Cover Image
close this bookDistrict-based Health Insurance in sub-Saharan Africa (Institut Tropical - Tropical Institute, Antwerp, Belgium, 1998)
View the document(introductory text...)
View the documentIntroduction
close this folder1. The Bwamanda scheme: Rationale and results
View the documentThe Bwamanda hospital insurance scheme1
View the documentResults of the insurance scheme
View the documentIs the Bwamanda insurance scheme a success?
View the documentThe conditions of success
close this folder2. The impact of the Bwamanda scheme on hospital utilisation patterns
View the documentIntroduction
View the documentMethods and sources of data
View the documentResults
View the documentDiscussion
close this folder3. The social perception of the Bwamanda scheme
View the documentIntroduction
View the documentMethods
View the documentResults
View the documentDiscussion
close this folder4. The Masisi health insurance scheme
View the documentIntroduction
View the documentMasisi health district
View the documentHospital admissions under the Masisi insurance scheme
View the documentAdverse selection and moral hazard
close this folder5. The Murunda hospital insurance scheme
View the documentHealth services in Murunda
View the documentFinancing of health care
View the documentThe Murunda insurance scheme
View the documentEffectiveness, efficiency and equity
View the documentReferences



The overall admission rate in Bwamanda hospital (39.2‰ district inhabitants) is in line with rates observed in other African rural hospitals (Petit and Van Ginneken 1995; Van Lerberghe et al 1992). On average the admission rate for the insured is three times higher than for the non-insured. The incremental utilisation of hospital inpatient services by the insured, however, is by no means a homogenous phenomenon: it is substantial for patients admitted to surgery and maternity, and rather marginal for patients in internal medicine.

The utilisation differentials between insured and non-insured admission rates for the different hospital departments analysed (surgery, internal medicine/gynaecology, paediatrics) are the most substantial for the insured communities located within a range of about 35 km from the hospital Overall, the reduction of financial barriers to hospital utilisation the insurance scheme does not seem to be a strong enough boost to overcome geographical barriers. However, the ratios of admission rates insured: non-insured indicate that the relative increase is only slightly influenced by distance. Hence, in relative terms, the higher hospital utilisation among insured is distributed in a grossly similar extent in both nearby and remote communities.

The analysis of distance decays in hospital admissions for the tracers of justified utilisation gives a different picture. For this type of problems the insurance scheme helped to overcome geographical barriers to hospital utilisation to the point of little residual influence of distance, and certainly far less than was observed in the spatial analysis of all admissions confounded.
This effect is particularly pronounced in the case of caesarean sections and strangulated hernias. These indicators of justified high priority hospital use are both life-threatening conditions in the event of no treatment. The same phenomenon is also observed, but to a far lesser extent, for uncomplicated hernias the tracer for justified but non-priority hospital use.

These findings suggest that the Bwamanda scheme succeeded in increasing hospital use among the insured for these sub-populations of patients considered in high need of hospital-based care, which provides an indication of the effectiveness of the scheme. Conversely, unmet need is likely to remain in the populations that did not subscribe to the scheme. These findings also indicate that the higher justified priority hospital use is not confined only to the communities living nearby to the hospital, which constitutes an indicator of the equity level of the scheme

The pattern of hospital utilisation for caesarean sections/symphysiotomies makes it possible to quantify the effect of the insurance scheme. It can be reasonably assumed that both populations of respectively insured and non-insured have a similar need for caesarean sections. If that is correct, the data show there remains substantial unmet need in the non-insured populations, and especially in the more remote non-insured communities. The marked difference observed between average caesarean section rates, in respectively insured and non-insured population, gains further significance in a context where indications for caesarean section are very stringent and where it is essentially performed on maternal indications. These findings strongly suggest that the observed deficit in caesarean sections has led to a number of obstetrical disasters in the non-insured population. Similar conclusions were found in studies measuring the coverage of obstetric interventions need in another district in Congo (Van Lerberghe et al 1988; Van Den Broek et al 1989) and in Morocco (De Brouwere et al 1996).

The deficit in caesarean sections among non-insured in the period 1991-1996 can be estimated by multiplying the number of expected pregnancies in the non-insured population (N=11,351) with the difference between the caesarean section rate among the insured (CRins= 1.97%) and the non-insured (CRnins = 0.74%). This gives a deficit- caesarean sections that should have been done but were not- of about 140 caesarean sections in the five-year period 1991-1996: an average of 28 per year.

