

0.4***

0.6

o***

17.1*

0.7**

8.3***

1.2***

Differences between pairs of hospitals are identified as follows: ' p i 0 .05, M p s 0 .01 , • • • p S 0.001

Data from Sanghvi et al. (1994)

Table 2. Annual costs of breastfeeding promotion activities (1992)

Annual costs

No. births

Annualcosts/birth

Incrementalcost/birth

Saving/birth:

Compared withcontrol

Compared withprc-programmebaseline

Net incrementalcost/birth:

Compared withcontrol

Compared withpre-programmebaseline

Brazil1

Programme Control

18 620 10 388

1623 1188

11.47 8.74

2.73

0.72

2.36

2.01

0.37

Honduras2

Programme Control

11 516 1 579

3895 10 590

2.96 0.15

2.81

0

2.81

Mexico3

Programme Control

18 896 410

7010 4790

2.70 0.09

2.61

2.34

0.28

1 Brazil data are from Fiedler (1994). Fifty per cent of postnatal clinics are dedicated to breastfeeding activities, and salaries of nursesare used to compute costs of medical interns. Exchange rate USSl .00 = 22 000 cruzeiros.2 Honduras data are from Phillips (1993). Exchange rate US$1.00 = 5.7 lempiras.3 Mexico data are from Phillips (1993a). Exchange rate US$1.00 = 3.09 New Pesos

162 Susan Horton et al.

Table 3. Breastfeeding status in programme and control groups (%)

Breastfeedingstatus

Not breastfed

Partiallybreastfed

Exclusivelybreastfed

Total

Brazil1

Programmen=193

23.7

32.9

43.4

100

Controln=161

35.4

44.7

19.9

100

Honduras2

Programmen = 144

0.6

56.7

42.7

100

Controln=245

4.1

73.7

22.2

100

Mexico3

Programme Controln=175 n = 147

42.3(42.2)

51.9(51.9)

5.8(5.9)

100

51.0(53.6)

47.6(45.1)

1.4(1.3)

100

1 Brazil data are for age 3 months, from Lutter et al. (1994).2 Honduras data are for age 2 months. From Lutter et al. (1994) (breastfeeding practices of non-working women only).3 Mexico data are for age 4 months, from Perez-Escamilla et al. (1993). Adjusted prevalences, after controlling for maternal attributesfound to be associated with breastfeeding duration, are in parentheses, from Perez-Escamilla et al. (1995).

populations, the nature and magnitude of differencesdo not suggest a systematic bias that would alter theconclusions of this paper.h

Differences in breastfeeding behaviour were trans-lated into percentage reduction in diarrhoea mortality,ARI mortality, and diarrhoea morbidity using relativerisk ratios as shown in Table 4. Differences in thebreastfeeding promotion activities were estimated tohave had a substantial, favourable impact on diar-rhoeal disease incidence and mortality and on ARImortality, and, consequently, on DALYs gained.

The net cost and effectiveness data are combined inTable 5 to give the cost per case of diarrhoea averted,per diarrhoeal death averted and per DALY gained.The DALY estimate combines diarrhoea and ARIdeaths averted.

In Table 6, cost-effectiveness comparisons of diar-rhoea control interventions are shown using ouranalysis of breastfeeding promotion and estimates inthe literature for oral rehydration therapy (ORT),cholera immunization, measles immunization, rota-virus immunization and promotion of hygiene.

Discussion and conclusionsUsing a "quasi-experimental, opportunistic design tomeasure effects on breastfeeding practices, this studyhas documented the nature and level of impact thatcan be expected from field programmes that modifybreastfeeding promotion practices in maternity ser-vices. It also identifies the resource implications of

achieving these changes. This has enabled us to pro-ceed with making some judgements about the relativeefficiency and attractiveness of investing in hospital-based breastfeeding promotion compared with otherhealth investments. In this section, we first compareour findings with previous estimates in the literatureof the cost-effectiveness of breastfeeding promotion,followed by a discussion of the relative cost-effectiveness of breastfeeding promotion when com-pared with other health interventions.

