The effects of user fees on quality and utilization of primary health-care services in Afghanistan: a quasi-experimental health financing pilot study in a post-conflict setting.

THE INTERNATIONAL JOURNAL OF HEALTH PLANNING AND MANAGEMENT

Int J Health Plann Mgmt 2013; 28: e280–e297. Published online 23 September 2013 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/hpm.2178

The effects of user fees on quality and utilization of primary health-care services in Afghanistan: a quasi-experimental health financing pilot study in a post-conflict setting Laura C. Steinhardt1*, Krishna D. Rao2, Peter M. Hansen3, Sahibullah Alam4 and David H. Peters1 1

Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland USA 2 Public Health Foundation of India, Delhi, India 3 GAVI Alliance, Geneva, Switzerland 4 Ministry of Public Health, Kabul, Afghanistan

SUMMARY Background After the fall of the Taliban regime, most clinics in Afghanistan were charging fees to patients. The government invested in monitoring and evaluation systems for its newly rebuilt primary care system, but little was known about the effects of user fees. This study was undertaken to provide evidence on user fees’ effects on quality and service utilization and to help inform development of health financing policy and strategy. Methods A quasi-experimental health financing pilot study was implemented in 2005. Fortyseven facilities were randomized to implement a standardized user fee intervention, offer free services, or serve as controls, continuing current cost-sharing systems. Revenues were co-managed by staff and community leaders for facility improvement. Baseline and follow-up facility assessments, exit interviews, and household surveys, as well as routine data were used to evaluate user fee effects over 2 years. Results Observed and perceived quality improved at most facilities but did not differ by study group. Utilization increased in all groups, but the increase was 682 to 748 visits per month larger in facilities randomized to free services compared with those randomized to fees or controls (p < 0.01). Conclusion User fees demonstrated few beneficial effects and slowed the rate of increase of service utilization in Afghanistan. In 2008, the government abolished primary care fees, citing results of this study. Copyright © 2013 John Wiley & Sons, Ltd. KEY WORDS:

user fees; quality; care seeking; Afghanistan; primary care

INTRODUCTION Afghanistan’s health sector has made remarkable progress since the fall of the Taliban regime in 2001, despite problems of insecurity and continued opium *Correspondence to: L. C. Steinhardt, Department of International Health, Johns Hopkins Bloomberg School of Public Health, 615 N. Wolfe St., Suite E8130, Baltimore, MD 21205, USA. E-mail: [email protected]

Copyright © 2013 John Wiley & Sons, Ltd.

PRIMARY HEALTH-CARE SERVICES IN AFGHANISTAN

e281

production. The Ministry of Public Health (MoPH), in collaboration with donors and non-governmental organizations (NGOs), established a basic package of health services (BPHS) in 2003 that was rapidly rolled out across the country. In most provinces and districts, delivery of the BPHS, which is almost exclusively donor financed, at an average of $4 per capita (Loevinsohn and Sayed, 2008), is contracted out to national and international NGOs. The MoPH retains an oversight role over policy, financing, and service delivery and directly provides services in three provinces. With donor assistance, the MoPH has invested in a rigorous monitoring and evaluation system for the BPHS: annual nation-wide facility surveys and dissemination of a balanced scorecard to service providers (Peters et al., 2007b) have contributed to substantial gains in the quality of primary care services in Afghanistan (Hansen et al., 2008a, 2008b, 2008c). There is a lack of information on the scope and quality of the private sector in health, which is not well organized and tends to be concentrated in urban areas (Sabri et al., 2007) Although the BPHS was implemented in 2003, there was no formal policy on cost sharing during the first few years of service delivery. Most NGOs contracted to provide BPHS services initially charged some type of user fee for outpatient visits, typically around 5–10 Afghanis (approximately $0.10–0.20). A national survey of health facilities found that 70.4% of primary care facilities charged user fees in 2004, rising to 84.3% in 2007 (Johns Hopkins University/Indian Institute for Health Management Research, 2008). Fees were not standardized, and some NGOs charged for drugs, whereas others did not. There was little data on how much money was raised through fees, what was being done with collected fees, and what effect fees had on utilization and service quality. The first post-Taliban national health financing policy for Afghanistan was approved in 2007 by the MoPH. Although acknowledging that Afghanistan will depend on donor funding in the medium term, the policy stipulated that increased government revenues for health and other alternative sources of financing should be pursued. The policy postponed the decision on whether to allow user fees until further study. A National Health Accounts survey conducted for the year 2008–2009 found that health expenditures were estimated at $42 per capita annually, three-quarters of which were estimated to come from households; most household health spending went towards retail health goods (37.5%), private outpatient facilities (27.5%), public hospitals (21.2%), and private hospitals (9.3%) (Government of the Islamic Republic of Afghanistan Ministry of Public Health, 2011). Only 3.8% of households’ out-of-pocket health expenditures were estimated to be spent on outpatient care at public health facilities (Government of the Islamic Republic of Afghanistan Ministry of Public Health, 2011).1 This study was designed to inform government decisions about health financing. The objective of the health financing pilot (HFP) study, initiated in 2004–2005, was to compare alternative health financing mechanisms on their ability to: (1) increase funding available for the health sector; (2) provide incentives to improve quality of 1

It should be noted that the national health accounts study cited here took place in 2008–2009, the year after the MoPH banned user fees in the public sector for outpatient services. Therefore, household expenditures on outpatient care in the public sector might have been higher during the time that the health financing pilot study presented here took place. Copyright © 2013 John Wiley & Sons, Ltd.

Int J Health Plann Mgmt 2013; 28: e280–e297. DOI: 10.1002/hpm

e282

L. C. STEINHARDT ET AL.

care; (3) improve financial access to care; and (4) increase community ownership and local accountability of the health facility. The primary objectives of this paper are to describe the effects of user fees on key outcomes: observed facility structural quality of care, perceived quality of care, and utilization of outpatient services. Additional outcomes included out-of-pocket health expenditures and sources of financing for health care.

