The impact of the introduction of user fees at a district hospital in Cambodia

Transcription

1 doi: /heapol/czh036 Health Policy and Planning 19(5), HEALTH POLICY AND PLANNING; 19(5): Oxford University Press, 2004; all rights reserved. The impact of the introduction of user fees at a district hospital in Cambodia BART JACOBS 1 AND NEIL PRICE 2 1 Enfants&Développement, Phnom Penh, Cambodia and 2 Centre for Development Studies, University of Wales, Swansea, UK Proponents of user fees in the health sector in poor countries cite a number of often interrelated rationales, relating inter alia to cost recovery, improved equity and greater efficiency. Opponents argue that dramatic and sustained decreases in service utilization follow the introduction of user fees, highlighting evidence that user fees reduce service utilization when they fail to result in improved quality of care and/or when services are priced higher than those charged by private health care providers. Utilization of public health services in Cambodia is low. Supply-side factors are significant determinants of such low public sector utilization, including low official salaries of service providers (forcing many to seek additional income in the private sector), and operations budgets which are erratic and often insufficient to cover running costs of service delivery outlets. The Cambodia Ministry of Health (MOH) encourages user fee schemes at operational district level. By allowing revenue to be retained at the health facility level, the MOH aims to improve health care delivery and consequently service utilization through increased salaries to health facility staff and increases in operations budgets. This case study of the introduction of user fees at a district referral hospital in Kirivong Operational District in Cambodia, using the findings from empirical research, examines the impact of user fees on health-careseeking behaviour, ability to pay and consultation prices at private practitioners. The research showed that consultation fees charged by private providers increased in tandem with price increases introduced at the referral hospital. It further demonstrates for the first time that we are aware of from the available literature that the introduction and subsequent increase in user fees created a medical poverty trap, which has significant health and livelihood impact (including untreated morbidity and long-term impoverishment). Addressing the medical poverty trap will require two interventions to be implemented immediately: regulation of the private sector, and reimbursing health facilities for services provided to patients who are exempted from paying user fees because of poverty. A third, longer-term initiative is also suggested: the establishment of a social health insurance mechanism. Key words: user fees, poverty, private sector, Cambodia, health financing Introduction The international development discourse presents numerous, often interrelated, rationales for the introduction of user fees in the health sector, relating inter alia to cost recovery, improved equity and greater efficiency (Price 2002). A major argument for the generation of revenues through cost-recovery strategies based on user fees relates to covering operational costs: Facilities that retain revenues generally performed substantially better than facilities that sent all their revenues to the treasury (Shaw and Griffin 1995, p. 26). With regards to equity, it is claimed that universal free health care reinforces inequitable distribution of resources in that it provides better access to services in wealthy urban areas at the expense of poor and rural populations (Birdsall 1989; Foreit and Levine 1993). Proponents of this view claim that user fees avoid the provision of subsidies to those who can afford to pay all or some of the costs, and in doing so, free-up funds to pay all or part of the costs of services for those less able to pay. The dominant theme in the efficiency rationale for user fees is that charging attaches value to a service, i.e. it increases demand by increasing perception of quality and deterring unnecessary use of health care systems. Some advocates of user fees claim that free services reduce utilization because of inefficiencies leading to quality and time costs borne by the users, and because of the low value ascribed to free services (Lewis 1986). Foreit and Levine (1993) argue that selective user fees encourage clients to use appropriate service delivery outlets, by charging higher prices at tertiary level (hospital outpatient clinics) than at primary level (health posts). Correct pricing schemes, according to Shaw and Griffin (1995), thus signal to consumers to use health resources effectively and efficiently, and serve as a warning that those who choose to bypass the referral system will be required to pay the full cost of the service. Opponents of user fees refer to studies indicating dramatic and sustained decreases in service utilization following their introduction (Waddington and Enyimayew 1989, 1990; Mbugua et al. 1995; Mwabu et al. 1995). Creese (1991) provides extensive evidence that user fees divert those who cannot pay to other sources of health care or away from the health care system. Other studies highlight how user fees

2 The impact of user fees 311 reduce service utilization when they fail to result in substantial and sustained improvement in the quality of care and/or when services are priced higher than those charged by private health care providers (Audibert and Mathonnat 2000; Chawla and Ellis 2000; Ha et al. 2002). Despite very poor health indicators (maternal and infant mortality rates of 437 per live births and 95 per 1000 live births respectively [Ministry of Health Cambodia 2002]), public health service utilization is low in Cambodia with 0.3 annual contacts per capita (Ministry of Health Cambodia 1999). Supply-side factors are significant determinants of such low public sector utilization, including the low official salaries of service providers (often in the range of US$15 30 per month) that force many public health professionals to seek additional income in the private sector, and operations budgets which are erratic and often insufficient to cover the running expenses of service delivery outlets. The Cambodia Ministry of Health (MOH) encourages the development of user fee schemes at operational district level, with all revenue but 1% being retained at the health facility. The overarching rationale of the MOH is that user fees will improve health care delivery and consequently