An extensive WHO review was made in 1998 (henceforth called WHO Study) concerning 82 non-profit health insurance schemes for people outside formal sector employment in developing countries (Bennett et al. 1998). It was observed that very few of these schemes covered large populations or even covered high proportions of the eligible population. From a subset of 44 of the schemes, the median value of the percentage of the eligible population covered was 24.9%; 13 schemes had a coverage rate below 15%, and 12 schemes had a coverage rate above 50%.
Further information became available since 1998 (Table 2). Low percentages of enrolment were observed in a study on five CHIs in East and southern Africa (Musau 1999). In four schemes, enrolment percentages vary between 0.3% and 6.5% of the target population; one scheme is very small with 23 members of a target population of 27 cooperative society members. In Rwanda, a project was launched, establishing 54 CHIs in three districts in July 1999 (henceforth called Rwanda Project). By the end of the first year of operation, the enrolment rate reached in the three districts was 7.9% (88 303 members of a total target population of 1 115 509) (Schneider & Diop 2001). Another study was made in nine West and Central African countries (Atim 1998) (henceforth called WCA Study) on 22 CHIs. From the available information on beneficiaries and target population, one CHI in Benin reached an enrolment rate of 24% in 1998, whereas another achieved an enrolment rate of 8%. The target population in these CHIs was 13 000 and 7300 respectively. In two CHI schemes in Ghana and Mali, 53% and 25% of the target population of 25 000 and 200 000, respectively, was covered. And in Senegal, one CHI reached a coverage rate of 26% after 3 years of operation whereas another achieved an enrolment rate of 82%; the target population was 13 650 and 1200 respectively. A study of four of 16 CHIs in the area of Thiès (Jütting 2001) in Senegal (henceforth called the Thiès Study) indicated that in the year 2000, the average household enrolment percentage in these villages was 68%, with enrolment rates varying between a minimum of 37.4% and a maximum of 90.3%.
It is equally interesting is to study the enrolment over time. Sometimes, there is evidence about reductions in enrolment rates, which beg for better understanding. For instance in the Maliando Mutual Health Organization in Guinea-Conakry, subscription dropped from 8% to 6% of the target population mainly because of huge disappointment with the quality of care offered at health centre level (Criel & Waelkens 2003). However, membership rates might be low in the beginning, but might increase as the performance of the CHI convinces the population that subscribing may be profitable. One study on the Bwamanda Hospital Insurance Scheme in the D.R. Congo shows that in 1986 when the scheme was established, 32 600 people or 28% of the district population joined within 4 weeks. Over the years, membership climbed to 66% in 1993 and seems to have stabilized at 61% in 1997 (Criel 1998). Another study on the Lalitpur Scheme in Nepal shows that population coverage in the target areas rose from 19–20% in 1983 to 27–48% in 1995 (Harding 1996). Few studies, however, offer a long-term view of CHI.
A variety of factors influence people's decision to join the schemes given the voluntary character of CHI. Affordability of premiums or contributions is often mentioned as one of the main determinants of membership. A number of schemes in the WHO Study had addressed the issue of affordability. For instance in the Nkoranza Scheme in Ghana, the estimated cost of contributions varied from 5% to 10% of annual household budgets (Atim 1998). It was recognized that such contributions could be a financial obstacle to membership.
The technical arrangements made by the scheme management may influence people's perception of personal benefits. One example is the unit of enrolment. In the WHO Study, almost half of the schemes surveyed had the family as the unit of membership, a measure introduced to avoid the problem of adverse selection. In the Rwandan Project Study, large households with more than five members had a greater probability to enrol in the CHIs than others did (Schneider & Diop 2001). The explanation given is that contributions were kept flat, irrespective of household size up to seven members; the average contribution per household member was therefore less than for smaller families, inducing greater enrolment.