We were not able to locate any other studies whichhave measured directly the cost and effectiveness ofbreastfeeding promotion programmes in hospitals. In-deed there are few credible economic evaluations offield programmes of any breastfeeding promotion in-tervention. In a review of potential diarrhoeal diseasecontrol strategies,5'40 cost estimates of breastfeedingpromotion activities were made, though these wereretrospective reconstructions of the value of likelyinputs for programmes whose impacts were mea-sured, rather than direct measurements of costs.That exercise generated estimates for breastfeedingpromotion (one of which was hospital-basedpromotion) ranging from (US$ 1992) $13 to $100(median $60) per diarrhoea! episode averted and bet-ween $538 and $14 448 (median $1344) per diar-rhoeal death averted. These are substantially higherthan our new estimates.

Part of the difference is explained by the generouscost estimates for breastfeeding promotion activitiesin the hypothetical exercise - $7 per birth - more

Breastfeeding and priority setting 163

Table 4. Mortality and morbidity impacts of differences in breastfeeding status

Location ofprogramme

Brazil

Honduras

Mexico:unadjustedadjusted

Percentage reductionDiarrhoeal ARImortality mortality

a

27.1

25.8

10.913.5

b

15.9

10.2

8.310.5

in:Diarrhoealmorbidity

c

18.7

13.9

8.510.7

Diarrhoealcases averted

d

561

417

255321

Per 1000ARI

deaths averted

e

1.59

1.02

0.831.05

birthsDiarrhoeal

deaths averted

f

3.66

3.48

1.471.82

DALYgained

g

171

146

7593

Notes:a,b,c are derived from Table 3, using the following relative risks equations given in the text:

Relative risk of diarrhoeal mortality is from Victora et al. (1987), as follows: compared with exclusive breastfeeding =partial breastfeeding = 4.2, no breastfeeding = 14.2

Relative risk of ARI mortality is from Victora et al. (1987), as follows: compared with exclusive breastfeeding = 1,partial breastfeeding = 1.3, no breastfeeding = 3.4

Relative risk of diarrhoeal morbidity is from Martines (1988), as follows: compared with exclusive breastfeeding = 1,partial breastfeeding = 1.65, no breastfeeding = 4.22

d = 3 x c (3 diarrhoea cases per chiloVyear in children under 6 months)e = 10 x b (10 ARI deaths per 1000 children/year in children under 6 months)f = 13.5 X a (13.5 diarrhoeal deaths per 1000 children/year in children under 6 months)g = (e + 0 x 32.5 years (32.5 disability-adjusted life years gained per infant death prevented)

Adjusted estimates for Mexico reflect breastfeeding prevalences estimated when controlling for other confounding variables.

Table 5. Cost-effectiveness of hospital-based breastfeeding promotion

Birth1Net cost (USS 1992) per

Diarrhoeal case Diarrhoeal death DALY gained11

avertedb avertedc

Brazil (current)

Brazil (historical)

Honduras

Mexico:unadjustedadjusted

2.01

0.37

2.81

0.280.28

3.58

0.66

6.74

1.100.87

549

101

807

190153

11.78

2.17

19.21

3.753.00

" From Table 2

b a X 1000/column f in Table 4

c a x 1000/column d in Table 4

d a X 1000/column g in Table 4

164 Susan Horton et al.

Table 6. Comparison of the cost-effectiveness of interventions for diarrhoea control

Intervention1 Cost per case averted(USS 1992)

range median

Cost per death averted(USS 1992)

range median

Notes

Measlesimmunization

Breastfeedingpromotion:from formulafrom noformula

Oralrehydrationtherapy

Rotavirusimmunization

Choleraimmunization

Promotion ofhygiene

4-80

0.65-1

3.50-6.75

NA

4-40

121-1949

7-672

9

1

NA

7

234

13

89-1554

100-200

550-800

1344-13 440

188-1882

1445-22 458

NA

192

150

2688

296

2688

NA

Hypothetical estimates built up from individual cost compo-nents. Assumes addition of measles to existing EPI programme.Does not take into account current high coverage and thushigher incremental costs to raise coverage further.

Based on actual programme costs. Multiple benefits (e.g. ART)not included.

Hypothetical programme. Median represents USS 1.34 per casetreated and for every 20 cases, one death averted.

