Setting

SNEHA is a secular, Mumbai-based non-profit organization addressing four major areas of public health in urban slums: child health and nutrition, maternal and newborn health, sexual and reproductive health, and prevention of violence against women and children. The Aahar acute malnutrition programme is located in Dharavi, one of the largest urban slums in Asia with an estimated population of 700,000 to over 1 million [13]. Under-five mortality rates in slum areas are estimated at 27 per 1000 live births [14]. The prevalence of severe acute malnutrition (measured by Weight-For-Height <-3SD) is high: 4% of infants under two years old (0–24 months old, 95% CI: 2.7–6.0) suffer SAM [14, 15]. In a subsequent survey, 20% of infants under two years old were wasted (weight-for-height/length z <-2, 95% CI: 17.7, 22.2), 4.6% severely wasted (weight-for-height/length z <-3, 95%CI: 3.5, 5.6), and 18.8% were stunted (height-for-age z score <-2, 95%CI: 16.7, 20.9) [12] (MUAC and oedema were not used).

Intervention characteristics

The CEA focused on one year during which the intervention was fully operational in five sectors of Dharavi. The project aimed to prevent and treat acute malnutrition in infants aged 0–3 years (more information on the project has been published in [2, 3]). Specifically, the Aahar acute malnutrition programme team identified six key result areas to reach these objectives: (1) prevent and treat acute malnutrition in infants under 3 years old; (2) increase in optimal breastfeeding practices among lactating mothers; (3) improve complementary feeding practices in infants 6 months to 3 years of age; (4) reduce infections and increase referrals and treatment of illnesses; (5) completion of immunizations; and (6) improve coverage of services provided by the ICDS. The programme monitoring system, using real time data collected via the mHealth platform, was based on these objectives and monthly reports against indicators were generated to track progress, identify gaps, and build adaptive strategies for optimal program performance. The indicators included were: 1) the total cumulative number of infants under three and pregnant women screened into the program, 2) the total number of pregnant women, SAM, MAM and Normal infants currently in the program, 3) the total number of infants who had left the program and reasons for leaving, 4) the population coverage of growth monitoring, 4) the number of home-based counseling visits to SAM infants, MAM infants, infants under 6 months, and pregnant women, 5) the number of SAM infants in OTP, 6) the number of SAM infants regularly consuming MNT. The mHealth platform used CommCare, a mobile application developed by Dimagi, USA [16].

The Aahar acute malnutrition programme was implemented in partnership with ICDS where 30 Anganwadis (one local ICDS child care centre per population of 1000) covered a total population of 30,000. ICDS standard activities in the communities included growth monitoring (to track weight-for-age), food distribution (take-home-rations), breastfeeding advice, complementary food advice and referral for immunization. In collaboration with the Anganwadis and government health workers, the programme was delivered by Aahar acute malnutrition programme community health workers who facilitated participation of caregivers and their infants younger than 3 years. Their responsibilities included facilitating growth monitoring of weight-for-height with ICDS, counselling, referrals, and support access to treatment. They received intensive training and were monitored and supervised closely by field supervisors based on real time performance tracking.

The programme had two components: treatment and prevention of acute malnutrition.

Decision tree model

A decision tree model (Fig 1) described the pathways along which a child might proceed if admitted to the treatment and prevention programme or if benefitting only from ICDS standard care.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Decision tree CMAM and prevention for Aahar acute malnutrition programme.

Cured, SAM: Improved to MAM or normal (Weight-for-Height z >-3 SD (WHO 2006 growth references) for at least 1 month over the 1 year period extended by the 3 following months as change in nutritional status may be delayed) Cured, MAM: Improved to normal (Weight-for-Height z >-2 SD for at least 1 month within 12 months extended by the 3 following months as change in nutritional status may be delayed) Failure to recover, SAM: Remained SAM. Failure to recover, MAM: Moved from MAM to SAM, or remained MAM over the 1 year period. Failure to recover, Normal: Moved from normal to MAM or SAM over the 1 year period. Default, SAM, MAM or Normal: Turned 3 years old, migrated, or was screened incorrectly. Relapse, SAM: Recovered but relapsed to SAM over the 1 year period. Live, Children in the programme (SAM, MAM or normal): Based on programme data Live, Children not in the programme (SAM no treatment, MAM no treatment or normal): Based on data from survey data.

