Construction of a scale measuring inpatients opinion on quality of care

Transcription

1 International Journal for Quality in Health Care 1999; Volume 11, Number 6: pp Construction of a scale measuring inpatients opinion on quality of care LAURENCE SALOMON 1, ISABELLE GASQUET 1,2, MOUNIR MESBAH 3 AND PHILIPPE RAVAUD 1 1 Evaluation and Quality Department, Medical Policy Direction, Assistance Publique Hôpitaux de Paris, 2 National Institute of Health and Medical Research, Villejuif, 3 Institut Universitaire et Professionnel, South Brittany University, Vannes, France Abstract Objective. To develop a reliable and valid measure of patient opinions on quality of hospital care. Design. Issues of importance to patients and possible scale items were generated by literature review and non-structured interviews of patients, former patients, health care providers and researchers. Semi-structured interviews with inpatients and pilot studies were conducted to modify or remove ambiguous questions and reduce skewed responses. A study was then made to select from these questions relevant items and variables correlated to patient evaluation of quality of care. A principal-components analysis was performed to select items and assess construct validity. Cronbach s α coefficients were calculated to estimate the reliability of the scale. Time reliability and concurrent validity were also considered. Setting. An 800-bed French short-stay teaching hospital in Paris. Study participants. Five-hundred and thirty-four consecutive patients hospitalized in eight medical and surgical wards. Results. A 26-item scale was developed. Component analysis indicated two subscales: medical information and relationship with staff and daily routine. Levels of reliability were satisfactory: Cronbach s α coefficient exceeded 0.87 for overall scale and subscales. Concurrent validity and time reliability were also satisfactory. Multivariate analysis showed that, taking into account patients and hospitalization characteristics linked to scores (age, health status, number of hospitalizations, comorbidity, time since diagnosis, admission pattern, private patient and difficulties reported by staff ), these scores differed among departments. Conclusion. A reliable, valid measure of inpatients opinions on quality of care has been developed in a French hospital and variables that have to be taken into account to compare hospital departments have been selected. Items selected in the scale emphasized the importance that patients give to receiving medical information. Keywords: hospitalization, inpatient, patient satisfaction, psychometric testing Patient satisfaction is increasingly recognized as an important health care providers and regulators to locate and solve dimension of quality of care [1 3]. From the hospital s problems that can improve the quality of care [11,12]. This perspective, the patient s view of care is important to both increased importance of measuring patient satisfaction raises clinical and management staff and patient satisfaction is critical questions about the quality and validity of a patient predictive of future behaviour such as compliance with satisfaction measurement system. In addition, patient expectations treatment and intent to return for care [4 8]. and priorities vary among countries and are highly During the last decade, a number of measures of perceived related to cultural background and to the health care system quality of care and patient satisfaction have been developed [10 15]. The French health care system clearly differs from with varying evidence of validity (i.e. the requirement for it others: it provides universal coverage, both public and private to measure what it purports to measure) [9]. structures and a free choice of physicians [16]. These dif- Former studies have shown that asking patients direct ferences justify the development of a specially adapted questionnaire. detailed questions about what happened rather than asking Nevertheless, although the assessment of inpatient general questions about satisfaction can locate the problems satisfaction became mandatory in France in April 1996 [17], and so enable them to be solved [2,10]. a validated questionnaire adapted to the French health system The results of satisfaction surveys are increasingly used by has not yet been developed. Address correspondence to Isabelle Gasquet, INSERM U472, 16, avenue Paul Vaillant Couturier F-94807, Villejuif, France. Tel: Fax: gasquet@vjf.inserm.fr 1999 International Society for Quality in Health Care and Oxford University Press 507

