OUP user menu

Determinants of primary care service quality in Afghanistan

Peter Meredith Hansen, David H. Peters, Anbrasi Edward, Shivam Gupta, Aneesa Arur, Haseebullah Niayesh, Gilbert Burnham
DOI: http://dx.doi.org/10.1093/intqhc/mzn039 375-383 First published online: 17 September 2008

Abstract

Objective To identify factors associated with service quality provided by agencies implementing a basic package of health services in Afghanistan.

Design Cross-sectional survey of outpatient health facilities, health workers, patients and caretakers.

Setting Primary health care facilities in every province of Afghanistan.

Main outcome measures Composite scale measuring the quality of clinical processes in four areas: patient histories, physical examinations, communication and time spent with patient.

Results No difference in service quality was observed between male and female providers or between male and female patients, but when both the provider and patient were female quality was much higher. Overall, the quality of care at non-governmental organization and government-managed health facilities did not differ, but the poor received higher quality care at non-governmental facilities than at government facilities. Doctors provided higher quality care than lower level providers. Provision of six or more supervisory visits in the last 6 months was associated with higher service quality. Training doctors in integrated management of childhood illness was not associated with quality, but when lower level health workers received such training the quality of patient–provider communication was higher. Other recurrent inputs and geographic remoteness are not associated with the quality of care provided.

Conclusions The government's strategy to form partnerships with non-governmental organizations has led to higher quality of care provided to the poor. This represents a promising start in the reconstruction of Afghanistan's health system and provides useful evidence to other countries striving to increase access to quality care for the poor.

Keywords
  • quality of care
  • Afghanistan
  • equity
  • non-governmental organizations
  • supervision
  • integrated management of childhood illness

Introduction

Afghanistan is a poor country recovering from decades of strife. When the transitional government was established in 2002, Afghanistan had a decimated health infrastructure and very low levels of access to health services [1]. The infant mortality rate, <5 years of age mortality rate, maternal mortality ratio and total fertility rate in Afghanistan are among the highest in the world [2].

In order to achieve rapid expansion of the geographic scope of basic health services in the reconstruction period, the government of Afghanistan defined a role for the Ministry of Public Health centered largely around the stewardship function and contracted local and international non-governmental organizations to deliver a basic package of health services in most areas of the country, with the government acting as direct provider of services in some areas [3].

The geographic scope of health services increased rapidly starting in 2003, with contracted agencies able to establish service delivery systems within a relatively short timeframe [4]. By the middle of 2004, administrative and financial arrangements to deliver the basic package were extended to districts in which 77% of the population of Afghanistan resides [4].

Evidence from other developing countries shows that the benefits of health funding tend to accrue to urban and relatively wealthy populations, while poor and remote populations tend to have lower access to quality health services [5, 6]. Identification of variations in service quality enables managers and policymakers to identify ways to improve service quality and monitor whether current strategies are meeting stated goals for a pro-poor health system that meets the needs of women and children and remote populations.

The objectives of this study are to describe the level of quality of care provided by agencies implementing basic health services in Afghanistan and identify factors associated with variations in quality.

Methods

This study utilizes data derived from an assessment of health services conducted in Afghanistan in 2004. The assessment included a random sample of up to 25 health facilities implementing a basic package of health services in each province, stratified by health facility type: Basic Health Centers, Comprehensive Health Centers and District Hospitals. Each has different staffing levels and provides different sets of services [3].

In each sampled facility, a supervisor implemented a facility assessment instrument that measures infrastructure, staffing, service capacity, management processes and availability of equipment, drugs and supplies. Survey teams interviewed up to four health workers and conducted five direct observations of patient–provider interactions involving patients <5 years of age, and five involving patients aged ≥5 years. An exit interview was conducted with each patient whose consultation was observed, or the caretaker if the patient was a child. The final sample included 617 health facilities in all 33 provinces, 1553 health workers, 5719 direct observations of patient–provider interactions and 5597 exit interviews.

Outcome variable: quality of care

A composite scale of quality of care was developed using four measures from the clinical consultation: (i) patient histories, (ii) physical examinations, (iii) communication and (iv) time spent with patients. These measures—which are based on indicators included in the Ministry of Public Health's routine monitoring system, the Afghanistan Health Sector Balanced Scorecard—were developed through formative research and consultative processes involving service providers, officials from the Ministry of Public Health and content experts from the government and technical and donor agencies [7]. The Ministry of Public Health uses these and other indicators on the Balanced Scorecard to clarify strategy, facilitate decision-making and monitor progress in implementation of services and achievement of desired outcomes [7, 8].