User fees in context Since their widespread introduction in the 1980s, user fees for publicly provided health services have become common in many developing countries, but continue to be a contentious policy issue. Proponents of fees have argued that fees help to raise revenue for typically under-budgeted health sectors and to improve the quality of health services if retained at local levels and used to re-invest in health facilities or drug supplies (Akin et al., 1987). Opponents of user fees claim that they suppress utilization of beneficial health services and are an inequitable way to finance health care (Ridde, 2003; Gilson and McIntyre, 2005), as they are especially harmful for the poor and can decrease their access to health services (Nyonator and Kutzin, 1999; Gilson et al., 2001). In settings where fees have been removed, utilization has increased (Lagarde and Palmer, 2008), and utilization increases in Uganda following fee removal were pro-poor (Wilkinson et al., 2001; Burnham et al., 2004; Gilson and McIntyre, 2005; Nabyonga et al., 2005). A review of multiple user fee experiences found that their impact on revenues, quality, and access to health care is highly dependent on the context in which they are implemented, in addition to their design and implementation (Peters et al., 2007a). Although user fees have been shown to decrease utilization of services, in a few cases, they have led to utilization increases, when combined with large improvements in quality and when the initial quality of services was very low (Litvack and Bodart, 1993; Diop et al., 1995; Rao and Peters, 2007). Despite a handful of examples, there is no clear evidence on whether user fees, when retained at health facilities and used to improve quality, can create quality gains sufficient to offset utilization declines following fee introduction. In many cases, fees were introduced as part of broader reforms that also included quality improvement initiatives, making it difficult to isolate the independent effects of user fees. There is also little empirical evidence on user charges specifically in postconflict settings, although community health funds (pre-payment schemes) have proven to be successful after several years in Rwanda. NGOs operating in postconflict or emergency settings sometimes charge fees, but some researchers have recommended against fees in complex emergency and “transitional” settings (Humanitarian Practice Network, 2004), where the majority of the population is poor and implementing effective waiver systems does not warrant administrative costs (Meessen et al., 2006). Copyright © 2013 John Wiley & Sons, Ltd.



e283

METHODS Study design: the health financing pilot study in Afghanistan Three interventions were tested in the HFP study using a quasi-experimental design: (1) a user fee arm, with a flat fee for services, and a percentage charge of wholesale drug price; (2) a free services arm, where services were free at the point of use, which was considered to be an intervention, given that the majority of BPHS facilities were charging some type of fees at the time of implementation; and (3) a community health fund arm, where households in the catchment area were asked to (voluntarily) pre-pay a given amount in exchange for access to the pilot health facility for a very small co-payment (1 Afghani, equivalent to $0.02 US), while non-members paid an elevated user charge. The community health fund arm was discontinued early—after less than 1 year of operation—because of low enrollment for various reasons, including insufficient community mobilization; its evaluation is reported elsewhere (Islamic Republic of Afghanistan. Ministry of Public Health, 2007; Rao et al., 2009). At user fee facilities, all revenues collected were retained at the facility and co-managed by a committee of facility staff and community members to improve the quality of health services (e.g., replenishing drug stocks, repairing facility infrastructure and equipment, providing small staff incentives) and to expand access to care in the community (e.g., transportation reimbursement for poor patients to come to the facility). At facilities collecting fees prior to the user fee pilot, most revenues were sent back to the managing NGO to supplement their overall budget. All promotive and preventive care, including vaccination, well-baby visits, antenatal care, deliveries, postnatal care, and family planning services, was systematically exempted from user fees. Tuberculosis-Directly Observed Therapy, Short Course provision and emergency care were also free of charge across all arms of the pilot. Fee waiver cards were provided to very poor and female-headed households, identified by community leaders and groups, enabling all household members to access care for free at user fee facilities. Approximately 14.5% of households in the immediate catchment area of facilities received waiver cards (Steinhardt and Peters, 2010). Fee levels were kept intentionally low and allowed to vary within a suggested range for each facility type: between 2 and 5 Afghanis ($1 US= 49 Afghanis in 2004–2007) for basic health centers (BHCs), between 2 and 8 Afghanis for comprehensive health centers (CHCs), and between 5 and 10 Afghanis for district hospitals (DHs); and a medications fee equivalent to between 15% and 50% of the wholesale drug cost (mean charge = 6.5 Afghanis for 2.4 drugs). Staff members from NGOs and pilot facilities were trained on the user fee design and how to establish user fee committees of community members and facility staff. The committees then finalized the service and drug fee levels based on catchment area socio-economic characteristics and decided how revenues would be spent. An information campaign was carried out in the catchment area to inform residents of any changes in fees. Ten of the 34 provinces in Afghanistan participated in the HFPs. Services in each province were managed by one NGO (or by the MoPH, in the case of three provinces). In each province, five primary care facilities—BHCs or CHCs—were nominated by the service provider on the basis of readiness to participate (e.g., full Copyright © 2013 John Wiley & Sons, Ltd.


e284


staff capacity, relatively good administrative systems). Two of the five facilities were randomized to one intervention (e.g., user fees) and two to a second (e.g., free services), and the fifth facility served as a control, continuing whatever cost sharing system the NGO in that province had in place. When possible (in six provinces), the DH close to one user fee pilot facility also participated in the intervention, to send price signals to community members that encouraged use of lower levels of care. A total of 47 pilot facilities were included: 27 user fee facilities (including six DHs), 10 free services facilities, and 10 control facilities (Table 1). Evaluation research design and data collection Baseline data were collected in May to September 2004, implementation of the pilot projects began between July to November 2005, and follow-up data collection for the final evaluation took place in April through August 2007. Several sources of data were used, including the following: baseline and follow-up facility assessments, patient exit interviews and catchment area household surveys, and administrative data from the health management information system (HMIS). The facility assessments and patient exit interviews were part of an annual nationwide survey of BPHS performance (National Health Services Performance Assessment), conducted since 2004 with technical assistance from a third party to monitor quality of care. At each facility, trained surveyors gathered details on standardized indicators of structural quality of care, such as drug availability and equipment functionality. Exit interviews were conducted using systematic random sampling of five caretakers of patients under 5 years and five patients of ages 5 years and older at each facility. Among other questions, patients/caretakers were asked to rate their agreement, using a four-point Likert scale, with several statements about various aspects of facility quality of care. Table 1. Number of health-financing pilot facilities in selected provinces and cost sharing at baseline Randomized to: Province Balkh* Badghis Sar-i-Pol* Wardak Farah Samangan* Nimruz* Kapisa* Parwan* Panjshir† TOTAL

