Towards patient-centered health services in India a scale to measure patient perceptions of quality

Transcription

1 International Journal for Quality in Health Care 2006; Volume 18, Number 6: pp Advance Access Publication: 29 September /intqhc/mzl049 Towards patient-centered health services in India a scale to measure patient perceptions of quality KRISHNA DIPANKAR RAO 1, DAVID H. PETERS 2 AND KAREN BANDEEN-ROCHE 3 1 Johns Hopkins University, Department of International Health, New Delhi, India, 2 Johns Hopkins Bloomberg School of Public Health, and 3 Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA Abstract Objectives. (i) To develop a reliable and valid scale to measure in-patient and outpatient perceptions of quality in India and (ii) to identify aspects of perceived quality which have large effects on patient satisfaction. Design. Cross-sectional survey of health facilities and patients at clinics. Setting. Primary health centers, community health centers, district hospitals, and female district hospitals in the state of Uttar Pradesh in north India. Main outcome measures. Internal consistency, validity, and factor structure of the scale are evaluated. The association between patient satisfaction and perceived quality dimensions is examined. Results. A 16-item scale having good reliability and validity is developed. Five dimensions of perceived quality are identified medicine availability, medical information, staff behavior, doctor behavior, and hospital infrastructure. Patient perceptions of quality at public health facilities are slightly better than neutral. Multivariate regression analysis results indicate that for outpatients, doctor behavior has the largest effect on general patient satisfaction followed by medicine availability, hospital infrastructure, staff behavior, and medical information. For in-patients, staff behavior has the largest effect followed by doctor behavior, medicine availability, medical information, and hospital infrastructure. Conclusions. The scale developed can be used to measure perceived quality at a range of facility types for outpatients and inpatients. Perceived quality at public facilities is only marginally favorable, leaving much scope for improvement. Better staff and physician interpersonal skills, facility infrastructure, and availability of drugs have the largest effect in improving patient satisfaction at public health facilities. Keywords: India, patient perceptions, patient satisfaction, quality of health care A critical challenge for health services in developing countries is to find ways to make them more client-oriented. Indifferent treatment of patients, unofficial payments to providers, lack of patient privacy, and inadequate provision of medicines and supplies are common, yet are rarely acknowledged by traditional quality assessment methods. Assessing patient perspectives gives users a voice, which, if given systematic attention, offers the potential to make services more responsive to people s needs and expectations, important elements of making health systems more effective [1]. Studies have shown that health care utilization, a long-standing concern for many developing countries, is sensitive to user perceptions of quality [2 7]. For these reasons, patient perceptions of health services are now an important part of quality assessment in health care. The few studies on user perceptions conducted in developing countries have shown that patients are able to evaluate structural, process, and outcome measures of quality [2,8,9]. Surprisingly, little research has been done on patient perceptions of quality in India, most of which has been confined to the family planning field [10]. Whereas some local studies have attempted to measure patient perceptions of health care quality, the validity and reliability of these scales are unknown. The applicability of these instruments to the Indian context is questionable given that they were developed in other cultural contexts [2]. The first objective of this study is to develop an instrument to measure patient perceptions of health care quality in India and evaluate its reliability, validity, and dimensionality. The second objective is to identify aspects of perceived quality which have large effects on patient satisfaction. The study was conducted at district hospitals, community health centers, and primary health centers in the north Indian state of Uttar Pradesh. Address reprint requests to Krishna Dipankar Rao, Johns Hopkins University, Department of International Health, E-13/6 Vasant Vihar, New Delhi-57, India. krao@jhsph.edu International Journal for Quality in Health Care vol. 18 no. 6 The Author Published by Oxford University Press on behalf of International Society for Quality in Health Care; all rights reserved 414

