Development and testing of a 25-item patient satisfaction scale for Black South African diabetic outpatients

Although there is general agreement that patient satisfac­ tion is an integral component of service quality, there is a paucity of South African research on reliable and valid satisfaction measures and the effects of health status on satisfaction. A 25-item patient satisfaction scale was de­ veloped and tested for evaluating the quality of health care for black diabetic outpatients. It was hypothesised that: (1) the underlying dimensions of patient satisfaction were interpersonal and organisational; and (2) patients in poor health would be less satisfied with the quality of their care than patients in good health. The questionnaire was ad­ ministered to 263 black outpatients from Pretoria Aca­ demic Hospital and Kalafong Hospital. Factor analysis was conducted on the patient satisfaction scale and three factors, accounting for 71 % of the variance, were extracted. The major items on Factor I were helpfulness, communi­ cation, support and consideration, representing the inter­ personal dimension. Factors II and III were mainly con­ cerned with service logistics and technical expertise, with the emphasis on waiting time, follow-up and thorough­ ness of examination. The three factors had excellent reli­ ability coefficients, ranging between 0.82 (technical), 0.85 (logistics) and 0.98 (interpersonal). Multiple analyses of co-variance showed that patients in poor general health were significantly less satisfied with the logistical (p = 0.004) and technical (p = 0.007) quality of their care than patients in good health; patients in poor mental health were significantly less satisfied with the interpersonal quality of their care (p = 0.05) than patients in good mental health. These findings provided support for both hypotheses and sug­ gested that patients in poor health attend to different as­ pects of their care than patients in good health. Of more importance to clinical practice, the results endorsed the need for a multidisciplinary health team comprising nurse/ social worker (Factor I: support, communication), health service managers (Factor II: service logistics) and physi­ cian (Factor III: technical expertise) to enhance treatment Ab stra k Ten spyte daarvan dat dit algemeen aanvaar word dat pasiëntsatisfaksie ‘n integrale komponent van die gehalteversekering van gesondheids-dienslewering is, is daar weinig navorsing gedoen ten opsigte van die betroubaarheid en geldigheid van m etings vir satisfaksie en die effek van die gesondheidstoestand op satisfaksie in Suid-Afrika. ‘n Vyf en twintig (25) item pasiëntsatisfaksieskaal is ontwikkel en getoets vir die evaluering van die gehalte van gesondheidsorg op swart pasiënte met diabetes. Twee hipoteses was gestel naam lik: (1) die onderliggende dim ensies van pasiëntsatisfaksie is organisatories en interpersoonlik van aard, en (2) pasiënte met swak gesondheid is minder tevrede met die gehalte van sorg as die met goeie gesondheid. Twee honderd drie en sestig (n = 263) buitepasiente van die Pretoria Akademiese en Kalafong Hospitaal het die vraelys voltooi (vir ongeletterde pasiënte is ‘ntolk gebruik). Faktor analise van die pasiëntsatisfaksieskaal toon dat 71 % van die variansie deur drie faktore verklaar kan word. Die hoof items in Faktor I wat die interpersoonlike dim ensie verteenwoordig was hulpvaardigheid, kom m unikasie, ondersteuning en inagnem ing. Faktore II en III was hoofsaaklik gemoeid met die logistiek van dienslewering en tegniese kundigheid (mediese/geneesheer). Die klem het hier geval op wagtyd, opvolg en deeglikheid van die ondersoek. Hierdie 3 faktore het uitstekende betroubaarheids koëffisiënte getoon naamlik: 0.82 (tegnies), 0.85 (logistiek) en 0.98 (interpersoonlik). Veelvuldige analise van die kovariansie toon dat pasiënte met ‘n swak algemene gesondheid beduidend minder tevrede was met die logistiese (p=0.004) en tegniese (p=0.007) aspekte van sorg, in vergelyking met pasiënte wie goeie gesondheid geniet het. Pasiënte met swak geestesgesondheid was beduidend minder tevrede met die gehalte van sorg ontvang in die interpersoonlike dimensie (p=0.05) in vergelyking met pasiënte met 68 Curationis August 2002 outcome for diabetic patients. It is recommended that: (1) further research is conducted on this patient satisfac­ tion scale with diverse populations in different settings to complement and validate the scale for generalised use in South Africa; (2) the scale is used to collect information on patient satisfaction before and after implementing an intervention to improve the quality of health care, and (3) measurement of health status is an essential adjunct to assessment of patient satisfaction. Key w o rd s : Em path y, com m unication, logistics, expertise, satisfaction , health status goeie geestesgesondheid. Hierdie bevindings verleen steun aan beide die gestelde hipoteses, asook dat pasiënte met swak gesondheid, ander behoeftes het ten opsigte van hulle sorg, as die met goeie gesondheid. Van meer waarde in die kliniese praktyk is dat hierdie bevindings die noodsaaklikheid van ‘n multidissiplinêre gesondheidsspan bestaande uit ‘n verpleegkundige, maatskaplike werker (Faktor I: ondersteuning en kommunikasie), bestuurders van gesondheidsdienslewering (Faktor II: logistiek van diens) en geneesheer (Faktor III: tegniese kundigheid) beklemtoon, om die pasiënt uitkomste van die behandeling van diabetes te bevorder. Dit word aanbeveel dat: (1) verdere navorsing gedoen word om hierdie pasiën tsatisfaksieskaal te valideer vir ander bevolkingsgroepe en om standighede, om dit meer algemeen bruikbaar te maak in Suid-Afrika, (2) die skaal gebruik word om inligting te verkry tov pasiëntsatisfaksie voor en na im plem entering van in tervensies om gesondheidsorg te verbeter, en (3) die meting van die pasiënt se huidige gesondheidstoestand, is ‘n noodsaaklike aanvullende aspek in die evaluering van pasiëntsatisfaksie. Sleutelw oorde: Em p atie , kom m unikasie, logistiek, kundigheid, satisfaksie, gesondheidstoestand