Hypothetical programme. Assumes 100% coverage, with 80%vaccine efficiency.

Hypothetical programme. Assumes 100% coverage, with 85%vaccine efficiency.

1 Cost-effectiveness estimates for all interventions except breastfeeding promotion are from Martines J, Phillips M and Feachem RGA(1993) 'Diarrhcal Diseases' in Disease Control Priorities in Developing Countries (ed.) Jamison et al. (1993), World Bank. 1982 USShave been adjusted to 1992 USS assuming an average annual inflation rate of 3%.

than twice any of the costs per birth actually measuredin our study. The rest is explained by the conservativeassumption in that study that the reduction in non-breastfeeding infants (30%, very similar to themedian 33% we measured in our study) translatedsolely into increases in partial, but not exclusive,breastfeeding, while our study found that there weresubstantial increases in exclusive breastfeeding inHonduras and Brazil.

The differences in cost-effectiveness among the threecountries (in large measure explained by the savingscomponent) are discussed in more detail in Phillipset al.;18 however, it is worth noting here that therange of cost-effectiveness estimates obtained in thisanalysis is indicative of expected values for varyingpolicy contexts. Clearly, when breastfeeding promo-tion is accompanied by a shift from substantial for-mula use to almost no formula feeding (Mexico,Brazil historical), the cost-effectiveness of the pro-gramme is likely to be high. The cost-effectivenessis lower when formula has already been removed(Honduras, Brazil current).

With reference to other interventions, our estimatessuggest that breastfeeding promotion in hospitalscompetes very closely with measles and rotavirusvaccination as the most efficient option for diarrhoealdiseases control. This is dramatically obvious incases where hospitals still use formula (e.g. Mexico,Brazil historical). Breastfeeding promotion ismarkedly more cost-effective than ORT and choleraimmunization even when formula has already beenwithdrawn (e.g. Brazil current and Honduras).

In considering priorities for future investments,hospital-based breastfeeding promotion is likelyto be relatively even more attractive in terms ofexpansion since, unlike the vaccination options, thereare many areas without promotion programmes inhospitals and initial start-up or establishment costsare not large.18 Furthermore, where programmesalready exist, effective coverage is still generallylow and the scope for expansion before reaching thepoint of diminishing returns, or encountering resis-tant pockets in the population, is likely to beconsiderable.

Breastfeeding and priority setting 165

Using the DALY indicator, broader comparisons withother kinds of health interventions are possible(Figure 1). They reveal that investments in breast-feeding promotion are among the most cost-effectivehealth interventions available. This applies not onlyto the case where formula is used prior to the in-tervention, but also to investment in education andsupport of mothers, even when reductions in formulause have already occurred.

This is an important finding. Altering routines hasbeen an appealing option for policy-makers keen, inthe current climate of financial constraint, to iden-tify savings potential. What our results suggest is thatto stop at the stage of changing hospital routines forformula feeding without completing the next step ofestablishing comprehensive support and educationalactivities for mothers, especially in the post-partumperiod, would be to miss out on a highly attractivehealth investment.

The budgetary requirements for maintaining breast-feeding promotion programmes are modest, and thepayoffs large. At a net incremental cost ranging fromabout 30 to 40 US cents per birth, programmesstarting with formula feeding in nurseries and matern-ity wards can reduce diarrhoea cases for approxi-mately US$0.65 to US$1.10 each, prevent diarrhoeadeaths for US$100 to US$200 each and reduce theburden of disease for approximately US$2 to US$4

per DALY. Maternity services that have alreadyeliminated formula can, by investing from US$2 toUS$3 per birth, prevent diarrhoea cases and deathsfor US$3.50 to US$6.75 per case, and US$550 toUS$800 per death respectively, with DALYs gainedat US$12 to US$19 each.

We have considerable confidence in our conclusionsnot only because of the way in which key outcomes(breastfeeding practices) and costs were measureddirectly rather than imputed, but also because of theconservative approach we employed in measuring im-pact.' Modest assumptions were adopted wherethere was room for doubt about any values (for ex-ample, in the choice of mortality risks associated withbreastfeeding practices, and by assigning a zero valueto benefits from breastfeeding beyond 6 months ofage). Several possibly important impacts wereexcluded altogether, such as the longer-term effectsof promotion on breastfeeding practices with subse-quent babies, the effect on birth spacing, and theeffects on maternal and infant emotional health anddevelopment.