The patient history and physical examination index includes eights items, while the communication index contains nine items (Table 1). Both indices have a range from 0 to 100 and equal weighting for individual items. Time spent with patients is a dichotomous variable, indicating whether the provider spent at least 10 min with the patient, a target established by the Ministry of Public Health and incorporated into its routine monitoring system [7].

View this table:
Table 1

Elements of quality of care measured

Element of qualityScoring scheme
(1a) Patient history and physical examination (for children <5 years of age)Health worker completes what proportion of the following steps:
-greets the patient or caretaker at beginning of consultation
-asks about nature of complaint
-asks duration of complaint
-checks patient's age
-weighs child
-asks whether child is able to drink or breastfeed
-checks child's palms
-check's the child's immunization card
(1b) Patient history and physical examination (for patients ≥5 years of age)Health worker completes what proportion of the following steps:
-greets the patient or caretaker at beginning of consultation
-asks patient's age
-asks about nature of complaint
-asks duration of complaint
-asks about other complaints
-asks about previous treatment for same condition before coming to the facility
-examines some part of the patient's body, either by close inspection or contact
-door was closed or screen was drawn to ensure privacy
(2) CommunicationHealth worker completes what proportion of the following steps:
-tells patient or caretaker name of disease
-explains causes and course of disease
-explains what precautions or home nursing care to take
-tells patient or caretaker name of medicine, if medicine given
-tells patient or caretaker how medicine should be taken, if medicine given
-explains what adverse reactions might be expected and what to do about them, if medicine given
-tells patient or caretaker what signs or symptoms should prompt return to clinic
-tells patient when to return for check-up or go to another facility
-asks if patient or caretaker has any questions
(3) Time spent with patientHealth worker spends at least 10 min with patient

The composite scale of quality, which has a range from 0 to 100, was constructed by averaging the scores on the patient history and physical examination index, the communication index and time spent with patient, with the latter scored as 100 if the provider spent ≥10 min with the patient and 0 if not.

Independent variables

Regional characteristics

A measure of remoteness was included to assess whether the level of physical accessibility of a facility affects the ability of an agency to deliver high quality health services, and whether populations in remote areas have equal access to high quality care compared with populations living in less remote areas. This variable is a measure of the time required to travel by motor vehicle (or horse where no roads were available) from the provincial capital to the center of the district in which the facility is located. Remoteness may affect the ability of an agency to deliver high quality services by making it difficult to recruit and retain skilled staff, maintain equipment and stocks of drugs and supplies, make salary payments and provide supervision and training.

Province was included in each model as a control variable.

Health service characteristics

Type of implementing agency is a dichotomous variable that measures whether the Ministry of Public Health or a non-governmental organization is the direct provider of services. Given the heavy emphasis on use of non-governmental organizations to rapidly expand service delivery in Afghanistan, it is important to assess the level of quality at non-governmental facilities, relative to government facilities. Facility type is included to determine whether service quality varies by level of the primary care system.

Four types of recurrent inputs—provision of in-service training, timely salary payments, supervisory visits and training in integrated management of childhood illness—are included to determine whether increasing inputs leads to higher quality of care.

Two health worker characteristics are included to determine whether quality of care varies by health worker type and sex. Health workers were divided into two groups: doctors and others, a residual category that includes assistant doctors, nurses and a small number of other health workers.

Patient characteristics

Two patient characteristics are included to assess whether poor and female patients receive equitable levels of service quality, compared with non-poor and male patients, respectively. An asset index was developed to measure household wealth status, using household assets and dwelling characteristics recorded during exit interviews [9]. A dichotomous variable was created, with the poorest quintile from the national distribution of scores on the asset index defined as poor patients.

Data analysis

Statistical analysis was conducted using Stata 9. The clustering of observations at the facility level was accounted for through Taylor linearized variance estimation [10]. Multiple linear regression models were fit using ordinary least squares to compare differences in levels of quality across groups. Model fit was evaluated using summary measures of fit and examining residual plots. Multicollinearity was assessed by estimating the Variance Inflation Factor. All variables had <2% missing values, except for two items in the asset index that had 2–4% missing values. To assess whether the determinants of quality vary by patient age, age-stratified models were estimated. Interaction terms were included in the regression models and results for statistically significant interaction terms are presented through cross-tabulations to facilitate interpretation.

Results

Bivariable results are summarized in Table 2 and multivariable results are summarized in Table 3. The strongest association with quality identified involved patient and provider sex. Holding other variables constant, no difference in service quality was observed between male and female providers or between male and female patients, but when both the provider and patient are female service quality is much higher (9.47, P < 0.001).