2 Patient-centered health services in India Researchers have found it useful to differentiate between general patient satisfaction and patient perceptions of quality [11 13]. Patient satisfaction reflects the extent to which expectations of service standards have been met and is typically operationalized by asking patients about general satisfaction with care received. Perceptions of quality record patient ratings about specific aspects of service quality [11,12]. Satisfaction reflects personal preferences much more than ratings of specific aspects of quality. Furthermore, ratings of specific aspects of quality offer much more actionable information for quality improvement than general satisfaction with services. Both constructs, general satisfaction with care and the patient perceptions of quality, are used in this article. Data and methods Questionnaire development The first stage of questionnaire development involved a qualitative study of patients to get an insight into issues that are important to them [14]. In-patients and outpatients were systematically sampled for in-depth interviews at one district hospital, community health center, and primary health center located outside Lucknow city in Uttar Pradesh. The in-depth interviews covered topics related to perceptions about the hospital, difficulties faced by patients, and ways to improve hospital services. Aspects of service quality which were identified as important included staff and doctor behavior, drug quality and availability, financial cost of treatment (including unofficial payments to providers), hospital infrastructure, and availability of electricity and water. Scale items pertaining to these areas were developed. The second stage involved a comprehensive review of studies on patient perceptions of quality, and potential scale items from these studies were selected [8,9,15]. Results from both efforts led to an initial list of scale items. In the third stage, this list was reviewed by public health experts and a subset of items selected for pretesting. Separate sets of items were created for in-patients and outpatients. In the fourth stage, the selected scale items were translated from English into Hindi, the principal language of Uttar Pradesh. The translation was verified by experts. Each scale item had an associated 5-point Likert-type scale ranging from a score of 1 for completely disagree to 5 completely agree, with 3 being the neutral position. The fifth stage of questionnaire development consisted of pre-testing and refining the preliminary questionnaire. Because of low literacy rates in the population, a visual representation of the Likert scale was needed. Three types of scales were tried a chapati (Indian bread) scale, a smiley face scale, and a money scale. The most easily understood and acceptable scale was a pictorial money scale: one rupee (completely agree), 75 paise (agree), 50 paise (neither agree nor disagree), 25 paise (disagree), and zero paise (completely disagree). Before beginning the interview, the interviewer would first explain the money scale to the patient. He would then read each item statement to the patient who would indicate his or her level of agreement with the statement by pointing to a money amount on the sheet. The initial in-patient scale contained 36 items and the outpatient scale 26 items. Sampling and data collection This study was part of a larger evaluation of the Uttar Pradesh Health Systems Development Project (UPHSDP) conducted between April and September In the first stage of sampling, project and control facilities were selected. In the second stage, a convenience sample of patients within these facilities was selected. A total of 117 facilities from 28 districts were brought under UPHSDP, which included 28 district hospitals, 25 female district hospitals, 28 community health centers, and 36 primary health centers. Project facilities were stratified by the four regions of Uttar Pradesh. From each region, one project district hospital and its associated female district hospital were randomly selected. In the larger eastern and central regions, two district hospitals and their associated female district hospitals were randomly selected. In each region, a similar number of control district hospitals and their associated female hospitals were randomly selected from non-project facilities. The project community and one primary health center in the district of the selected district hospital were automatically included in the sample. Controls for these facilities were selected from within the same district. An equal number of control sites were not available in each district, and the final sample included 12 district hospitals (six projects and six controls), 12 female district hospitals (six projects and six controls), 17 community health centers (seven projects and 10 controls), and 13 primary health centers (six projects and six controls), a total of 54 facilities (26 projects and 28 controls). A convenience sample of outpatients was drawn at all facilities and in-patients from the district and female district hospitals. Each interviewer was given a target of 60 outpatients at district hospitals, 30 outpatients at female district hospitals and community health centers, and 20 outpatients at primary health centers. Thirty in-patients were to be sampled at district and female district hospitals. Outpatients were sampled as they exited the health facility. A convenience sample of inpatients at district and female district hospitals were selected from the general ward if they had spent at least one night in the facility. Verbal consent was taken from patients before interviewing, but no record of refusals was kept. The target sample size was achieved at 94% and 83% of the facilities for outpatients and in-patients, respectively. A total of 1869 outpatients and 611 in-patients were interviewed. Data were collected by the Academy of Management Studies, Lucknow, India. Statistical methods Patients who responded to <50% of the items and items with >50% non-responses were dropped from the analysis. Missing values were imputed using the mean of the individual s responses rounded to the nearest integer. The initial 36-item in-patient and the 26-item outpatient scales had most items in common. This facilitated merger of 415

