Cost effectiveness of CMAM in Malawi

Government of Malawi guidelines

Summary of published research¹

A recent study assessed the cost-effectiveness of community-based management of acute malnutrition (CMAM) to prevent deaths due to severe acute malnutrition among children under-five in Dowa District in Malawi.

Nutrition surveys conducted since 2002 indicate that the prevalence of severe acute malnutrition (SAM) in Dowa is generally at or close to 1% (range 0.4 to 1.1% based on seven district-wide surveys). Dowa district in central Malawi was the site of the first large-scale CMAM pilot, initiated by Valid International and Concern Worldwide in partnership with the Dowa District. Since that time, the Dowa programme has operated increasingly under the management of the District Health Office, who have reportedly budgeted all CMAM costs from their own budget for the 2010 fiscal year. Outpatient Therapeutic Programme (OTP) services for SAM are available at all 21 health centres in the district and linked via referral to inpatient therapeutic programme (ITP) services at the district's three hospitals. Malawi's CMAM protocols are in line with international standards. In 2007, 2896 children with SAM were admitted to the Dowa OTP. Over half (59%) of admissions were oedema cases, 20% were wasted according to mid upper arm circumference (MUAC) or weight-for-height criteria and the remainder were transfers/ returning defaulters/other. CMAM is now included in Malawi's national health protocol and implementation has expanded from just two of Malawi's 28 districts in 2004, to 21 in 2008.

The study describes the incremental cost effectiveness of the CMAM programme in Dowa district of Malawi for 2007. It places the findings within the broader context of other child health economic evaluations that have been undertaken in Malawi or neighbouring countries.

Decision tree

The study authors used a decision tree to model the cost-effectiveness of CMAM integrated into existing health services versus existing health services with no CMAM, for the period January to December 2007 in Dowa district (see Figure 1). At year end, 2780 children had been discharged from the OTP (including 85 transfers to ITP). The decision tree is built around this known exit total and outlines the possible treatment pathways for SAM beginning with the primary decision to implement CMAM (scenario 1) or not implement CMAM (scenario 2), and branching at additional decision nodes until a terminal node of either dead or alive is reached. The proportion of SAM cases assigned to each branch was based on data collected from the Dowa CMAM programme from January to December 2007 and key assumptions regarding mortality outcomes and uptake of health services among non-enrolled children. The respective costs and effects (in terms of DALYs1) were calculated and aggregated for each treatment pathway. The difference in costs and effects between the two scenarios was used to estimate the incremental cost per DALY averted. DALYs are the sum of the present value of future years of life lost due to premature death (YLL) plus the present value of future life years lived with a disease or injury (YLD).

Reflecting the CMAM programme structure, children in the model exit the OTP in one of four ways: cured, died, defaulted/non-recovered, or referred to the ITP due to complications. A child defaults after missing two consecutive fortnightly visits. A child is discharged as non-recovered if he has not reached discharge criteria after his fourth month in the programme.

The defaulter and non-recovered categories have been combined for simplicity, assuming a similar mortality risk. Once children referred from the OTP to ITP are treated for complications, they either (a) return to the OTP to exit via one of the OTP exit categories above, or (b) die, either while under ITP care or after defaulting from ITP, or (c) live after defaulting from ITP.

The model also accounted for additional mortality risk faced post-discharge during the study year by all children exiting cured, defaulted and non-recovered. SAM cases not covered by (i.e. not enrolled in) the CMAM programme in scenario 1 or who live in areas where no CMAM is implemented (scenario 2) are assumed to have two treatment options: non-CMAM care or no treatment. Non-CMAM care includes any treatment for underlying illnesses received at a health centre (excluding therapeutic feeding) and any traditional inpatient therapeutic services received at one of the three hospitals (including therapeutic feeding).