Charges at baseline

User fees

Free

Control

Services

Drugs

3 2 3 2 2 3 3 3 3 3 27

2 2 2 2 2 10

1 1 1 1 1 1 1 1 1 1 10

Y Y Y Y Y Y N N N N

N N Y N Y N N N N N

*In these provinces, the district hospital was also included in the user fee arm, to rationalize the referral system; this yields three user fee facilities in these provinces as opposed to two. † Panjshir province implemented only user fees at three facilities and does not have a district hospital. Copyright © 2013 John Wiley & Sons, Ltd.



e285

Further details on the National Health Services Performance Assessment methodology are provided elsewhere (Peters et al., 2007b). A household survey was conducted for each pilot facility, whereby two villages with greater than 100 households each and within 90-min walking distance were randomly selected. With the use of a random start, up to 25 households were surveyed using the nearest-next door method (Turner et al., 1996) if they had a woman 18 years or older with a child 3 years or younger, for a total of 50 households across two villages, per pilot facility. The household survey included questions on illness, care seeking behavior, money spent on treatment, perceived quality of care at the pilot facility, and household characteristics. In the follow-up household surveys, a third village was added in the catchment areas of free services and control facilities to increase the number of surveyed households to 75 at these facilities. All survey forms were edited for completeness and consistency and double-entered into a database. Although the pilot design was quasi-experimental, with interventions randomized to pre-selected facilities within a province, the pilot was implemented in 10 provinces, which are managed by different NGOs and were found to have different fee structures at baseline. Because of the different historical practices of the NGOs, in some areas, the intervention entailed the removal of user fees, whereas in other areas, the intervention involved introducing user fees, and in still others, it involved a change from one type of user fee to the pilot user fee design. User fee amounts before the HFP study were comparable with amounts implemented during the pilot study, although only one of the 10 provinces had a charge for drugs, in addition to a consultation fee. Furthermore, comparison of the study groups at baseline revealed several differences at facilities randomized to free services, in terms of lower facility awareness, perceived affordability, and care seeking, despite randomization of facilities within provinces (data available from authors). Analysis Facility structural quality of care was measured by summing binary responses to questions in the facility assessment across four main quality domains: cleanliness/ need for facility repairs, equipment functionality, drug availability, and facility infrastructure. The structural quality index had a possible range of 0 to 31 points. Patients’ and household members’ perceived quality of care was assessed by summing responses to eight four-point, Likert-scale questions, asking respondents to rate their level of agreement with statements about various aspects of quality (e.g., cleanliness of facility, trust in skills/abilities of health workers, ease of obtaining medicines, and other questions about quality), which appeared to be easily understood by patients during pilot testing of the survey (Hansen et al., 2008a, 2008b, 2008c). Difference-in-differences (DD) linear and logistic regression models were used to test the statistical significance of changes in selected quantitative outcomes, by study group. The advantage of a DD approach is that baseline characteristics, including levels of the outcome of interest, do not need to be the same at baseline, as the method compares differences over time in treatment groups to those in the control group (Buckley and Shang, 2003). DHs randomized to user fees were analyzed Copyright © 2013 John Wiley & Sons, Ltd.


e286


separately from BHCs and CHCs randomized to fees, as hospitals were only included in the user fee arm and differ in many ways (staffing, services, size of catchment area population) from primary care facilities. In addition to the primary predictors (study group, time, and the interaction of study group and time), additional independent variables were included in DD regressions if they were significant (p < 0.05) in bivariate analyses or were hypothesized to be strong confounders, as follows: Y ¼ b0 þ b1 Post þ b2 UF þ b3 UF-DH þ b4 Free þ b5 UFPost þ b6 UF-DHPost þ b7 FreePost þ b8 Province þ b9 ProvinceTime þ b10 Confounders þ e where Post = post-financing intervention time period, UF = user fee BHCs and CHCs, and UF-DH = user fee-DH facilities. The coefficients of interest include b5 (difference over time in the change among primary care facilities randomized to user fees compared with those randomized to be controls), b7 (difference between primary care facilities randomized to free services compared with primary care controls), and b7–b5 (difference in the change over time between primary care facilities randomized to free services and those randomized to user fees). Analyses were conducted using data from only facilities that had data at both baseline and follow-up periods (n = 40 for facilitylevel outcomes; n = 41 facilities for patient and household-level outcomes), as six facilities had to be dropped at follow-up because of security concerns. For the household survey responses, sensitivity analyses were conducted using only common villages between baseline and follow-up (69 of 107 (64.5%) of villages, containing 77.4% of households surveyed). For the HMIS analysis, data for all facilities not participating in the pilot study in a given province were added to the control facility arm, because their cost-sharing arrangements had not changed during the pilot period, resulting in a total of 98 facilities (24 user fee, 10 free services, and 55 control, with one province omitted from the analysis because it lacked pre-pilot HMIS data). For each facility, the average monthly number of visits was calculated for the 1-year period prior to pilot implementation and for the 1-year period after pilot implementation, resulting in two data points per facility: one before and one after. The DD regression model specified earlier was used to test whether the changes in visit volumes over time differed significantly between the two intervention groups. Standard errors were adjusted for stratified sampling and clustering at the facility level (for patient exit interviews) and village level (for household surveys) using the Taylor-series linearization, in STATA 10.0 (Lehtonen and Pahkinen, 1995; StataCorp, 2007).

