Predicting and comparing patient satisfaction in four different modes of health care across a nation

ARTICLE IN PRESS

Social Science & Medicine 63 (2006) 1671–1683 www.elsevier.com/locate/socscimed

Predicting and comparing patient satisfaction in four different modes of health care across a nation Annemieke P. Bikkera, Andrew G.H. Thompsonb, a

Independent researcher School of Social and Political Studies, University of Edinburgh, Adam Ferguson Building, George Square, Edinburgh EH8 9LL, Scotland

b

Available online 6 May 2006

Abstract This study aims to inform strategic policy makers and managers about the value of general population surveys by determining and comparing dimensions of satisfaction in four different health services in Scotland: general practice, domiciliary care, outpatients and inpatients (including day cases). The research design involved secondary data analysis of a national telephone survey conducted to inform the development of a national health plan. The database was created using a stratified quota sample of 3052 people of 16 years and above resident in Scotland in 2000. The main outcome measures investigated were overall measures of patient satisfaction with each type of service. Principal components analysis was used to determine the dimensions. Interest was in the extent to which patients, many of whom were the same (having used more than one service), evaluated different services in similar ways, as well as those factors specific to each service. Using logistic regression, the results demonstrate that interpersonal care and information, and desired improvements in service were universal and key explanatory dimensions in all services, followed by a combination of access, physical facilities, time and quality of food, depending on relevance to the service. These factors, particularly interpersonal care and information, distinguished well the highly satisfied from the others, with age providing further discrimination between nonhospital patients, while gender added to discrimination between inpatients. In conclusion, despite the limitations of telephone interviews, it is feasible to ask about several services at the same time and for the answers to reflect common underlying dimensions of evaluation found in more exhaustive research within each service. These factors offer a set of summary measures by which services can be easily evaluated at a strategic level and point to where efforts to increase patient satisfaction can be maximised. r 2006 Elsevier Ltd. All rights reserved. Keywords: Patient satisfaction; General practice; Domiciliary care; Outpatients; Inpatients; Population surveys; Scotland; UK

1. Introduction

Corresponding author. Tel.: +44 131 651 1562;

fax: +44 131 650 6546. E-mail address: [email protected] (A.G.H. Thompson). 0277-9536/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2006.03.022

There has been a steadily growing policy interest in the UK in ways in which patients can increase their involvement in the planning, delivery and evaluation of health care, from the demands of voluntary and community activists in the 1970s, through the development of a consumerist agenda

ARTICLE IN PRESS 1672

A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

in the 1980s and early 1990s, to a more citizenfocussed set of policies in the last few years. The modernisation agenda of stakeholder engagement and inclusiveness of the New Labour Government since 1997 was initially set out in the White Paper ‘The New NHS’ (Department of Health, 1997) and ‘A First Class Service’ (Department of Health, 1998). Subsequent to the devolution of responsibility for health care to the constituent nations of the UK in 1999, each one has developed similar but distinctive policies in relation to making progress with the citizen involvement agenda. In Scotland, the policy direction was set out initially in ‘Designed to Care’ (Scottish Office, 1997) and since devolution in the Health Plan for Scotland, ‘Our National Health: a plan for action, a plan for change’ (Scottish Executive Health Department, 2000), which place patients as key partners in health and social care development. More recently, ‘Partnership for Care’ (Scottish Executive Health Department, 2003) reinforces this notion by equating the priority of patients’ views with clinical standards and financial performance. In order to inform the Health Plan for Scotland, two commercial polling companies, MORI Scotland and System Three, were commissioned to carry out a survey of public perceptions and experiences (as patients) of the NHS in Scotland (Scottish Executive Central Research Unit, 2001). It is this dataset that forms the basis for the analysis presented here. Patient evaluation of health services has long been seen as a legitimate and necessary part of the patient involvement project for a variety of reasons, depending in part on the viewpoint of the particular stakeholder group. In the UK some of the earliest work has been in groups that have included academic social scientists concerned with sociological (Cartwright, 1964), psychological (Raphael, 1967) and managerial understandings (Moores & Thompson, 1986); health care professionals in nursing (McGhee, 1961), hospital medicine (Hill, Bird, Hopkins, Lawton, and Wright (1992), hospital surgery (Meredith, Emberton, & Devlin, 1993), and general practice (Baker, 1990); users’ organisations (Jones, Leneman, & Maclean, 1987); and managers and policy makers involved in the policy developments referred to above, beginning with the Royal Commission on the NHS (Gregory, 1978) and accelerated by the recommendations of the Griffiths Report on managing the NHS (Department of Health and Social Security, 1983). The reasons for this high level of interest have been manifold,