P asiënte m et sw ak geestesgesondheid was beduidend minder tevrede met die gehalte van sorg ontvang in die interpersoonlike dimensie (p=0.05) in vergelyking met pasiënte met outcome for diabetic patients.It is recommended that: (1) further research is conducted on this patient satisfac tion scale with diverse populations in different settings to complement and validate the scale for generalised use in South Africa; (2) the scale is used to collect information on patient satisfaction before and after implementing an intervention to improve the quality of health care, and (3) measurement of health status is an essential adjunct to assessment of patient satisfaction.

Introduction
There is general agreement that patient satisfaction is an inte gral component of service quality (Carr-Hill 1992: 236;Sitzia andWood 1997: 1829), since expanded definitions of health service quality make explicit mention of patient satisfaction (Fitzpatrick 1991: 888).It has been proposed that the effective ness of health care is determined, to some degree, by satisfac tion with the services provided.Support for this viewpoint has been found in studies that have reported a satisfied pa tient is more likely to utilise health services (Larsen and Rootman 1976:30), comply with medical treatment (Kincey, Bradshaw and Ley 1975: 564) and continue with the health provider (Baker 1990:489).
Various studies have shown that satisfaction is related to tech nical and interpersonal competence, more partnership build ing, more immediate and positive non-verbal behaviour, more social conversation, courtesy, consideration, clear communi cation and information, respectful treatment, frequency of con tact, length of consultation, service availability and waiting time (Hall, Roter and Katz 1988:657;Sikosana 1994:269;Singh, Mustapha and Haqq 1996:255).
Single measures of satisfaction are not recommended as con sumers of health care usually report high levels of satisfaction with very little variation in responses (Carr-Hill 1992: 244).Surveys in the United Kingdom (UK), Zimbabwe, Saudi Arabia and South Africa have reported that over 80% of respondents state that they are satisfied with the quality of health care they receive, despite deficiencies in manpower, skills, equipment and facilities; lengthy waiting times; and extremely short con sultations (Carr-Hill 1992: 240;Community Agency for Social Enquiry 1995:17;Mansourand Al-Osimy 1996: 312;Sikosana 1994:270).Although these findings may reflect a reluctance to complain about services, acceptance of low standards of care, or even low levels of expectation concerning free health care; such high levels of assent, undifferentiated across populations, cast doubt upon single measures of satisfaction with health care (Carr-Hill 1992:240;Sitzia andWood 1997:1830).
Several self-response questionnaires have been developed and tested in the US and UK (Baker 1990: 436;Hulka, Zyzanski, Cassel and Thompson 1970: 431;Ware, Snyder, Wright and Davies 1983: 252).The problem with using self-response ques tionnaires in South Africa is the high proportion of function ally illiterate persons in the general population, which makes the validity of such measures problematic.A more promising approach to measuring patient satisfaction comes from Indo nesia, where focus group discussions were held to identify aspects of the service that were meaningful to patient satisfac tion (Bemhart, Wiadnyana, Wihardjo and Pohan 1999: 991).Fourteen factors were identified in this study.Respondents were dissatisfied with the lack of medicine, privacy during ex amination, the cleanliness of the facility and communication about their condition.It would appear that communication of information is the most frequent source of dissatisfaction in developing and developed countries (Carr-Hill 1992: 240).A recent South African study developed and tested a 20-item satisfaction scale for evaluating family planning services 69 Curationis August 2002 (Westaway, Viljoen and Chabalala 1998:5-6).Satisfaction was based on interpersonal factors such as friendliness and en couragement, and the organisational factors of choice, change and provision of different methods.Respondents were least satisfied with follow-up, maintenance of contact, availability of the service and waiting times; similar to previous findings on dissatisfaction (Bemhart, et al. 1999: 994;Carr-Hill 1992: 237).
