This study comprised a patient-completed, anonymous 6-page questionnaire collecting a range of details including aspects of the organisation of care, perceived needs, perceived process, and outcome of consultations with the participating GPs.
Sampling frame
A database containing the mean deprivation scores of all GP practices in the west of Scotland was made available by the Information and Statistics Division of NHS Scotland. Deprivation data was extracted on practices in 4 health board regions in the west of Scotland; Greater Glasgow, Argyll and Clyde, Lanarkshire, and Ayr and Arran. Ethical approval for the study was obtained from each of the 4 health boards. The 'low deprivation' group of practices invited to participate in the study were selected from practices in the lower quartile of deprivation scores for the 4 regions combined. In the high deprivation groupings, practices were selected from those in the lower quartile of the combined deprivation scores of practices in the three health board regions out-with Greater Glasgow, and those in the lower quartile of deprivation scores within Greater Glasgow. This was necessary because of the concentration of severe deprivation within Greater Glasgow.
Because patient enablement scores have been shown to be influenced by practice size [3], the sampling frame was limited to medium-sized Practices (3-4 Partners). Only non-training Practices (i.e., those that are not accredited for training GP Registrars) were included. All medium-sized practices (3 to 4 GP principals) in the upper or lower quartile of deprivation (based on a multiple index-of-deprivation score used nationally) in 4 health board regions in the west of Scotland were mailed letters that explained the details of the study and asked the practice to nominate 1 GP to participate. 26 GPs from 26 Practices agreed to participate in the study, from 70 eligible practices approached across the 4 health board areas, giving an overall recruitment rate of 37% (36% in the high deprivation groups and 38% in the low deprivation groups). We have reported some further details on the sampling frame in previous papers [16, 20].
Patients
Consecutive, unselected patients were asked by the reception staff if they would be willing to complete a questionnaire when they arrived for their consultations. This self-completed, anonymous questionnaire collected details of perceived needs, perceived process, and outcome of consultations with the participating GPs. Individual demographic, socio-economic, and health details were also collected [16, 20].
The process of the consultation was assessed using the Consultation and Relational Empathy Measure (CARE Measure), a validated tool which seeks to capture the patient's perception of the doctor's empathetic understanding and action in the consultation [20]. Continuity of care was assessed by asking if the patients were seeing their usual doctor, and if so, how well they felt they knew the doctor, on a 5 point scale as used previously [21]. Consultation length was recorded by the doctor. The outcome of the consultation was measured using the patient enablement instrument (PEI) [1, 2].
The questionnaire also recorded age, gender, marital status, children, employment status, and educational level, type of living accommodation, ethnicity, and postcode. The latter was used to calculate individual deprivation scores. General health information included the General Health Questionnaire (GHQ-12) [21], general health over the previous twelve months, and any long-term illness, health problem, or disability. The number and type of chronic diseases was also recorded as described previously [22].
The reason for making the consultation ('new problem', 'long-standing problem' or 'both new and old problems'), how many problems the patient wished to discuss, confidence that the doctor would be able to help (patient expectations), whether the patient hoped to receive a new prescription, and the GHQ-12 and socio-economic and demographic details were all recorded before the consultation [16]. A complex consultation was defined as one in which a patient wanted to discuss a psychological or social problem (with or without a physical problem) as previously explained [16]. GHQ-caseness (meaning significant psychological distress) was defined as having a cut-off of 4 or above as previously described [16, 20]. The other items relating to the process and outcome of care, and general health were completed straight after the consultation. The completed questionnaires were then collected in a sealed box at the reception area.
The patient response rate to the questionnaire overall was 70% (70% in high deprivation group, 71% in low deprivation group). Although data was not collected on the 30% of consulting patients who chose not to participate in the study, we have reported the distribution of participating patients per practice, as a percentage of the distribution of deprivation (in quartiles) of all patients registered with that practice and shown that here was a reasonably equitable spread of deprivation scores of participating patients, suggesting that the patients who declined to participate were not substantially skewed towards the most deprived end of the spectrum [20].
Data analysis
Important patient, consultation, and system variables collected in the patient survey were treated as potential confounding factors for enablement and were evaluated for their association with PEI by use of multi-level modeling [19] that took into account the hierarchical nature of the data (patients within GPs). Because of the skewed nature of the PEI, we carried out the analysis by way of binary logistic regression, with PEI scores categorized as below average versus average or above. Univariate analysis was done first for all potential confounding factors. Those factors with P < 0.25 in univariate analysis were then analyzed by multivariate analysis with use of a forward stepwise selection strategy. In general, the process added the most significant confounding factor to the model at each step and continued until all confounding factors that made a significant (P < 0.05) contribution were in the model.
The models are fitted by the method of Markov Chain Monte Carlo (MCMC) algorithm of MLn for Windows software package (Version 2.02, Institute of Education, University of London, U.K.). The Deviance Information Criterion (DIC) diagnostic statistic, which is a generalization of the Akaike Information Criterion (AIC), is used to assess the statistical significance of the estimates at 5% level of significance.