The timing of collecting the contributions may also matter for membership. From the WHO Study, it was observed that schemes in urban areas were more inclined to establish monthly or quarterly contributions so as to match the income patterns of urban informal sector workers. Annual contributions, collected at the time of harvest of cash crops, seem to be prevalent among schemes in rural areas (Bennett et al. 1998). However, in some schemes, such as the ORT Health Plus Scheme (OHPS) in the Philippines (Ron 1999), payment schedules were held flexible, with monthly, quarterly or semi-annual payments. Other schemes link the time of payment of the contribution with a suitable event in the community. For instance, burial societies in Uganda use their monthly meetings for the collection of premiums, either for the first-time members or for those who renew their membership (Carrin et al. 2001).
Trust in the integrity and competence of the managers of the CHI may also have an effect on enrolment. The existence of entry points in the community, such as a micro-credit scheme, a development co-operative or other social groups, may facilitate the establishment of CHI. If such existing initiatives have won the population's trust (van Ginneken 1999a), it may become easier to start up a CHI. For instance, the development co-operative in Bwamanda, initiated by the local Catholic mission, transformed into an integrated development project at the end of the 1960s (Centre de Développement Intégré, CDI). The CDI gradually improved agricultural activities in the area. This resulted in fairly stable economic conditions in the Bwamanda region throughout the 1970s and 1980s, which has enhanced the capacity and willingness of the population to enrol in the Bwamanda Scheme initiated by the CDI.
Trust can be enhanced when people see that their preferences matter. When the scheme administrators tend to be responsive to the community's preference, people's overall satisfaction with the community scheme's services is likely to increase. An important amount of evidence was recently reported by the ILO in a study about the role of CHI in the extension of social protection (Baeza et al. 2002) (henceforth called ILO Study). A total of 258 community-based health schemes were reviewed. Of 100 schemes with information, 57 schemes included participation of the community related to the benefit package. And in 51 schemes of 104 with information, the community was a partner in discussing the level of the premiums (Baeza et al. 2002). Trust was also considered as a factor in the development of health insurance among informal workers in Dar es Salaam, Tanzania. Informal sector workers constituted their own associations, which proved to constitute a good basis for building trust among members. Subsequently, health insurance was easier to develop (van Ginneken 1999b).
The quality of care offered through the CHI is another factor to be considered. The latter was highlighted in an evaluation of the Maliando scheme in Guinea-Conakry (Criel & Waelkens 2003). Focus group discussions were organized with 137 persons sampled from the member and non-member population. Participants referred to rapid recovery, good health personnel, good drugs and a nice welcome at the health facility as the most important features of quality. When membership was discussed specifically, lack of quality of care was cited as the most important cause of non-enrolment.
It is also important to see whether community health insurance is accessible across different population groups. One conclusion from the WHO Study was that very few schemes reached the vulnerable population groups, unless government or others facilitated their membership through subsidies (Bennett et al. 1998). In the Thiès Study, income appeared to be a significant factor in explaining enrolment. Belonging to lower and upper income terziles decreased and increased enrolment respectively. When households classified themselves into poor and non-poor, it also appeared that the self-reported poor had a lower probability to join CHI than higher income households (Jütting 2001; Jakab et al. 2001). One way to increase insurance membership for poor households is to introduce exemptions. Yet, only a minority (13) of the 44 schemes surveyed in the WHO Study had exemption policies to allow the poor households to join. In one of the three districts in the Rwandan Project, attention was paid to this particular issue: in Kabutare, the local church paid for the contributions of about 3000 orphans and widows with their family members. Contributions are also generally levied as flat sums, which is a disadvantage for the poorest: flat contributions are regressive, a flat rate contribution as a percentage of income being higher for poor than for the non-poor.
Related to policies to increase access of the poor to CHI, most schemes can be qualified as deficient. One scheme from the start introduced a pro-poor policy is the Gonosasthya Kendra (GK) Scheme in Bangladesh, differentiating contributions according to one of four socio-economic groups (the ‘destitute’, ‘poor’, ‘middle class’ and ‘rich’). For instance, contributions for the destitute were one-tenth of the contribution proposed to the highest income category. Contributions and other payments by households were minimized by using subsidies transferred to the scheme either from GK's own commercial ventures or from international sources. An important finding is that the membership rates among the two lowest socio-economic groups are substantially higher than in the other groups. However, after 15 years of operation of the GK scheme 20% of the ‘destitute’ group and more than half of the ‘poor’ group had still not been reached. The contribution levels and other payments are still said to be too excessive especially for the ‘poor’ as well as the lower middle income group of the ‘middle class’ (Desmet et al. 1999).