It is true that cost-effectiveness may differ in othercontexts. In groups with a high proportion of womenin the formal work force and without maternity leaveprotection, the impact of such programmes may belower;42 in areas where mortality and morbidityrisks are different, the proportion of cases and deaths

100

10

iS

0.01

0.001

0.0001

Chemotherapy for tuberculosis

$100/DALY

J $1000/DALY10,000 1000 100 10 1

Cost per intervention or per Intervention-year (dollars, log scale)

0.10

BF1 - programmes moving from formula feeding to no formula and education (Brazil historical, Mexico)BF2 - programmes investing in education and promotion after formula removal (Honduras, Brazil current)

Figure 1. Benefits and costs of health and nutrition interventions

166 Susan Horton et al.

averted and DALYs gained will change. In general,however, our results are likely to be broadly relevantto many developing country settings: they are basedon interventions managed locally and implementedunder normal field conditions, and not on controlledfield or clinical trials, or pilot programmes withspecial external backing. The interventions wereimplemented in populations of poor, urban womenwith little education, a description which would fita large proportion of developing country women. Theinterventions were implemented across a spectrumof rather different basic breastfeeding patterns,ranging from short durations (Mexico) to longer dura-tion but very early supplementation (Honduras).

Countries where exclusive breastfeeding is not widelypracticed in the early months, and where a signifi-cant proportion of women give birth in healthfacilities,.should give priority to the promotion ofbreastfeeding in these facilities. There are bothmedical and economic arguments for establishingroutines propitious to breastfeeding such as rooming-in and removal of formula feeding, and there isevidence from this study that investment in educa-tion and support can generate important changes inbreastfeeding practices and that this is a highly effi-cient way of improving health status.

Endnotes* Evidence of an association of feeding mode with risk of mor-

tality in infants comes from epidemiologic studies undertaken inseveral countries in Asia and Latin America. Breastfed infants areless likely to develop diarrhoea, complications such as severedehydration when they do get diarrhoea, and have a lower riskof dying. Deaths due to respiratory infections are also lower. Therisk of death is progressively worsened with a reduction in theamount of breastfeeding, from exclusive breastfeeding to partialbreastfeeding to no breastfeeding at all; even the addition of juices,water and teas increases the risks. The impacts are most severein early infancy, decline substantially after 6 months, and disappearafter 12 months. The effects accrue from reduced exposure tocontaminated bottle-feeds and from maternal anti-infective agentspassed through breastmilk.

b Mothers were not eligible if they or their infants wereseriously ill, were resident outside the city or did not wish toparticipate. In Brazil, criteria for exclusion also included birthweight <2000 grams, certain medications (lithium, thyroid andchemotherapy drugs), or if they planned to give their infant upfor adoption.

c In separate analyses reported elsewhere, the increase inmedian duration of exclusive breastfeeding in Mexico, Brazil andHonduras,20 and any breastfeeding for Mexico19 was estimatedthrough multivariate survival analyses. The results show that dif-ferences attributable to programme impact are statistically signifi-cant for exclusive breastfeeding durations in Honduras and Brazil,

and for any breastfeeding (but not exclusive breastfeeding) inMexico.

d The benefits of breastfeeding programmes in the earlyneonatal period are not captured in this study. These are substan-tial, as differences in in-hospital breastfeeding practices were largein Mexico and Brazil (historical), and differences in relative risksof mortality and morbidity due to breastfeeding practices areseveral-fold higher at younger ages.4*25

e An earlier review of several studies suggested that infants inthe first 6 months of life who do not receive any breastmilk areabout 25 times more likely to die and infants partially breastfedabout 9 times more likely to die than those exclusively breastfed.5

These rates are considerably higher than the ones we use.f The savings generated from instituting rooming-in have not

been included because detailed, reliable information was notavailable. A study conducted by Valdes et al.39 compared separaterecovery rooms for mothers and newborns postpartum recoverywith mothers and newborns kept together in the same recoveryroom. Results indicated a 34% reduction in personnel time, andwhen $2000 was included for capital investments to make changesat start-up, depreciated over 10 years at 5%, cost of care permother/infant pair declined from $3.57 to $3.05 (1992 US$). Ourestimates of net incremental costs are likely to be higher becausethese savings were not included.