service utilization by providing funds for increased salaries for health facility staff and increases in operations budgets. A prerequisite to the introduction of user fees is community participation, which has to be ensured by establishing Health Centre Co-Management and Feedback Committees comprised of elected community members (Ministry of Health Cambodia 1997a,b). Prior to initiation of a user fee scheme, approval has to be obtained from the MOH (Ministry of Health Cambodia 1997a,b), although the scheme can be piloted prior to applying for official recognition (Ministry of Health Cambodia/UNICEF 2000). Health financing schemes must be transparent to the local community, can only be developed in consultation with community representatives, and a system for identification of the poorest for exclusion from payment has to be established (Ministry of Health Cambodia 1997b). This case study of the introduction (and subsequent increase) of user fees at a district referral hospital in Kirivong in Cambodia draws on empirical research to examine the impact of user fees on health-care-seeking behaviour, ability to pay and consultation prices at private practitioners. The case study demonstrates that the introduction and subsequent increase in user fees created a medical poverty trap, which has the potential to lead to untreated morbidity, reduced access to care, long-term impoverishment and irrational drug use. Background Kirivong Operational District (KOD) is located in Takeo Province, in the southeast of Cambodia, bordering Vietnam. It consists of four administrative districts with 31 communes and 290 villages, with a total population of (1998 census), whose main economic activity is subsistence farming supplemented by fishing and gathering. Two of the administrative districts are located in the Tonle Bassac Delta; because of widespread irrigation through canals and flooding during the rainy season, these two districts have different farming patterns to the two administrative districts on higher ground, which have very limited irrigation. KOD has 20 health centres and an 80-bed referral hospital, and is run through a contracting-in agreement whereby a non-governmental organization (NGO) is responsible for its management and administration. KOD has a flourishing private health sector. A census conducted in October 2002 (Jacobs, unpublished data) indicated that there were 511 private practitioners operating in KOD, of whom 75 were qualified and operating from their homes where they also ran a small pharmacy (henceforth referred to as qualified private practitioners ). At the time of our study these qualified practitioners were not required to register with the MOH. Most outlets for prescription drugs, however, are small shops run by unqualified personnel who sell other retail products such as rice and sugar. There were 336 such vendors identified. An additional 50 persons were found to sell prescription drugs at local markets, while an equal number reportedly went from house to house to sell and administer injections. Lay persons selling prescription drugs are referred to as drug sellers in the text. All curative and preventive health care in the public health sector was, until November 2001, officially provided free at the point of delivery. Health staff were paid (by the contracted-in NGO) a monthly salary supplement of US$ In November 2001, it was decided to pilot user fees at the referral hospital. Prices for services were based on the findings from a cross-sectional survey of women with children undertaken in July 2001, regarding willingness and ability to pay for curative services (Jacobs, unpublished data). 1 Prices were initially kept low, as hospital management recognized that quality of care was poor (defined by the inability to provide services such as obstetrics and surgery). The user fee scheme was piloted for 5 months, following which a final scheme based on higher prices was elaborated and implemented, at the same time that obstetric and surgery services were introduced. To formulate the final scheme, 151 patients or their carers were interviewed during the piloting phase (see research methods below): 115 patients aged more than 5 years paid fees while 36 children of 5 years or less were admitted free of charge. No differences in health-seeking behaviour or ability to pay were observed between the two groups, nor was there a statistical correlation between the total expenditure on health care and ability to pay. Mutual assistance relations were found to be common whereby only 10% of borrowers obtained money from private lenders and 31% paid interest. These observations and the commencement of obstetric and surgery services were considered appropriate to justify an increase in the user fee prices when the final scheme was introduced. An overview of the user fee price schemes during the pilot and implementation phases is provided in Annex 1. Methods Randomly selected patients admitted at KOD Referral Hospital were interviewed using a piloted pre-coded

3 312 Bart Jacobs and Neil Price questionnaire administered before the introduction of the pilot user fee scheme, during the pilot phase, and after the implementation of the final user fee scheme. For patients aged less than 18 years, the accompanying carer was interviewed. Questions concerned personal details, place of residence and distance from the hospital, care-seeking behaviour and its associated costs, income forgone by patients and/or carers during hospitalization, and ability to pay for both hospitalization and any consultation costs in the private sector. The latter was defined as having sufficient cash to pay for all direct costs incurred with the illness episode without having to borrow or sell assets. Data were analyzed using the statistical package Epi-Info 6.04b and differences in proportions were compared using the χ 2 test. Significance was determined at 5%. For skewed data a non-parametric test was used. Data were stratified according to period of interview: before introduction of user fees (baseline), during the piloting period (pilot) and following the increase of user fees (implementation). To calculate total costs of hospitalization, direct costs (transport, consultation charges at private providers, hospitalization costs) were used. 2 Indirect costs were estimated by questioning interviewees regarding daily amount of revenue lost by patients and carers during the hospitalization period. This question was only asked during piloting and implementation, and the respective findings were used to assess the influence of seasonality on ability to pay all costs related to hospitalization without having to borrow or sell assets. Data regarding the monthly number of patients admitted at the hospital were derived from the facility s Health Information System (HIS) reports. We further assessed whether the introduction of user fees had an effect on the case mix of patients being hospitalized. For this purpose we selected the five most common conditions reported in the HIS during the three phases. These data were also used to calculate hospital mortality rates, by dividing the total number of patients admitted by the respective number of deaths. Tuberculosis patients and deaths due to tuberculosis were not included. 