2 L. Salomon et al. The purpose of this study was to develop a reliable, valid To assess concurrent validity, the questionnaire also enquired measure of inpatients evaluation of care suitable for largeacteristics about socio-demographic and health care charmeasure scale department evaluation in France. (medical care insurance cover, number of previous hospitalizations, perceived health status), medical and psychological outcomes, intent to use again or recommend the Materials and methods department to others. Finally we included six analogous visual scales about the quality of care (for example technical quality of nursing care ) which measured the extent to which patients Questionnaire development expectations were met (from did not meet my expectations We used qualitative methods to identify issues concerning at all to far exceeded my expectations ). inpatient care that were important to patients, as well as Some questions were completed by physicians: these were possible scale items. Using individual unstructured interviews, on medical characteristics (severity at admission, time since we interviewed a total of 20 people, both former patients diagnosis, comorbidity), admission and discharge pattern, and patients who were hospitalized at the time of the interview. department of hospitalization, stay in intensive care unit, length of the hospital stay, private or non-private patient, We also compiled a list of items related to satisfaction therapeutic or relational difficulty during hospitalization. with inpatient care from foreign questionnaires [2,9,11,13, 14,18 31], French regulations [32] and an ethics code for physicians [33]. An analysis of the literature also allowed Data collection us to identify factors related to patient satisfaction (socio- The study was conducted in 1997 at Cochin Hospital (Asdemographic, medical and health care characteristics) [10,14, sistance Publique Hôpitaux de Paris), a French public, adult, 24,26,34]. short-stay teaching hospital of beds. Data were collected We chose an indirect approach in which patient opinion in four medical departments (internal medicine, rheumatology, on quality of care is inferred from the choice of answers [5, cardiology, gastro-enterology), and four surgical departments 24,25]. For example, a positive answer to physicians answered (urology, abdominal surgery and two orthopaedic surgery all my questions would reflect a good quality of care. departments). A questionnaire including these possible items was subapproached All adult patients hospitalized for at least 24 hours were mitted to health care providers (physicians, surgeons, psyresearch the day before discharge by independent, trained chiatrists, nurses and social workers) and researchers assistants. They explained the purpose of the study, (epidemiologists, biostatisticians and psychometricians) in invited patients to take part and gave them a questionnaire. order to assess content validity. Some of the patients inpatients. A printed form of these instructions was also given to terviewed during questionnaire development were asked to Patients completed the anonymous questionnaire comment on its content and suggest additional issues or alone, and returned it to the assistants in a sealed envelope; questions. The questionnaire was thereafter administered to the envelopes were collected in a box. Clinical data for all a sample of 50 inpatients from several medical and surgical patients, including those who did not complete the patient departments of two hospitals to determine basic com- questionnaire, were collected from the physician in charge. prehension and to further ensure content validity. Questions Patients were approached until a sample of 60 patients which were confusing, ambiguous or gave skewed responses was obtained in each of the eight departments. A total of were removed or rewritten. 534 patients were included (59 71 per department). We compiled a bank of items to be selected for a scale. To assess reliability over time (test retest) a sample of 37 To avoid an acquiescent bias, we included positively and patients was given a second questionnaire in a stamped negatively-worded items [9,35]. These 59 items constituted envelope to be completed and mailed back 1 week later. the main part of the questionnaire; they were meant to evaluate different aspects of inpatient care: medical practice, Statistical strategy medical information to patients and close relatives, respect for Selection of the items patients, availability and attentiveness of nurses, psychological We first performed a principal-components factor analysis and social support, staff support in daily routine, continuity with varimax rotation on the correlation matrix, in order to and co-ordination of care, and discharge management (see separate distinct dimensions and assess construct validity [37]. Appendix). We