Under the CMAM-implemented scenario, the probability of being enrolled in CMAM was based on the results of a CMAM coverage survey conducted in Dowa in early 2008. Mortality outcomes for children during their average treatment period of six weeks were known from OTP and ITP data, with the exception of defaulters/non-recovered. Additional mortality faced post-discharge was captured by adding 2.4% to the known OTP mortality rate of 1.0%. This 'background' mortality rate was derived from the overall under-five mortality ratio for Malawi in 2006 (120/1000 live births) divided by five to represent the study year in which 1/5th of the denominator's live underfive births would have occurred. The mortality rate for OTP and ITP defaulters/non-recovered was conservatively assumed to be the same as those receiving no treatment. The mortality rate among children referred to the ITP within the CMAM-implemented scenario (scenario 1) was derived from the only available dataset-a database combining exit data for all ITP children, including but not differentiating those referred in from the OTP. To estimate specifically the outcomes of OTP referrals, the relative proportion for each exit category from the combined ITP database was applied to the total known referrals from OTP to ITP (n=85). The resulting numbers were used to represent the ITP outcomes in the model.

The mortality rate for non-CMAM care (in scenario 1 and 2) was assumed to be the same as that for the CMAM programme's ITP component (11% as described above) plus the 2.4% annual background mortality rate. The mortality rate for children seeking no treatment (in scenario 1 and 2) was assumed to be 18.1%. This is based on a prospective cohort study conducted in northern Malawi which found the mortality rate among severely wasted children under-five (MUAC <110mm) to be 181.8 per 1000. The aggregate annual mortality rate for SAM cases covered by CMAM was 4.4% (123/2780), for the CMAM-implemented scenario as a whole it was 11.9% (809/6796) and for the CMAM not implemented scenario, 17.1% (1160/6796).

Estimating relative cost

To estimate the relative cost of each treatment pathway in the model, two unit costs were calculated: the average cost per child treated in CMAM and the average cost per child treated in non-CMAM care.

The no treatment arm was assumed to have zero cost. CMAM costs included those incurred by Concern for the 2007 calendar year and those budgeted by the government for the 2007/2008 fiscal year. All were converted into 2007 US$ at 140 Malawi Kwacha per $ and 0.73 Euro per $. Concern covered 90% of the total cost of the programme. Ready to use therapeutic food (RUTF) accounted for the largest portion of total costs, followed by Concern administrative and direct staff costs.

Concern's CMAM expenditure was tracked through its financial accounting system and split between capital (cars, motorbikes and computers) and recurrent costs (all remaining costs). Purchases prior to 2007, all made in US$, were inflated to 2007 costs by the US Consumer Price Index. Car and motorbike costs were annualised over five years and computers over three years, using a 3% discount factor. The cost of RUTF was the total reported spent by Concern for Dowa for 2007, covering the cost of purchase, transport from the producer's factory in central Malawi and warehousing.

Concern administration costs were comprised largely of an allocation from Concern Malawi's support staff and office costs, with the addition of support transport costs incurred at the Dowa and Lilongwe offices. Recurrent transport costs included the running costs of three shared cars allocated based on logs recording total kilometres driven for CMAM purposes.

Government costs for CMAM included those for OTP and ITP. Total government OTP costs were estimated using allocations from the budget for 1 July 2007 to 30 June 2008.

Allocations were made from annual staff salaries at 21 health centres (10% of one nurse, one medical assistant and four health surveillance assistants), district staff salaries (20%, 10% and 5% of one district nutritionist, two maternal and child health coordinators, and one health management information system officer, respectively), the district health budget's operating costs (1%) and the district drug budget (1%). The CMAM portion of the district's operating costs and drug budget were both allocated by multiplying the estimated proportion of health centre staff involved in OTP (15%) by the average proportion of total health centre days spent working in the OTP (10%) by the proportion of the total district health budget spent on health centres (66%).

Actual ITP costs (all covered by the District Health Office) were difficult to determine from district records. Further, it was not possible to distinguish ITP costs for children referred in from OTP (CMAM costs) vs self-referrals (non-CMAM costs). For this reason, a unit cost per child treated in the ITP was calculated using an assumed average stay of 7 days and an estimated cost per bed per day in a tertiary hospital in Malawi.