View this table:
Table 2

Bivariable results for dimensions of quality of care

(1a) Patient history and physical examination (<5 years)(1b) Patient history and physical examination (≥5 years)(2) Communication(3) Time spent with patient
NMean score (0–100)P-valueNMean score (0–100)P-valueNMean score (0–100)P-valueN% >9 minP-value
Facility type
 Basic health center123746.80.58150467.80.63227336.90.35283415.50.18
 Comprehensive health center97448.1117269.2220034.2225518.5
 District hospital19747.725669.949436.450012.8
Implementing agency
 Ministry of Public Health43247.40.9753667.70.7498535.70.86101617.80.59
 Non-governmental organization193747.3234668.4439035.4447416.4
Travel time
  <3 h190946.70.06231567.50.01432135.70.73441016.60.92
 3–6 h37250.245773.484335.387015.7
  >6 h12749.315070.728238.128917.8
Health worker type
 Other80445.80.0597565.7<0.01167435.60.90174718.80.13
 Doctor160148.2167370.2315835.8318915.2
Health worker sex
 Male221947.40.92252868.20.83485735.70.44497816.40.49
 Female13447.139268.753934.153818.4
Patient's age
  <5 years240847.4229936.00.56233918.0<0.01
  ≥5 years265368.5253935.5260515.1
Patient sex
 Male124047.30.7498268.10.68209935.00.15217416.40.98
 Female114547.5164968.6270436.1273416.3
Wealth status
 In least poor quintiles163347.90.44187868.00.02338036.00.45343316.50.84
 In poorest quintile41347.045871.482335.085216.2
Salary payments are current
 No104646.70.29115968.60.93211238.10.02216015.10.28
 Yes136247.9149468.5272634.0278417.5
Supervision frequency
  <6 Visits in 6 months91148.20.28101467.10.15181435.30.67188515.60.52
  ≥6 Visits in 6 months149746.9163969.4302436.1305917.0
Supervision book
 Recommendations not written97447.90.45107368.80.75193437.50.10200317.90.31
 Recommendations written143447.0158068.3290434.6294115.5
In-service training received
 No110047.30.92122667.90.45223936.80.30227017.80.28
 Yes130847.4142769.1259934.9267415.4
IMCI training received
 No152647.00.35168267.70.17306933.8<0.01317514.80.05
 Yes88248.197170.0176939.2178719.4
  • Note: P-values are for comparison of quality scores for categories of each characteristic using analysis of variance; IMCI, integrated management of childhood illness.

View this table:
Table 3

Linear regression coefficients for quality of care

Parameters (reference category)All clientsClients <5 yearsClients ≥5 years
Facility characteristics
 Facility type
  Comprehensive health center (Basic health center)−0.17−0.08−0.30
  District hospital (Basic health center)−1.06−2.52−0.22
Non-governmental organization facility (Ministry of Public Health)0.442.64−0.85
 Travel time from provincial capital to district of facility
  3–6 h (<3 h)1.252.610.65
   >6 h (<3 h)0.290.52−0.90
 Health worker is doctor (other)2.97*2.343.77**
 Health worker is female (male)−3.60−1.78−6.44
 Salary payments are current (not current)0.670.551.03
 ≥6 Supervision visits in 6 months (<6 visits)2.49*0.893.78**
 Recommendations from supervision written in book (not written)−0.83−1.06−0.68
 Health worker received in-service training in last 3 months (not received)−0.83−0.64−1.48
  ≥1 Health worker in facility received training in integrated management of childhood illness (none received)2.69
Patient characteristics
 Sex is female (male)0.20−1.240.57
 Wealth status is poor (non-poor)−4.80*−3.74−4.38
Interaction terms
 Both health worker and patient are female (one or both are male)9.47***4.0310.87*
 Non-governmental organization facility and patient is poor (government facility or patient non-poor)6.14**3.057.54*
  • Note: *P < 0.05; **P < 0.01; ***P < 0.001.

Holding other variables constant, poor patients receive lower service quality than non-poor patients (−4.80, P < 0.05). The effect of wealth status on service quality is, however, modified by type of implementing agency. An interaction term measuring the combined effect of type of implementing agency and household wealth status is statistically significant, with quality of care higher when the implementing agency is non-governmental and the patient's household is in the poorest quintile (6.14, P < 0.01).