including pressures to democratise public service provision through greater public accountability (Barnes, 1997; Hogg, 1999), a growth in consumerism in public policy (Avis, Bond, & Arthur, 1995), concerns about treatment concordance (Wright, 1993), professional ethics (Grol, 2001), and the drive to improve the quality of care with respect to treatment outcomes and social acceptability of the processes (Coulter, 1991). However, the way in which services have been subjected to user evaluation has generally been limited to measures of satisfaction, sometimes labelled subjective indicators, and/or patient reports, believed by some to be objective indicators (Ware, Snyder, Wright, & Davies, 1983; Wensing and Elwyn, 2002). As Crow et al. (2002) make plain, patient satisfaction has been a dominant strand, with their search of 7 electronic databases uncovering more than 270,000 ‘hits’ in articles published mainly in the English language incorporating this theme over the period 1980–1998. Despite the continuing concern and criticism that patient satisfaction is a poorly theorised concept (Williams, Coyle, & Healy, 1998), it has become a major source of feedback to the NHS at the national level in England (Airey et al., 1999) and at lower organisational and professional levels across the UK. The Scottish Executive has pursued a different route, encouraging local surveys while sponsoring an ad hoc, national population survey of the major NHS services. While a large volume of research has grappled with the meanings and determinants of satisfaction, the growth in routine systems of incorporating patients’ views demands that attention be paid to what might be of value within these measures. This will hold true whether it is based on surveys of the general population for use by policy makers and managers at the system or programme level, or of specific services to indicate where health care practitioners and operational managers might effectively focus their efforts to improve quality in a patient-centred way. Relatively little research has investigated in a deeper analytical way comparisons between services to explore whether patient evaluations, especially by those who use several services within a particular time period, follow similar lines of rationality across services, or whether different contexts provoke divergent criteria. A notable exception is the work of Williams and Calnan (1991), who used a self-completion postal questionnaire to compare the criteria for assessing

ARTICLE IN PRESS A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

general practice, community dentistry and inpatient hospital care. Their conclusions were that, while different contexts offer specific and distinct issues of concern to patients, there were common predictors of global satisfaction in all three settings, labelled professional competence and patient-professional relationships, with age offering further explanation for satisfaction. In line with several other scholars (Jenkinson, Coulter, Bruster, Richards, & Chandola, 2002; Thompson, 1986) they also argued for attention to specific dimensions of care rather than the more general, global outcome measures. Nevertheless, while in accord with this view, Fitzpatrick (2002) also sees value in including global questions as a way of identifying the most important issues for health care providers, in addition to the specifics of care. The increasing interest at the strategic level in population-level surveys of patient satisfaction requires firmer evidence of its value across services and whether the convergence in evaluation exists across a broader range of services in different localities. Given the wide variety of instruments used to measure satisfaction it would also be very useful to identify the common dimensions that patients use in their evaluations in a more robust way. Virtually all measures of satisfaction either exhibit distributional problems of skew and kurtosis, leading to biased estimates of population parameters, or require careful attention to the level of measurement. Thus, techniques that rely on assumptions of interval measurement, such as Pearson’s correlation coefficient or multiple regression, require careful handling, especially with relatively small sample sizes, when the questions use ordinal (e.g. Likert) or nominal (yes/no) answer formats. Therefore, this study was designed to determine the underlying dimensions of patients’ perceptions of quality, as indicators of satisfaction, within the four different public health services of general practice, domiciliary care, outpatient hospital care and inpatient hospital care. The context was set to be within the jurisdiction of one National Health Service, NHS Scotland, where we could be sure of a common policy framework. This would then offer the possibility of exploring the degree of evaluative convergence across services, as well as providing a robust basis for determining some of the likely key explanatory variables of the measures of global satisfaction for each service, as suggested by extant literature. In addition to the expected dimensions,

1673

many studies have pointed to the importance to explanation of certain socio-demographic variables, including age, which consistently shows older people as more satisfied (Crow et al., 2002), gender, which is inconclusive as to whether men or women are more satisfied (Crow et al., 2002), education, which tends to show the higher educated as being more critical, although less pronounced in the UK (Sitzia and Wood, 1997), and health status, which generally shows lower satisfaction among those in poorer health (Cohen, 1996; Crow et al., 2002). Methods Sample This study provides a secondary analysis of a dataset created from a survey of 3052 adults’ opinions of the NHS in Scotland (Scottish Executive Central Research Unit, 2001), sampled from those aged 16 years and above resident in Scotland. Stratified quota sampling reflected the characteristics of the population regarding age, gender, social class and housing tenure (as a partial proxy for wealth and social class). The quotas were based on the Scottish Household Survey (Scottish Executive Central Research Unit, 2000), which was used to weight the respondents’ profile to improve the representativeness of the sample to the national population profile. In fact, the degree of weighting was relatively small, with most adjustment required for gender (reducing the female proportion from 59% to 53%) and age (increasing the proportion of 16–24 year olds from 9% to 13%). Potential participants for the structured telephone interviews were contacted using Random Digit Dialling (RDD). Interviews were conducted over a 25-day period (28 September–22 October 2000) using Computer Assisted Telephone Interviewing, with an average interview length of 19 min. The number of questions asked depended on the number of NHS services used in the past year, which varied between 14 items for those who used no services and 195 items for 4 services, with a mean per respondent of 85. A set of descriptive, weighted statistics for each question item can be found from the Scottish Executive Central Research Unit (2001) report. Questionnaire The questionnaire consisted of 211 items, chosen to reflect the recent organisational changes being