RESULTS Revenues raised from user fees Analysis of administrative data showed that user fees were able to raise limited amounts of revenue, in part, because charges were set intentionally low to remain affordable to the catchment area population. On average, BHCs received $103.7 per month from Copyright © 2013 John Wiley & Sons, Ltd.



e287

registration, medication, and other fees, CHCs an average of $211.6 per month, and DHs an average of $697.1 per month, representing an average of 3.4% of total facility operation costs (including direct and indirect costs) (Johns Hopkins University/Indian Institute for Health Management Research, 2008). Funds were typically spent on items related to small maintenance and repairs at user fee facilities, such as paint and generator fuel; occasionally on investments in larger infrastructure projects, such as construction of new walls, rooms, or walkways; and in the community, such as transportation for referral of poor patients to higher levels of care (Johns Hopkins University/Indian Institute for Health Management Research, 2008). Observed structural quality Among the 40 facilities (18 user fee-non-DHs; 5 user fee-DHs; 8 free services; and 9 controls) that were included in analysis of observed structural quality, facility structural quality of care increased significantly in all groups from baseline to follow-up (Figure 1), on average by 7.1 points. This increase was greatest in the free services group (+11.7 points) compared with the user fee-non-DH (+5.4), the user fee-DH (+5.2), and the control groups (+7.6), although the unadjusted difference was only significant compared with the user fee-non-DH group, p = 0.04. A series of progressively larger DD linear regression models was fit to examine the primary relationship of interest, the effect of study group on measured structural quality over

Note: Possible range for quality score: 0 to 31; each line represents one facility. 0=Baseline; 1=Follow-up. Source: Baseline and follow-up facility assessments.

Figure 1. Change in facility structural quality scores from baseline to follow-up, by study group. Note: Possible range for quality score: 0 to 31; each line represents one facility. 0 = baseline; 1 = follow-up. Source: Baseline and follow-up facility assessments Copyright © 2013 John Wiley & Sons, Ltd.


e288


time, controlling for province and the interaction of province and time (Table 2). Differences existed among the provinces’ facility quality scores at baseline, and their rates of improvement over time differed as well, with the average provincial increase in the structural quality index ranging from 3.3 points in one province (Wardak) to

Table 2. Regression coefficients from four linear regression DD models of facility structural quality Model 1

Model 2

Study Province* group*time time Constant Post Study group (reference = controls) User Fee-non-DH User Fee-DH Free Services User Fee-non-DH*Post User Fee-DH*Post Free Services*Post Province (reference = Balkh†) Kapisa† Parwan Wardak Samangan† Farah† Nimruz Saripol Panjshir Kapisa*Post Parwan*Post Wardak*Post Samangan*Post Farah*Post Nimruz*Post Saripol*Post Panjshir*Post R-squared Adj. R-squared N Linear combination of (Free Services*Post User Feenon-DH*Post)

19.4*** 7.6**

19.8*** 8.7***

0.3 2.6 2.4 2.2 2.4 4.2

0.39 0.33 80 6.3*

8.0** 0.8 3.8 2.5 3.9 0.3 3.7 0.7 4.7 6.3 12.0** 2.3 5.1 5.2 4.2 0.9 0.63 0.53 80 Not applicable

Model 3 Study group*time + province

Model 4{ Study group*time + province* time

20.4*** 7.6***

19.3*** 9.8*

0.1 2.7 2.1 2.2 2.4 4.2

0.0 3.2 0.1 1.9 3.5 0.3

5.7** 1.8 1.6 1.3 1.1 2.9 1.6 0.9

0.54 0.43 80 6.3*

8.0* 1.3 4.3 2.5 3.3 0.6 3.4 1.2 4.7 6.2 11.8** 2.3 4.3 4.5 3.5 0.6 0.66 0.52 80 2.2

Source: Baseline and follow-up facility assessments. Includes free services facilities (in addition to user fee and control); non-starred provinces contain only user fee and control facilities. { Final model, with robust standard errors, using the Huber-White Sandwich method (White, 1980). N = 40 facilities at baseline and 40 at follow-up. *p < 0.05, **p < 0.01, ***p < 0.001. †




e289

15.0 points in another (Farah). When the effect of province over time was included in the final model, the difference in improvement between free services and user fee facilities decreased from 6.3 to 2.2 and was no longer significant (Table 2). The variable for province best represents the effect of the managing service provider (national and international NGOs, or the MoPH for three pilot provinces) on service delivery quality in the province. Models of the direct difference (quality score at follow-up–quality score at baseline) were compared with the DD models and led to similar inferences as DD models; namely, that province was significantly related to the difference between follow-up and baseline, but the study group was not. For the structural and perceived quality of care outcomes, DD tobit models were also run, as the data were censored (at 31 and 32, respectively) at follow-up, because of a ceiling effect. These models yielded similar conclusions. Perceived quality of care Patients’ and households’ perceived quality of care. The mean of patients’ perceived quality scores (possible range: 8 to 32) was 26.2 at baseline and increased by 1.3 points on average to 27.4 at follow-up, but did not differ significantly by study group. A series of DD linear regression models indicated that the study group was not significantly related to changes in patients’ perceived quality over time and that province was the only variable significantly related to the change in quality (data available from authors). DD regressions using households’ perceived quality scores confirmed that the effect of province over time on perceived quality was much greater than that of study group. Perceived quality increased from baseline to follow-up by 1.7 and 2.3 points more in the non-DH user fee and the free services facility catchment areas, respectively, compared with the control facility catchment areas, adjusting for other covariates (p < 0.05 for both), but there was no significant difference between these two arms (non-DH user fee and free services) in terms of changes in perceived quality (Table 3). Results did not change substantively if only overlapping villages (those surveyed at both baseline and follow-up) were analyzed. Use of curative care Analysis of the HMIS data revealed a 102% increase in outpatient department (OPD) visits in the free services arm, compared with a 33% increase in control facilities and 22% in the primary care user fee facilities, comparing the average monthly number of visits 1 year before to 1 year after pilot implementation. Regression models indicated that the average monthly OPD visits at free services facilities increased significantly more than visits at control and user fee facilities after HFP implementation, p < 0.05 (Table 4). Interaction terms between baseline fees, study group, and time indicated that those facilities randomized to free services that had previously charged both service and medication fees (i.e., two facilities in Farah province) showed especially large increases from baseline to follow-up of between 1694 and 2286 additional OPD visits per month on average, compared with all types of user fee facilities and control facilities, p < 0.001 (Table 5). The DD regression model indicated that the increase in visits at user fee facilities with no previous fees was significantly smaller (on average by 592 fewer visits per Copyright © 2013 John Wiley & Sons, Ltd.