focused on interpersonal aspects of medical Items with substantial loadings ([ 0.50) on only one factor and nursing care, as previous studies have demonstrated that were to be retained. Stability of these factors was assessed these aspects of care are better predictors of global satisfaction with principal-components factor analysis for different groups than non-medical ones (living arrangements, amenities...) of patients (male or female) and structures (medical or [10,14,15,27,36]. For each item (for example I received clear surgical). information about possible side-effects of my treatment ) At the end of this step, we aimed to maximize the reliability patients rated their opinion (absolutely, quite, not quite, not of the scale with a minimum number of items. To avoid at all, not applicable) and the importance they assigned to redundancy, we removed items for which selection in the this aspect of care (essential, very important, important, not important). scale would not statistically increase their reliability [38]. Cronbach s α coefficient was calculated for each dimension 508

3 Inpatients opinions of quality of care [39]. Items were considered redundant if their removal did more often had a severe disease at admission, a comorbidity, not decrease the coefficient. At each step, the item whose a longer mean length of stay, a difficulty reported by staff removal gave the maximum Cronbach s α coefficient for the during hospitalization and less often returned home after remaining items was selected and the coefficient recalculated discharge. Respondents did not differ from non-respondents for the remaining items in a step-by-step process, until according to sex, time since diagnosis of the main disease, two items remained. This statistical strategy guarantees the type of department (surgical or medical), stay in an intensive unidimensionality of the final set of items [38]. care unit, and hospitalization status (private patients or not). The clinical relevance of items and the percentage of patients who considered the item essential or very important were also considered and discussed before deciding on their Item selection selection. This procedure allowed the construction of a scale After conducting principal-components analysis on the 59 consisting of the remaining items grouped by dimensions main items, a robust two-factor solution emerged. These two (subscales). factors accounted for 34.9% of measured variance (27.5% We calculated component scores and total score by scoring for the first factor, 7.4% for the second) (see Appendix). questions from 0 to 3 (3 always representing the best practice), The first factor, medical information, included 21 items summing these scores and expressing them as a number from with loadings [ 0.50 ( ). Because of their clinical 0 to 100. relevance, we selected two further items concerning the information provided to relatives and discharge management, Evaluation of the scale despite lower loadings (0.41 and 0.46 respectively). These questions were also retained because they were considered By calculating Cronbach s α coefficient, we estimated the to be very important or essential by more than 90% of internal consistency or reliability of the scale and each compatients (Table 2). ponent. Test retest correlation was estimated with intraclass The second factor, relationship with staff and daily routine, correlation coefficients and with weighted κ coefficients included 17 items with loadings [ 0.50 ( ). Two performed for each item. further items were selected despite lower loadings (0.33 and Concurrent validity was assessed using one-way analysis 0.35): one question concerning physical pain was chosen of variance (ANOVA). It tested the association between quality of care scores and global questions measuring different because 91% of the patients considered pain management known attributes of patient satisfaction (e.g. intention to especially important (Table 2). Another question, concerning return to the same hospital). bedside staff meetings and clinical teaching, was also selected Correlation between scores and analogous visual scales because it is a particular practice in French hospitals: the measuring the patient s expectations were measured with doctor involved in the patients care visits them at their Pearson s correlation coefficient. bedside on a daily basis, accompanied by staff and students. Sociodemographic, medical and hospital-stay charponents analysis, 23 items for the factor medical information Thus 42 items were selected following the principal-comacteristics related to the patient opinion scores were first selected with a univariate ANOVA (P level=0.1). A multiple and 19 items for the factor