ARTICLE IN PRESS 1674

A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

pursued under the modernisation agenda, of which 177 items were used for this secondary analysis. The variables that formed the core of this analysis related to users’ last experiences within the previous 12 months regarding various aspects of the services of general practice, domiciliary care, outpatients, and inpatients, and to the background variables of gender, age, education and self-reported health status of the respondents. In addition, questions relating to global views of quality of each service were included. For each service the questions were grouped in the questionnaire under the following headings: access, waiting, service experience (where respondents were asked to grade statements rated on a 5point scale of perceptions, where 1 ¼ poor to 5 ¼ excellent), and aspirations for future improvement (rated on a 5-point scale, where 1 ¼ not at all important for improvements, to 5 ¼ essential to make improvements). Moreover, an item measuring the overall satisfaction score of each service was included on a 5-point Likert scale from very satisfied (5) to very dissatisfied (1). Most of the questions in the survey allowed for a ‘Don’t Know’ (DK) response, which was not treated as part of the measurement scale and consequently treated as missing in the analysis. In order to make the scales uniform between variables and to simplify interpretation of findings, many items were recoded such that scales would be in the same direction, with the lowest rating having the lowest number. In addition, education was regrouped from 10 into 3 categories (no formal qualifications, formal school qualifications, postschool higher qualifications). Analysis This secondary analysis is based on the original, unweighted socio-demographic distribution of the achieved sample, since we were not attempting to estimate the true population parameters for each questionnaire item, but rather to explore the interrelationships between variables. In order to assess the structure of the dataset, each item was examined for normality of distribution, outliers and missing data, within which the number of DK responses was given particular attention. As it was deemed important to see whether factors expected from previous research would emerge in this dataset, we employed exploratory factor analysis to reduce the questionnaire items to

a coherent set of variables that specified the dimensions of user satisfaction related to each service. We used principal components analysis (PCA)1 with varimax rotation in order to facilitate the identification of the variables with the factors, although we also confirmed that almost identical structures emerged using principal axis factoring and oblique rotation. The cases with missing values (including DK) were excluded listwise2 to ensure comparability of those at each stage of the analysis, although it did reduce the sample size. Imputation was not deemed appropriate since the majority of DK responses expressed non-attitudes, due to inexperienced issues rather than ambivalence; e.g. disabled facilities; public transport; children. The factorability of the set was assessed by visually examining the correlation matrix, while the Kaiser–Meyer–Olkin (KMO) and Barlett test of sphericity statistics were used to test empirically whether the data were likely to factor well. The measure of sampling adequacy (MSA) was inspected in the anti-image correlation matrix to weed out any variables with a value of o0.5. The sets of factors for each service were based on a combination of criteria, including Kaiser’s criterion, Cattell’s scree plot, the proportion of explained variance, and above all, through examining whether the factors could be interpreted meaningfully. The cut-off value of 0.3 was used as a guide to determine whether a factor loading was high enough in order to consider any given variable as a defining part of that factor. Logistic regressions were employed to examine which dimensions of user satisfaction identified above and which user characteristics had an effect on overall satisfaction with each service. Multicollinearity among the predictors was tested by the variance-inflation factor. The most discriminating variables within the final model were chosen using the criteria related to the goodness of fit of the overall model, namely the reduction in the log likelihood value (
2LL), the Hosmer and Lemeshow value, the significance of the logit coefficients 1

While PCA assumes normality of data distributions, departures from this are not a problem if used for description rather than hypothesis testing (Tabachnick & Fidell, 2001, p.588). Thus, factor scores will reflect the skew and kurtosis of the original variables. Nonetheless, correlations will be lower between dissimilar distributions than they should be, which could degrade the factor structure. 2 Listwise deletion excludes cases with missing values for any variables within the specific analysis. It is the most stringent assumption and it can reduce the sample size considerably.

ARTICLE IN PRESS A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

using the Wald coefficient, and the percentage explained correctly in the classification table. The final model was investigated in a hierarchical, blockentry manner, entering all the factors after controlling for the socio-demographic variables. The four dependent variables were severely negatively skewed, having over 50% of the responses in the ‘very satisfied’ category and a further 30% or more in the ‘fairly satisfied’ category. In order to circumvent some of the analytical problems incurred by such distorted distributions, we transformed the dependent variables into dichotomies, and a distinction was made between being very satisfied (coded 1) and not being very satisfied (coded 0). While this limits the analysis, it is a technique used by others to deal with problems of skew and limited sample size (Jackson, Chamberlin, & Kroenke, 2001). It has also been strongly recommended by Collins and O’Cathain (2003) that ‘very satisfied’ and ‘satisfied’ should not be collapsed into one category, as is commonly done, due to the distinct meanings embodied in each of these levels of evaluation. The reliability of the findings of the factor analyses and the logistic regressions were checked and verified by jack-knifing methods through redoing the analyses on random sub-samples of 60% of the original sample, because no other comparable data were available for independent verification. Results Sample characteristics Of the 3052 in the total sample3, 635 (20.8%) had not used any of the four services in the previous year. Within the user-group, three-quarters had used the GP service, almost 10% had received domiciliary care, just over a third had attended the outpatient clinic, and nearly 14% had been a hospital inpatient (of whom 37% were day pa3

The gross numbers approached to participate and nonrespondent characteristics were not available, meaning that the response rate could not be directly estimated. This is a common problem in quota sampling, where the next available person is chosen when the intended respondent is unavailable or difficult to trace. In this way the intended sample size is reached using randomly chosen respondents, albeit effectively excluding an unrecorded proportion. A similar study in Northern Ireland (DHSSPS, 2003) has estimated the true response rate to be 65% of potential respondents.