https://doi.org/10.1371/journal.pone.0205688.g001

Infants exited the programme in one of four ways: cured, death, default, or failure to recover. Infants of normal nutritional status benefitted from the prevention component. Infants with SAM and MAM who were not part of the programme and received standard ICDS services only were assumed to have the same mortality rates as untreated infants. The normal treatment path for MAM and SAM leading to ‘cured’ was “screened, receiving services as described in the treatment section” (1. in decision tree MAM / SAM ‘receiving services’). In some cases, MAM and SAM infants followed other paths that did not lead to ‘cured’ due to refusal of home visits and MNT consumption (2. in decision tree ‘not receiving full services’). The ‘normal’ infants received services for prevention (1. Normal infants receiving services). Screened normal infants were measured (weight and height) monthly after the initial screening and caretakers were counselled on optimal feeding practices and followed while the child was <6 m and if her mother was breastfeeding. Nevertheless, some of the screened normal infants did not receive all prevention services (2. Normal infants not receiving services); they were not measured if they were unavailable at the time of measurement or if their caregivers refused to have them weighed. Some of these infants only benefited from the community events in the area such as breastfeeding promotion. Finally, infants less than 3 years old who were not part of the programme were assumed to benefit from ICDS standard care only and were normal, MAM receiving no treatment or SAM receiving no treatment.

Analytical costing

The analysis aimed to value the total costs of community treatment of SAM and MAM and prevention of acute malnutrition. An activity-based cost (ABC) analysis was used to estimate the total costs by cost centre. We identified cost centres by grouping activities that were related and validated them with SNEHA staff (S1 Table). They included all costs incurred by SNEHA and the households of infants with SAM. The treatment and prevention programme was evaluated for one year from February 2013 to January 2014 when the programme was fully operational. Costing was done using programme yearly costs, key documents including budget and financial reports, and key informant interviews with SNEHA staff (financial, programme, administrative, monitoring). The yearly costs of the treatment and prevention programme were the basis for the cost centres, with time allocation used to split staff salaries and associated costs across cost centres. Time allocation grids were designed based on interviews with SNEHA staff (community organiser, programme coordinator, programme officers, trainers). Activities for each profile were identified based on the post terms of references and were amended as necessary by the staff during an interview. The time spent was as well estimated on a daily and monthly basis. Interviews with the staff managers were conducted to check activities and time estimate. Cost centres allocation was then done using an allocation grid designed in collaboration with the financial team.

We grouped supervision costs in one cost centre to allow for comparison with other studies using similar grouping. All costs were expressed in local currency, and converted from Indian Rupees to US dollars using the July 2014 exchange rate when the analysis was conducted (US$ = INR 60.14). We used US dollars in order to compare our findings with those of previous studies.

Baseline and endline survey costs were not included. We assumed a 2-year shelf life for fixed assets. One-time setup costs were allocated at 50% (the residuum allocated to the programme covering other areas). We were not able to value the cost of ICDS standard care and assumed zero cost in the costing. The estimated costs of the programme therefore represent the incremental cost of adding such a programme to standard ICDS activities.

Outcome assumptions

Outcome data for infants in the treatment and prevention programme were collected by the mHealth platform (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Treatment and prevention outcomes for Aahar acute malnutrition programme programme for girls and boys (admission, cured, death, default and relapse rate).

https://doi.org/10.1371/journal.pone.0205688.t001

Over the time of the study, 12,362 (49% female) infants were enrolled in treatment or prevention programmes, (8980 prevention, 3382 treatment). Of the infants admitted to the treatment programme, 1888 were cured (56%: 60% cured in infants with SAM and 55% cured in infants with MAM), 1056 (31%) failed to recover, 227 (8%) defaulted, 8 (0.2%) died and 203 (6%) did not receive full services (either receiving home visits but not being weighed, or not receiving home visits and not being weighed). Of the 1888 infants who were cured, 75 (18% of the SAM cured) relapsed into the SAM category during the study period of one year (Table 1). Here it is important to note that, unlike in a CMAM programme, the cured infants were not discharged and readmitted when screened for SAM again. These infants were still being monitored even after being considered as cured. Mortality ratios were 2% for the prevention and treatment programme, 1.9% for prevention only, 2.4% for treatment only (Table 1).