The household's geographical location is a second determinant of inequality in access. For instance, in the GK scheme, membership among the two lowest socio-economic groups appeared to be related to distance: up to 90% of that target population from nearby villages subscribed, whereas only 35% did so for the target population in the distant villages (Desmet et al. 1999). In the Rwandan Project Study, it was also found that households who lived <30 min from the participating health facility had a much larger probability to enrol in the CHIs than those who lived farther away (Schneider & Diop 2001).
A few case studies refer to measures introduced to reduce the impact of distance on enrolment and utilization. In the RAHA Scheme in India, a sliding scale of co-payments was established, decreasing according to distance (RAHA 1990; Bennett et al. 1998). A similar principle was established in the Bwamanda Scheme (Criel & Kegels 1997). However, although enrolment was seen to increase among the population living at the greatest distance from the affiliated hospital, utilization of the insured hospital care did not increase. For this reason the sliding scale was later abandoned.
Ratio of prepaid contributions to healthcare costs
From the WHO Study, information about the prepayment ratio, but through household contributions only, was available for 24 CHIs (Bennett et al. 1998). Thirteen CHIs had a ratio ≤60%. This means that assuming there would be no subsidies or grants from sources other than households (which is unlikely), the share of out-of-pocket payments (co-payments or user fees) in health expenditure would be 40% and higher.
In our framework, it is important to identify all stakeholders that contribute to the prepayment of health care, including central and local government, enterprises and donors. Communities indeed do not necessarily have to finance all healthcare costs and may draw on the financial inputs coming from the above-mentioned stakeholders. In the case of CHI, most schemes strive to cover only the portion of care that is currently recovered through user fees; hence the relevance of having information about the aggregate prepayment ratio. For six of 13 CHIs mentioned above, sufficient information was available to assess the out-of-pocket payments that are finally incurred by households. In four of those schemes, these out-of-pocket payments were in the 40–70% range, so that many households are likely to be subject to excessive out-of-pocket payments. A similar finding comes from the Mutec Health Centre in Mali (Atim 1998), where in 1996 the household prepayment ratio proved to be 15%, and user fees for medical visits and drugs accounted for 85% of health expenditure. It is likely that in such cases catastrophic payments will be incurred by certain families. It is obvious however that if the CHI enrolment rate in a given community is low, which is often the case, user fees will inevitably remain the main source of funding of health care in that same community.
Some schemes are performing particularly well in terms of the prepayment ratio. In the Bwamanda Scheme (Moens 1990), the total prepayment ratio amounted to 80.3%; the household prepayment ratio was 58%, and 22.3% came from subsidies and gifts. The co-payments of scheme members and user fees of non-members amounted to 8.7% and 11% of hospital expenditure respectively. There is also recent information (Ranson 2002) from the Self-Employed Women's Association (SEWA) health insurance scheme in Gujarat (India). SEWA is an organization of self-employed women and workers in the informal sector, a cooperative bank being one its major initiatives (Krause 2000). In 1992, SEWA started to offer health insurance. By 1999–2000, this scheme had 23 314 members. From data on hospital-related claims between mid-1994 and mid-2000, it was found that this scheme had an important impact on the occurrence of catastrophic spending. In this study, ‘catastrophic spending’ occurs when a patient consumes more than 10% of the person's annual household income on health care. It was found that without the insurance, hospital care would have been catastrophic for 35.6% of patients. However, as a result of the SEWA health insurance, the latter percentage was able to drop to 15.1%. We submit that this reduction was made possible by the relatively high prepayment ratio (76%) and by the inclusion of costly inpatient care in the benefit package.