1675

tients). These statistics and the gender, age and health status variables are shown in Table 1 for each service. Most participants (42% of user-group) had used only one of the services in the previous year, while 24% had used two, 10% had used three and just under 4% had used all four services. Initial data analysis The Kolmogorov–Smirnov (K–S) test failed to identify any normal distributions among the variables (2.697oK–So21.003), which could not effectively be normalised through transformations due to severe negative skew (and related outliers) especially visible in the global satisfaction questions and severe negative kurtosis in the aspiration questions. Through the use of w2 tests, the DK responses were checked for overall randomness and whether they formed systematic relationships either with participants possessing certain socio-demographic characteristics, suggesting sample bias, or with certain questionnaire items, indicating that the phrasing used or issues explored by certain questions may require scrutiny. Some 2% (7261) of responses were recorded as DK for reasons other than non-use of a service, affecting 94% of the items in the analysis. However, most items exhibiting high levels of DK (10%+) concerned non-use, such as facilities for disabled people. While there was a statistically significant, positive association, with high power, between age and DK responses, the low effect size (W) suggests it may not be practically meaningful (w2 ¼ 15.809, df ¼ 6, po0.005, W ¼ 0.095, power0.05 ¼ 0.86). There was no statistically significant relationship between DK and gender, education or self-reported health status. Thus, there would not appear to be any obvious bias through the exclusion of these cases from each analysis. An important further consideration is the difference between those respondents who used any of the four services under consideration and those who used none, who would effectively be excluded from the analysis. There was a statistical difference for gender (w2 ¼ 52.323, df ¼ 1, po0.005, W ¼ 0.29, power0.05 ¼ 0.99), indicating that more men than women had not used any of these services, which may partly account for the lower proportion of willing male respondents to the survey. No clear statistical difference (p ¼ 0.065) was found between the age profile of users and non-users. When age was controlled by gender, a significant association

ARTICLE IN PRESS 1676

A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

Table 1 Characteristics of the respondents, broken down by the non-users and users of the NHS in the last year and by each service n ¼ 3052

All respondents

Users

Non-users

General practice

Domiciliary care

Outpatients

Inpatients

Users

n

%a

n

%

n

%

n

%

n

%

n

%

Response rate

635

21

2417

79

2299

75

296

10

1050

34

416

14

Gender Male Female

350 285

55 45

891 1526

37 63

834 1465

36 64

81 215

27 73

385 665

37 63

167 249

40 60

Age 16–24 25–34 35–44 45–54 55–64 65–74 75+

77 110 160 117 92 68 17

11 17 25 18 15 11 3

202 459 516 440 391 288 121

8 19 21 18 16 12 5

192 445 494 420 372 270 106

8 19 22 18 16 12 5

25 72 54 31 40 36 38

8 24 18 11 14 12 13

84 169 213 193 192 133 66

8 16 20 18 18 13 6

40 80 78 74 70 46 28

10 19 19 18 17 11 7

Health status Good Fairly good Not very good

507 116 12

80 18 2

1286 777 351

53 32 15

1227 733 337

53 32 15

102 91 103

34 31 35

429 376 244

41 36 23

142 140 134

34 34 32

Educational status Post-school higher qualifications Formal school qualifications No formal qualifications

207 203 186

35 34 31

767 880 611

34 39 27

734 844 568

34 39 27

77 114 82

28 42 30

312 385 285

32 39 29

107 156 118

28 41 31

a

All percentages rounded to nearest integer.

was found for men (g ¼ 0.151, po0.001) showing increasing use with age, but not for women. A Mann–Whitney U test revealed that non-users had proportionately a higher self-reported state of health than service users (z ¼
12.665, po.005), with 80% of the former in good health, compared to 53% of the latter. A similar test for education failed to reveal a statistically significant association (z ¼
0.873). In summary, the initial data analysis showed that a number of considerations need to be kept in mind when analysing and interpreting the results. The lack of normality in the global satisfaction measures suggests that multiple regression would not be a good choice for analysing the relationships with the hypothesised predictor variables. Moreover, while deletion of cases with DK responses does not seem to affect the randomness of the data, it does diminish the sample size (2417) and it causes even more distortion in the gender disparity, increasing the number of female respondents to 63% of the total.

Deriving the dimensions of health care Table 2 shows the results of the factor analyses for each of the four services. Absolute values under 0.3 are omitted to facilitate interpretation. Five factors were derived for the GP service (explaining 60% of the variance), three factors for domiciliary care (76%), four factors for outpatients (61%), and five factors for inpatients (68%). Table 3 shows the interpretative labels given to the extracted factors, which can be surmised to represent the underlying latent dimensions of ideal expectations (Thompson & Sun˜ol, 1995) in factor 1 and user satisfaction in the other factors. The most explanatory factor for all services consists of the aspirational items relating to ‘improvement of service’, explaining 25% or more of the total variance. Additionally, the second highest factor, labelled ‘interpersonal care and information’, is part of all services. The factors with high loadings from items measuring ‘physical facilities’ and ‘access’ appeared in three of the four sets. The factor