DALY assumptions

DALYs are a measure of overall disease burden, expressing the number of healthy life years that are lost due to illness and death. Here, DALYs combined years of life lost (YLL) due to premature mortality related to SAM/MAM and years lived with acute malnutrition (YLD). For each node in the decision tree, either YLL or YLD were calculated using standard formulas for calculating DALYs based on YLD = number of prevalent cases * disability weighting * duration of the SAM episode and YLL = number of fatal cases * (residual life expectancy at mean age at death). Assumptions for calculation of YLL and YLD are described in Table 2. The probability and uncertainty of each outcome were calculated based upon programme data on mortality and recovery rates (Table 2) using similar methods to those of previous studies [5, 6]. We decided not to use discounting and age weighting in the base case scenario due to their controversial nature. Age weighting implies that the value of life depends on age while discount rate refers to the annual loss of value in percent. Nevertheless, to make the findings comparable to other studies [5, 6], we used the same parameters for age weight and discount rate as in these studies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Assumptions used to estimate Disability-Adjusted Life Years (DALYs) based on programme data and other evidence from previous CEA of nutrition studies.

https://doi.org/10.1371/journal.pone.0205688.t002

We used the mHealth programme data when possible. However, for age at death of infants with SAM, the small number of cases could not give us a reliable average. Thus, we estimated the age of death as the age at SAM + the duration of SAM episode based upon a previous study [5]. We used the expected mortality ratio in the population when infants exited the programme or received only ICDS standard care (2.6% in NFHS-3 [11]). For untreated, non-responder and defaulter SAM and MAM case fatality ratios (7.6% for SAM and 3.4% for MAM) were estimated based on Pelletier et al.’s study [23] averaging the death rate for SAM and MAM respectively from 9 studies [17–26]. The case fatality rate used for untreated SAM is lower than the ones used in other studies; 18.1% in Wilford [6], 20.7% in Puett et al. [5], 18% in Bachmann [4]. This can be explained by the fact that they were based on MUAC rather than WFH, with MUAC being a better predictor of mortality than WFH [30–32]. For normal infants who left the programme, death rate was based on infant mortality rate in slums (2.6% in NFHS-3 [11]). Following Bulti et al. [33], the number of SAM deaths averted was calculated as excess mortality * proportion of treated SAM cases cured by the programme * number of SAM cases treated by the programme.

We used the disability weighting for severe wasting established by the Global Burden of Disease 2010 project [28]. For the levels of services received for infants with SAM, we used the confidence interval to reflect lower and higher disability weight. We used the same disability weight for infants with MAM as there were no data in the literature, but also used the lower confidence limit to reflect lower disability weight.

Data analysis and sensitivity/uncertainty analysis

Estimates of DALYs averted were calculated for prevention and treatment compared to standard ICDS care. The incremental cost-effectiveness ratio (ICER) for the treatment and prevention programme versus the ICDS standard care alone was calculated as the estimated cost per DALY averted as no ICDS related cost was included for both options. Based on the 2001 recommendation of the Commission on Macroeconomics and Health, the World Health Organization classifies interventions as ‘highly cost-effective’ for a given country if they avert a DALY for less than per capita gross national income (GNI) or gross domestic product (GDP), and cost-effective if they avert a DALY for less than 3 times GNI or GDP. In India, GNI per capita was 1570 US$ in 2013 [34].

Sensitivity analyses were conducted for base, worst and best case using assumptions in Table 2 and a +/-25% range. The worst case scenario was based on the 25% CI with all inputs varying together and presented the least favourable results.

Uncertainty analysis was undertaken to develop scenarios with Dirichlet probabilities assigned to each possible outcome. The Dirichlet distribution was the natural choice for modelling the uncertainty in the transition probabilities of the decision tree models, because each node could result in two or more mutually exclusive outcomes. The cost uncertainty analysis was conducted by assigning cost centre to outcomes in the decision tree. The average costs for the individual decision tree nodes were used.

Calculations of DALYs, costs, estimated cost per DALY averted, and related uncertainties were performed in R version 3.4.0 (R Core Team 2017) [35]. Uncertainty analyses were based on 10,000 iterations, and 95% uncertainty intervals (UI) were defined as the 2.5th and 97.5th percentiles of the resulting uncertainty distributions.

Potential risk of bias was addressed by taken into account uncertainty in the decision tree parameters using a Monte Carlo-based uncertainty analysis. Furthermore, we performed sensitivity analyses by altering the main mortality scenarios, and by performing alternative social weighting scenarios.

Ethical approval

Ethical clearance was granted by the ethical committee of Bandra Holy Family Hospital and Medical Research centre in Mumbai. We complied with the Principles of Ethical Practice of Loughborough University.