Prepayment does not have to rely on households exclusively. As explained above, financial contributions can come from other sources as well, such as central or local governments, or local and international donors. In doing so, one may obtain a prepayment ratio that is high enough to ward off the negative impact of out-of-pocket payment. In the GK Scheme in Bangladesh, the real out-of-pocket payments were much lower than initially thought from simply inspecting the household prepayment ratio of 12% of recurrent expenditure. International subsidies and an internal subsidy from GK's commercial venture represented 50% and 14% of expenditure respectively. Finally, out-of-pocket expenditure by the GK members and non-members represented 8% and 16% of health expenditure in the GK scheme respectively (Desmet et al. 1999).
From the ILO Study it appears that most of the schemes (90 of the 136 for which information was available) do not bear the bulk of the financial risk (Baeza et al. 2002). Schemes may only cover a small part of the cost of the benefit package from members’ contributions. It is observed that in most of those cases (69 of 85 for which information was available), central and local government cover the larger part of the cost of health services. Again, central government together with others are the most important financiers in seven cases (Baeza et al. 2002). The latter results are not totally surprising, as central and/or local governments are the legal or de facto owners of CHIs in 61 (or 33%) of 184 cases that had sufficient information. These results, among others, lead the ILO study to conclude that most of the CHIs are in fact ‘entry points’ to larger pooling arrangements (Baeza et al. 2002). CHI may also be understood as an institutional mechanism for organizing risk pooling, thereby explicitly or implicitly using funds from both public and non-public sources. Thus, the role that CHIs could play in universal coverage strategies is clearly a subject for further policy research.
Practice of risk-pooling
From the ILO study, the evidence related to the size of the risk pool is that of 85 CHIs for which information was available, a majority (47) has <500 members. Only 14 schemes had more than 10 000 members (Baeza et al. 2002). Expansion of these groups may be problematic. Trust among the members is an important factor of the viability of such associations (Meessen et al. 2002). Trust is built on knowledge of each other, which initially at least tends to keep the groups small. An important element is also the availability of information among potential members of a CHI. The importance of information in shaping trust in the management of CHI schemes was clearly established in the Maliando Scheme in Guinea-Conakry (Criel & Waelkens 2003). Flows of information can in fact be considered as a form of social capital (Ray 1998). Adequate knowledge about how people behave vis-à-vis health insurance, in particular concerning moral hazard behaviour, should in principle help potential members decide to enrol or not. Geographical proximity enhances the information flows between people (De Weerdt 2002), and therefore is likely to help voluntary risk-sharing arrangements such as CHI. The latter may partly explain why pooling of risks across populations from geographically separated villages in a number of counties in the Rural Cooperative Medical Schemes (RCMS) project in China (Carrin et al. 1999) proved to be difficult to achieve in a short period of time.
The small scale of a CHI not only implies poor financial viability and danger of bankruptcy, it also has implications on the managerial capacity. Small schemes cannot set aside the financial resources needed to hire professional management. Managers are often voluntary members and may lack the skills as well as the time to improve the performance of the scheme. The need for administrative and management capacity was stressed in the WCA Study. The cases reviewed show a shortage of skills that are specific to CHI, such as the setting of contributions, collection of contributions and compliance, determination of the benefit package, marketing and communication, contracting with providers, management information systems, and accounting.
Several alternative strategies exist for greater risk pooling aiming at protecting schemes from bankruptcy and sustaining the financial protection of insured households. A first possible measure to protect CHIs against unexpected high-level expenditure is that of reinsurance: a scheme buys insurance with a re-insurer in order to avoid the risk of financial insolvency when expenditures are exceptionally high, for example due to an epidemic or a catastrophe involving a large number of members (Dror 2001). A more fundamental measure, however, is to expand the size of the risk pool. Larger risk pools can be achieved via assistance to management of small-scale schemes or via the establishment of a federation (Mills 1998) or network (Steinwachs 2002) of CHIs.