3 K. D. Rao et al. the two scales into a single scale for in-patients and outpatients. Principal component analysis was used to determine scale dimensionality. The importance of a component was evaluated by examining both scree plots and the contribution of each component to total variance ( 5%). Maximum likelihood factor analysis with varimax rotation was then applied with the principal component analysis results guiding the number of factors to be extracted. Items with substantial loadings ( 0.40) on a single factor were retained. This process was repeated till all items had substantial loadings. For each patient, responses to the final perceived quality scale items were averaged to generate overall perceived quality scores ranging from 1 to 5, with 5 being the highest quality rating. Similarly, item responses in each dimension were averaged to generate dimension-specific scores for each patient. To check if the factor analysis results were sensitive to within-facility correlation, each scale item was regressed against facility dummy variables and the residual factor analyzed. Internal consistency reliability of the scale was measured using Cronbach s alpha. Construct validity refers to the degree of agreement between what is theorized about the construct and its operational form. It was evaluated by examining the degree of congruence between expected scale dimensionality and what was operationally achieved. The questionnaire on general patient satisfaction contained three questions related to general satisfaction with the care received: 1. Overall how satisfied are you with the services at this hospital? 2. How satisfied are you with the services you received at this hospital compared with what you paid? 3. Are you completely satisfied with your treatment? Responses to these questions were on a 5-point scale with higher values indicating greater satisfaction. For each patient, these were averaged to generate scores, which ranged from 1 to 5. Economic status of the sampled patients was estimated by constructing a wealth index from the patient s asset ownership, as described in detail elsewhere [16,17]. The patient s household asset scores were ranked according to standardized asset scores made up of similar items from the National Family Health Survey (NFHS) 1998/99 for Uttar Pradesh, which is representative of Uttar Pradesh s population [18]. To estimate the relative influence of the different perceived quality dimensions on general patient satisfaction, we regressed general satisfaction scores on the standardized score of the perceived quality dimensions, individual level characteristics (age, sex, caste, economic status, and urban or rural place of residence), and self-reported waiting time. Of interest is the magnitude and significance of the standardized perceived quality dimension regression coefficients. These are interpreted as the increase in the mean general patient satisfaction score due to a unit increase in the perceived quality score, controlling for other factors. The NFHS classified head of households into Scheduled Castes and Tribes, Other Backward Castes (OBCs), and Others. Scheduled Castes, Schedule Tribes, and OBCs are castes, tribes, and communities that the Government of India officially recognizes in the schedules of its Constitution as socially and economically backward [18]. The Other category includes those not belonging to a deprived group. Because patients in the study are clustered in facilities, observations within facilities will likely be correlated because of unobserved facility level effects. To adjust for this, we used multilevel regression modeling with patients in level 1 and facilities at level 2. The unobserved facility level effects were considered as fixed parameters [19]. Model fit and assumptions were checked using residual plots. The statistical package STATA 8.2 [20] was used for all statistical analysis. The regression analysis used the XTREG module with fe option in STATA 8.2. Results Patient participation in the study was high. The full range of responses was observed on each scale item, although responses were generally skewed towards higher values. Thirty-two (2%) of the 1869 sampled outpatients who responded to 50% of the scale items were dropped from the analysis. Item non-response was low (<8%) on most outpatient items. However, three items had non-responses >10%, the highest being 35%. No in-patient responses were dropped. The final sample included 1837 outpatients and 611 in-patients. Table 1 summarizes the background characteristics of the sampled patients. The structure of patient quality perceptions Principal component analysis suggested that both the inpatient and outpatient scales had five important dimensions. Furthermore, factor analysis on each of these scales indicated that similar items loaded on each dimension in both scales giving the dimensions