• Since the proportion and nature of employment in the twogroups of women giving birth at these hospitals was different, andinsufficient detail on the nature of employment was available tocontrol for confounding through multiple regression analysis, theseestimates were made for non-working women only. Eighteen percent of the women at the control hospital worked compared with61 % of women at the programme hospital. It is very likely thatthe impacts of breastfeeding promotion in Honduras, and conse-quently cost-effectiveness, would be significantly less amongworking women.

h To test the effect of unadjusted versus adjusted prevalence,we estimated mortality, morbidity and DALYs using projectedprevalence for any breastfeeding for Mexico from Perez-Escamillaet al.19 The results showed a greater impact of the programme,with lower cost per unit of effectiveness. See Tables 4 and 5.

' Our estimates of mortality and morbidity reduction are con-siderably lower than those estimated, for example, on a population-wide basis by Monteiro et al.41 for programme effects in SaoPaulo cky.

References1 Victora CG, Barros FC. Review Report: Diarrheal diseases

among children under five years in the Americas 1980-1992.Report commissioned by UNICEF. Universidade Federal dePelotas. 1993.

2 Popkin BM, Adair L, Akin JS, Black R, Briscoe J, Flieger W.Breast-feeding and diarrheal morbidity. Pediatrics 1990;86(6): 874-80.

3 Brown KH, Black RE, de Romana GL, de Kanashiro HC.Infant-feeding practices and their relationship with diarrhealand other diseases in Huascar (Lima), Peru. Pediatrics 1989;81(1): 31^M).

4 Victora CG, Smith PG, Vaughan JP et al. Evidence for pro-tection by breast-feeding against infant deaths from infectiousdiseases in Brazil. The Lancet 1987; 2(8554): 319-21.

3 Feachem RG, Koblinsky M. Interventions for the control ofdiarrheal diseases among young children: promotion ofbreastfeeding. Bulletin of the World Health Organisation 1984;62(2): 271-91.

Breastfeeding and priority setting 167

6 Dimond HJ, Ash worth A. Infant feeding practices inKenya, Mexico and Malaysia: the rarity of the exclusivelybreastfed infant. Human Nutrition: Applied Nutrition 1987;4 1 : 51-64.

7 Perez-Escamilla R. Breastfeeding in Africa and the LatinAmerican and Caribbean region: the potential role of urban-ization. Journal of Tropical Pediatrics 1994; 40: 1-6.

8 Adair LS, Popkin BM, Guilkey DK. The duration ofbreastfeeding: how is it affected by biological, sociodemo-graphic, health sector and food industry factors? Demography1993; 30(1): 63-80.

9 Stewart JF, Popkin BM, Guilkey DK, Akin JS, Adair L,Flieger W. Influences on the extent of breastfeeding: a pros-pective study in the Philippines. Demography 1991; 2J8(2):181-99.

10 Winikoff B, Castle M, Laukaran V (eds). Feeding infants infour societies: causes and consequences of mother's choices.Westport, Conn.: Greenwood Press, 1988.

11 Margen S, Melnick V, Neuhauser L, Rios E. Infant feedingin Mexico: a study of health facility and mothers' practicesin three regions. Nestle Infant Formula Audit Commission.1991. Washington DC.

12 WHO/UNICEF. Protecting, promoting and supporting breast-feeding: the special role of maternity services. A joint WHO/UNICEF statement. Geneva: World Health Organization,1989. •

13 Winikoff B, Baer EC. The obstetrician's opportunity:translating 'breast is best' from theory to practice. AmericanJournal of Obstetrics and Gynecology 1980; 138; 105-15.

14 Burkhalter BR, Marin PS. A demonstration of increased ex-clusive breastfeeding in Chile. International Journal ofGynecology and Obstetrics 1991; 34: 353-9.