Table 4 shows further breakdowns of differences in service quality received by the poor and non-poor. For patient history and physical examinations of patients aged ≥5 years, the poor receive a higher level of service quality than the non-poor in non-governmental facilities (73.2 vs. 67.6, P < 0.01), while there is no difference in government facilities. The quality of provider–patient communication was lower for the poor than the non-poor in government-managed facilities (28.5 vs. 37.8, P < 0.01), while there is no difference in non-governmental facilities.

View this table:
Table 4

Quality scores by type of implementing agency and household wealth status

Facilities managed by governmentFacilities managed by non-governmental organizations
Non-poorPoorP-valueNon-poorPoorP-value
Patient history and physical examinations (<5 years)49.044.30.1047.547.60.99
Patient history and physical examinations (≥5 years)68.963.00.0967.673.2<0.01
Communication37.828.5<0.0135.236.40.40
Time spent with patients16.917.20.9416.416.00.84
Index of quality clinical processes (patients <5 years)48.947.20.2148.548.50.94
Index of quality clinical processes (patients ≥5 years)52.349.70.1151.052.20.11
Index of quality clinical processes (all patients' ages)50.748.40.0449.850.50.25
  • Note: Statistically significant associations are italicized.

The level of quality is higher when the provider is a doctor, compared with when the provider is not a doctor (2.97, P < 0.05). Higher frequency of supervision visits is associated with higher service quality (2.49, P < 0.05), but having written recommendations provided from the last supervision visit is not.

Remoteness, facility type, provision of timely salary payments and in-service training were found not to be associated with quality.

Results from the under-five model show that a term measuring receipt of integrated management of childhood illness training is not statistically significant. Further breakdowns of results related to integrated management of childhood illness in Table 5 show that receipt of such training by lower level health workers, but not doctors, is associated with higher quality in communication (34.8 vs. 48.6, P < 0.01). Cross-tabulations by patient wealth status show that among consultations involving poor patients, the provider spent at least 10 min with the patient in 24.7% of cases where integrated management of childhood illness training was provided, compared with 12.8% of cases where such training was not provided (P = 0.02). Among consultations involving patients not in the poorest quintile, there was no observed difference in time spent per patient by receipt of training.

View this table:
Table 5

Quality scores of patients <5 years of age by receipt of training in integrated management of childhood illness, by provider type

≥1 Doctor at facility received training≥1 Assistant doctor or nurse at facility received training
NoYesP-valueNoYesP-value
Patient history and physical examination47.446.30.5947.249.60.29
Communication36.234.10.4934.848.6<0.01
Time spent with patient >9 min18.512.90.1617.424.10.20
  • Note: Statistically significant associations are italicized.

The R2 was 0.26 for the all clients model, 0.29 for the under-five model and 0.27 for the five-and-older model.

Discussion

Results show substantial variation in the level of quality of care provided by primary care facilities in Afghanistan, and room for improvement is observed in all areas, especially patient histories and physical examinations of patients <5 years of age, communication and time spent with patients.

The lack of female providers has long been seen as an access and comfort issue for female patients in Afghanistan [10] – this study provides evidence that the lack of female providers is also a quality of care issue for female patients. Existing literature provides little evidence on the effect of patient and provider sex on quality of care. One study from Paraguay found that female doctors provide higher quality care than male doctors [11], but no study identified in the literature based on observed measures of quality investigated the interaction between patient and provider sex. The relationship between patient and provider sex and service quality is likely to be highly context specific. Female providers in Afghanistan may be more sensitive to the needs of female patients, and adult females can interact freely with each other, whereas sex discordant adults face constraints in their interactions that may hinder the ability of a provider to deliver high quality care.

Results from this study showing that non-governmental implementers—but not governmental implementers—provide equitable quality of care to the poor, and provide evidence on an important issue not frequently addressed in the literature. A study from Mexico based on respondents' self-reports of antenatal procedures completed by the provider found that service quality in government facilities is lower in poor areas than in wealthier areas [12]. In Delhi, India, private doctors in wealthy neighborhoods were found to be more competent than doctors in poor neighborhoods [13] and in Tanzania the same pattern was found for public sector doctors [5]. In Indonesia [14] and Paraguay [15], no difference was found in the level of public sector service quality between rich and poor areas. No study was identified in the literature that measures equity in service quality received by the poor in non-profit non-governmental facilities compared with traditional public sector services. In addition, no study was identified that uses direct observation of patient–provider interactions to examine the effect of wealth status on the level of service quality received.

The mechanism explaining the higher performance of non-governmental facilities compared with government facilities in delivering quality care to the poor is not clear. Non-governmental organizations holding contracts that are closely monitored for quality and equity [7, 8], and which in some cases contain performance bonuses, may be more highly sensitized and motivated to provide quality services to the poor. Further investigations are needed to explain the mechanisms behind this finding and to determine whether the provision of equitable levels of service quality to poor and remote populations persists over time.