Most studies on satisfaction have found that older patients report higher levels of satisfaction than younger patients (Carr-Hill 1992:237;Cohen 1996Cohen :1085;;Sitzia andWood 1997: 1835;Williams and Calnan 1991:712).However, the relationship be tween age and satisfaction is confounded by health status or health-related quality of life (HRQOL).Whilst Williams and Calnan (1991: 712) found no significant relationship between health status and satisfaction in either primary or hospital care settings, Cohen (1996Cohen ( : 1088) ) reported that pain and psychoso cial health status, adjusting for age, were significantly related to lower satisfaction with health care.Cohen's findings sug gest that patient satisfaction is susceptible to change in re sponse to organisational, clinical and interpersonal treatment.
Health status was measured by the general (5 items) and men tal health (5 items) sub-scales from the 20-item abbreviation of the Rand Medical Outcomes Study (Stewart, Hays and Ware 1988: 733-735).Scores on the sub-scales are transformed lin early to 0-100, where 0 and 100 are assigned to the lowest and highest possible scores, respectively.The cut-off point for defining poor general health is a score of 70 or lower; for poor mental health the cut-off point is a score of 67 or lower (Stewart, Hays and Ware 1988: 728).High scores denote better general and mental health.In a recent South African study with black diabetic patients (Westaway, Viljoen and Rheeder 1999: 215), reliability coefficients ranged between 0.79 (mental health) and 0.81 (general health); in the respectable range according to Arias and de Vos (1996:65).Based on previous South African research (Westaway, et al., 1998), a review of the literature (Andaleeb, 2001;Bemhart, et al., 1999;Carr-Hill, 1992;Cohen, 1996;Fitzpatrick, 1991;Singh, etal., 1996;Williams and Calnan, 1991) and interviews with 20 patients, a 25-item satisfaction scale was designed to measure the characteristics of providers and the service with regard to interpersonal and organisational dimensions of satisfaction.
A study on service quality perceptions and patient satisfac tion in Bangladesh (Andaleeb 2001(Andaleeb : 1364) ) found five service quality dimensions: responsiveness (caring, helpful); assur ance (skilled staff, competence); communication (explanation of tests, answering questions); discipline (cleanliness of facil ity) and baksheesh (no services without tips).With the excep tion of baksheesh, many of the items used to measure service quality were similar to those used to measure patient satisfac tion (Westaway, et al. 1998:6).It would appear that the dimen sions of health care quality (technical aspects of care, interper sonal relationships and the amenities of care) are virtually iden tical to the dimensions of patient satisfaction.In South Africa, there is a paucity of satisfaction measures for specific populations.In addition, no local studies have inves tigated the relationship between health status and patient sat isfaction.Therefore, a local patient satisfaction scale was de veloped and tested.

Research Design
A cross-sectional, analytical research design was used for the study.Based on previous research on patient satisfaction, two hypotheses were tested: 1.
The underlying dimensions of patient satisfaction are interpersonal and organisational; 2.
Patients in poor health are less satisfied with the in terpersonal and organisational dimensions of their care than patients in good health.

The Questionnaire
A structured questionnaire, with a consent form, was designed to obtain information on: demographic variables (age, gender, years of schooling, marital status and employment status); health status (general health and mental health); and satisfac tion with the interpersonal and organisational dimensions of the service.