DALYs and cost effectiveness analysis

The number of DALYs averted by the programme was 15,016 (95%UI: 12,246–17,843) in the base scenario (9,479 worst case– 20,697 best case) (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Disability-Adjusted Life Years (DALYs), DALYs averted, and cost per DALY averted under different mortality scenarios (base, best and worst) for the community based treatment and prevention programme versus ICDS standard care only.

https://doi.org/10.1371/journal.pone.0205688.t003

This is the difference between the number of DALYs for the programme for base of 8,912 (95%UI: 7,494–10,425) and the infants less than 3 years old outside the programme (ICDS standard care imaginarily applied to the same cohort of individuals (i.e., the DALYs if those children would have undergone standard care), respectively 23,928 (95%UI: 21,610–26,406). The number of SAM deaths averted was 27 for base, (17 worst case– 37 best case). The estimated cost per SAM death averted was US$12,630 (US$27,724 worst case–US$9,713 best case). The main contributors to the averted DALYs were the normal infants who had been prevented from becoming malnourished (10,961 DALYs averted), the MAM infants (2,311 DALYs averted) and SAM infants (1,743 DALYs averted) (details in S3 Table). The fact that these are the main contributors are explained by the difference in malnutrition prevention and in assumed mortality for non-treatment between the 2 arms.

The ICER of implementing the treatment and prevention programme in addition to existing ICDS health services was, for base, estimated US$23 per DALY averted (95%UI:19–28) (US$36 worst case—US$16 best case). The scenario analysis using 3% time discount rate (S4 Table) shows that the number of DALYs averted was 6,658 (95%UI: 5,437–7,915) compared to 15,016 (95%UI: 12,246–17,843) with no time discounting. The ICER of implementing the treatment and prevention programme in addition to existing ICDS health services was, using a 3% time discount rate US$51 per DALY averted (95%UI: 43–62). When using 3% time discount rate and age weighting, the number of DALYs averted was 7,380 (95%UI: 6,393–9,307) and the ICER was US$43 per DALY averted (95%UI: 36–53).

Effectiveness (health outcomes of the programme)

Of the 3382 infants enrolled in the programme, 1888 were cured (treatment proportion of 56%, of which 76% had SAM and 24% had MAM). The infants who were not cured (44%) included 1056 (31%) who did not recover, 272 (8%) who defaulted, and 203 (6%) who did not receive full services (these infants did not benefit from growth monitoring). These performances are low in terms of cure and default rates according to SPHERE minimum standards and previous studies (Table 4). The SPHERE minimum standards are evidence-based and represent sector-wide consensus on best practice in humanitarian response.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Performance indicators comparison of Aahar acute malnutrition programme with SPHERE and other CEA studies of nutrition interventions.

https://doi.org/10.1371/journal.pone.0205688.t004

Regarding the cure rate, comparison with previous studies should be made carefully as these used a criterion of 15% weight gain, while the Aahar acute malnutrition programme used WHZ ⩾ − 2. This is in line with the new WHO guidance issued in 2013 on acute malnutrition treatment. Compared to previous programmes using the old criteria of 15% weight gain, the use of WHZ ⩾− 2 as a recovery criterion explains the lower recovery rate. Aguayo et al. (2014) have shown that use of WHZ ⩾− 2 reduced recovery rates approximately twofold (17.5%) compared with the use of weight gain ⩾ 15%. The programme cure rate was in line with another large-scale CMAM programme in Bihar state, India that reported a 57% cure rate using the new WHO criterion [36]. The reasons for this low cure rate might be low food security at household level or the fact that infants may have been considered to have recovered too early. A suggested action might be to consider interventions at the household level to increase food security for vulnerable groups [37].

Cost effectiveness and DALYs

The estimated cost of US$23 (95%UI: 19–28) per DALY averted suggests that the model of treatment and prevention is highly cost effective based on the WHO GDP per capita threshold (Commission on Macroeconomics and Health 2001) (in India, GNI per capita was 1570 US$ in 2013). This is the same as reported in CMAM or community based treatment programme studies in countries with similar GNI per capita (Bangladesh: US$1010, Zambia: US$1810 in 2013), US$ 27 (dollar value in 2015) per DALY averted for community treatment of SAM in Southern Bangladesh [5], and US$58 (dollar value in 2015) in Zambia [4]. The programme had similar cost effectiveness to other priority child health interventions and the high caseload of children admitted to the programme contributed to the high number of DALYs averted.

Cost analysis

The overall programme cost was US$ 335,126 This is less than reported in Malawi with a similar number of admissions (US$ 470,703). The difference is explained by the high cost of RUTF in Malawi (US$ 148,519) versus (US$ 20,857) in Aahar acute malnutrition programme. In terms of cost centres, day care centre costs represented the greatest relative expense (14%), followed by supervision and follow-up (12% each). The importance of the day care centre to costs, and the fact that it was related only to infants with SAM and MAM should be taken into consideration in cost structure improvement: might the programme be as effective without day care centres? The estimated cost per death averted was US$13,977, much higher than reported in other studies (US$ 869 in [5], US$1760 in [4]) is mainly due to the lower cure rate.