15 Perez-Escamilla R, Segura-Millan S, Pollitt E, Dewey KG.Effect of the maternity ward system on the lactation successof low-income urban Mexican women. Early Human Develop-ment 1992; 3 1 : 25-40.

16 Horton S. Unit costs, cost-effectiveness and financing of nutri-tion interventions. PHN Working Papers WPS 952.Washington DC: World Bank, 1992.

17 Sanghvi TG, Lutter CK, Perez-Escamilla R, Rivera A, TeruyaK et al. The coverage and quality of breastfeeding promo-tion through maternity services in three Latin American coun-tries. Report for USAID. Mimeo report ISTI/LAC HNS.Washington DC, 1994.

18 Phillips M, Sanghvi TG, Fiedler J, Lutter CK, Perez-EscamillaR et al. Factors affecting the cost-effectiveness of breastfeedingpromotion through hospitals. Report for USAID. Mimeoreport ISTI/LAC HNS. Washington DC, 1994.

19 Perez-Escamilla R, Lutter CK, Wickham C. Effectiveness ofbreastfeeding promotion efforts in Mexico. Report forUSAID. Mimeo report ISTI/LAC HNS. Washington DC,April 1995.

20 Lutter CK, Perez-Escamilla R, Segall A, Sanghvi T, TeruyaK, Rivera A. The effect of hospital-based breastfeeding pro-grams on exclusive breastfeeding in three Latin Americancountries. Report for USATD. Mimeo report ISTI/LAC HNS.Washington DC, 1994.

21 World Bank. World Development Report 1993: Investing inHealth. Published for the World Bank. New York: OxfordUniversity Press, 1993.

22 Jamison DT, Mosley W H , Measham AR, Bobadilla JL .Disease control priorities in developing countries.Published for the World Bank. New York: Oxford Univer-sity Press, 1993.

23 Sanghvi TG, Perez-Escamilla R, Lutter CK, Rivera A, TeruyaK, Segall-Correa A. What urban low-income women in LatinAmerica say about plans, practices and reasons for terminatingexclusive breastfeeding. Report for USAID. Mimeo reportISTI/LAC H N S . Washington DC, 1995.

24 Victora C G , Smith PG, Vaughan JP , Nobre LC, Lombardi Cet al . Infant feeding and deaths due to diarrhea. AmericanJournal of Epidemiology 1989; 120(5): 1032-41.

25 Mata L, Carvajal JJ, Garcia M E et al. Effect of hospital andfield interventions on breast-feeding, health and survivalof infants. Paper presented at the Western HemisphereNutrition Congress, Miami, Florida, August 7 -12 , 1983.Monograph.

26 Martines J. The interrelationships between feeding mode,malnutrition and diarrheal morbidity in early infancy amongthe urban poor in southeastern Brazil. 1988, Doctoral thesis,University of London.

27 Bern C , Martines J, de Zoysa I, Glass RI. The magnitude ofthe global problem of diarrheal disease: a ten-year update.Bulletin of the World Health Organisation 1992; 70(6):705-14.

28 de Zoysa I, Rea M, Martines J. Why promote breastfeedingin diarrhoeal disease control programs? Health Policy andPlanning 1991; 6: 371-9.

29 Huilan S, Zhen GL, Mathan MM et al. Etiology of acute diar-rhea among children in developing countries: a multicentrestudy in five countries. Bulletin of the World Health Organisa-tion 1991; 69(5): 549-55 .

30 DHS n , 1991 for Northeast Brazil, Macro Systems, Columbia,Maryland, USA.

31 National Family Health Survey, FHI, Honduras 1991.32 D H S I, 1987 for Mexico, Macro Systems, Columbia,

Maryland, USA.33 Mills A, Drummond M. Value for money in the health sector:

the contribution of primary health care. Health Policy andPlanning 1987; 2(2): 107-28.

34 Phillips M, Mills A, Dye C. Guideline for cost-effectivenessanalysis of vector control. Monograph published by the Panelof Experts on Environmental Management for Vector Con-trol (PEEM), W H O / F A O / U N E P . 1993 WHO/CWS/93 .4 .Geneva.