Significant loadings in bold
health were less satisfied with the quality of their care than patients in good health.

Demographic Information
The questionnaire was administered to 263 patients (96 from Pretoria Academic Hospital and 167 from Kalafong Hospital).There were 174 females and 89 males aged between 16 and 89 years (average age = 53.5 years, sd = 13.9).Twenty three per cent had no formal schooling, 28% some primary level school ing, 38% some high school and 11 % had completed high school.
The average number of years of schooling was 6.3 (sd = 4.1).Age was significantly related to schooling (r = -0.28,p = 0.01), indicating that older patients had limited educational opportu nities.

Health Status
The average score on the general health sub-scale was 53.3 (sd = 37.7) and 75.8 (sd = 24.0) on the mental health sub-scale.Fifty per cent of the patients reported poor general health whereas only 28% re ported poor mental health.General health and mental health were significantly related (r = 0.34, p < 0.01).The two sub scales had very good reliability coefficients (Arias and de Vos 1996:65), ranging between 0.88 (mental health) and 0.95 (gen eral health); even higher than 0.79 (mental health) and 0.81 (general health) found previously by Westaway, et al. (1999: 215).Age and employment status were significantly related to general health (p = 0.01).As was found previously (Stewart, et al. 1988: 729;Westaway, et al. 1999: 213), older patients and those who were unemployed had poorer general health than younger patients and those who were employed.

Patient Satisfaction
The sample size of 263 patients fulfilled Boyle's (1985:50) mini mum sample size of 250 subjects and Nunnally's (1978: 421) minimum criterion for factor analysis (10 persons per item).A direct solution (principal components analysis) was the first step in analysing the 25-item scale (Nunnally 1978: 357-367).
Only items with communality estimates (common factor vari ance) $ 0.30 were taken into consideration, as items with unique Factor II contained 7 significant loadings.The most important items were waiting time (0.73) and follow-up (0.72), applicable to service logistics.Factor III contained 3 significant loadings and appeared to be concerned with technical expertise such as thoroughness of examination (0.87), privacy (0.82) and medi cine received (0.76), Table 1.Two items (cost of attendance and convenience) did not meet the factor loading criterion of $ 0.50 (Child, 1970;Nunnally, 1978).According to the factor pattern, Hypothesis 1 received considerable support, since 23 out of the 25 items had loadings $.0.50.
The rotated coefficients were used to generate a weighted score for each of the three sub-scales (interpersonal, logistics and technical) and overall scale.Mean scores, standard devia tions, range and reliability coefficients for the three sub-scales and overall scale are shown in Table 2.The reliability coeffi cients were 0.98 (interpersonal), 0.85 (logistics), 0.82 (techni- variance (specific variance + error variance) > 0.70 tend to be unreliable (Child 1970: 42).In order to ascertain significant loadings at the 1% level, loadings $ 0.50 were examined (Child 1970: 36-38;Nunnally 1978:423).All communality estimates exceeded the criterion of 0.30 and ranged between 0.47 and 0.91.The Kaiser-Meyer-Olkin measure of sampling adequacy of the number of items was 0.92, in the marvellous range ac cording to Kaiser (1974: 34), and confirmed that factor analysis was the correct procedure for the data.
cal) and 0.96 (overall scale).Inter-correlations among the sub scales are shown in Table 3.The three sub-scales were signifi cantly related to each other (p < 0.001).
All items exceeded the convergent validity criterion for their scale and ranged between 0.83 and 0.94 (interpersonal), 0.58 and 0.68 (logistics) and 0.58 and 0.73 (technical).Two items (service availability and convenience) were slightly less than the discriminant validity criterion.
Based on the eigenvalues, the percentage of the total variance accounted for by the different factors and the scree plot (spa tial representation of the factors), three factors, accounting for 71.1% of the total variance, were extracted.An orthogonal (VARIMAX) rotational solution was conducted to minimise the number of variables with high loadings on a factor and achieve simple structure.Factor 1 contained 13 significant loadings.The major items were: communication (0.91), sup port (0.89), consideration (0.88), friendliness (0.84), helpfulness (0.84) and encouragement (0.84), Table 1.Factor I seems to represent a combination of empathy and communication (in terpersonal dimension).
Multiple analyses of co-variance (MANCOVAs) were used to tease out general and mental health group effects.The ad justed mean scores (controlling for mental health) were signifi cantly lower for the poor general health group on service logis tics (p = 0.004) and technical expertise (p = 0.007).The ad justed mean scores (controlling for general health) were sig nificantly lower for the poor mental health group on the inter personal dimension and overall scale (p = 0.05), Table 4.These findings provided full support for Hypothesis 2, since patients in poor general and mental health were less satisfied with the interpersonal and organisational aspects of their care than pa tients in good general and mental health.