relationship with staff and daily linear regression analysis was performed in order to assess routine. the specific relationship between these selected variables and The step-by-step Cronbach selection procedure allowed the overall score (dependant quantitative variable). us to remove redundant items. By following this procedure, All analyses were performed using SAS software (6.12 we could have removed further items at this step, especially version). from the factor relationship with staff and daily routine, without statistically affecting the scale. Nevertheless questions identified by patients as especially important were also selected: information provided to close relatives, discharge Results organization and pain management. Therefore, 13 items were Non-respondent characteristics selected from the factor medical information and 13 items from the factor relationship with staff and daily routine. Among the 534 patients who were approached, the response rate was 80.2 %. Among the 106 non-respondents (19.8%), 54 patients refused to answer, 11 could not participate due Psychometric properties of the final scale to a language barrier and 41 for reasons of medical incapacity A principal-components factor analysis on this final 26-item (severe asthenia, cognitive troubles etc.). Among the 428 scale showed the same two factors (Table 2), which accounted patients who agreed to complete the questionnaire, comfor for 42.3% of the variance (31.7% for the first factor, 10.6% pletion rates for each item ranged from 93.2 to 100%. the second). Mean scores were 71.9 (SD=21.9) for Respondents were compared to non-respondents with the subscale medical information, 77.7 (SD=15.5) for the regard to demographic, medical and hospital-stay char- subscale relationship with staff and daily routine, and 74.8 acteristics (Table 1). Non-respondents were significantly older (SD=15.9) for the overall score. and were more frequently admitted in emergency than respondents. Reliability of the scale was assessed with Cronbach s α Compared with respondents, non-respondents coefficient values. These were 0.88 for the dimension medical 509

4 L. Salomon et al. Table 1 Demographic, medical and hospital-stay characteristics of patients (collected from physicians) Total Respondent Non-respondent... (n=534) (n=428) (n=106) P-value Demographic characteristics Mean age (years±sd) 58.2 (17.8) 55.5 (17.1) 69.6 (16.3) Sex (% female) NS Characteristics of the main disease <6 months since diagnosis (%) NS Mild severity at admission (%) Comorbidity (%) Hospital care characteristics Emergency admission (%) Mean length of stay (±SD) 8.8 (7.8) 8.5 (7.8) 10.2 (7.7) 0.05 Stay in intensive care unit (%) NS Surgical department (%) NS Discharged to home (%) Private patient (%) NS Difficulty reported by staff (%) NS, Not statistically significant (P> 0.05). Student t-test for age and length of stay and χ 2 test for other variables. information, 0.87 for the dimension relationship with staff The multiple linear regression model showed that, when and daily routine and 0.90 for the overall scale. variables linked to patients evaluation of care were taken Intraclass correlation coefficients between test (scores at into account, four variables were significantly associated with discharge) and retest (scores 1 week later) were 0.59 for the a high score: department, patient s perceived health status, total score, 0.54 for the medical information score and 0.74 difficulty reported by staff, private patient. A tendency to a for the relationships with staff and daily routine score. less favourable opinion of the care was observed for patients Values of weighted κ coefficients ranged from 0.32 (for the admitted in emergency and for patients previously hospitalized item: information about warning signs to look for ) to 1 (for many times (Table 5). the item nurses were too overworked to take care of me ). Kappa coefficients were especially low ( ) for items related to information about treatment or about care after Discussion discharge (warning signs to look for, follow-up, activities and discharge management). Values of the κ coefficients were This questionnaire is the first scale about patient-reported satisfactory (> 0.6) for the other items. experience of care constructed from a study of a population There were significant associations between the scores and of patients living in France and adapted to the French hospital variables chosen for concurrent validity (all P Ζ 0.001): intent system. It has satisfactory reliability and validity. to use or to recommend the department in the future, The questionnaire has satisfactory internal reliability with psychological and physical outcomes (Table 3) and scores a Cronbach