In order to make comparisons, we grouped costs in comparable categories to other studies.

In this configuration, administration was the main cost item (39%), followed by technical support (29%), management (13%) and day care centres (8%). The high administrative share accrued because administrative costs included central office costs. The second greatest cost component was technical, at 29%, lower than in other programmes (39% in Zambia) [4]. Management costs represented 13% and were lower than in other studies: 53% in Southern Bangladesh [5], 34% in Zambia [4], and 51% in Malawi [6]. This suggests that the programme was not management-heavy. It relied on a light management structure spread across levels and on a close and successful partnership with ICDS and the Municipal Corporation of Greater Mumbai which facilitated implementation of the programme. RUTF contributed only 4% to total cost, versus 24–40% in other CMAM programmes [4–7, 38]. This low cost is explained by the fact that the RUTF was locally produced in Mumbai at low cost compared to other programmes that had to import or transport RUTF at a higher price (US$38 per child vs US$40—US$73 in [5, 7]) (1.5 cup on average per day *56 average number of days for treatment *30 INR cost per cup = 2520 INR or $38). Household costs (US$ 0.72 per child admitted) were low compared to previous studies (US$29 for domiciliary care per child in Dhaka) [8]. Since many of the other studies covered medicine and other costs for cases of SAM, this could be probably explained by the fact that household were not able to dedicate additional resources such as increasing food expenditure. The source of other cost savings could be the leverage and utilization of all public health facilities (use of local health centres and the outpatient therapeutic programme and the stabilisation centre at the tertiary Hospital).

Limitations

Programme performance is routinely evaluated based upon six organisational indicators, but not upon SPHERE indicators. For this analysis, SPHERE standard indicators had to be recalculated to allow for comparison with previous studies.

The analysis used zero to value the cost of ICDS standard care due to unavailability of cost data of the programme. This is a limitation as it would have been more useful to reflect the full cost of the intervention (ICDS + Aahar acute malnutrition programme).

A potential bias in the cost calculation could be due to either the overestimation or underestimation of time spent by staff members on each activity.

Recommendations

The strengths of the programme are cost-effective prevention and treatment of acute malnutrition in a large caseload of infants, leading to a high number of DALYs averted by complementing India’s national ICDS programme. Areas of improvement in the programme performance have been identified related to the low cure rate and further research should be done to investigate if the poor performance for cure rate is due to the service delivery (e.g. the non-standard definition of discharge as cured, the use of WHZ versus MUAC, the lack of adherence to RUTF) or to household level factors (e.g. food insecurity).

Various levels of service were provided within the community based treatment and prevention programme. This is not typically the case in CMAM programmes. Although weighing infants can give valuable information to the caretaker, the fact that some of these infants have SAM means that without appropriate treatment they will not recover and face a high risk of dying. In addition, the high number of infants who did not respond to the treatment provided means that SNEHA could explore how the programme could improve the quality of the services and standardize the level of services by better promotion of the intervention. This could be linked to improving the acceptability of locally made RUTF. Since this research, Action Against Hunger (ACF) has worked in collaboration with the production unit to improve its quality and bring it up to WHO standards. This should lead to improvement in organoleptic qualities and better acceptability by children.

The programme day care centres accepted only infants with MAM and SAM, but not all MAM and SAM were able to attend and to benefit from them and they constituted a dominant cost item. A day care centre is similar to a crèche: infants are looked after by trained staff with a remit to include cognitive stimulation. This type of day care centre incorporates aspects of supplementary feeding centres and a crèche structure that have been referred to in the literature as having positive impacts on infants’ nutritional status if well-operated (in Nepal [39]; in Brazil [40]). SNEHA had decided at the time of our assessment to phase out the day care component because of its cost.

Policy implications

Our research provides more evidence on the cost structure at scale of a community based treatment and prevention programme. The evidence suggests that such a programme can cost-effectively treat and prevent SAM and MAM in infants. This knowledge can assist policy-makers in prioritizing interventions within national budget constraints. Financial and human resources necessary to ensure scale-up of this programme in these contexts must be mobilized and appropriately allocated. Decision makers at global, national and local levels can use the evidence presented here to promote community based prevention and treatment of acute malnutrition as an integrated component of primary health care packages and nutrition programmes in the large number of contexts broadly similar to Mumbai slums.