ARTICLE IN PRESS A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

interpreted as ‘quality of food’ was only appropriate for inpatients, while ‘physical facilities’ and ‘access’ were inappropriate for domiciliary care. Time featured in both non-hospital services. These hypothetical predictor variables will now be examined for each service in turn. Predicting global satisfaction in all four services The results of the logistic regression of the dimensions of user satisfaction and the sociodemographic variables influencing the global satisfaction with each of the services are presented in Table 4. Multicollinearity among all the predictors was very low in all services, as tested by the variance-inflation factors (max VIF ¼ 1.43). For each service a test of the model against a constantonly model was statistically reliable (po0.0001), implying that the predictor variables as a set reliably predict overall satisfaction. The variance accounted for ranged from 53.3% to 63.3%. The dimension of aspirations for improvement is included in all models, with significant odds ratios (po0.001) of less than 1.0, indicating an inverse relationship with satisfaction. Predicting global satisfaction with general practice All four dimensions of user satisfaction are included in the model, and the odds ratios show that ‘interpersonal care and information’ has the highest influence, followed by time, access, and physical facilities. Also, the background variable of age is included, with the positive coefficient showing that older patients are more likely to be very satisfied. The classification table reveals a prediction success rate of 84% for very satisfied users and 75% for not very satisfied users, an overall success rate of 80%, showing a 26% improvement on a random estimate (i.e. without this model). Predicting global satisfaction with domiciliary care The only dimension of user satisfaction included in the model is ‘interpersonal care and information’, while age also provides some prediction, showing that older patients are more likely to be very satisfied. The classification table showed a prediction success of 95% for very satisfied and 74% for not very satisfied users, giving a total success rate of 88%, a 19% improvement on a random estimate.

1677

Predicting global satisfaction with outpatients Three satisfaction dimensions are included, with the odds ratios showing that ‘interpersonal care and information’ once again has the highest influence, followed by physical facilities and finally access. As shown by the confidence interval, age is marginally non-significant in this service. The classification table showed a prediction success of 82% for very satisfied and 80% for not very satisfied users, giving an overall success rate of 81%, an improvement of 30% compared to a random estimate. Predicting satisfaction with inpatients Three dimensions of user satisfaction are included in the model and the odds ratios show that ‘interpersonal care and information’ has the highest influence, followed by quality of food and finally physical facilities. This time gender is included as a predictor, but age is not significant. The model indicates that more women are very satisfied with the service for inpatients, although it shows high variability in the estimate. The classification table showed a prediction success of 87% for very satisfied and 80% for not very satisfied users, giving an overall success rate of 83%, an improvement of 33% compared to a random estimate. Discussion The aim of this investigation was to determine and compare dimensions of patient satisfaction in the four selected services, in a way that would offer statistically robust measures and analyses. In common with Williams and Calnan (1991), we found both divergent dimensions which were specific to the contextual factors of particular services, as expected, and convergent factors across all services. These latter, statistically based dimensions included aspirations for service improvement, and interpersonal care and information. The first is likely to be an inverse reflection of what was perceived to be less than highly satisfactory and therefore signifies logical rationality rather than a focus for action. However, the second dimension flags a crucial focus for attention by both professionals and managers intent on improving quality of care from the patient perspective. It is comparable to a combination of the convergent dimensions of professional competence and patient–professional

Service experiencec Ease of getting through to surgery by phone Availability of convenient appointment times Availability of convenient public transport for visitors Having ramps to give access to disabled people Convenience of visiting times Choice of food Choice of when to eat Convenience of car parking Convenience of public transport routes Comfort of waiting room/area Quality of food Level of noise in ward Having child friendly facilities in waiting room General standard of cleanliness Availability of toilets Doctor/nurse having enough knowledge of situation Having privacy respected Being treated with dignity

Waiting Waiting time to get an appointment Time spent in waiting room to see doctor Waiting for GP/nurse to arrive

c

0.74 0.58 0.69

0.34

0.30

0.47

0.68 0.67 0.69

0.62 0.62 0.57

0.81 0.64 0.75

0.38


0.31

60.6% (552) 1 2 3 27.6 17.1 10.4 0.96 0.92 0.80

0.73

0.85 0.77 0.80

0.88

3 8.0 0.76

0.63

0.31

0.72


0.50

(175) 2 31.9 0.92

0.41

76.4% 1 36.5 0.98

0.33

0.60 0.67

0.47

5 4.9 0.62

Outpatients

0.49

0.56 0.72


0.60
0.36

0.50

4 7.0 0.77

Domiciliary care

0.51

59.5% (1273) 1 2 3 25.3 14.8 7.5 0.96 0.93 0.81

Total variance explained (n) Factor Variance explained by factor (%) Cronbach’s ab

Accessc The ease/difficulty of getting to surgery/hospital Convenience of surgery/clinic opening hours Distance to hospital from home

General practice

Questionnaire item

Table 2 Internal variable structure and loadings after principal components analysis with varimax rotationa

0.41 0.68

0.55 0.32
0.58

4 5.5 0.63

0.84 0.67 0.77

0.42

0.43

0.32

0.78

0.43 0.72 0.67

67.6% (274) 1 2 3 25.4 19.7 11.2 0.96 0.93 0.89

Inpatients

0.30

0.67 0.61

0.48

4 5.8 0.83

0.37

0.77


0.64

0.56

5 5.6 0.60

1678 A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

ARTICLE IN PRESS


0.32


0.33
0.32
0.30
0.36
0.35


0.42

0.58 0.78 0.73 0.85 0.88 0.89 0.88 0.83 0.82 0.83 0.82
0.32


0.39


0.49

0.52 0.66


0.54


0.35


0.44
0.57

0.50

0.65

0.63 0.52

0.80 0.75 0.72 0.80 0.77

0.91 0.94 0.95 0.89 0.92 0.87 0.91 0.91

0.83

0.88 0.88 0.84 0.87 0.88

0.89


0.45

0.52
0.47 0.79 0.78 0.89 0.91 0.91 0.90 0.86
0.31 0.86 0.87 0.88 0.77

0.63

0.60

0.53

0.76

0.85 0.81 0.78 0.84 0.80 0.57


0.49


0.35

0.62 0.67 0.84 0.86 0.89 0.88 0.82 0.85
0.30 0.86 0.86 0.78 0.74

0.57

0.38

0.37

0.57 0.33

0.49

0.84 0.83 0.81 0.85 0.79 0.71 0.62

b

Variables are presented in order as they appeared in the telephone schedule. Questions were asked for each service only where they were appropriate. Reliability scores are presented for those variables that are shown as having a loading Xj.3j for each factor, as in the table. c These are the labels used in the original questionnaire.

