Xuxi Zhang

organized. This number allowed us to select professionals with varied professions who had been actively involved in the UHCE approach. In Manchester, two actively involved trained assistants were interviewed. In total: 26 older persons, four informal caregivers and 22 professionals were included (7 nurses, 4 general practitioners/physicians, 2 physical/occupational therapists, 2 social workers, 2 trained assistants, 1 physical education teacher, 1 pharmacist, 1 volunteer, 1 care manager, 1 municipality officer). Supplementary Table S2 described the numbers by city. The focus group discussions and in ‐ depth interview were recorded, transcribed into the local language and translated into English if applicable. Data analysis Quantitative data were summarised using descriptive statistics (frequencies, means and percentages). Characteristics of persons included at follow ‐ up and persons who dropped out were compared by means of chi ‐ square tests for categorical variables and one ‐ way ANOVA for continuous variables. We further compared characteristics of persons who enrolled in a specific care ‐ pathway (falls, polypharmacy, loneliness and frailty) with persons who did not enrol in that care ‐ pathway but had an indication to receive that care ‐ pathway. For this purpose, multilevel random ‐ intercept logistic regression was used because data was clustered by city. 37 We built 4 separate models for each care ‐ pathway in which we analysed the association of independent variables age, sex, living situation, education level, function and mental health with dependent variable non ‐ enrolment. We corrected the effect estimates for all factors as well as clustering effects by city. We considered a P ‐ value of 0.05 or lower to be statistically significant. All quantitative analyses were performed using SPSS version 23.0 (IBM SPSS Statistics for Windows, Armonk, NY: IBM Corp). For the qualitative data from the focus groups and logbooks, the thematic analysis method was used. 38 Focus group transcripts were read multiple times by CF and XZ and meaningful overarching themes and subthemes were identified and summarized in a coding table. The coding table was discussed among the authors and refined. Subsequently, applicable quotes from the transcripts were entered into the coding table and categorized according to subtheme. Overarching themes confirmed topics in the topic guide that was used for the focus groups: benefits/barriers of the health assessment, benefits/barriers of the care ‐ pathways, and recommendations for improvement. Meaningful subthemes emerged from the raw data (e.g. under the overarching theme ‘barriers of the care ‐ pathways’, subthemes ‘mistrust’ and ‘embarrassment’ emerged). Qualitative logbook data on the older person’s reason(s) for not enrolling into any care ‐ pathways were coded into meaningful themes in an excel sheet by CF. Subsequently, XZ coded the data into the themes developed by CF. Coding by the two authors was compared and disagreements in coding were discussed and resolved. If necessary, themes were refined by discussion between authors. Rigour The design of this study was based on a theoretical framework for process evaluations of public health interventions developed by Stecklar and Linnan. 22 Using an established theoretical framework in the development and reporting of complex interventions improves 110 Chapter 5