in each scale the same interpretation [14]. This facilitated combining the two scales into a single 23- item scale. Principal component analysis indicated that the combined scale also had five important dimensions. Factor analysis of the combined scale indicated that a few factor loadings were not clean because of some items loading substantially on more than one factor and some items not loading on any factor. These items, which were dropped, pertained to patient privacy, location of the hospital, financial cost of treatment, and respect from the doctor. This process was repeated twice and left a final set of 16 items in the combined scale. The five dimensions explained 73% of the variance in the final set of items. Similar results were achieved across different facility and patient types indicating consistency of the construct [14]. The factor analysis results were also unaffected by within-facility correlation. Table 2 summarizes the final scale items and their descriptive statistics. The final scale items and their rotated factor loadings are summarized in Table 3. The first factor is labeled medicine availability and includes two items, which deal with the availability and ease of obtaining drugs at the clinic. The second factor has been labeled medical information, and the three items loading on it deal with the information given to patients by their physician about the cause of their illness, 416

4 Patient-centered health services in India Table 1 Background characteristics of sampled in-patients and outpatients 1 Variable Outpatients In-patients Uttar Pradesh 2... Sample size Age (years) 29 (17.31) 31 (15.87) Male (%) Urban (%) Schedule Caste (%) Schedule Tribe (%) Other Backward Caste (%) Other caste (%) Patient asset score 0.97 (2.78) 0.72 (2.75) 1 Standard deviations are shown in parentheses. 2 Data from Uttar Pradesh National Family Health Survey-2, Table 2 Descriptive statistics of final scale items Scale items Mean (SD) Strongly disagree (%) Disagree (%) Neutral (%) Agree (%) Strongly agree (%)... This hospital has all the medicines needed by you 3.46 (1.21) You are able to get all the necessary medicines easily 3.82 (1.26) The doctors gave you advice about ways to avoid illness 2.15 (1.63) and stay healthy The doctor gave you complete information about 3.08 (1.78) your illness The doctor gave you complete information about 3.07 (1.77) your treatment Hospital workers talk politely 3.97 (0.86) Hospital workers are helpful to you 3.82 (0.95) You are given enough time to tell the doctor everything 4.44 (0.92) Doctors listen carefully to what you have to say 4.53 (0.89) The doctor checks patients properly 4.12 (1.37) The doctor is always ready to answer your questions 4.11 (1.15) The doctor gave you adequate time 4.26 (1.02) The cleanliness of the hospital is adequate 3.41 (0.90) The condition of the toilets are good 2.87 (1.13) Drinking water is easily available in the hospital 3.53 (1.26) This hospital has all the requisite amenities 3.57 (0.90) treatment, and health advice. The third factor is labeled staff behavior and includes two items, which reflect the hospital staff s helpfulness and courtesy towards patients. The fourth factor is labeled doctor behavior, and the five items loading on this factor capture different aspects of the doctor s interpersonal behavior such as time given by the doctor, his/her responsiveness to the patient s concerns, and the doctor s examination of the patient. The fifth factor is labeled hospital infrastructure, and the four items which load on it capture various aspects of the hospital s infrastructure such as its cleanliness and availability of conveniences like drinking water and other amenities. Table 4 summarizes the mean scores for general satisfaction, overall perceived quality, and the different perceived quality dimensions. Scale reliability and validity Internal consistency reliability. The general satisfaction and perceived quality scales have alpha coefficients >0.7 (Table 4). For both in-patients and outpatients, item rest correlations on the general satisfaction scale, overall perceived quality scale, and subscales ranged from 0.19 to Similar results were obtained for in-patients and outpatients across facility types. Content validity. Content validity of the perceived quality scale was repeatedly assessed throughout the process of selecting scale items. As described previously, scale items were identified from a review of the literature and interviews with patients and public health professionals. Construct validity. On the basis of the initial qualitative assessment in Uttar Pradesh and a review of the literature, the 417

5 K. D. Rao et al. Table 3 Final scale items and rotated factor loadings Scale items Factor 1 medicine availability Factor 2 medical information Factor 3 staff behavior Factor 4 doctor behavior Factor 5 hospital infrastructure... This hospital has all the medicines needed by you You are able to get all the necessary medicines easily The doctors gave you advice about ways to avoid illness and stay healthy The doctor gave you complete information about your illness The doctor gave you complete information about your treatment Hospital workers talk politely Hospital workers are helpful to you You are given enough time to tell the doctor everything Doctors listen carefully to what you have to say The doctor checks patients properly The doctor is always ready to answer your questions The doctor gave you adequate time The cleanliness of the hospital is adequate The condition of the toilets are good Drinking water is easily available in the hospital This hospital has all the requisite amenities The boldface values is to highlight the factor loadings that load on specific factors. 418