3 Table 1. Patient profile and health seeking behaviour Results Patients profile The total number of patients/carers interviewed was 101 at baseline, 151 during the pilot phase and 152 during implementation. The respective median age of interviewed patients (or their carers) was 24 years (range 3 months 47 years), 21 years (range 2 months 80 years) and 21 years (range 4 months 78 years). The literacy rate and proportion of patients/carers who were farmers remained statistically similar across the three surveys (Table 1). The socioeconomic status of respondents (using landownership 4 and possession of a motorbike as proxies), however, changed considerably, whereby the proportion of landless interviewees decreased from 16% during baseline to 7% during piloting and 5% during implementation. The proportion of interviewees who possessed a motorbike increased by 50%: from 17% during baseline to 25% in the two study periods thereafter. At baseline (i.e. when services were provided free of charge) interviewees resided on average 6km from the hospital (range 0 30), statistically similar to the 8km (range 0 45) observed during the user fee pilot phase. During final scheme implementation, however, interviewees resided significantly further from the hospital: 12.5km (range 0 55) (Kruskal- Wallis, difference between piloting and implementation, p < 0.001). During implementation, 76% of patients resided within the administrative district where the hospital is located, compared with 93% during piloting and 85% at baseline. Care-seeking and prices of private practitioners When there were no official charges at the hospital, only 20% of patients consulted private providers before presenting at the hospital. This proportion increased to 54% during piloting, and 73% during implementation (Table 1). The proportion of interviewees who initially consulted drug Variable Baseline Pilot Implementation n=101 n = 151 n = 152 p-value n (%) n (%) n (%) df = 2 Literate 42 (42) 69 (46) 61 (40) 0.4 Farmer 79 (78) 103 (68) 119 (78) 0.1 Owns land 85 (84) 140 (93) 144 (95) 0.01 Owns motorbike 17 (17) 38 (25) 38 (25) 0.04 Resides within administrative district of hospital 86 (85) 140 (93) 116 (76) <0.001 Initial consultation at private sector a : 20 (20) 82 (54) 111 (73) <0.001 Traditional healer 4 (4) 2 (1) 0 Drug seller 6 (6) 48 (32) 38 (25) <0.001 Private qualified practitioner 10 (10) 29 (19) 73 (48) <0.001 Consulted provider in Vietnam 0 3 (2) 0 a During baseline, four patients consulted additional private providers before going to hospital: one a drug seller and three private qualified practitioners. During piloting, three consulted additionally a traditional healer, one a drug seller and four private qualified practitioners. At implementation, all went to the hospital following failed consultation at the private sector.

4 The impact of user fees 313 sellers increased from 6% at baseline to 32% during piloting, and reduced slightly to 25% during implementation. The respective figures for consultations at private qualified practitioners were 10%, 19% and 48%. These figures indicate that the majority of patients who initially consulted private practitioners opted mainly for drug sellers during piloting but switched to private qualified providers during implementation. Table 2 provides an overview of the reasons mentioned for initially consulting private practitioners by all interviewees, and for those who consulted drug sellers and private qualified practitioners. Geographical proximity of the private provider was the reason mentioned most frequently for initially consulting such a practitioner during the three study periods, but was most pronounced during piloting compared with the other periods: 80% versus 45%, respectively. Proximity to provider was most mentioned for consulting drug sellers during piloting (90%) and least mentioned for consulting private qualified practitioners during implementation (32%). One in three interviewees (35%) initially going private during baseline did so because of trust in the consulted provider. This was mentioned by only 6% during piloting and not at all during implementation. During baseline and piloting, about 5% of persons who initially consulted the private sector did so because of advice from peers. This rose to 19% during implementation. The proportion of interviewees seeking care initially in the private sector because of seriousness of the condition was 10 15% during the first two study periods. Following implementation of the final user fee scheme, 37% reported going first to private providers because of the seriousness of the condition. This was mentioned considerably more by interviewees consulting qualified practitioners (45%) than by interviewees going to drug sellers (21%). Distance from the hospital was inversely correlated with the initial contact being with this facility. During piloting 50% (29/58) of interviewees residing within 5km came straight to the hospital, compared with 30% (28/93) of interviewees living further away (p < 0.001). The respective figures during implementation were 32% and 11% (p < 0.001). Also during this period, interviewees reportedly knowing the hospital user fee prices beforehand were significantly more likely to first contact the public sector than those who did not know the prices: 34% versus 13%, respectively (p = 0.006). The former proportion, however, was significantly lower than the 60% (39/65) observed during piloting (p = 0.001). The means of reported costs for initial consultation with private sector providers (all categories) were R (US$5.8) at baseline, R (US$11.3) during piloting, and R (US$13.8) during implementation. These prices, however, are not adjusted for disease/condition or type of provider. Table 3 shows that the average consultation cost at drug sellers doubled from US$2.7 during baseline to US$5.4 during piloting and implementation. The treatments provided by private qualified practitioners tripled from US$6 during baseline to US$20 thereafter. Table 2. Main reasons for initially consulting a private provider Reason Baseline Pilot Implementation n (%) n (%) n (%) All Drug Private All Drug Private All Drug Private providers sellers qualified providers sellers qualified providers sellers qualified Close to home 8 (40) 4 (67) 4 (40) 66 (80) 44 (92) 21 (72) 48 (43) 25 (66) 23 (32) Cheap 1 (5) 1 (10) 1 (1) 1 (1) Trust in provider 7 (35) 4 (40) 5 (6) 1 (2) 3 (10) Advised to go there 1 (5) 1 (17) 3 (4) 1 (3) 21 (19) 5 (13) 16 (22) Seriousness of condition 3 (15) 1 (17) 1 (10) 8(10) 3 (6) 4 (14) 41 (37) 8 (21) 33 (45) Total