It is interesting to compare the data on abdominal hernia surgery with data collected in non-government organisations (NGO) supported district hospitals in East Africa (Nordberg 1984). These institutions are indeed fairly comparable to the Bwamanda hospital. The Bwamanda data for uncomplicated and strangulated hernias in the total population (i.e. insured and non- insured) are roughly similar to the minimum needs estimated in East Africa, whereas the data for the insured population largely exceed the estimated needs (see Table 10). Indeed, in this population, the rate of surgery for strangulated hernias (justified priority hospital utilisation) is 2.1 times higher than the minimum need estimated by Nordberg, and the rate of surgery for uncomplicated hernias (justified non-priority hospital utilisation) is 1.8 times higher. These findings suggest either that the need for this type of surgery would be much higher in Bwamanda than in East Africa, or that the estimates for East Africa are too conservative.

Table 10 Comparison of data on uncomplicated and strangulated hernias

Uncomplicated hernias

Strangulated hernias


Insured population



Non-insured population



Total population



Range in 5 NGO supported hospitals in East Africa, according to Nordberg, 1984



Estimated minimum need according to Nordberg,



The analysis of spatial concentration of admissions through the coefficients of localisation and location quotients, is coherent with the findings above. The systematically lower CL for the insured populations suggest that the insurance has decreased localisation and led to a more homogenous distribution of hospital admissions across health centre areas. Northcott qualifies such an effect as "an equalising trend in distribution pattern" (Northcott 1980). The ratios of coefficients of localisation non-insured: insured indicate that this equalising trend is most pronounced for surgical-obstetrical problems and lowest for internal medicine, gynaecology and paediatrics. The pat- tern of spread in LQi, which is systematically more narrow for the insured patient populations, especially where caesarean sections are concerned, indicates a more equal distribution of admissions over the 17 health centre areas. If equal hospital utilisation for equal need is considered as a proxy for equity, then it is possible to conclude that the insurance scheme has improved equity in the district.

The health problems for which the equalising trend is most pronounced are those that are tracers of priority justified care. Why is this so? Part of the explanation is that the hospital is the only institution in the district that can offer a reasonably effective technical solution (i.e. surgery) for these, often urgent, problems, and people are aware of that. Moreover, abdominal hernia repair and interventions related to motherhood generally score quite high on the Bwamanda community's own priority list. Yet another explanation may be the fact that for some health problems it is easier to anticipate future utilisation of health services. This is the case, for instance, for deliveries and for many of the surgical problems. The insurance scheme may thus have preferentially attracted people whose present or future health status is such that health service utilisation in the nearby future becomes reasonably 'foreseeable' or 'predictable'. Such a phenomenon of adverse selection of women in childbearing age has been clearly documented in the case of the Masisi hospital insurance scheme in eastern Congo (former Zaire) which will be discussed later in greater detail. Adverse selection may thus, at least theoretically, have contributed to the high impact of insurance on hospital utilisation for obstetrical and surgical problems. In Bwamanda however, member- ship can only take place on a household-, and not on an individual basis. This has probably reduced the occurrence of adverse selection and kept it within reasonable limits.

The equalising trend decreases when it comes to health problems that are less predicable, less urgent, less life-threatening, more ill-defined, and for which the hospital's answer often is of relatively limited effectiveness12.

12 At least compared to what can be done at health centre level or to what is available in the community itself.

The length of stay is only slightly (and not significantly) longer for the insured inpatient population. This finding may indicate that the insurance scheme has not been a cause of inefficiency in hospital use. However, the data do not allow to discard the possibility that non-insured leave the hospital at too early a stage. It would be wise to assess average lengths over much longer periods, and for each ward separately, before jumping to hasty conclusions. In addition, it would be appropriate to systematically appreciate the outcome of hospital stays for the same patient population in order to know whether non-insured leave the hospital prematurely or not.

The bed census13 was yet another instrument to measure whether the insurance scheme affected efficiency of hospital use. The data from the Bwamanda bed-census unfortunately do not allow to establish whether there is a significant difference in inappropriate bed occupation between insured and non-insured self-employed inpatients. The results of the subjective hospital justification assessments, on the other hand, indicate that virtually all admissions were considered appropriate by local staff criteria both in insured as in non-insured inpatients.

13 An interesting finding of the bed-census data was the indication of a substantial proportion of inappropriate bed occupation in the overall inpatient population: 43% of the inpatients investigated did not match any of the positive criteria justifying bed occupation on the day of the census This proportion is much higher than the ones observed in similar studies carried out in Zambia (Buvé and Foster, 1995) and South Africa (Henley et al 1991, Zwarenstein et al 1990) The proportion of inappropriate bed occupation was 13% in the Monze district hospital (Zambia), and 20% and 29% respectively in two different teaching hospitals of the Republic of South Africa. This result is surprising given the strong tradition of rational and efficient use of resources in Bwamanda which has been reported to be amongst the best performing districts in Zaire (USAID/Kinshasa, 1987)