e290


Table 3. Linear regression models for OPD visits, by study group, over time (1) Basic Constant Post Study group User Fee-non-DH User Fee-DH Free Services User Fee-non-DH*Post User Fee-DH*Post Free Services*Post Baseline fee levels (reference = none) Service + Meds Fee Service Fee User Fee*Service + Meds Fee*Post User Fee*Service Fee*Post Free Services*Service + Meds Fee*Post Free Services*Service Fee *Post Adjusted R-squared N

(2) Final Model

851.2*** 284.6***

827.7*** 284.6***

144.4 1528.0*** 93.8 66.1 100.1 682.4**

143.7 1522.7*** 69.5 297.7* 100.1 688.2*

0.41 194

131.9* 123.5 592.1* 254.1 1300.6** 443.2 – 194

Source: HMIS data. Note: Up to 10 additional non-pilot BHCs/CHCs per province added to the control group to augment sample size. Final model uses Huber-White robust standard errors (White, 1980). N = 97 facilities at baseline (1 year pre-HFP) and 97 at follow-up (1 year post-HFP). *p < 0.05, **p < 0.01, ***p < 0.001.

month) compared with user fee facilities that previously charged service and medication fees, p = 0.03. User fee facilities with no previous fees also increased by significantly less (by 298 fewer visits per month, on average) compared with control facilities, p = 0.03, and to free services facilities that remained free (by 985 fewer visits per month), p = 0.007. Household survey results indicated that 19.1% of household members at baseline had an illness or injury in the previous 30 days that was serious enough to warrant care outside the home. Among those ill/injured, 84.1% overall sought care, and this did not vary significantly by study group or over time. On average, 49.3% of those seeking care at baseline went to the pilot facility first. All study groups experienced an increase from baseline to follow-up (17.3 percentage points overall) in the percentage of sick household members seeking care first at the pilot facility. This increase was largest in the free services group (27.5 percentage points), followed by the non-DH user fee facilities (20.0 percentage points), the DH user fee facilities (11.2 percentage points), and then the control facilities (7.8 percentage points; Table 4). However, despite having the largest increase from baseline to follow-up, the free services group still had the lowest proportion seeking care there first at follow-up, seven percentage points lower than control facilities and 13 percentage points lower than user fee-non-DH facilities. A logistic regression DD model indicated that the change in the odds of seeking care first at the free services group over time Copyright © 2013 John Wiley & Sons, Ltd.



*p < 0.05, **p < 0.01, ***p < 0.001.

Selling assets or land (when going to pilot facility) (%)

Selling assets or land (%)

Mean paid, SD, in Afs.

Median paid, in Afs.

Paying nothing (%)

Seeking care first at pilot facility (if seeking) (%)

n

Difference Baseline Follow-up Difference Baseline Follow-up Difference

Follow-up

Baseline Follow-up Difference Baseline Follow-up Difference Baseline Follow-up Difference Baseline

Baseline Follow-up 53.0 73.0 20.0 9.0 13.0 4.0 200 25 -175 399.0 672.1 264.4 680.9 -134.6 5.0 6.0 1.0 1.9 3.4 1.5

871 1089

UF-non-DH

50.3 61.5 11.2 12.2 6.9 -5.3 200 110 -90 463.4 744.3 324.1 551 -139.3 14.5 3.5 -11.0 9.4 2.3 -7.2

175 335

UF-DH

32.1 59.6 27.5 9.5 49.6 40.1 300 20 -280 443.5 536.6 370.9 771 -72.6 10.8 1.5 -9.3 4.4 0** -4.4

464 787

Free

Table 4. Care seeking patterns and payments among ill household members seeking care, by study group and time period

59.3 67.0 7.7 9.6 23.9 14.3 150 17 -133 414.4 693.6 222.8 554 -191.6 2.6 4.9 2.3 1.7 1.3 -0.4

464 758

Control

49.3 66.6 17.3 9.5 24.8 15.3 200 27 -173 418.7 654.4 288.7 665.5 -130.0 6.6 4.6 -2.0 2.9 2.0 -0.9

1974 2969

Total PRIMARY HEALTH-CARE SERVICES IN AFGHANISTAN

e291

Int J Health Plann Mgmt 2013; 28: e280–e297.

DOI: 10.1002/hpm

e292


Table 5. Coefficient values for linear DD regressions of households’ perceived quality (1) Without (2) Without province study group Constant Post Study group (reference = controls) User Fee-non-DH User Fee-DH Free Services User Fee-non-DH*Post User Fee-DH*Post Free Services*Post Self-Reported Average Walking Time (min.) Standardized Wealth Score Province (reference = Balkh) Kapisa Parwan Wardak Samangan Badghis Farah Nimruz Saripol Panjshir Kapisa*Post Parwan*Post Wardak*Post Samangan*Post Badghis*Post Farah*Post Nimruz*Post Saripol*Post Panjshir*Post Adjusted R-squared† N (households) Linear combination of (Free Services*Post User Fee-non-DH*Post)

25.88*** 1.18 0.72 1.39 1.31 2.09* 0.51 2.44 0.01 0.30*

0.03 3517 0.35

(3) Final model

26.17*** 2.76***

26.55*** 3.86***

0.02***

0.40 0.91 1.11 1.74* 1.08 2.28* 0.02***

0.14 7.51*** 0.28 0.15 2.03* 4.38** 4.76*** 0.46 1.84* 2.55*** 9.48*** 2.31 1.22 4.60*** 1.27 1.89* 3.93*** 3.86*** 2.61** 0.30 3517 Not applicable