33 Creese A and Parker D (eds) with M Phillips and R Robert-son. Cost analysis in primary health care programs.WHO/UNICEF. 1990. Oxford: Oxford Medical Publications.

36 Phillips M. The nature and cost of breastfeeding promotionactivities in Hospital General and Hospital de la Mujer,Mexico City. Report for USAID. Mimeo report ISTI/LACH N S . Washington DC, 1993.

37 Phillips M. The nature and cost of breastfeeding promo-tion activities in three Honduran hospitals. Report forUSAID. Mimeo report ISTI/LAC HNS. Washington DC,1994.

38 Fiedler JL . The cost of the breastfeeding program inGuilherme Alvaro hospital, Santos, Brazil. Report forUSAID. Mimeo report ISTI/LAC HNS. Washington D C ,1994.

39 Valdes V, Perez A, Labbok M , Pugin E, Zambrano I, CatalanS. The impact of a hospital and clinic-based breastfeedingpromotion program on a middle-class urban environment.Tropical Pediatrics 1993; 39: 142-51 .

40 Phillips MA, Feachem RG, Mills A. Options for diarrhoealdiseases control. The cost and effectiveness of selected in-terventions for the prevention of diarrhoea. Report for W H O .London School of Hygiene and Tropical Medicine. 1987.

168 Susan Horton et al.

41 Monteiro C, Rea M, Victora C. Can infant mortality bereduced by promoting breastfeeding? Evidence from Sao Paulocity. Health Policy and Planning 1990; 5(1): 2 3 - 9 .

42 Lutter CK, Percz-Escamilla R, Sanghvi T , Rivera A. Mater-nal employment and exclusive breastfeeding: the effect ofusual versus current employment. Report for USAID. Mimeoreport ISTI/LAC H N S . Washington D C , 1994a.

AcknowledgementsThe detailed and thoughtful comments of Dr Barry Popkin (Univer-sity of N. Carolina), Dr Karhryn Dewey (University of Califor-nia at Davis), and Dr Robert HomiV (University of Pennsylvania)greatly helped in strengthening this paper. We also wish to thankDr Jose Martines (WHO/CDD) and Dr Philip Musgrove (WorldBank/PHN) for their comments on an early draft, concerningmethodology development. We are grateful for the encouragementand support of Karen Nurick (USAID) and Tom Bossert (LACHNS). This analysis would not be possible without the collabora-tion of the breastfeeding programme organizers and data collec-tion teams led by Dr Keiko Teruya (Guilherme Alvaro Hospital,Santos), Dr Lilian Dominguez (Tegucigalpa, Honduras), and DrsHoracio Reyes Velasquez, Armando Montano and Adolfo Her-nandez (Hospital General, Mexico City). The study was fundedby the US Agency for International Development, contract No.LAC-O657-C-0O-0051 (LAC Health and Nutrition Sustainability)with URC and ISTI.

BiographiesSusan Horton, PhD, is Professor of Economics at the Universityof Toronto Institute for Policy Analysis.

Tina G Sanghvi, PhD, is Nutrition Advisor at the InternationalScience and Technology Institute in Arlington, Virginia, USA.

Margaret Phillips, MS, MPH, is a consultant in public health andhealth economics, based in the UK.

John Fiedler has a PhD in economics. He works for the Interna-tional Science and Technology Institute at Sturgeon Bay, Wisconsin,USA.

Rafael Perez-Escamilla, PhD, is Assistant Professor in Nutritionin the Department of Nutritional Sciences at the University ofConnecticut, Storrs, USA.

Chessa Lutter has a PhD in nutrition. She works for WELLSTARTInternational in Washington DC.

Ada Rivera, MD, MPH, is Chief of Community and PreventiveMedicine at the Social Security Institute, Ministry of Health, SanPedro Sula, Honduras.

Ana Maria Segall-Correa, MD, MPH, is Assistant Professor ofEpidemiology at the University of Campinas, Sao Paulo, Brazil.

Correspondence: Susan Horton, University of Toronto, Institutefor Policy Analysis, 140 George Street, Suite 707, Toronto, M5S1A1, Canada.