Impact on child mortality of removing user fees: simulation modelBMJ 2005; 331 doi: https://doi.org/10.1136/bmj.331.7519.747 (Published 29 September 2005) Cite this as: BMJ 2005;331:747
[posted as supplied by the author]
Modelling changes in child mortality, following fee abolition
The number of under-five deaths that could be prevented from user fee abolition is estimated by combining evidence on the impact key health interventions have in reducing child mortality,2,3 with analysis of fee abolition’s potential to increase the proportion of the population benefiting from these interventions. This analysis can only give a first estimate of the likely impact of fee abolition, since country context will determine the exact nature of the changes observed. In particular, a population’s response to fee abolition may not be identical in different settings: experience of the health system; the implementation process (of fee removal); and the political, socio-economic and cultural environment also matter.
We first developed a classification system for key child survival interventions. This classifies 26 interventions according to whether their use is expected to increase following fee abolition; and, if so, by what magnitude. The intervention selection is comprehensive, incorporating all interventions for which there is strong evidence of proven efficacy.2
This grouping is based primarily on the expected magnitude of price reduction following fee abolition, but also incorporates health promotion effects, with interventions classified into six groups. See the figure in the article for details. Note that it is based on representative prices across the region, rather than using exact figures from any one country.
It is assumed that this intervention grouping applies to all 20 African countries analysed. This is a necessary assumption given the lack of appropriately detailed data on the fee structures of the countries analysed. However, in some countries (or regions within countries), a flat rate fee may be charged. We therefore also undertook sensitivity analysis to reflect this situation.
Importantly, the model assumes improvements in access following fee removal even if countries currently have waiver or exemption mechanisms for under-five children and pregnant women (as is the case, for example, in Tanzania). This is because there is widespread evidence that these have generally been ineffective.4 This is also consistent with the experience in Uganda, where before abolition, policies to protect under-fives and pregnant women existed, yet after abolition under-five’s use of services still increased significantly.5
In the second stage of model development, this intervention classification system is combined with evidence from Uganda, South Africa, Madagascar and Kenya on more generalised changes in utilisation of health services following user fee abolition, to produce estimates of expected utilisation rate changes for each of the 26 interventions. This was done by adjusting the generalised utilisation changes from these four countries downwards or upwards, according to each intervention’s classification.
Two basic scenarios were explored: the Ugandan experience (analysed separately because of its more detailed data), and other post-fee abolition studies. For both, the intervention-specific estimates of utilisation rate changes equal the median (across studies) observed change in utilisation for the intervention proxy (for example, ‘curative care’ as a proxy for the specific intervention antimalarials), multiplied by a factor of 0.5-1.5. The weighting factor used depended on the group classification of each intervention (group 3: 0.5; group 4a: 0.75; group 4b: 1; and group 4c: 1.5). This factor range approximates variation in the price reduction for the 26 interventions analysed, with those interventions experiencing the largest price decrease following fee abolition multiplied by a higher factor.
Sensitivity analysis – simulating a response to fee abolition where flat rate fees are in place (as discussed above) – was carried out by applying median observed utilisation changes without any factor adjustment. That is, the adjustment factor was simply set to one for all interventions. Importantly, sensitivity analysis estimates showed that the results were not overly sensitive to the precise weighting factors used (see the results section of the main text).
Expected higher increases in utilisation for the poor, as compared with general population estimates, were also calculated in the two basic scenarios, based on widespread evidence that the poor (and near poor) are the most responsive to price changes and have higher rates of illness, giving four scenarios in total. In particular, the poor are assumed to be at least 60% more responsive than the general population to price changes across all interventions, based on evidence from Cote d’Ivoire and Burkina Faso11,12 (a conservative estimate, compared with more anecdotal evidence elsewhere in Africa13).
The table in the article provides the observed utilisation data from these studies and the estimates of intervention-specific changes that were derived:
In a third stage, these estimated utilisation increases for different health interventions were converted into plausible reductions in under-five mortality. This was done by inputting estimates of expected proportional increases in the coverage of each intervention from 2003 levels into the updated Bellagio child survival impact model. This model estimates child mortality effects by bringing together estimates of under-five mortality levels by country and cause, national coverage data for all interventions of proven efficacy, and cause-specific efficacy estimates for each intervention.2,3
Potential reductions in child mortality can then be modelled by increasing coverage for each of these interventions. Taking antimalarials as an example, current national coverage levels for antimalarials are adjusted upwards, using estimates derived for antimalarial utilisation rate increases following fee abolition (e.g. a 27.3% increase for the Uganda general population scenario, and higher in the other three scenarios, as shown in the table). The Bellagio model combines this increased coverage following fee abolition, with efficacy estimates of antimalarials in reducing malaria deaths and the actual number of under-fives dying from malaria (by country) to estimate child mortality effects. Note that the estimated proportional increases were applied to current coverage levels, expressed on an odds scale, so that the new coverage levels could never exceed 100% and the coverage of interventions already at very high levels would improve less (in absolute terms) than that of interventions currently at modest levels of coverage.
Analysis was restricted to the 20 African countries with over 50,000 child deaths per annum and with user fees in place as of 2003. The model also assumes that increased contacts with health facilities occur when a child is sick and at risk of dying: a seemingly plausible assumption given that travel and other non-health care costs will remain even after fee abolition.
- This Week In The BMJ Published: 29 September 2005; BMJ 331 doi:10.1136/bmj.331.7519.0-c
- Education And Debate Published: 29 September 2005; BMJ 331 doi:10.1136/bmj.331.7519.762
- Letter Published: 13 October 2005; BMJ 331 doi:10.1136/bmj.331.7521.904-c
- News Roundup [abridged Versions Appear In The Paper Journal] Published: 22 June 2006; BMJ 332 doi:10.1136/bmj.332.7556.1470-a
- Maternal deaths from suicide must be tackled, say expertsBMJ December 07, 2016, 355 i6585; DOI: https://doi.org/10.1136/bmj.i6585
- Andrew Wakefield calls Trump “on our side” over vaccines after meetingBMJ December 05, 2016, 355 i6545; DOI: https://doi.org/10.1136/bmj.i6545
- The future of NHS dentistry in ScotlandBMJ December 05, 2016, 355 i6466; DOI: https://doi.org/10.1136/bmj.i6466
- South Africa begins first HIV vaccine trial in seven yearsBMJ December 01, 2016, 355 i6501; DOI: https://doi.org/10.1136/bmj.i6501
- Better evidence for smarter policy makingBMJ December 01, 2016, 355 i6399; DOI: https://doi.org/10.1136/bmj.i6399
- Proactive community case management and child survival in periurban Mali
- Trends and social differentials in child mortality in Rwanda 1990-2010: results from three demographic and health surveys
- Improving equity by removing healthcare fees for children in Burkina Faso
- WHO wants safer medicines for children
- Health in Africa: It's time for good guidelines on health financing practice
- Removing user fees for primary care in Africa: the need for careful action