5 314 Bart Jacobs and Neil Price Table 3. Average costs of consultations with drug sellers and private qualified practitioners Provider consulted Baseline Pilot Implementation US$ (No. of patients) US$ (No. of patients) US$ (No. of patients) Drug seller 2.7 (7) 5.9 (49) 5.4 (38) Private qualified practitioner 6.0 (13) 18.9 (33) 20.6 (73) Case mix of patients and mortality rate Table 4 provides an overview of the patient case mix for the five most common diseases/conditions of patients admitted at the hospital during the 3 months of each of the study periods. The proportions of admitted patients suffering from malaria remained similar during the baseline and piloting but decreased during implementation. There were no changes observed in proportions of patients admitted for diarrhoea/dysentery and respiratory infection, and for women who delivered at the facility. The proportion of admitted dengue patients halved after baseline. The proportion of other conditions that are not specified in the HIS increased from 35% during baseline and piloting to 50% during Table 4. Patient case mix and mortality implementation. The mortality rate was 6.6/1000 admitted patients during baseline and increased to 13.6/1000 admissions thereafter. Total cost for hospitalization Table 5 and Figure 1 provide an overview of the total direct costs associated with hospitalization during the three study periods. The total average direct cost per illness episode per patient was US$3.2 during baseline, US$10.9 during piloting and US$19 during implementation. Costs of consultations at private practitioners formed the major proportion of total direct costs during the study periods: 72% during baseline, 67% during piloting and 59% during implementation. Disease/condition Baseline Pilot Implementation n (% of total) n (% of total) n (% of total) Malaria 91 (10) 73 (9) 42 (4) Diarrhoea/dysentery 80 (9) 100 (12) 111 (11) Respiratory infections 182 (20) 156 (19) 228 (22) Dengue 122 (13) 51 (6) 45 (4) Deliveries 85 (9) 93 (11) 85 (8) Other a 308 (34) 282 (35) 518 (50) Total patients (non-tb) Deaths (non-tb) Death/1000 patients a includes conditions not registered in the health information system. direct costs in US$ baseline piloting final 2 0 Private practitioners Transport Hospitalisation Figure 1. Direct costs incurred during baseline, pilot and implementation phases

6 The impact of user fees 315 Table 5. Direct costs associated with hospitalization during baseline, piloting and implementation (US$) Baseline (n = 101) Pilot (n = 151) Implementation (n = 152) Total Per % of Total Per % of Total Per % of spent patient total spent patient total spent patient total Private providers Transport Hospital Total Note: The baseline data do not include unofficial charges at the hospital, which according to Keller and Schwartz (2001) are likely to have been significant. Hospitalization fees at least officially were nil during baseline and amounted to 25% and 34% of the total direct costs during piloting and implementation, respectively. Transport costs accounted for 28% of total direct costs during baseline and became 7 8% during the periods thereafter. Ability to pay Despite the provision of free hospital care, 40% of respondents in the baseline reported having insufficient cash to cover all expenses incurred. This was significantly higher for those who initially sought care at the private sector than for those who did not: 65% versus 33% (p = 0.01), although such difference was no longer observed during piloting and implementation. During the pilot phase, however, 60% of interviewees were unable to cover all expenses, despite the fact that 25% were exempted from hospital fees (Table 6). During implementation, the proportion of interviewees unable to cover all expenses reduced to 41%, with only one patient exempted. Respondents in the pilot phase survey who knew hospital prices beforehand reported significantly more availability of sufficient cash than those not knowing the prices: 55% (36/66) versus 28% (24/85) (p < 0.001). During implementation, the proportions were 74% (73/99) and 30% (16/53) respectively (p < 0.001). During implementation, 66% of interviewees knew the hospital prices beforehand, significantly higher than the 43% during piloting (p < 0.001). The majority of interviewees who reported having insufficient cash resorted to borrowing to cover the direct costs of hospitalization. The remainder sold an animal to obtain money. Of those borrowing, the majority secured loans from relatives: 61%, 60% and 63% during baseline, piloting and implementation, respectively. Of people borrowing throughout the three periods, only 30% paid interest. 5 At baseline, interviewees borrowed together US$253, a ratio of 0.77:1 for borrowed money to direct costs. The respective amounts and ratio were US$1398 and 0.84:1 during piloting; and US$1542 and 0.53:1 during implementation. During baseline and piloting none of the borrowers reported difficulty repaying the loan. During implementation, however, 8% of borrowers or 3% of all interviewees reported having to sell land to repay their loan. At the pilot phase, patients and carers reported losing a total of US$1768 (US$12 per patient) because of being unable to work as a result of hospitalization. During implementation a total of US$346 (US$2.3 per patient) was reportedly lost due to inability to work. Effects of introducing and increasing user fees on monthly number of admissions Figure 2 presents a one-year overview of the number of paediatric cases and total number of patients admitted at the hospital during July 2001 July The average monthly total number of admissions was 287 at baseline. It fell slightly during the first 2 months following the start of the user fee pilot phase, whereupon it returned to its previous level.