a

Aspirations for improvementc Acceptable waiting time for appointment Ease of getting through to surgery by telephone Availability of convenient appointment times Availability of convenient public transport for visitors Having ramps to give access for disabled people Convenience of visiting times Choice of food Convenience of car parking Choice of when to eat Convenient public transport routes Quality of food Comfort of the waiting room Level of noise in ward Having child friendly facilities in the waiting room General standard of cleanliness Availability of toilets Doctor(s)/nurse(s) having enough knowledge of situation Having privacy respected Being treated with dignity Being listened to by the doctor(s)/nurse(s) Being encouraged to ask questions Having enough time with the doctor/nurse Being given information about treatment Usefulness of given information Speed of results from tests Discharge arrangements from hospital

Being listened to by doctor(s)/nurse(s) Being encouraged to ask questions Having enough time with doctor/nurse Being given information about treatment Usefulness of given information Speed of results Discharge arrangements from hospital


0.76


0.72


0.43
0.77


0.55
0.39


0.40


0.54

A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

ARTICLE IN PRESS 1679

ARTICLE IN PRESS A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

1680

Table 3 Interpretative labels of the factors for the services Factor

1

General practice Domiciliary care Outpatients Inpatients

Improvement Improvement Improvement Improvement

2 of of of of

service service service service

Interpersonal Interpersonal Interpersonal Interpersonal

care care care care

relationships found in Williams and Calnan’s study (1991). The context-specific issues, however, were remarkably similar across services, with physical facilities and access featuring in all except the domestic environment, time being a feature in the non-hospital settings, and quality of food being specific to inpatient care. All of these dimensions have been uncovered previously in various studies (Crow et al., 2002) and therefore exhibit construct validity. The importance of the interpersonal care and information dimension is further reinforced by the service-specific analyses, where it provides the bulk of the explanation and where improvement is likely to have the most impact on overall service satisfaction. This adds further support to the importance of the psychosocial aspects of therapeutic relationships (Williams & Calnan, 1991), and it would seem to support the third of Fitzpatrick’s (1984) models of satisfaction that posits that patients express satisfaction in relation to their observation of health care professionals’ affective behaviour and skills in communication. This also reflects earlier work by Ben-Sira (1976), which concluded that the interest and affective behaviour by general practitioners was key to patient satisfaction. Nonetheless, Fitzpatrick and Hopkins (1983) warn against an over-emphasis on the affective and interpersonal aspects, as they might be instrumental evaluations, depending on the stage in the illness career. In terms of the socio-demographic variables, age features in primary and community services, but offers little prediction for hospital services. Gender provides important prediction in inpatients, with women more likely to express high satisfaction, although the low precision in the estimate reduces its reliability. Despite there being an indication that lower educational attainment might help predict high inpatient satisfaction, generally education and self-reported health status do not appear to have any predictive value for any global service evaluation.

and and and and

information information information information

3

4

5

Physical facilities Time Physical facilities Quality of food

Time

Access

Access Physical facilities

Access

We need to consider potential weaknesses in our conclusions, albeit not all within our control, before making prognostications about its utility to policy makers and programme managers. Quota sampling, even with stratification and weighting, is prone to non-random bias, exacerbated by the exclusion of mobile phones in RDD. Without more information about the non-responders we are unsure of the effect, except that we are more concerned here with explaining the interrelationships between variables, rather than population generalisation. The problem of missing (largely DK) responses offers little bias to this analysis through their exclusion. Although the service evaluations are for the previous 12 months, problems of memory and intervening events can distort patients’ perceptions of service delivery. Telephone interviewing brings other biases well known in social research, including the likelihood of more positive responses (Burroughs, Waterman, Cira, Desikan, & Claiborne Dunagan, 2001). There is the danger of a serial effect in consecutive questions due to proximity in the interview schedule, such that groupings of questions appear to be highly correlated and get forced into similar factors as an artefact of the instrument. Nonetheless, the high correlation between the perceptions and aspirations for different aspects of each service leads us to conclude that this was not a problem. It has been argued that clinical characteristics of patients are important and should be adjusted through case-mix controls when measuring outcomes (Powell, Davies, & Thomson, 2003). However, the survey did not seek such information and there is no clear evidence about its effect on satisfaction. The distributions of the survey variables were distinctly non-linear and non-normal, which creates difficulties in making generalisations, although we took this into account in selecting the statistical techniques. We adopted techniques which are less susceptible to some of the measurement problems and we benefited from a large sample of patients, who answered questions in a logically

 po0.05,  po 0.001.