The resulting total CMAM cost was then divided by the total programme exits (2780) to arrive at the CMAM unit cost. The average cost per child treated with non-CMAM care was based on the assumption that one in four children seeking non-CMAM care accessed ITP treatment (using the same ITP unit cost, above), while the remaining three accessed the equivalent of three clinic visits and three courses of drugs.

Cost-effectiveness of CMAM

The total cost of providing CMAM (US$470,703) and non-CMAM (US$23,394) treatment for SAM in Dowa district in 2007 (scenario 1) was US$494,097. The total estimated cost for treatment services where no CMAM programme was implemented (scenario 2) was considerably lower at US$39,714. However, according to the model, there were 342 fewer deaths in the CMAMimplemented scenario than in the CMAM not implemented scenario, which at a 3% discount rate, equates to 10,883 DALYs averted.

Child in Dowa OTP, May 2007, Concern Worldwide

The incremental costs and effects between the two options were combined to estimate an incremental cost-effectiveness ratio (ICER).The ICER of implementing CMAM in addition to existing health services was US$42 per DALY averted (or US$1,365 per life saved). Using the worst case scenario for all data gives an ICER of US$493 per DALY averted. In the best case it is US$11 per DALY.

This study shows that the implementation of CMAM as an addition to the existing standard health services in Dowa district in rural Malawi in 2007 was a cost-effective decision. The ICER of US$42 per DALY averted is very close to the findings of an analysis using similar methods for an urban CMAM programme in Lusaka, Zambia, which estimated a cost of US$41 per DALY. The current study complements the Lusaka findings by demonstrating that CMAM is also cost-effective in a rural setting, where population density is lower and transport costs presumably higher.

The resulting CMAM ICER is within the general range of cost-effectiveness ratios estimated for other priority child health care interventions in Africa, including community or facility-based case management of lower acute respiratory infections (US$398), integrated management of childhood illness (US$38), universal salt iodisation (US$34-36), iron fortification (US$66-70) and insecticide-treated bed nets for malaria prevention (US$11 for sub- Saharan Africa).

The results of this study are expected to be relevant and generalisable to CMAM programmes in broadly similar contexts in sub- Saharan Africa, treating a comparable caseload and distributed across a similar network of health facilities. A number of additional contextual factors, however, will affect the degree to which the results can be generalised. This study model was not designed to assess variations in all of these, including the effect of changes in caseload or its three main determinants: SAM prevalence, population density and programme coverage. Any extrapolation of this study's results must therefore consider the following. First, higher levels of SAM than the 1% recorded for Dowa in 2007 are common in emergency contexts. Second, Malawi's population density of 158 people/km2 is higher than the majority of Sub-Saharan African countries. Third, while Dowa's CMAM programme coverage was generally within the range seen in other CMAM programmes, variations are possible.

It is generally assumed that the ICER for CMAM could be reduced as a result of economies of scale achieved through higher coverage and/or more cases treated, as the unit cost for fixed costs, such as administration expenditure per child treated, would decrease. Economies of scale are less likely to be achieved through reductions in RUTF costs. Further, RUTF costs will rise in almost direct proportion to the number of children treated, mediated only slightly by their length of stay. Transport and staff costs will also increase proportionally with the number of admissions, although to a lesser extent. It is also expected that some cost efficiencies will be gained as CMAM is further integrated into existing Ministry of Health systems resulting in parallel Concern costs, particularly administration and staff, being reduced.

The authors believe and conclude that the study's findings are indicative and relevant to a large number of settings where SAM is found. As such, they make a considerable contribution to the evidence available for effective health and nutrition programming decisions. Decision makers at global, national and local level are therefore urged to include CMAM as an integrated component of primary health care packages and nutrition programmes in the large number of contexts broadly similar to Dowa or Lusaka.

¹Wilford. R (2011) et al, Cost effectiveness of communitybased management of acute malnutrition in Malawi. Health Policy and Planning, p 1-11. doi:10.1093/heapol/czr017

²Disabilty-Adjusted Life Years.