7 316 Bart Jacobs and Neil Price Table 6. Ability to pay costs of hospitalization Variable Baseline Pilot Implementation (n = 101) (n = 151) (n = 152) N (%) N (%) N (%) Proportion exempted from payment 101 (100) 37 (25) 1 (0.7) Unable to pay all costs 40 (40) 91 (60) 62 (41) Resort to borrowing 36 (36) 84 (56) 62 (41) Following the increase of user fee prices (at implementation), the number of admissions fell by a quarter for the first 2 months, whereupon it returned to its previous level. The monthly number of paediatric cases remained unaffected by user fees: an average of 63 per month before user fees and 66 following their introduction. During piloting, however, children aged less than 60 months were exempted from payment. At implementation, when user fees were introduced for this age group (April 2002), the average number of monthly admissions decreased for 2 months before recovering. Discussion The main aim of introducing user fees at the hospital was to stimulate its staff to provide quality care through increasing their salaries. It has been observed elsewhere in Cambodia that considerable improvement in quantity and quality of services resulted from staff incentive payments derived from user fees (van Damme et al. 2001; Meessen et al. 2002; Ministry of Health Cambodia/Swiss Red Cross/WHO 2002; Soeters and Griffith 2003; Akashi et al. 2004). An additional objective behind the decision to subsequently increase user Total Paediatric fees during the implementation of the final scheme at KOD Referral Hospital was the pursuit of financial sustainability, in line with the national policies on health financing (Ministry of Health Cambodia 1997b). Significant planning went into the design and implementation of the user fee scheme at KOD, including the cross-sectional survey regarding willingness and ability to pay for public health services, and the piloting of the user fees in tandem with an assessment of patients ability to pay the newly introduced fees. Prior to increasing user fees (the implementation of the final user fee scheme), efforts were directed towards improving the quality of care at the referral hospital through strengthening the managerial capacity of its staff, and their diagnosis and treatment competence. Fee collection practices were limited to a single payment, irrespective of the duration of hospitalization or additional diagnostic tests. The fee scale was simple to facilitate the public s ability to remember prices. These price schedules were disseminated to the public by the use of flyers through village authorities and pagoda volunteers. Following the increase in user fees during implementation, significantly more patients came from districts other than July Aug Sep Oct Nov Dec Jan Feb Mar Apr May June July Introduction of user fees Increase user fees Figure 2. Total number of patients and number of paediatric patients admitted at the hospital (July 2001 July 2002)

8 The impact of user fees 317 KOD: 24% versus 7 15% before. Consequently, these patients tended to reside significantly further from the hospital than those patients hospitalized during the user-fee pilot phase: 12.5km versus 8km respectively. These changes most likely result from introducing surgery at the hospital, as patients and carers would have been less likely to be referred to the Provincial hospital where prices were much higher. Furthermore, the introduction of surgery probably resulted in a public perception of better quality of care, resulting in a willingness to travel further distances for services. There were, however, considerable changes in the socioeconomic profile of patients following the pilot phase of user fees, as defined by the two variables of landownership and ownership of motorcycles. The proportion of interviewees who were landless decreased from 16% during baseline to 7% following introduction of user fees. Landlessness is a major indicator of poverty, and households without land have much less income than those who have it (Ministry of Planning 1999; Chan and Sarthi 2003). For their socioeconomic stratification of patients in Sotnikum, Cambodia, Hardeman et al. (2004) utilize ownership of a motorbike as an indicator for medium rich and rich households. Following the introduction of user fees, the proportion of patient/carers who owned a motorbike increased from 17% to 25%. Although only based on two socio-economic indicators, these data suggest an alteration of the socio-economic profile of patients following the introduction and increase of user fees, suggesting that a proportion of the poor and vulnerable have been deterred from seeking care at the hospital. The introduction and subsequent increase of user fees appear to have led more patients initially to seek care from the private sector. Whereas 20% and 54% of interviewees during baseline and piloting, respectively, first consulted private providers, 73% did so during implementation. The mostmentioned reason for initially consulting private practitioners was geographical proximity during all the study periods. This was especially pronounced during the pilot phase (80%), when most patients who initially went private consulted drug sellers. These findings are in line with Mbugua et al. (1995), who observed a shift from public fee-charging facilities to the private sector in Kenya following the introduction of user fees. As pointed out by Ensor and Cooper (2004), selection of provider requires a multitude of complex choices to be made. Factors that play a role are the perceived quality of care offered, physical and financial accessibility of services and knowledge of services offered. The limited knowledge of the population about appropriate health care may force them to seek advice from nearby private providers as a perceived cost-saving measure (Asenso-Okyere et al. 1998), as indicated in KOD by the shift of initial consultations from the hospital to drug sellers during piloting. In line with observations by Paphassarang et al. (2002) in neighbouring Laos, such drug sellers offer treatments based on ability to pay, overcoming uncertainties related to costs of treatment in our case, hospitalization. During implementation, however, most interviewees initially consulting private providers went to qualified providers. Only 32% of such interviewees mentioned proximity to home as a reason for consulting such providers. Instead, the most reported reason for consulting qualified private practitioners was the seriousness of their condition, 45%, compared with 10 14% during the previous periods. Whereas 35% of interviewees initially going to private practitioners mentioned trust in provider as a reason for this consultation at baseline, none mentioned this during implementation. During implementation, 19% of interviewees initially consulting private providers sought advice from their peers compared to 5% during free care and piloting. These data suggest that uncertainties related to the cost of hospitalization were most pronounced following the increase of fees at that facility and forced patients to seek care further from home at unfamiliar private providers, following consultation of their peers. The fact that seriousness of condition was the second most mentioned reason during implementation may indicate that patients delayed seeking care (due to uncertainties of cost of treatment), as observed by Paphassarang et al. (2002) and Asenso-Okyere et al. (1998). However, not all observed changes in health-care-seeking behaviour can be explained by uncertainties related to price. For example, during piloting, interviewees who knew the hospital prices beforehand were significantly more likely to first contact the public sector (60%) than those who did not know the prices (35%). During implementation, the respective figures were 34% and 13%, a considerable decrease, which suggests that patients may have had difficulties meeting the financial requirements for paying the hospitalization fees. Private practitioners appeared to increase their fees in tandem with the price increases introduced at the referral hospital. The average price per treatment obtained from drug sellers doubled from US$2.7 during baseline to US$5.4 thereafter, and the price per treatment from private qualified practitioners tripled from US$6 to about US$20 respectively. These price increases suggest profiteering by private providers, who operate without supervision or regulation and abuse the patients limited knowledge of appropriate health care (Mills et al. 2002). It may also indicate that patients delayed seeking health care, resulting in their presenting with more serious conditions that required more expensive treatment. Patient case-mix and mortality data at the hospital suggest that the price increase at private practitioners was mainly due to delayed care seeking. The case mix for the five most common conditions of patients, for example, altered little over the study periods, but the overall mortality rate for admitted patients increased from 6.6/1000 during baseline to 13.6/1000 thereafter. This increase in mortality 6 may be due to delayed care seeking because of uncertainties related to the total costs of treatment (at private and public facilities), but may also result from unsuccessful initial treatment in the private sector. The HIS, however, did not provide data to allow us to assess whether patients presented with more severe conditions at pilot and implementation phases than at baseline.