s α coefficient > 0.80 for the two subscales and provided by analogous visual scales (Table 4) for the overall scale. Validation studies usually rely on The quality of care scores ranged from 68.9 to 77.1 principal-components analysis and Cronbach s α coefficient according to departments (P=0.02). The score was higher to assess internal reliability. We decided that other criteria for patients with a scheduled hospitalization (P=0.002), no should be measured to assess the quality of the scale, such comorbidity (P=0.03), patients with chronic diseases (P= as the removal of redundant items. The quality of our selection 0.05), private patients (P=0.04), oldest patients (P=0.06), process was assessed by concurrent validity: there was a patients who thought they were in good health (P=0.03), strong relationship between scores and questions measuring patients who have been hospitalized often (P=0.01) and for several known attributes of patient satisfaction (health outwhom no difficulty occurred during the hospitalization (P= come, intent to use or recommend the ward, patient ex- 0.01). Other patient characteristics (sex, nationality, pro- pectations). fessional status, marital status, medical care insurance cov- Our statistical approach was mainly an exploratory one: erage, severity of disease) and hospital-stay characteristics we did not perform tests to confirm our results. This approach (stay in intensive care unit, length of stay, discharge pattern) could be enriched with an explanatory approach, using the were not linked to the patients opinion on the quality of LISREL technique, for example [40]. This type of analysis care score. could be performed in the future using larger set of data. 510

5 Inpatients opinions of quality of care Table 2 Factor loadings produced by principal-components analysis (varimax rotation), importance and opinion ratings for the 26 items selected in the scale Factor loadings... Importance 3 Opinion 4 Factor Factor 2 2 % % I received clear information about: Symptoms The purpose of the tests The results of the tests The purpose of treatments (drugs, operation) The possible side-effects of these treatments Warning signs to look for When to resume activities after discharge Medical follow-up Physicians answered all my questions I was involved in discharge management I could identify the physician in charge of me My relatives were fully informed I was involved in the information session for my relatives I received enough help in my daily routine I received enough help with meals I received enough help with washing I received enough help with going to the toilet I received enough help for my psychological problems I saw nurses as often as I wished Nurses were too overworked to take care of me I was upset by physicians bed-side statements There was good coordination in the department There was a good atmosphere in the department Staff discussed medical issues at my bedside, ignoring me There was enough privacy during medical care Everything possible was done to relieve my pain Medical information subscale. 2 Relationship with staff and daily routine subscale. 3 Percentage of patients who considered the item essential or very important. 4 Percentage of patients who responded absolutely to the items. Lastly, we determined that quality of care scores were chronic diseases) have a higher opinion of the care provided strongly related to departments, regardless of patient and than the other patients [2,14,26]. The same result is observed hospitalization characteristics. These results tend to indicate for patients in good health (perceived health status, lack of that the measurement tool is more sensitive to departments comorbid disease) [2,14], for private patients [34] or patients levels of performance than to patients characteristics or to with planned admissions [34]. modalities of hospitalization as shown in other studies [14, Test and retest scores were correlated, although intraclass 26]. Patients reported that quality of care was statistically correlation coefficients were moderate (especially for the different between departments, as observed by another study medical information subscale). Examination of the item-to- [27], yet rather homogeneous ( on a scale of 0 100). item agreement (κ coefficients) showed that this instability Therefore it is important that each department could identify was located on items related to treatment information and its weak points in order for it to implement specific and patients knowledge of their post-discharge medical care. targeted actions to improve quality of care. These data indicate satisfactory time reliability and may reflect As reported in previous studies, we found that the patients the fact that patients receive much medical information who have had multiple contacts with the care system (elderly in the hour preceding discharge. The intraclass correlation patients, those with multiple hospitalizations, or patients with coefficients would probably have been higher if the retest 511