0.13 7.37*** 0.49 0.03 1.96* 4.90** 5.18*** 0.63 2.00* 2.56** 9.22*** 2.73 1.57 4.55*** 2.24 1.21 4.28** 4.18*** 2.51* 0.31 3517 0.53

Source: Baseline and follow-up household surveys. Note: Standard errors account for complex survey design using Taylor Linearization Series. Models do not include highly outlying points (n = 24), diagnosed by high standardized residuals. † Adj. R-squared derived from models where standard errors do not account for complex survey design. *p < 0.05, **p < 0.01, ***p < 0.0011.

was not significantly different from that in the user fee or control groups (data available from authors). Out-of-pocket expenditures The median expenditure on health care declined in all groups from baseline to follow-up, on average by 173 Afghanis. Missing values for total expenditures were Copyright © 2013 John Wiley & Sons, Ltd.



e293

6.1% at baseline and 0.4% at follow-up, and these observations were dropped from analysis. Median expenditures declined the most in the free services group, by 280 Afghanis. Results were similar when analyzed for only those seeking care at the pilot facility, but median amounts in each group were lower and the percent paying nothing in the free services group at follow-up was 92.8% (data not shown). It is also notable that although selling assets or land to pay for health care was an uncommon occurrence (decreasing from 5.6% to 3.4% overall), there were no cases of this among the free care groups at follow-up, whereas 3.4% and 1.3% of those going to primary care user fee facilities and control facilities, respectively, reported selling assets or land to pay for their care.

DISCUSSION This pilot study is one of the few comprehensive studies that used a quasiexperimental design to assess how user fees affect quality and utilization of services, and one of the first rigorous studies to examine the impact of user fees in a postconflict setting. Contrary to what several previous studies have shown in other settings (Soucat et al., 1997; Audibert and Mathonnat, 2000), user fees had little effect on observed or perceived quality. In other settings where user fee introduction or increases have been linked to differential improvements in quality, the starting level of quality was also very low, and the health system had typically been in decline (Litvack and Bodart, 1993). Although the amount of revenue generated by user fees in the HFP was small in an absolute sense, the proportion of operating costs it was able to cover was comparable with the average of 5% that a review of national user fee programs found (Creese and Kutzin, 1995). In Afghanistan, it is likely that the potential benefits of user fees are negligible compared with other mechanisms to improve services and hold service providers accountable, particularly the annual dissemination and use of a balanced scorecard of health service performance (Hansen et al., 2008a, 2008b, 2008c). In the 10 provinces where the HFP study took place, one NGO (or the MoPH in three provinces) was responsible for service provision in each province, and performance monitoring using a balanced scorecard and follow-up of poor performance by the MoPH was carried out twice per year, resulting in substantial improvements in measured quality indicators (Hansen et al., 2008a, 2008b, 2008c). Unlike quality changes, where the managing service provider seemed to have a greater effect than the financing intervention, there was a differential increase in facility utilization according to fee structure. Making primary care services free was associated with large increases in the proportion of sick household members who sought care there first when ill. A facility’s starting point—in terms of prior fees charged—was important in determining subsequent changes in use with changes in fees. Facilities randomized to user fees that did not charge fees before the pilot experienced smaller increases in utilization compared with user fee facilities with both prior service and medication fees, and compared with facilities randomized to remain free. Additional sources of data (not analyzed in this study) from in-depth interviews and focus group discussions with facility staff and community leaders indicated that the primary reason for increased use when services are free is that they are more Copyright © 2013 John Wiley & Sons, Ltd.


e294


affordable. However, facility staff and community leaders also believed that community members tended to overuse free services even when they are not ill, placing undue burdens on facility staff, and they felt user fee revenues could provide an important source of discretionary income at the facility (Islamic Republic of Afghanistan. Ministry of Public Health, 2008; Arur et al., 2010). Nearly half the ill household members seeking care in free services catchment areas paid nothing at follow-up, compared with less than a quarter of households in other study group catchment areas. Households living near free services facilities experienced the largest decline in median amounts paid for care, from baseline to follow-up, compared with other catchment areas. None of the households going to free services facilities at follow-up had to sell assets or land to pay for care, whereas those in other study groups did, a coping strategy that can lead to medical poverty (McIntyre et al., 2006). Afghanistan still has high health-care needs and low levels of health services utilization, so it is important that use of health services be encouraged as the system— and trust in it—is rebuilt. Any barriers to health-care utilization, however small, should be removed, and continued attention should be given to ensuring that facility quality continues to improve. The balanced scorecard has helped shine a spotlight on performance in service provision and highlight key areas for quality improvement (Hansen et al., 2008a, 2008b, 2008c), whereas user fees have shown no demonstrable effect on observed or perceived quality. In May 2008, the Afghan MoPH made a policy decision to eliminate user fees at all BPHS facilities, based in part on the results of the HFP study (Ministry of Public Health. Islamic Republic of Afghanistan, 2008). Shortly thereafter, facilities ceased charging fees at all levels of the BPHS. In light of this policy decision, the MoPH should consider three major recommendations, based on experiences with removing fees in other countries (Gilson and McIntyre, 2005; Yates, 2009). First, there must be additional funding and flexibility provided to service providers, to cope with additional demand anticipated for health services. For example, this was successfully dealt with in Uganda through a one-time cash infusion and ongoing budget increases for health, following the abolition of user fees at public facilities in 2001 (Burnham et al., 2004; Nabyonga-Orem et al., 2008). Second, continued monitoring will help ensure that the policy of free services does not yield unintended negative consequences. Key indicators, such as utilization levels (for both preventive and curative care), and quality indicators, such as drug stock-outs, waiting times, and patient perceptions of quality, should continue to be carefully monitored. Continued feedback should be gathered from health workers to assess their experiences and address their concerns, which is especially important as many health workers in Afghanistan as well as elsewhere tend to view user fees favorably and may be dissatisfied with their removal (Burnham et al., 2004; Gilson and McIntyre, 2005). Finally, it is critically important that the MoPH and NGOs explore possibilities for providing discretionary income to facilities, for example, through dedicated funds to be co-managed by facility staff and community members. This would help facilities retain some of the potential benefits of user fees without incurring the potential negative effects. Copyright © 2013 John Wiley & Sons, Ltd.