Doctors appear to provide higher quality of care than other health workers—a finding consistent with studies from other settings [12, 13]—but Afghanistan cannot afford to train only doctors and most female doctors are unlikely to accept rural postings. Provision of low quality clinical services is not necessarily due to a lack of knowledge – many providers do not put into practice knowledge they already possess [5, 11]. Interventions that involve training other types of health workers and increasing provider application of knowledge need to be tested. The biggest changes in service quality may come from training lower level providers rather than those that already have years of training and altering incentives for health workers to expend greater effort. Studies from other settings have demonstrated that lower level health workers can achieve a large impact in improving health when the approach involves families and communities, sufficient training and support and appropriate referrals to higher levels [16, 17].

The only recurrent input found to be associated with quality in the all-ages model was frequency of supervision. In line with previous observations [18], increasing inputs is not tantamount to improving quality of care – improving clinical processes clearly requires more than provision of salary payments, supervision and training.

Observations from primary care settings in other countries indicate that supervisors tend to give little attention for improving the quality of clinical services and most supervisors lack training in supervision and well developed skills in observing health worker performance in providing care [19]. Few studies have systematically examined the quality of supervisory visits and the distribution of time devoted to different aspects of service delivery [20]. Supervisors in Zimbabwe were observed to spend <5% of their time on patient care issues [20]. Slightly more than half of health workers interviewed as part of the current study (54.7%) reported that the supervisor observed service provision during the last supervisory visit, and slightly more than one quarter (26.7%) reported that the supervisor shared technical information.

Use of a structured supervisory checklist was found to be associated with higher quality care in the Philippines, and the frequency of supervision visits was associated with higher health worker performance in the group using the structured checklist but not in the control group [21]. A similar tool was developed and adopted by the Ministry of Public Health for use in facilities across Afghanistan in 2006, but no such tool was available to supervisors at a national level prior to data collection for this study in 2004.

Previous studies from other settings have found that provision of integrated management of childhood illness training is associated with higher quality care for patients <5 years of age [2224]. Most previous studies have, however, been conducted on a small-scale demonstration basis; this association may be harder to demonstrate under conditions of rapid expansion of services at a national level [25]. Results from this study indicate that providing integrated management of childhood illness training to lower level health workers can lead to measurable improvements in quality of communication at the level of a national health system, even under conditions of rapid expansion of health services in a post-conflict setting.

The existing literature contains little evidence on the differential benefits of integrated management of childhood illness training by patient wealth status. An assessment in rural Tanzania found that inequalities decreased for some but not all health outcomes in areas where integrated management of childhood illness was implemented [26]. The current study demonstrates that provision of such training is associated with more time spent per patient for patients in the poorest quintile, but not for patients from less-poor households. This indicates that the poor may benefit the most from training health workers in integrated management of childhood illness. The reason for this is not clear, and should be investigated further. One possible explanation is that integrated management of childhood illness may be particularly effective in targeting the burden of disease afflicting the poor, at least in this study setting.

A major limitation of this study is the lack of empirical assessment of validity and reliability of the measures. The measures used in this study were developed through extensive consultative and formative research processes, with involvement of national and international technical experts and managers of primary care services in Afghanistan, and have been found to be useful indicators as part of a Balanced Scorecard approach implemented by the Ministry of Public Health [8]. High face validity is, however, no guarantee of construct or content validity or reliability [27].

This study was conducted early in the period of reconstruction of Afghanistan's health system. The finding that the poor and non-poor receive equal levels of service quality from non-governmental implementers and that lower level and remote facilities provide the same level of service quality as higher level and less remote facilities indicates that the government's strategy of partnering with non-governmental organizations to implement a basic package of health services has led to higher quality care for the poor. Further research is required to identify the mechanisms responsible for the associations described in this study, but this represents a promising start in the reconstruction of Afghanistan's health system and provides useful evidence to other countries striving to increase access to quality care for the poor.

Funding

This study was funded by a contract between the Afghanistan Ministry of Public Health and the Johns Hopkins Bloomberg School of Public Health, in collaboration with the Indian Institute of Health Management Research.

Acknowledgements

The contributions of the Johns Hopkins University and Indian Institute of Health Management Research Third Party Evaluation team and the Ministry of Public Health's Monitoring and Evaluation Department are gratefully acknowledged, along with the helpful comments from the editors, the anonymous reviewers, Dr. Cesar Victora and Dr. Philippe Bonhoure.

References

View Abstract