6 L. Salomon et al. Table 3 Relation between satisfaction scores and health outcomes, intent to return to or to recommend the department (ANOVA) Subscale 1 1 Subscale 2 2 Overall scale... n Mean (±SD) Mean (±SD) Mean (±SD) Health status at discharge 3 Worse (23.9) 45.3 (42.2) 53.1 (28.8) Stable (23.6) 76.1 (16.2) 72.1 (16.7) Somewhat better (21.1) 73.4 (15.8) 70.4 (16.2) Much better (20.3) 82.6 (11.8) 79.9 (13.2) Psychological status at discharge 3 Worse (21.9) 52.3 (18.7) 50.2 (14.1) Stable (21.1) 77.9 (13.8) 74.0 (14.7) Somewhat better (21.3) 75.0 (15.0) 72.8 (15.6) Much better (18.7) 84.2 (10.9) 82.6 (12.0) Intention to return to the department if necessary 3 No (21.2) 51.4 (23.8) 51.8 (15.8) Yes, probably (22.0) 70.4 (13.5) 67.7 (15.3) Yes, definitely (20.3) 82.7 (11.6) 79.5 (13.3) Intention to return for outpatient care 3 No (21.5) 53.0 (24.5) 51.6 (16.0) Yes, probably (10.9) 71.9 (14.4) 68.4 (16.0) Yes, definitely (7.5) 81.0 (12.8) 77.9 (14.1) Intention to recommend the department 3 No (20.3) 50.1 (23.1) 46.8 (11.8) Yes, probably (23.3) 70.0 (14.7) 67.1 (16.3) Yes, definitely (18.5) 82.9 (11.2) 80.0 (12.3) 1 Medical information subscale. 2 Relationship with staff and daily routine subscale. 3 P < Table 4 Relationship between expectations scores provided by analogous visual scales (from did not meet expectations, rated 0, to far exceeded expectations, rated 10) and satisfaction scores Subscale 1 1 Subscale 2 2 Overall scale... r 3 r 3 r 3 Overall quality of hospital care Technical quality of nursing care Technical quality of medical care Quality of information Human qualities of nurses Human qualities of physicians Medical information subscale. 2 Relationship with the staff and daily routine subscale. 3 Pearson s correlation coefficient; all are statistically significant (P< 0.001). had been conducted before discharge, which appeared to be impossible considering the length of our questionnaire and the fatigue of most patients. The questionnaire had good acceptability, as the response rate was 80% and the completion rate per question was very high. Content validity was ensured by the process of questionnaire development: issues identified during the qualitative phase that were important to patients were also rated as important in the questionnaire. In addition, to increase content validity, the importance assigned by patients was an additional criterion of item selection. Quality of care scores could have been modified by several 512

7 Inpatients opinions of quality of care Table 5 Relationship between demographic, medical and hospital-stay characteristics and overall satisfaction score (multiple linear regression) 1... DF F-value 2 P-value Department (n=8) Private patient (yes versus no) Perceived health status compared with people of the same age group (better versus similar versus worse) Difficulty with patient reported by staff (yes versus no) Emergency admission (yes versus no) Age (18 59 years versus [60 years) Number of previous hospitalizations (0 to versus 2 to 3 versus 4 and more) Comorbidity (yes versus no) Time since diagnosis (<6 months versus >6 months) 1 r 2 =0.26 (global measure of fit of the model). 2 Value of multiple regression test for each variable. DF, degrees of freedom. potential biases. First, in spite of many procedures to preserve anonymity and confidentiality, patients may have overrated Acknowledgements several questions thinking that this might influence sub- The authors thank Professor J. Lellouch (National Institute sequent care, which probably overestimated the score re- of Health and Medical Research, Unit 472: Epidemiology lationship with staff and daily routine. Secondly, respondents and Biostatistics) for his thorough review of the drafts of were healthier that non-respondents, and healthier patients this manuscript. We also thank the Evaluation and Quality had significantly higher scores (probably because they re- Department (Medical Policy Direction, AP-HP) and Dr P. quired less assistance, felt less discomfort, and were less Durieux (Department of Public Health, Cochin Hospital, affected by omissions of care). This difference may have AP-HP) for their logistic support and various Departments induced an over-estimation of patient opinion on quality of of Cochin Hospital, AP-HP, Paris, for their participation: care. Third, because patients receive much medical in- Internal Medicine (Prof. Sicard), Rheumatology (Prof. Douformation in the hour preceding discharge, the information- gados), Cardiology (Prof. Weber), Gastro-enterology (Prof. related score was probably underestimated. Couturier), Urology (Prof. Debré), Abdominal Surgery (Prof. The final scale contained two factors: medical information Chapuis) and Orthopedic Surgery (Prof. Kerboull and Prof. and relationship with staff and daily routine. The second Tomeno). factor contained various items described as usual components of patient satisfaction in other surveys. The dimension medical information was preponderant and accounted for 31.7% of variance. The importance given to patients information References is not so preponderant in English-speaking satisfaction scales 1. Kravitz RL, Callahan EJ, Paterniti D et al. Prevalence and [5,15,19]. The importance given to medical information is sources of unmet expectations for care. Ann Intern Med 1996; not surprising in France where physicians are culturally 125: reluctant to give precise information to their patients. This attitude is now being questioned under the influence of recent 2. Cleary PD, Edgman-Levitan S, Roberts M et al. Patients evaluate legal modifications and case law [32,33]. The burden of proof their hospital care: a national survey. Health Affairs. 1991; 10: is now on the doctor rather than on the patient. This result supported our choice of developing a specific French 3. Kelstrup A, Lund K, Lauritsen B et al. Satisfaction with care instrument. reported by psychiatric inpatients. Acta Psychiatr Scand. 1993; 81: This scale measures inpatients opinions of care. It was found to be valid and reliable and it provides a score related to hospital structures. Further studies are now needed to 4. Fitzpatrick R. Surveys of patient satisfaction: I. Important assess the possibility of generalizing our results. general considerations. Br Med J 1991; 302:

8 L. Salomon et al. 5. Hall JA, Dornan MC. Meta-analysis of satisfaction with medical 22. Rubin HR. Patient evaluations of hospital care: a review of the care: description of research domain and analysis of overall literature. Med Care 1990; 28 (suppl.): S3 S9. satisfaction levels. Soc Sci Med 1988; 27: Delbanco TL. Enriching the doctor-patient relationship by 6. Wolinsky FD, Steiber SR. Salient issues in choosing a new inviting the patient s perspective. Ann Intern Med 1992; 116: doctor. Soc Sci Med 1982; 16: Baygley KB, Grunkemeier GL, Lansky DJ. Researching the 24. Fitzpatrick R. Surveys of patient satisfaction: II Designing a success of treatment in patients terms. Med Care. 1995; 33 questionnaire and conducting a survey. Br Med J 1991; 302: (suppl.): S226 S Weingarten SR, Stone E, Green A et al. A study of patient 25. Cohen G, Forbes J, Garraway M. Can different patient satsatisfaction and adherence to preventive care practice guidelines. isfaction survey methods yield consistent results? Comparison AmJMed1995; 99: of three surveys. Br Med J 1996; 313: Ross CK. A comparative study of seven measures of patient 26. Hall JA, Dornan MC. Patient sociodemographic characteristics satisfaction. Med Care 1995; 33: as predictors of satisfaction with medical care: a meta-analysis. Social Sci Med 1990; 30: Bruster S, Jarman B, Bosanquet N et al. National survey of hospital patients. Br Med J 1994; 309: Carmel S. Satisfaction with hospitalization: a comparative ana- lysis of three types of services. Soc Sci Med 1985; 21: Tamblyn R, Benaroya S, Snell L et al. The feasibility and value of using patient satisfaction ratings to evaluate internal medicine 28. Oberst MT. Patients perceptions of care: measurement of residents. J Gen Intern Med 1994; 9: quality and satisfaction. Cancer 1984; 10 (suppl.): Bark P, Vincent C, Jones A, Savory J. Clinical complaints: a 12. Stump TE, Dexter PR, Tierney WM et al. Measuring patient means of improving quality of care. Qual Health Care 1994; 3: satisfaction with physicians among older and diseased adults in a primary care municipal outpatient setting. An examination of three instruments. Med Care. 1995; 33: Hulka BS, Zylanski SJ. Validation of a patient satisfaction scale. 13. Nelson EC, Larson C. Patients good and bad surprises: how Theory, methods and practice. Med Care 1982; 10: do they relate to overall patient satisfaction? Quart Rev Bull 31. Health Service Research Group, A guide to direct measures of 1993; 19: patient satisfaction in clinical practice. Can Med Assoc J 1992; 14. Charles C, Gauld M, Chambers L et al. How was your hospital 146: stay? Patients reports about their care in Canadian hospitals. 32. Inpatient s Charter attached to ministerial instruction n Can Med Assoc J. 1994; 150: May 6th Paris, France. 