Table 4 Evaluation of outcome measure processes

Practice characteristics (list size, location, deprivation)

Prior to randomisation

Practice and Association of Public Health Observatories website

6 practices

Categorical, nominal/ordinal.

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Patient age, gender, ethnicity, self-reported health status

Prior to index consultation

Pre-consultation postal questionnaire

180 patients

Categorical, nominal/ordinal.

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences (random effect on cluster, adjustment for practice location).

Patient deprivation data from patient postcodes

Following return of patient pre-consultation questionnaires and consent forms

Practice records mapped to the Index of Multiple Deprivation

180 patients

Continuous (IMD scale)

Mean and standard deviation, to report data descriptively.

Linear hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

GP age, gender, ethnicity, time since qualification

Prior to index consultation

GP practices and General Medical Council GP registry

18 GPs

Categorical, nominal/ordinal

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Patients’ preferences for involvement in decision-making.

Prior to index consultation

Patient pre-consultation postal questionnaire

180 patients (90 per arm)

Ordinal (6 point Likert scale).

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Ratings of shared decision-making during the consultation from an observer perspective.

During data analysis

Assessment of video’d consultations by two trained researchers using the OPTION(5) score ([18] [19];)

18 GPs, 180 patients

(9 GPs and 90 patients per arm)

Continuous (OPTION score 0-100%)

Mean and standard deviation, to report data descriptively.

Linear hierarchical modelling to estimate between group differences* (random effect on cluster, adjustment for practice location).

Patient-reported rating of involvement in decision-making about their healthcare

Immediately following the index consultation

Patient post-consultation questionnaire—using collaboRATE score ([18] [19];)

180 patients (90 per arm)

Continuous (collaboRATE score 0–100%)

Mean and standard deviation, to report data descriptively.

Linear hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Patient and GP scores compared using logistic regression modelling (patient scores as outcome, GP scores as explanatory variable).

Patient-reported rating of feeling satisfied with the healthcare received

Immediately following the index consultation

Patient post-consultation questionnaire

180 patients (90 per arm)

Categorical, ordinal (3 point Likert scale)

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Patient-reported rating of having trust in the GP they saw

Categorical, ordinal (3 point Likert scale)

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Patient-reported rating of enablement

Discrete (PEI score 0–12)

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

GP-reported rating of their involvement of the patient in decision-making about their healthcare

Immediately following the index consultation, after confirming patient consent for each aspect of data collection.

GP questionnaire using adapted collaboRATE ([18] [19];)

18 GPs (9 per arm)

Continuous (collaboRATE score 0-100%)

Mean and standard deviation, to report data descriptively.

Linear hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Patient and GP scores compared using logistic regression modelling (patient scores as outcome, GP scores as explanatory variable).

Patient contacts in a 28-day period following the index consultation, including the nature of contact with the GP surgery, the hospital admissions, A&E attendances. If patient moved away within 28 days (i.e. lost to follow up)

Approximately 12 weeks after index consultation (to allow time for contacts to be recorded in the notes)

Case note review by two researchers

180 patients, (90 per arm)

Count

Median and range, to report data descriptively.

Poisson hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Deaths within a seven day period following the index consultation; death within 28 days (i.e. did not have full study follow-up).

Approximately 12 weeks after index consultation (to allow time for contacts to be recorded in the notes)

Case note review by two researchers

180 patients, (90 per arm)

Count

Median and range, to report data descriptively.

Poisson hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Documented decision outcomes from the index consultation, e.g. starting/stopping/changing medication, referrals and investigations

Approximately 12 weeks after index consultation (to allow time for contacts to be recorded in the notes)

Case note review by two researchers

180 patients, (90 per arm)

Binary (yes/no) variables for each type of change

Frequencies, to report data descriptively.

Logistic hierarchical modelling to estimate between group differences^a (random effect on cluster, adjustment for practice location).

Participant experiences of the intervention, participants experiences of the study

Following receipt of participant post-consultation questionnaires and consent forms

Interviews with the participants from practices assigned to the intervention

9 GPs, 15 patients

Audio-recordings for qualitative analysis

Both deductive and inductive approaches to thematic analysis