9 318 Bart Jacobs and Neil Price The total average direct costs associated with hospitalization were US$3.2 at baseline, although the real direct costs are likely to have been higher due to charging of unofficial fees by hospital staff (Keller and Schwartz 2001), despite none of the interviewees reporting such practice. In line with observations elsewhere in Cambodia, charging unofficial payments at Kirivong hospital decreased following the introduction of user fees (Wilkinson et al. 2001a; Soeters and Griffith 2003; Akashi et al. 2004; Hardeman et al. 2004). Increased salaries for staff and contractual agreements forbidding unofficial charges, along with the establishment of facility-level financial committees (responsible for service fees and salary supplements), installation of suggestion/complaint boxes in visible places and staff on duty wearing identification badges, have all facilitated this reduction in unofficial charges in Cambodia (van Damme et al. 2001; Ministry of Health Cambodia/Swiss Red Cross/WHO 2002; Soeters and Griffith 2003; Akashi et al. 2004; Hardeman et al. 2004). A combination of these approaches was introduced at the KOD hospital, along with the establishment of a staff disciplinary committee. Nevertheless, the reported total direct costs increased to US$10.9 during piloting and US$19 during implementation. The major proportion (59 72%) of direct costs related to consultations at private practitioners, whereas hospitalization fees accounted for 0 34% during the observation periods. The proportions of interviewees who were unable to pay all costs related to hospitalization without resorting to borrowing or selling assets were similar during baseline and implementation (40%). During piloting, significantly more interviewees reported an inability to pay (60%), while 25% were exempted from user fees in comparison with the implementation phase when only one interviewee was exempted. Several factors may influence the ability to pay for all costs associated with hospitalization. For example, the data suggest that prior knowledge of the user fees influences the need to borrow money to pay for services. During piloting, 43% of interviewees knew the prices beforehand, significantly lower than the 66% observed during implementation. Interviewees who knew the fees in advance were significantly more likely to have sufficient available cash to cover all expenses (74%) than those who did not know the fees beforehand (31%). They were also significantly more likely to initially consult public health providers (34%) than those who did not know the prices (13%), reducing total expenditure. Being acquainted with prices beforehand tackles uncertainty in knowing how much a hospitalization will cost and may consequently reduce the need for borrowing (Nyonator and Kutzin 1999). The difference in ability to pay between pilot and implementation when the socio-economic status of interviewees was similar is likely to be due to seasonality: the pilot phase was conducted during the peak period of agricultural activities, when households have less cash than after harvest when crops have been marketed. Opportunity costs per interviewee (for patients and carers) were US$12 at pilot and US$2.3 during implementation. The ratios of borrowed money to direct costs were 0.84:1 and 0.53:1 respectively. As pointed out by Chan and Sarthi (2003), Cambodian farmers are highly dependent upon credit during the planting season. The proportion of interviewees who could pay hospital fees without borrowing cannot, however, be taken as representative of the wider population because it understates the proportion of the population that was deterred by the fees (Mwabu et al. 1995). As suggested by the decrease in landless interviewees and an increase in motorbike owners from baseline to pilot, a major proportion of the most vulnerable may have been deterred from seeking care at the hospital. The failing exemption system may have contributed to this. Only 31% of borrowers had to pay interest, which is similar to the proportion observed during baseline and piloting. However, 3% of respondents in the implementation phase survey reportedly had to sell land in order to repay the loan, a major cause for impoverishment in Cambodia (Oxfam GB 2000). The introduction and subsequent increase in user fees at the hospital has created what may be termed a medical poverty trap, which occurs when user fees are introduced at public health facilities concurrently with an increase in outof-pocket expenses for private sector services (Whitehead et al. 2001). The main effects of a medical poverty trap fall into four categories: untreated morbidity, reduced access to care, long-term impoverishment, and irrational use of drugs leading inter alia to drug resistance. In line with findings from other countries (de Béthune et al. 1989; Wilkinson et al. 2001b), hospitalization rates declined following increase of user fees. The phased approach, the simultaneous quality improvement and intensive efforts to disseminate information on the user fee scheme allowed hospitalization rates to recover following a 2-month decline. Addressing the impact of the medical poverty trap described above will require two interventions to be implemented immediately at operational district level: regulating the private sector and exempting the poor from user fees. A third, longer-term initiative might also justify serious consideration, namely the establishment of a social health insurance mechanism. The research undertaken at KOD shows that the greatest proportion of costs incurred during an illness episode was due to consulting private providers. Currently, within the context of its strategic 5-year plan for , the Cambodia Ministry of Health (2002) envisages taking steps to develop a legislative and regulatory framework for the private health sector. The private sector in Cambodia, as in many other low-income countries, is unregulated: providers emphasize curative over preventive services, treatment is often according to ability to pay, best practice is often ignored, etc. Integrating the private health sector into an overall public policy framework remains a considerable and long-term challenge (Mills et al. 2002). In the short term, strategies should, therefore, focus on influencing healthseeking behaviour to promote initial consultation with public providers, and bringing first-line curative services closer to