Given the limitations to the size of the association due to voluntary management, assisting voluntary managers to carry out certain administrative tasks may promote expansion of the schemes. In Burkina Faso, several support organizations are set up to assist starting mutual health organizations. Some make professional staff members available to help voluntary scheme managers in daily management. One example is the Mutual Health Organisation of Bobo-Dioulasso that gets regular assistance of an employee of the support organization ‘Projet Houet-Comoé-Kénédougou’ for certain management tasks. The Network for Support to Mutual Health Organisations (Le Réseau d'Appui aux Mutuelles de Santé), another support organization, gives permanent support to the elected scheme managers of several CHIs for book keeping and administration (Zett 2001).
Another approach is to subcontract the management of the scheme to an umbrella organization. This approach combines the advantages of the strong group identity of existing schemes, with the advantages of professional management. Ownership remains with the small-scale associations. An example of this approach is the Mutual Society for Health Care in the Informal Sector (UMASIDA). This is a CHI scheme owned and operated by a group of co-operatives of informal sector workers in Dar es Salaam, Tanzania. One person is employed for the administration and a second ensures medical control of the claims. But this arrangement does not achieve a larger risk pool, as every group keeps its own fund and there is no risk sharing among various groups (Steinwachs 2002). In Ivory Coast, the Federation of Medical Mutual Health Organisations (Fédération des Mutuelles Médicales de Côte d'Ivoire) has set up a similar construction. The federation is formed by eight mutual health organisations created by eight enterprises. It covers about 38 000 beneficiaries (workers and their dependants). The management is delegated to an insurance company, Managed Care International. In this case, too, each mutual health organization has negotiated its own conditions and funds remain separate (Lichtenberger 2003).
Depending on the socio-economic context, keeping funds of different population groups separate may well be an initial but necessary phase in the development of CHI. For example, the RCMS in China organized different funds for different groups within the same rural area. One of the reasons is the growth of industry in these rural areas in China, basically establishing two professional groups, that of farmers and of workers. From research in 42 townships (Carrin et al. 1999), it was found that at least eight townships established separate accounts for farmers and workers. The benefit package was also found to be different, with that of workers better than that of farmers. Benefits were adapted to the financial situation of the two funds. It was said that enterprises and their workers were reluctant to have funds pooled, workers fearing that in a fully pooled system, they would have to pay a multiple of the farmers’ contribution. The absence of willingness to pool funds was exacerbated after workers judged that farmers’ declared income was far below their real income and that, therefore, their capacity to pay contributions was underestimated.
One step closer towards forming one common pool is the initiative of the Union of Mutual Health Organisations (Union Technique de la Mutualité Malienne, UTM) in Mali. The UTM created a CHI to which all kind of groups can subscribe: mutual health organizations, small and middle enterprises, saving and credit organizations, etc. (Ouattara 2002). The UTM ensures the management of the individual schemes that cannot however negotiate their own conditions: the package of benefits is the same for all schemes.
Real merging of small-scale groups is achieved in hospital-based schemes in Uganda where pre-existing groups such as dairy co-operatives of burial societies, rather than individual households, constitute the basis for enrolment (McCord & Osinde 2003). All these schemes have in common the sharing of management tasks with clear delimitation of the roles of the manager(s) of the small groups and the professional insurer (Walford et al. 2000). The managers of the constituting units carry out the tasks of community sensitization, collection of contributions, and control of possible misuse. The insurer or the umbrella organization, however, takes care of the technical aspects: actuarial and financial management, selection of providers and negotiation of contracts, follow-up of quality of care, etc., for which voluntary managers have neither the time nor the competence. Enrolment based on existing groups favours an extension of the pool in urban areas where the constitution of solidarity groups large enough to form one single may be problematic, but where there is abundance of all type of small associations. The same holds for rural areas, where distance between villages may hamper the creation of one single large group. To be viable however, the introduction of professional management may well require external subsidies.
The set-up of the Rwandan schemes is yet another way to create a larger risk pool. The schemes function as individual CHIs at the health centre level, while a federation at district level covers the costs of hospital care.