15. Abramowitz S, Coté AA, Berry E. Analyzing patient satisfaction: 33. New medical code of practice D. n September 6th a multianalytic approach. Quart Rev Bull 1987; 13: Paris. 16. Fielding JE, Lancry PJ. Lessons from France Vive la dif- 34. Doering ER. Factors influencing inpatient satisfaction with care. férence. The French health care system and US health system reform. J Am Med Assoc 1993; 270: Government order n , April 24th Reforming public and private hospitalisation. Journal Officiel April 24th 1996; Paris. 18. Davies AR, Ware JE. GHAA s consumer satisfaction survey and user s manual. Group Health Association of America Inc. 1988; 29 pp. Qual Rev Bull; 1983; 9: Ware Jr JE. Effect of acquiescent response set on patient satisfaction rating. Med Care 1978; 16: Cleary PD, McNeil BJ. Patient satisfaction as an indicator of quality care. Inquiry 1988; 25: Bernstein IH, Garbin CP, Teng GK. Applied multivariate analysis (chapter 12). New York: Springer Verlag, Carey RG, Seibert JH. A patient survey system to measure quality improvement: questionnaire reliability and validity. Med 38. Curt F, Mesbah M, Lellouch J et al. Handedness scale: How Care 1993; 31: many and which items? Laterality 1997; 2: Ware JE, Hays RD. Methods for measuring patient satisfaction 39. Cronbach LJ. Coefficient alpha and the internal structure of with specific medical encounters. Med Care. 1988; 26: tests. Psychometrika 1951; 16: Meterko M, Rubin HR. Patient judgments of hospital quality: 40. Houts SS. Use of LISREL in scale validation. Psychol Rep 1990; a taxonomy. Med Care 1990; 28 (suppl.): S10 S14. 67:

9 Inpatients opinions of quality of care Appendix. Initial principal-components analysis (varimax rotation) with the 59 items of the initial questionnaire... Factor 1 Factor 2 Variance explained Eigenvalue Medical Information I received clear information about: My disease Symptoms The purpose of the tests The results of the tests Possible pain related to tests The purpose of the treatments (drugs, operation...) Possibilities of treatment Possible side effects of these treatments The possible development of my disease Warning signs to look for When to resume activities after discharge A healthier way of life I was given the name of the drugs I took The physicians and nurses gave me contradictory information My relatives were fully informed I was involved in information sessions for my relatives Medical practice I was upset by physicians bedside statements I felt uncomfortable during bedside staff meetings Everything possible was done to relieve my pain I was involved in the choice of my treatment I underwent a thorough medical examination I saw the physicians as often as I wished The physician in charge of the ward took care of me I was encouraged to ask questions by the physicians Physicians were fully available Physicians answered all my questions I understood the explanations of the physicians Availability and attentiveness of nurses I saw the nurses as often as I wished The nurses were too overworked to take care of me I was encouraged to ask questions by the nurses The nurses translated physicians statements for me The nurses described to me the care they were about to give me The nurses were fully available Respect of patients Care providers entered my room without knocking My beliefs and culture were respected I was treated like a child There was enough privacy during medical care The nurses introduced themselves to me The physicians introduced themselves to me I was respected as a human being The staff discussed medical issues at my bedside, ignoring me

10 L. Salomon et al. Appendix continued... Factor 1 Factor 2 Psychological and social support I received enough help for my psychological problems I received enough help for financial and administrative problems The nurses and physicians paid attention to my psychological problems I received information on the possibility of meeting with a psychologist I felt uncomfortable during bedside staff meetings Staff support and daily routine I received enough help in my daily routine I received enough help with meals I received enough help with washing I received enough help with going to the toilet There was a good atmosphere in the department Continuity and co-ordination I received information on the possible length of my hospitalization There was good co-ordination in the department The ward schedule was explained to me I could identify the physician in charge of me I could identify the nurse in charge of me There was a good co-ordination between departments Discharge management The department will get in touch with my general practitioner I received clear information about medical follow-up I was involved in discharge management Items in italics were selected for the final scale. 516