10 The impact of user fees 319 the people by linking these to monthly outreach activities (and when resources allow, by establishing health posts). Exempting the poor from user fee payment is an important component of any viable user fee system (Mwabu et al. 1995). However, exemption mechanisms strain already limited administrative capacity, and are expensive to operate, with start-up costs particularly high: tested and low-cost models for identifying those who simply cannot afford to pay for health care are as rare today as they were a decade ago (Shaw and Griffin 1995, p. 42). Moreover, schemes aimed at targeting the poor with exemptions often miss the intended beneficiaries (Mills 1991; Willis and Leighton 1995). Gilson et al. (1995) note three main problems with targeting the poor. First, judging who should be eligible for exemptions faces a number of constraints. Means testing depends largely upon household income data, the paucity of such information leading to inappropriate or subjective eligibility (Mills 1991; Willis and Leighton 1995). As measures of poverty become more sophisticated, the costs of their calculation rise: budget constraints limit accuracy. Secondly, the administration and monitoring of exemptions is fraught with difficulties and requires an efficient administrative system, with clear guidelines on the application of exemptions (which are often missing, resulting in misallocation of benefits). Finally, insufficient information and stigma may prevent the poor from taking up exemptions. These three problems combine to make targeting expensive and ineffective: the expenses incurred in targeting may undermine the revenue generated from the implementation of user fees. Hence, either user fees are imposed with ineffective targeting, harming the poor, or the user fee initiative is financially unviable (Thomas et al. 1998, p. 51). The exemption system at Kirivong hospital is constrained by all the above problems, reflected in the fact that only one respondent in the implementation phase interviews was exempted from paying hospital fees. Furthermore, an exemption scheme requires operational arrangements which will ensure that exemptions do not result in loss of income for hospital staff. Fear of loss of income may result in refusal to grant exemptions to those most in need, and/or discrimination by staff against those unable to pay (Whitehead et al. 2001), thus exacerbating the stigma reported by Gilson et al. (1995). Reducing the risk of discrimination by staff will require mandating a third party to reimburse fees to the hospital. This is currently being piloted in Cambodia at several operational districts where non-governmental organizations are responsible for the reimbursement through the operation of an equity fund (Hardeman et al. 2004). Alternative and more sustainable approaches are being considered which will require exploration of existing social networks and development of strategies to stimulate the community cash contributions to protect the most vulnerable. In the longer term, in order to address the medical poverty trap, consideration needs to be given to changing from direct payments at the point of service delivery to a social health insurance system in which healthy, high-income groups subsidize health care for low-income groups (Nyonator and Kutzin 1999; Whitehead et al. 2001). Jütting (2003) has shown that social health insurance schemes can be viable in resource-poor settings, but because they continue to exclude the poorest, well-targeted subsidies may be necessary. In Cambodia such subsidies may take the form of reimbursement of the membership fee for the poorest by the equity fund. Endnotes 1 The cross-sectional survey interviewed a sample of 400 mothers with a child aged under 18 months in 40 villages. 2 Costs for private consultations or hospitalization were not adjusted for disease or condition. 3 During the study period, the initial treatment for TB patients was 2 months of hospitalization. By decree, TB patients are exempted from paying user fees at public health facilities. The hospital has 16 beds for TB patients. 4 With more than three-quarters of the population engaged in agriculture in Cambodia, land is the most important asset in rural areas (Ministry of Planning 1999, p. 16). 5 The remaining 70% were able to borrow from relatives, friends or neighbours without paying interest. 6 We recognize the problems of interpreting hospital death rates, which require adjustment for severity of illness, length of hospital stay, age, diagnosis, type of admission (Seagrott and Goldacre 1994; Jarman et al. 1999). References Akashi H, Yamada T, Huot E, Kanal K, Sugimoto T User fees at a public hospital in Cambodia: effects on hospital performance and provider attitudes. Social Science and Medicine 58: Asenso-Okyere WK, Anum A, Osei-Akoto I, Adukonu A Cost recovery in Ghana: are there any changes in health seeking behaviour? Health Policy and Planning 13: Audibert M, Mathonnat J Cost recovery in Mauritania: initial lessons. Health Policy and Planning 15: Birdsall N Pragmatism, Robin Hood and other themes: good government and social well-being in developing countries. New York: Rockefeller Foundation. Chan S, Sarthi A Facing the challenge of rural livelihoods: a perspective from nine villages in Cambodia. Phnom Penh: Cambodia Development Resource Institute. Chawla M, Ellis RP The impact of financing and quality changes on health care demand in Niger. Health Policy and Planning 15: Creese AL User charges for health care: a review of recent experience. Health Policy and Planning 6: de Béthune X, Shesoko A, Lahaye J-P The influence of an abrupt price increase on health service utilization: evidence from Zaire. Health Policy and Planning 4: Ensor T, Cooper S Overcoming barriers to health service access: influencing the demand side. Health Policy and Planning 19: Foreit K, Levine R Cost recovery and user fees in family planning. Policy Paper Series No.5. Washington, DC: The Futures Group. Gilson L, Russell S, Buse K The political economy of user fees with targeting: developing equitable health financing policy. Journal of International Development 7: Ha NTH, Berman P, Larsen U Household utilization and expenditure on private and public health services in Vietnam. Health Policy and Planning 17: Hardeman W, van Damme W, van Pelt M et al Access to health care for all? User fees plus a Health Equity Fund in Sotnikum, Cambodia. Health Policy and Planning 19: Jacobs B, Price N Community participation in externally