A last strategy for large pools at the local level may be worthwhile considering, i.e. that of aiming for larger risk pools from the very start on (Davies & Carrin 2001). In that case, there is also greater likelihood of having cross-subsidies between rich and poor households. Instead of focusing on village populations, for example, the population of an entire district could be targeted. This has proved successful in the case of the Bwamanda Scheme in the D.R. Congo and in the Nkoranza Scheme in Ghana where the enrolment was high. It was however not so in the case of the CHF established at district level in Tanzania where the enrolment remained low.
As an alternative to merging, it could be explored whether CHIs could not be interconnected via risk-adjustment or equalization mechanisms (Cutler & Zeckhauser 2000). Basically, the latter would bring about financial support for those CHIs that face more than average risks; this support would be financed via transfers from those CHIs that face lower than average risks. Thus CHIs in relatively poor areas with high health risks would be able to set contributions at an affordable level, in view of subsidies received via equalization mechanisms. In this respect, we refer to van den Heever (1997) who studied employer-based health insurance schemes in South Africa. He notices that, since the late 1980s, new employer-based medical benefit schemes for low-income and largely Black workers have been established, and that these are largely separated from the funding of medical benefits for high-income workers. In other words, virtually no cross-subsidy seems to exist, so that the health insurance benefits of the former schemes are relatively limited. Policy proposals were therefore made to establish an equalization mechanism across medical schemes, probably through a para-statal organization, in order to offer a similar basic package of health insurance benefits.
Practice of strategic purchasing
From the ILO Study, of 62 schemes for which information was available, 10 were only found to have adopted some form of strategic purchasing (Baeza et al. 2002). Information on 67 mutual health organizations in the WCA Study showed that strategic purchasing was not imbedded yet in management practice (Atim 1998). For example, only four schemes had introduced essential and generic drug policies.
One of the conclusions in the WHO Study was that, overall, benefit packages were only weakly defined (Bennett et al. 1998). Although some schemes defined exclusions, there was a tendency to include all available services at facilities participating in the CHIs. With this broad approach, enrolment rates among patients with pre-existing conditions, especially chronic illnesses, tended to be high. After financial review, some schemes had to redefine the benefit package, even excluding certain population groups such as the elderly and/or excluding patients with pre-existing conditions (Bennett et al. 1998). Another way to contain costs as a result of introducing a broad benefit package was to introduce strict gate-keeping and referral practices. The latter was the case of the Bwamanda Health Insurance Scheme and the Chogoria Hospital Scheme, whereby patients could only get access to (insured) hospital care after being referred by a primary healthcare centre.
The WHO Study noted however that some schemes gradually took a greater role in purchasing. This was the case of the UMASIDA scheme in Tanzania. This scheme has contracted with providers who respect a number of conditions, such as access to services of a qualified medical officer, the availability of maternal and child health services, adequate laboratory services, provision of health education and occupational health, use of essential drugs list and prescription by generic name, and engaging in appropriate record-keeping. The SEWA scheme in India also engaged in more active purchasing, learning from claims processing which clinics could provide adequate care at reasonable prices, and then encouraging members to use these.
The OHPS in the Philippines designed a benefit package consisting of ambulatory and inpatient care, prescribed drugs and basic ancillary services (Ron 1999). Primary health care was directly provided by salaried doctors and nurses. Hospital-based diagnostic and therapeutic services were purchased from a private non-profit hospital through a capitation contract. Hospital-based care could only be accessed after referral from a primary healthcare doctor.
In the Rwandan Project (Schneider et al. 2000, 2001a,b) efforts were undertaken to strategically purchase health services. At the health centre level, services covered include preventive and basic curative care by nurses, essential drugs, hospitalization at the health centre, and ambulance transfer to the district hospital. At the district hospital, a number of services were covered, but only after referral from the health centres. In two districts, the hospital services covered were: consultation with a physician, overnight stay and Caesarean section. In the third district, malaria cases (>5 years), paediatric cases (<5 years) and Caesarean sections were covered.