e295

It is worth noting several potential limitations of this study. Certain outcomes, such as changes in the use of preventive care, could not be assessed, given large amounts of missing data in the HMIS. It also became clear that despite randomization of facilities to the study arms, there were differences between the study groups at baseline that were only known after the allocation procedure. For example, fewer households in the free services catchment areas sought care first at the pilot facility or perceived it to be affordable, compared with other study groups. In addition, the user fee intervention was implemented somewhat differently across facilities, with the actual fee levels varying (within suggested ranges) and significant variations in how the revenue was spent. However, additional analysis, not shown here, found no significant relationship between user fee levels, expenditures, or quality improvements. Finally, when the clustering of households within villages, and patients within facilities, was taken into account in the DD analysis, the power to detect small to moderate differences in changes among the study groups (e.g., differential increases in utilization) was relatively low. Despite these limitations, the quasi-experimental research design with preintervention and post-intervention data collection is a notable strength of this study in a field that has tended to rely on observational evidence (Palmer et al., 2004). The study used multiple sources of data, attempting to triangulate information from different sources and gain more insight into the pilots’ effects. It is also one of the first dedicated studies to examine the impact of user fees in a post-conflict setting, where residents are very poor and many are gaining access to health services for the first time, across a broad range of outcomes, including quality and utilization.

CONCLUSIONS In health systems that are rebuilding and rapidly changing in a post-conflict setting, it may be more difficult for user fees to produce beneficial effects than they have occasionally demonstrated in other developing country settings. In this study, fees had no demonstrable effect on facility structural quality or on patients’ and households’ overall perceptions of quality, in a context where quality was closely monitored and steadily improving. Even nominal user fees may compromise financial access to care for some of the population in a post-conflict setting. Although the barriers posed by fees may be small for most households, particularly relative to other factors, such as geographic access in mountainous parts of Afghanistan, it is especially important in a context of high health-care needs to facilitate access to the greatest extent possible.

ACKNOWLEDGEMENTS This study was funded by a contract with the Afghanistan Ministry of Public Health and the Johns Hopkins University Bloomberg School of Public Health, in collaboration with the Indian Institute of Health Management Research. The authors also express their appreciation for the financial support (grant no. H050474) provided by the UK Department for International Development (DFID) for the Future Health Copyright © 2013 John Wiley & Sons, Ltd.


e296


Systems research program consortium. This document is an output partly funded from a project financed by DFID for the benefit of developing countries. The views expressed are not necessarily those of DFID. The authors declare they have no competing interests.

REFERENCES Akin JS, de Ferranti D. 1987. Financing Health Services in Developing Countries: an Agenda for Reform. T. W. Bank: Washington, DC, The World Bank. Arur A, Peters D, Hansen P, Mashkoor MA, Steinhardt LC, Burnham G. 2010. Contracting for health and curative care use in Afghanistan between 2004 and 2005. Health Policy Plan 25(2): 135–144. Audibert M, Mathonnat J. 2000. Cost recovery in Mauritania: initial lessons. Health Policy Plan 15(1): 66–75. Buckley J, Shang Y. 2003. Estimating policy and program effects with observational data: the “differences-indifferences” estimator. Pract Assess, Res Eval 8(24). Burnham GM, Pariyo G, et al. 2004. Discontinuation of cost sharing in Uganda. Bull World Health Organ 82 (3): 187–195. Creese A, Kutzin J. 1995. Lessons from Cost-Recovery in Health. Forum on Health Sector Reform. WHO/SHS/ NHP/95.5, World Health Organization. Diop F, Yazbeck A, Bitrán R. 1995. The impact of alternative cost recovery schemes on access and equity in Niger. Health Policy Plan 10(3): 223–240. Gilson L, Kalyalya D, Kuchler F, Lake S, Oranga H, Ouendo M. 2001. Strategies for promoting equity: experience with community financing in three African countries. Health Policy 58(1): 37–67. Gilson L, McIntyre D. 2005. Removing user fees for primary care in Africa: the need for careful action. BMJ 331(7519): 762–765. Government of the Islamic Republic of Afghanistan Ministry of Public Health. 2011. National Health Accounts Afghanistan 2008-2009: Preliminary Findings, USAID. Hansen PM, Peters DH, Edward A, et al. 2008a. Determinants of primary care service quality in Afghanistan. Int J Qual Health Care 20(6): 375–383. Hansen PM, Peters DH, Edward A, Niayesh H, Singh LP, Dwivedi V, Burnham G. 2008b. Measuring and managing progress in the establishment of basic health services: the Afghanistan health sector balanced scorecard. Int J Health Plann Manage 23(2): 107–117. Hansen PM, Peters DH, Viswanathan K, Rao KD, Mashkoor A, Burnham G. 2008c. Client perceptions of the quality of primary care services in Afghanistan. Int J Qual Health Care 20(6): 384–391. Humanitarian Practice Network. 2004. Cost-recovery in the health sector—continuing the debate. Retrieved