0.59

Constant

0.82 (0.58/1.18) 0.80 (0.52/1.23)

Education (Post-school higher qualifications ¼ 1) Formal school qualifications No formal qualifications (0.50/0.68) (4.82/7.60) (1.14/1.57) (2.00/2.82) (1.16/1.58)

0.88 (0.62/1.25) 1.02 (0.64/1.63)

Health status (Good ¼ 1) Fairly good Not very good

0.59 6.05 1.34 2.37 1.36

1.42 (1.00/2.13) 1.41 (0.38/5.16)

Odds ratios (95%CI) 1.17 (1.05/1.31) 1.33 (0.096/1.85)

Age Female

Factors Improvement of service Interpersonal care and information Physical facilities Time Access Quality of food

100.546 95.122 88.0 69.0 0.637 0.337

1035.161 611.378 79.8 53.5 0.536 0.367


2LL Model w2 % correct predictions Random estimate Nagelkerke R2 Hosmer & Lemeshow goodness of fit (sig)

0.74

0.99 (0.57/1.73)

0.33 (0.18/0.59) 11.42 (4.61/28.27)

0.66 (0.17/2.60) 0.40 (0.09/1.83)

1.15 (0.32/4.14) 1.94 (0.49/7.60)

Domiciliary care (n ¼ 158)

General practice (n ¼ 1192)

Table 4 Predictors of global satisfaction with each service

0.49

1.69 (1.33/2.17)

0.53 (0.42/0.67) 7.79 (5.25/11.58) 2.19 (1.67/2.87)

0.82 (0.46/1.46) 1.16 (0.58/2.29)

0.79 (0.45/1.38) 0.98 (0.51/1.88)

1.17 (0.99/1.38) 0.83 (0.50/1.37)

424.484 294.677 80.9 51.3 0.578 0.070

Outpatients (n ¼ 519)

0.12

1.11 (0.76/1.62) 2.05 (1.34/3.14)

0.44 (0.31/0.62) 7.15 (4.14/12.33) 1.97 (1.32/2.94)

1.75 (0.74/4.17) 2.17 (0.79/5.94)

2.21 (0.93/5.26) 0.82 (0.32/2.11)

1.16 (0.88/1.54) 2.84 (1.24/6.47)

198.646 150.684 83.3 50.4 0.600 0.188

Inpatient (n ¼ 252)

A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

ARTICLE IN PRESS 1681

ARTICLE IN PRESS 1682

A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683

consistent way, reflecting the findings from previous research in this field. In conclusion, we believe that this study demonstrates that, beyond the obvious differences in context, there is quite convincing evidence that patients evaluate different services in similar ways, mainly being concerned with interpersonal care and information. Age appears as a moderating variable for non-hospital patients, while gender offers prediction in the inpatient context. Policy makers and managers concerned with gaining an overview of patient satisfaction with health services, as a major and important perspective on quality and in line with the involvement agenda, can take heart from the use of such broadbrush surveys. Additionally, they have a clear focus on where improvements are likely to have the most impact on satisfaction. Nonetheless, these surveys operate at a relatively superficial level and more detailed and ultimately valuable understanding of the quality of service delivery is likely to be gained from the use of indepth studies of satisfaction within each service with which to inform professionals and operational managers. The psychosocial dimension of interpersonal care and information is multifaceted and difficult to unravel, requiring detailed understanding at the patient interface. Patient satisfaction is undoubtedly a complex concept, with relatively little agreement to date on how it can be effectively theorised. Nonetheless, surveys such as this offer the prospect of understanding where attention should be given in the meantime with the potential to improve the quality of care from the patient perspective. Acknowledgements Grateful thanks is given to the Scottish Executive Health Department Analytical Services Division for access to the dataset upon which this investigation has been based. The authors, however, take full responsibility for the views expressed here, including any errors of analysis or interpretation.

References Airey, C., Bruster, S., Erens, R., Lilley, S-J., Pickering, K., & Pitson, L. (1999). National surveys of NHS patients: General practice 1998. Leeds: NHS Executive. Avis, M., Bond, M., & Arthur, A. (1995). Satisfying solutions? A review of some unresolved issues in the measurement of