The CHIs in Guinea-Conakry (Criel & Waelkens 2003) have also introduced active purchasing by way of official contracts between schemes on the one hand and providers on the other. For instance, via a contract valid for 1 year, the Maliando Scheme in Yendé provides access for its members to pre-defined health services from the Prefectoral Hospital of Gueckédou; the services included emergency obstetric and surgical care for adults and paediatric care. In the same way, a contract was established with the Health Centre of Yendé in order to purchase a package of curative and preventive primary healthcare services. Emergency transport of patients to hospitals is also arranged for via a contract with a local transport association.
The provider payment mechanism is an important element of strategic purchasing. In the WHO Study, 42 of 60 schemes for which information was available used salaries and budgets as payment method (Bennett et al. 1998). These payment mechanisms are expected to be beneficial for cost containment. But they may also lead to rationing, as a result of the enforcement of hard budgets.
Fee-for-service payment was found the second most prevalent way of paying providers in this review (11 of 60 schemes). In the WCA Study, fee-for-service was found to be most frequent payment method (Atim 1998). Fees may be used to induce the performance of providers, certainly in a situation of under-provision of health services. In one study in Pereang District in Cambodia, fees were part of an incentive system to increase the quantity and quality of publicly provided care. It even appeared that patients’ out-of-pocket expenditure decreased with respect to the time before the establishment of the incentive system; the latter was the result of official fees being competitive vis-à-vis unregulated private healthcare prices and being associated with good quality of care (Soeters & Griffiths 2003). A major disadvantage of the fee-for-service method, however, is that it may induce providers to over-prescribe treatment, certainly when part of fees collected are used as additional remuneration for providers. There is the additional risk that this payment method provokes a reduction in demand for health services, especially among the poorest.
Capitation payment, which has built in incentives for providers to keep costs down, so far was used in a few schemes only. For example, it was used in the ORT scheme to pre-pay a contracted private non-profit hospital for hospital-based services to ORT members. In the Rwanda Project, capitation was introduced as the payment method for health centre services in the 54 village schemes. It is also stated by that project that this should give health centre providers incentives to increase preventive care (Schneider et al. 2001a).
Yet another element in purchasing is setting referral rules across echelons of the health system so as to realize efficiency gains. From the WHO Study it appeared that many of the hospital-based CHIs ignored first-line health care, while first-line healthcare-based CHIs underestimated the costs of referrals for hospital care. In the WCA Study, only two of 15 schemes whose benefit packages contained both primary and hospital care had introduced mandatory referral for benefits beyond the primary care level (Atim 1998). The Bwamanda Scheme only covers insured patients for inpatient hospital care if they are referred by one of the health centres in the district. Also in the Rwandan Project, the district hospital services that are part of the benefit package are only covered after health centre referral (Schneider et al. 2001a).
Finally, the establishment of a waiting or qualifying period before one can make effective use of insurance, is a device to help contain the effects of adverse selection on the overall costs of a CHI. While it is certainly desirable when people have the possibility to enrol in a CHI throughout the year, some restraint on immediate use of health care may be introduced. In the WCA study, of the six CHIs for which information was available, five had established a waiting period of 2–3 months (Atim 1998). In the ORT scheme people can sign up at any time but the waiting period for inpatient care is 2 months (Ron 1999).
A remark is in order about administrative costs of CHI. These costs obviously matter as they have a direct influence on the financial resources that eventually are available to purchase health care. From a selected number of schemes in the WHO Study, the ratio of administrative costs to total scheme revenue varied from 5% to 17% (Bennett et al. 1998). These ratios varied between 11% and 44% in a selection of seven CHIs from the WCA Study (Atim 1998). In the Rwanda Project, administrative costs represent 7% of total annual expenditure (Schneider et al. 2001a). Compared with West European health insurance funds, where there are important economies of scale and where administrative costs are generally about 5% of fund revenue (Bennett et al. 1998), administrative costs in several of the documented CHIs certainly are on the high side. It can be expected that the relative importance of these costs will decrease when the size of the risk pool increases and/or when CHIs would enter into a federation.