6 Patient-centered health services in India Table 4 Scale reliability and perceived quality and general patient satisfaction scores Scales Standard deviations are shown in parentheses. Cronbach s alpha Outpatient mean score In-patient mean score... General patient satisfaction (three items) (0.65) 3.74 (0.72) Overall perceived quality (all 16 items) (0.65) 3.71 (0.63) Perceived quality subscales Medicine availability (two items) (1.12) 3.23 (1.11) Medical information (three items) (1.52) 3.22 (1.48) Staff behavior (two items) (0.84) 3.90 (0.88) Doctor behavior (five items) (0.92) 4.51 (0.78) Clinic infrastructure (four items) (0.75) 3.42 (0.73) latent construct of patient perception of quality was theorized to be multidimensional. The factor analysis (Table 3) identified five distinct dimensions of perceived quality, which was in line with our expectation of perceived quality being a multidimensional construct. However, there are differences. Financial costs and bribes to staff did not emerge as a dimension of perceived quality, and the final scale does not have any items related to it. Furthermore, behavior of doctors and staff emerged as two separate dimensions. On the basis of these observations, it appears that the dimensionality of the perceived quality scale has reasonable congruence with what was theorized. Perceived quality and patient satisfaction Results from regressing general patient satisfaction scores on the different standardized perceived quality subscale scores and other variables are summarized in Table 5. Analysis of residuals indicated that model fit was generally good. All the perceived quality dimensions have a positive and statistically significant effect on overall in-patient and outpatient satisfaction, except for the hospital infrastructure for inpatients. The effect of the perceived quality dimensions on general satisfaction does not significantly differ between inpatients and outpatients, except for staff behavior, which is significantly lower for outpatients compared with in-patients. Waiting time of >30 minutes significantly lowers in-patient satisfaction. Discussion The first part of this study described the development of a 16- item scale that can be used to measure perceived quality at a range of facility types for both outpatients and in-patients in India. This scale can be used for routine quality evaluation by both hospital administrators and external evaluators. The analysis identified five distinct dimensions of perceived quality: (i) medicine availability, (ii) medical information, (iii) staff behavior, (iv) doctor behavior, and (v) hospital infrastructure. These dimensions provide information on the structure and process of care [21] and correspond quite well to the general themes from the initial qualitative study. The reliability and validity of the perceived quality scale were assessed in various ways, and the scale was found to have good reliability and validity across different patient types and facility levels. The high alpha coefficient of the overall quality perception scale (Table 3) suggests that the items in the final scale measure a common underlying construct. The characteristics of the sampled patients correspond well with those of the general population of Uttar Pradesh. For both outpatients and in-patients, scores on general satisfaction, overall perceived quality, and the perceived quality subscales (except for the doctor behavior subscale) were slightly better than average at all facility levels. This indicates that patient satisfaction and perceptions of quality at public health facilities are on average only marginally favorable, leaving much room for improvement. The higher scores on the doctor behavior subscale are likely due to acquiescence bias [22] and gratitude bias [23]. The regression results provide guidance about which aspects of perceived quality have the biggest impact on general patient satisfaction. For outpatients, doctor behavior has the largest effect followed by medicine availability, hospital infrastructure, staff behavior, and medical information. For in-patients, staff behavior has the largest effect followed by doctor behavior, medicine availability, medical information, and hospital infrastructure. In both cases, the interpersonal skills of the medical personnel and availability of medicines have a large influence on patient satisfaction. Furthermore, longer waiting time had a progressively larger negative effect on both in-patient and outpatient satisfaction. These are important areas in which health services in India require strengthening. However, it would be wrong to infer that some quality dimensions can be ignored in favor of others. These different dimensions are interdependent and improvements in one are likely to strengthen the other. In India and many developing countries, the excessive emphasis on service coverage and inputs in the provision of health services has ignored the needs of the very people for whom these health services exist. Incorporating patient views into quality assessment offers one way of making health services 419