patient satisfaction. Journal of Advanced Nursing, 22(2), 316–322. Baker, R. (1990). Development of a questionnaire to assess patients’ satisfaction with consultations in general practice. British Journal of General Practice, 40(341), 487–490. Barnes, M. (1997). Care, communities and citizens. Harlow: Addison, Wesley, Longman. Ben-Sira, Z. (1976). The function of the professional’s affective behavior in client satisfaction: A revised approach to social interaction theory. Journal of Health and Social Behavior, 171(1), 3–11. Burroughs, T. E., Waterman, B. M., Cira, J. C., Desikan, R., & Claiborne Dunagan, W. (2001). Patient satisfaction measurement strategies: A comparison of phone and mail methods. Joint Commission Journal on Quality Improvement, 27(7), 349–361. Cartwright, A. (1964). Human relations and hospital care. London: Routledge and Kegan Paul. Cohen, G. (1996). Age and health status in a patient satisfaction survey. Social Science & Medicine, 42(7), 1085–1093. Collins, K., & O’Cathain, A. (2003). The continuum of patient satisfaction—from satisfied to very satisfied. Social Science & Medicine, 57(12), 2465–2470. Coulter, A. (1991). Evaluating the outcomes of health care. In J. Gabe, M. Calnan, & M. Bury (Eds.), The sociology of the health service (pp. 115–139). London: Routledge. Crow, R., Gage, H., Hampson, S., Hart, J., Kimber, A., Storey, L., et al. (2002). The measurement of satisfaction with healthcare: Implications for practice from a systematic review of the literature. Health Technology Assessment, 6(32), 1–244. Department of Health and Social Security, (1983). NHS management inquiry. London: HMSO. Department of Health, (1997). The new NHS: Modern, dependable. London: The Stationery Office. Department of Health, (1998). A first class service. London: The Stationery Office. DHSSPS (Department of Health, Social Services and Public Safety), (2003). Public consultation survey to inform the DHSSPS regional strategy for health and social wellbeing, Research and Evaluation Services, DHSSPS. Belfast: DHSSPS. http://www.dhsspsni.gov.uk/publications/2003/ health_social_wellbeing.pdf (accessed 6/10/05). Fitzpatrick, R. (1984). Satisfaction with health care. In R. Fitzpatrick (Ed.), The experience of illness (pp. 154–175). London: Tavistock. Fitzpatrick, R. (2002). Capturing what matters to patients when they evaluate their hospital care. Quality and Safety in Health Care, 11(4), 306. Fitzpatrick, R., & Hopkins, A. (1983). Problems in the conceptual framework of patient satisfaction research: An empirical exploration. Sociology of Health and Illness, 5(3), 297–311. Gregory, J. (1978). Patients’ attitudes to the hospital service, Royal Commission on the National Health Service. Office of Population Censuses and Surveys. London: HMSO. Grol, R. (2001). Improving the quality of medical care: Building bridges among professional pride, payer profit, and patient satisfaction. Journal of the American Medical Association, 286(20), 2578–2585. Hill, J., Bird, H. A., Hopkins, R., Lawton, C., & Wright, V. (1992). Survey of satisfaction with care in a rheumatology

ARTICLE IN PRESS A.P. Bikker, A.G.H. Thompson / Social Science & Medicine 63 (2006) 1671–1683 outpatient clinic. Annals of Rheumatic Disorders, 51(2), 195–197. Hogg, C. (1999). Patients, power and politics: From patients to citizens. London: Sage. Jackson, J. L., Chamberlin, J., & Kroenke, K. (2001). Predictors of patient satisfaction. Social Science & Medicine, 52(4), 609–620. Jenkinson, C., Coulter, A., Bruster, S., Richards, N., & Chandola, T. (2002). Patients’ experiences and satisfaction with health care: Results of a questionnaire study of specific aspects of care. Quality and Safety in Health Care, 11(4), 335–339. Jones, L., Leneman, L., & Maclean, U. (1987). Consumer feedback for the NHS. London: King Edward’s Hospital Fund for London. McGhee, A. (1961). Patients’ attitudes to nursing care. Edinburgh: Livingstone. Meredith, P., Emberton, M., & Devlin, H. B. (1993). What value is the patients’ experience of surgery to surgeons? The merits and demerits of patient satisfaction surveys. Annals of the Royal College of Surgeons of England, 75(Suppl. 3), 72–73. Moores, B., & Thompson, A. (1986). What 1357 hospital inpatients think about aspects of their stay in British acute hospitals. Journal of Advanced Nursing, 11(1), 87–102. Powell, A. E., Davies, H. T. O., & Thomson, R. G. (2003). Using routine comparative data to assess the quality of health care: Understanding and avoiding common pitfalls. Quality and Safety in Health Care, 12(2), 122–128. Raphael, W. (1967). Do we know what patients think? International Journal of Nursing Studies, 4(3), 209–223. Scottish Executive Central Research Unit, (2000). Scottish Household Survey Annual Report: 1999. Edinburgh: The Stationery Office. Scottish Executive Central Research Unit, (2001). Public attitudes to the NHS in Scotland. Edinburgh: The Stationery Office.

1683

http://www.scotland.gov.uk/cru/resfinds/hcc2.pdf (accessed 6/10/05). Scottish Executive Health Department, (2000). Our national health: A plan for action, a plan for change. Edinburgh: The Stationery Office. Scottish Executive Health Department, (2003). Partnership for care. Edinburgh: The Stationery Office. Scottish Office, (1997). Designed to care. Edinburgh: The Stationery Office. Sitzia, J., & Wood, N. (1997). Patient satisfaction: A review of issues and concepts. Social Science & Medicine, 45(12), 1829–1843. Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). London: Allyn & Bacon. Thompson, A. (1986). The soft approach to quality of hospital care. International Journal of Quality and Reliability Management, 3(3), 57–65. Thompson, A. G. H., & Sun˜ol, R. (1995). Expectations as determinants of patient satisfaction: Concepts, theory and evidence. International Journal for Quality in Health Care, 7(2), 127–141. Ware, J. E., Snyder, M. K., Wright, R., & Davies, A. R. (1983). Defining and measuring patient satisfaction with medical care. Evaluation and Program Planning, 6(3–4), 247–263. Wensing, M., & Elwyn, G. (2002). Research on patients’ views in the evaluation and improvement of quality of care. Quality and Safety in Health Care, 11(2), 153–157. Williams, S. J., Coyle, J., & Healy, D. (1998). The meaning of patient satisfaction: An explanation of high reported levels. Social Science & Medicine, 47(9), 1351–1359. Williams, B., & Calnan, M. (1991). Convergence and divergence: Assessing criteria of consumer satisfaction across general practice, dental and hospital care settings. Social Science & Medicine, 33(6), 707–716. Wright, E. C. (1993). Non-compliance—or how many aunts has Matilda. Lancet, 342(8876), 909–913.