Introduction
In preparation for the quantitative and qualitative studies, we undertook an iterative expert consultation and rapid scoping of the literature to identify the following: factors that might be associated with the development of diabetes, diabetes outcomes and diabetes health-care interventions for people with SMI.
The aim was to create a ‘longlist’ of the range of factors potentially relevant to our study and map which explanatory factors had been theorised or empirically demonstrated to have a relationship with the outcomes. From this longlist, we considered which factors were feasible to explore, for example which variables were available in the CPRD or could be explored in an interview. We planned that this approach would assist in the construction of key explanatory, outcome and health-care intervention variables for the quantitative study, and shape the topic guides for qualitative interviews.
Methods
We iteratively searched and aggregated evidence from a range of sources between January and March 2018 using an inclusive approach to capture a broad range of potential factors. Systematic reviews and meta-analyses known to the research team provided a starting point. We also drew on primary quantitative studies, literature reviews, expert consultation and clinical guidelines, as well as results from a pilot interrogation of CPRD data of around 1000 people with SMI and diabetes. The list of sources was reviewed and added to by the study research team and the SSC. Additional sources were identified through iterative targeted searches of specific variables in key databases [e.g. MEDLINE, EMBASE™ (Elsevier, Amsterdam, the Netherlands), Cochrane Database of Systematic Reviews] for systematic reviews in the first instance, and, when these were not available, other study types. Only publications in the English language were considered.
Details of determinants (including social determinants), outcomes and interventions were extracted from each source to generate a longlist of potential variables. This was reviewed by the research team to ensure that relevant variables were included and appropriately labelled, and that duplicates removed or merged. Quality appraisal of sources was not undertaken. All evidence sources were imported into NVivo 12 (QSR International, Warrington, UK).
Tables containing each variable, type, titles and number of sources cited, and summary of justification (theory, research findings) were produced to inform plans for the quantitative and qualitative studies (see Appendix 2, Tables 28–31).
Under objective 7, we planned to explore which health-care interventions were associated with better diabetes outcomes for people with SMI and diabetes, subject to data availability in the CPRD and completeness of recording by practices. As these interventions needed to be from the UK, we identified candidate interventions from the following sources:
Diabetes and SMI QOF indicators that could be identified in CPRD data. The QOF is a national programme that offers financial incentives to general practices for meeting quality-of-care targets across a range of conditions, including SMI and diabetes.
109,110 These indicators include, for example, annual monitoring for key biological measurements such as blood glucose and blood pressure, dietary review, foot examination, retinal screening, and structured education and comprehensive care planning.
111Diabetes interventions recommended by NICE.
32Interventions identified from the systematic review and service user engagement completed in NIHR study 13/54/40,
112,113 which identified a number of diabetes-related indicators of primary care quality (e.g. diabetes screening, monitoring concomitant antipsychotic medications, BMI and weight loss, retinal and foot examination, and education about nutrition and physical activity).
Interventions identified from the DIAMONDS systematic review
114 and PPI consultation, and through ongoing consultation with the study research team, collaborators and steering committee, and the wider DIAMONDS research group, DIAMONDS VOICE panel, and virtual stakeholder network.
Results of our pilot interrogation of CPRD data carried out by the DIAMONDS research group to characterise the population and to develop and test clinical (Read code) lists in preparation for the study.
115Interventions identified from systematic reviews as being potentially effective in the UK for reducing inequalities in diabetes or SMI care and outcomes.
116–121
Results
Fifty-six sources were used to identify variables, as shown in . The full list can be found in Appendix 2, Tables 28–31.
Type of studies and number of related sources providing variables
The list of variables () was reviewed by the study research team to check for clinical significance, duplication or missing variables, and to determine which variables could be investigated in CPRD data. Any that could not be identified or that would potentially be biased were considered for inclusion in the qualitative interview topic guides.
Longlist of variables for exploration in the quantitative study
We used the variables list to construct explanatory (e.g. age, obesity, type and severity of SMI) and outcome variables (e.g. complications, cardiovascular control, mental illness relapses) and to shape the topic guides for qualitative interviews (see Report Supplementary Material 1). Although it was important not to be constrained by preconceived notions about participants’ experiences, being aware of the potential influence of, for example, social determinants of diabetes in people with SMI steered researchers towards initiating discussion around topics such as financial constraints and barriers to service access.