11 320 Bart Jacobs and Neil Price funded health programmes: lessons from Cambodia. Health Policy and Planning 18: Jarman B, Gault S, Alves B et al Explaining differences in English hospital death rates using routinely collected data. British Medical Journal 318: Jütting JP Do community-based health insurance schemes improve poor people s access to health care? Evidence from rural Senegal. World Development 32: Keller S, Schwartz JB Final evaluation report: contracting for health services pilot project. Phnom Penh: Asian Development Bank. Lewis MA Do contraceptive prices affect demand? Studies in Family Planning 17: Mbugua JK, Bloom GH, Segall MM Impact of user charges on vulnerable groups: the case of Kibwezi in rural Kenya. Social Science and Medicine 41: Meessen B, Van Damme W, Ir P, Van Leemput L, Hardeman W The New Deal in Cambodia: The second year. Phnom Penh: Médecins sans Frontières Cambodia. Mills A Exempting the poor: the experience of Thailand. Social Science and Medicine 33: Mills A, Brugha R, Hanson K, McPake B What can be done about the private health sector in low-income countries? Bulletin of the World Health Organization 80: Ministry of Health Cambodia. 1997a. Community co-management and co-financing in health centres: operational guidelines for pilot areas. Phnom Penh: Ministry of Health. Ministry of Health Cambodia. 1997b. National charter on health financing in the Kingdom of Cambodia. Phnom Penh: Ministry of Health. Ministry of Health Cambodia National Health Statistics Report Phnom Penh: Ministry of Health, Department of Planning and Health Information. Ministry of Health Cambodia Health sector strategic plan Phnom Penh: Ministry of Health. Ministry of Health Cambodia/Swiss Red Cross/World Health Organization Takeo Provincial Referral Hospital: Pioneering a health financing scheme. Phnom Penh: Ministry of Health. Ministry of Health Cambodia/UNICEF Introducing user fees at public health facilities in Cambodia: an overview. (Briefing paper) Phnom Penh: Ministry of Health. Ministry of Planning, Cambodia Cambodia Human Development Report 1999: Village economy and development. Phnom Penh: Ministry of Planning. Mwabu G, Mwanzia J, Liambila W User charges in government health facilities in Kenya: effect on attendance and revenue. Health Policy and Planning 10: Nyonator F, Kutzin J Health for some? The effect of user fees in the Volta region of Ghana. Health Policy and Planning 14: Oxfam GB Health and landlessness: A mini case study. Phnom Penh: Oxfam, Cambodia Land Study Project. Paphassarang C, Philavong K, Boupha B, Blas E Equity, privatization and cost recovery in urban health care: the case of Lao PDR. Health Policy and Planning 17: Price NL Service sustainability strategies in sexual and reproductive health programming. London: John Snow International. Seagroatt V, Goldacre M Measures of early postoperative mortality: beyond hospital fatality rates. British Medical Journal 309: Shaw RP, Griffin C Financing healthcare in Sub-Saharan Africa through user fees and insurance. Washington, DC: World Bank. Soeters R, Griffiths F Improving government health services through contract management: a case from Cambodia. Health Policy and Planning 18: Thomas S, Killingsworth JR, Acharya S User fees, selfselection and the poor in Bangladesh. Health Policy and Planning 13: Van Damme W, Meessen B, von Schreeb J et al Sotnikum New Deal, The first year. Phnom Penh: Médecins sans Frontières Cambodia. Waddington C, Enyimayew KA A price to pay: the impact of user charges in Ashanti-Akim district in Ghana. Part I. International Journal of Health Planning and Management 4: Waddington C, Enyimayew KA A price to pay: the impact of user charges in Ashanti-Akim district in Ghana. Part II. International Journal of Health Planning and Management 5: Whitehead M, Dahlgren G, Evans T Equity and health sector reforms: can low-income countries escape the medical poverty trap? The Lancet 358: Wilkinson D, Holloway J, Fallavier P. 2001a. The impact of user fees on access, equity and health provider practices in Cambodia. Phnom Penh: Ministry of Health. Wilkinson D, Gouws E, Sach M, Karim SS. 2001b. Effect of removing user fees on attendance for curative and preventive primary health care services in rural South Africa. Bulletin of the World Health Organization 79: Willis CY, Leighton C Protecting the poor under cost recovery: the role of means testing. Health Policy and Planning 10: Acknowledgements The authors would like to thank Jeff Sine for comments on an earlier draft of this article, and two anonymous reviewers for Health Policy and Planning. The project in Kirivong was implemented by Enfants&Développement, funded by the Cambodian Ministry of Health. The usual disclaimer applies: the opinions expressed herein are those of the authors and do not necessarily reflect the views of Enfants&Développement or the Government of Cambodia. We accept responsibility for any errors. Biographies Bart Jacobs currently manages the Kirivong Operational Health District in Cambodia. He holds an MSc in Health Planning and Development, and has worked extensively in Southeast Asia, East Africa, the former Soviet Union, and Australia as a researcher, consultant and manager in a wide variety of health projects including TB control, blood safety, STD/HIV control, social marketing, indigenous health, and health sector reform. His main areas of research and operational interest and expertise are in health-seeking behaviour and the development of locally appropriate public health interventions. Neil Price is Senior Lecturer and Deputy Head at the Centre for Development Studies, University of Wales Swansea. He holds a doctorate in social anthropology and has undertaken ethnographic fieldwork in Cambodia, the Caribbean, China, Kenya and Zambia. He has extensive experience of advisory and commissioned research with international development agencies throughout Sub-Saharan Africa, Asia, Latin America and the Middle East. His main areas of advisory expertise are in social, policy and institutional analysis in the health sector, and in the appraisal, monitoring and evaluation of HIV/AIDS and reproductive health programmes. Correspondence: Neil Price, Centre for Development Studies, University of Wales Swansea, Swansea, SA2 8PP, UK. n.l.price@swansea.ac.uk