11 August, 2007, from http://www.odihpn.org/report. asp?id=2677. Islamic Republic of Afghanistan. Ministry of Public Health. 2007. An experiment with community health funds in Afghanistan, Johns Hopkins University & Indian Institute of Health Management Research. Islamic Republic of Afghanistan. Ministry of Public Health. 2008. Final evaluation on health financing pilots: effects of user fees versus free services on primary care in Afghanistan. Kabul, Johns Hopkins University/ Indian Institute of Health Management Research,. Johns Hopkins University/Indian Institute for Health Management Research. 2008. Final evaluation report on the health financing pilots: the effects of user fees vs. free services on primary care in Afghanistan. Kabul, Ministry of Public Health. Lagarde M, Palmer N. 2008. The impact of user fees on health service utilization in low- and middle-income countries: how strong is the evidence? Bull World Health Organ 86(11): 839–848. Lehtonen R, Pahkinen EJ. 1995. Practical Methods for Design and Analysis of Complex Surveys. John Wiley & Sons: West Sussex, England. Litvack JI, Bodart C. 1993. User fees plus quality equals improved access to health care: results of a field experiment in Cameroon. Soc Sci Med 37(3): 369–383. Loevinsohn B, Sayed GD. 2008. Lessons from the health sector in Afghanistan: how progress can be made in challenging circumstances. JAMA 300(6): 724–726. McIntyre D, Thiede M, Dahlgren G, Whitehead M. 2006. What are the economic consequences for households of illness and of paying for health care in low- and middle-income country contexts? Soc Sci Med 62(4): 858–865. Meessen B, Van Damme W, Tashobya CK, Tibouti A. 2006. Poverty and user fees for public health care in low-income countries: lessons from Uganda and Cambodia. Lancet 368(9554): 2253–2257. Ministry of Public Health. Islamic Republic of Afghanistan. 2008. User fee ban at BPHS facilities. Ministry of Public Health. Nabyonga-Orem J, Karamagi H, Atuyambe L, Bagenda F, Okuonzi SA, Walker O. 2008. Maintaining quality of health services after abolition of user fees: a Uganda case study. BMC Health Serv Res 8: 102.


PRIMARY HEALTH-CARE SERVICES IN AFGHANISTAN Nabyonga J, Desmet M, Karamagi H, Kadama PY, Omaswa FG, Walker O. 2005. Abolition of costsharing is pro-poor: evidence from Uganda. Health Policy Plan 20(2): 100–108. Nyonator F, Kutzin J. 1999. Health for some? The effects of user fees in the Volta Region of Ghana. Health Policy Plan 14(4): 329–341. Palmer N, Mueller DH, Gilson L, Mills A, Haines A. 2004. Health financing to promote access in low income settings—how much do we know? Lancet 364(9442): 1365–1370. Peters DH, Garg A, Bloom G, Walker DG, Brieger WR, Rahman MH. 2007a. Poverty and access to health care in developing countries Int J Equity Health 1136: 161–171. Peters DH, Noor AA, Singh LP, Kakar FK, Hansen PM, Burnham G. 2007b. A balanced scorecard for health services in Afghanistan. Bull World Health Organ 85(2): 146–151. Rao KD, Peters DH. 2007. Quality improvement and its impact on the use and equality of outpatient health services in India. Health Econ 16(8): 799–813. Rao KD, Waters H, Steinhardt L, Alam S, Hansen P, Naeem AJ. 2009. An experiment with community health funds in Afghanistan. Health Policy Plan 24(4): 301–311. Ridde V. 2003. Fees-for-services, cost recovery, and equity in a district of Burkina Faso operating the Bamako Initiative. Bull World Health Organ 81(7): 532–538.


e297

Sabri B, Siddiqi S, Ahmed AM, Kakar FK, Perrot J. 2007. Towards sustainable delivery of health services in Afghanistan: options for the future. Bull World Health Organ 85(9): 712–718. Soucat A, Levy-Bruhl D, Gbedonou P, et al. 1997. Local cost sharing in Bamako Initiative systems in Benin and Guinea: assuring the financial viability of primary health care. Int J Health Plann Manage 12(Suppl 1): S109–135. StataCorp. 2007. Stata Statistical Software: Release 10. StataCorp LP: College Station, TX. Steinhardt LC, Peters DH. 2010. Targeting accuracy and impact of a community-identified waiver card scheme for primary care user fees in Afghanistan. Int J Equity Health 9: 28. Turner AG, Magnani RJ, Shuaib M. 1996. A not quite as quick but much cleaner alternative to the Expanded Programme on Immunization (EPI) Cluster Survey design. Int J Epidemiol 25(1): 198–203. White H. 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48(4): 817–838. Wilkinson D, Gouws E, Sach M, Karim SS. 2001. Effect of removing user fees on attendance for curative and preventive primary health care services in rural South Africa. Bull World Health Organ 79(7): 665–671. Yates R. 2009. Universal health care and the removal of user fees. Lancet 373(9680): 2078–2081.


The impact of user fees on health services utilization and infectious disease diagnoses in Neno District, Malawi: a longitudinal, quasi-experimental study.

Erratum to: The impact of user fees on health services utilization and infectious disease diagnoses in Neno District, Malawi: a longitudinal, quasi-experimental study.

Myth: user fees ensure better use of health services.

Perspectives on reproductive healthcare delivered through a basic package of health services in Afghanistan: a qualitative study.

Factors affecting maternal healthcare utilization in Afghanistan: secondary analysis of Afghanistan Health Survey 2012.

A study of consumer attitudes about health care: the control, cost, and financing of health services.

Tailored e-Health services for the dementia care setting: a pilot study of 'eHealthMonitor'.

Health services utilization of patients with vertigo in primary care: a retrospective cohort study.

Erratum to: 'Setting performance-based financing in the health sector agenda: a case study in Cameroon'.

EARLY: a pilot study on early diagnosis of atrial fibrillation in a primary healthcare centre.

Assessing the elimination of user fees for delivery services in Laos.

The effect of a virtual ward program on emergency services utilization and quality of life in frail elderly patients after discharge: a pilot study.

User fees advance in Congress.

Curing over-use by prescribing fees: an evaluation of the effect of user fees' implementation on healthcare use in the Czech Republic.

Capitation-Based Financing Hampers the Provision of Preventive Services in Primary Health Care.

Singapore's regional health systems-a data-driven perspective on frequent admitters and cross utilization of healthcare services in three systems.

The effects of a people-centred model on longitudinality of care and utilization pattern of healthcare services--Brazilian evidence.

Preserving and financing ambulatory health services.

The evaluation of sexual health services in a medical setting.

A qualitative study of staff perspectives of patient non-attendance in a regional primary healthcare setting.

Electronic health records and information portability: a pilot study in a rural primary healthcare center in India.

Biospecimen User Fees: Global Feedback on a Calculator Tool.

Utilization of healthcare services and renewal of health insurance membership: evidence of adverse selection in Ghana.

Feasibility and acceptability of the informant AD8 for cognitive screening in primary healthcare: a pilot study.