7 K. D. Rao et al. Table 5 Regression results of general patient satisfaction on standardized perceived quality dimensions and other variables General patient satisfaction score (dependent variable) Outpatients In-patients... Standardized staff behavior score ( ) 0.18 ( ) Standardized doctor behavior score 0.20 ( ) 0.16 ( ) Standardized medical advice score 0.06 ( ) 0.08 ( ) Standardized hospital infrastructure score 0.11 ( ) 0.05 ( ) Standardized medicine availability score 0.13 ( ) 0.11 ( ) Patient age (reference 0 20 years) Patient age years 0.02 ( ) 0.05 ( ) Patient age years 0.01 ( ) 0.08 ( ) Patient age 61 years 0.04 ( ) 0.08 ( ) Male (reference: female) 0.04 ( ) 0.08 ( ) Other caste (reference: deprived castes 1 ) 0.03 ( ) 0.03 ( ) Urban (reference: rural) 0.01 ( ) 0.04 ( ) Wealth group (reference lowest 40%) Wealth group: middle 40 60% 0.01 ( ) 0.11 ( ) Wealth group: highest 40% 0.05 ( ) 0.08 ( ) Waiting time (reference: 10 minutes or less) Waiting time minutes 0.10 ( ) 0.00 ( ) Waiting time 31 minutes 0.19 ( ) 0.18 ( 0.31 to 0.06) Constant 3.69 ( ) 3.69 ( ) Rho Observations Number of health facilities R-squared % confidence intervals in parentheses. 1 Deprived castes include Schedule Castes, Schedule Tribes, and Other Backward Castes. 2 Difference between outpatients and in-patients is statistically significant at 5%. more responsive to people s needs. It also gives users an opportunity to voice their opinion about their health services. While conducting this study, we found many instances in which patients were eager to record their concerns about the services they had received in the hope that some action would be taken. It is likely that the very act involving patients in evaluating their health services will make providers more sensitive and alert to patient needs. Acknowledgements This study was funded by the Reaching the Poor Program of the World Bank. The authors thank the Editor of this journal and the two anonymous reviewers for their valuable and insightful comments. References 1. World Health Organization. The World Health Report 2000 Health Systems: Improving Performance. Geneva: WHO, Haddad S, Fournier P, Machouf N, Yatara F. What does quality mean to lay people? Community perceptions of primary care services in Guinea. Soc Sci Med 1998; 47: Akin JS, Hutchinson P. Health-care facility choice and the phenomenon of bypassing. Health Policy Plan 1999; 14: Acharya BL, Cleland J. Maternal and child health services in rural Nepal: does access or quality matter more? Health Policy Plan 2000; 15 (2): Ager A, Pepper K. Patterns of health service utilization and perceptions of needs and services in rural Orissa. Health Policy Plan 2005; 20 (3): Choi KS, ChoWH, Lee S, Lee H, Kim C. The relationships among quality, value, satisfaction and behavioral intention in health care provider choice: a South Korean study. J Bus Res 2004; 5: Alden DL, Hoa DM, Bhawuk D. Client satisfaction with reproductive health-care quality: integrating business approaches to modelling and measurement. Soc Sci Med 2004; 59: Baltussen RMPM, Ye Y, Haddad S, Sauerborn RS. Perceived quality of care of primary health services in Burkina Faso. Health Policy Plan 2002; 17: Andaleeb SS. Service quality perceptions and patient satisfaction: a study of hospitals in a developing country. Soc Sci Med 2001; 52: Population Council. Improving Quality of Care in India s Family Welfare Programme the Challenge Ahead. New York: Population Council,

8 Patient-centered health services in India 11. Haddad S, Potvin L, Roberge D, Pineault R, Remondin M. Patient perception of quality following a visit to a doctor in a primary care unit. Fam Pract 2000; 17: Cleary PD. Satisfaction may not suffice! Int J Technol Assess Health Care 1998; 14: Duong DV, Binns CW, Lee AH, Hipgrave DB. Measuring client-perceived quality of maternity services in rural Vietnam. Int J Qual Health Care 2004; 6: Rao KD. Quality Improvement and its Effect on Patient Perceptions and Demand for Health Services Evidence from Uttar Pradesh, India. Dissertation. Baltimore, MD: Johns Hopkins University, David W, Lesley H, Marie-Anne D. Measures of Need and Outcomes for Primary Health Care. New York: Oxford Medical Publications, Filmer D, Pritchett LH. Estimating wealth effects without expenditure data or tears: an application to educational enrollments in states of India. Demography 2001; 38: Gwatkin DR, Rutstein S, Johnson K, Pande RP, Wagstaff A. Socioeconomic Differences in Health, Nutrition and Population. The World Bank: Washington, DC, Indian Institute of Population Sciences. National Family Health Survey (NFHS-2), , India. Mumbai: IIPS and ORC Macro, Wooldridge JM. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: The MIT Press, StataCorp. Stata Statistical Software: Release 8. College Station, TX: StataCorp LP, Donabedian A. The Definition of Quality and Approaches to its Assessment. Ann Harbor: Health Administration Press, Lewis-Beck MS. Basic Measurement. New Delhi: Sage Publications, Thi PLN, Briancon S, Empereur F, Guillemin F. Factors determining in-patient satisfaction with care. Soc Sci Med 2002; 54: Accepted for publication 4 September