Discussion
Factor analysis of the patient satisfaction scale provided sub stantial evidence concerning the underlying constructs of sat isfaction.The interpersonal dimension, consisting of empathy and communication, was more important than the organisa tional dimension (logistics and technical expertise).In addi tion, these findings substantiated Weitzman's (1998: 385-390) model of health care quality, as the three factors included inter personal relationships (empathy and communication), techni-care than patients in good health, providing full support for hypothesis 2, and substantiating Cohen's (1996Cohen's ( : 1088) ) find ings.Patients in poor general health tended to be less satisfied with the organisational aspects of their care than patients in good health.In contrast, patients in poor mental health tended to be less satisfied with the interpersonal aspect of their care than patients in good mental health.These findings suggest that patients in poor health attend to different aspects of their care than patients in good health.Therefore, measurement of health-related quality of life is essential for assessing satisfac tion with the quality of care.Of major importance to clinical Overall scale 84.1 1.0 86.5 1.1 2.4 82.9 1.4 86.2 0.9 3.9' * p = 0.05, ** p < 0.01 cal aspects of care and the amenities of care, such as waiting time and the availability of a seat.The reliability coefficients and the multi-trait analyses of the three patient satisfaction sub-scales and overall scale provided substantial evidence on the reliability and validity of the scale.Moreover, the lack of demographic effects showed that this scale held particular promise for assessing the quality of health care from a patient perspective in diverse populations and set tings.
Patients in poor general and mental health were less satisfied with the interpersonal and organisational dimensions of their practice is the emphasis on interpersonal and organisational dimensions that can lead to greater satisfaction with the qual ity of health care.Overall findings endorse the need for a multidisciplinary health team comprising nurse/social worker (Factor I: support, communication); health service managers (Factor II: service logistics); and physician (Factor III: techni cal expertise) to enhance treatment outcome for diabetic pa tients.
It is recommended that: (1) further research is conducted on this patient satisfaction scale with diverse populations in dif-7 3 Curationis August 2002 ferent settings to complement and validate the scale for gener alised use in South Africa; (2) the scale is used to collect infor mation on patient satisfaction before and after implementing an intervention to improve the quality of health care, and (3) measurement of health status is an essential adjunct to assess ment of patient satisfaction.

Table 1 :
Two trained black female multilingual interviewers administered the questionnaire at Pretoria Academic Hospital and Kalafong Hospital.All black persons attending Pretoria Academic Hos pital and Kalafong Hospital for diabetes outpatient treatment, during the period November 1999 to July 2000, were asked to participate in the study.There were 263 patients who partici pated in the study; only 23 patients refused to participate due to time constraints.Ethical approval for the study was ob tained from the University of Pretoria's Ethical Committee.Permission for the study was obtained from the two Hospital Superintendents.The Nursing Service Managers and staff at Pretoria Academic Hospital and Kalafong Hospital were con sulted regarding the purpose of the study.Informed consent was obtained from all persons interviewed.Orthogonal (V A R IM A X) Rotational Solution for Patient Satisfaction In order to test Hypothesis 2, two groups were created, based on Stewart, Hays andWare's (1988: 728)recommended cut-off points: patients scoring # 7 0 (poor health group) and patients scoring > 70 (good health group) on the general health sub scale; and patients scoring # 67 (poor mental health group) and patients scoring > 67 (good mental health group).Multi ple analyses of co-variance, with Bonferroni t tests for multiple comparisons, were used to determine whether patients in poor 70

Table 2 :
M ean Scores, Standard deviations, Range and Reliability Coefficients on the Three Patientt Satisfaction Sub-Scales and Overall Scale

Table 3 .
Intercorrelation M atrix for the Three Patient Satisfaction Sub-Scales

Table 4 .
Adjusted M ean Scores (am ), Standard Errors (se) and F Tests for Patient Satisfaction: General Health (GH) and M ental Health (M H )