Results of in-depth interviews
Through in-depth interviews with patients/consumers, their reasons for being more willing to accept CFDS were identified and summarized. Examples included patients/consumers’ understanding level of CFDS, benefits of CFDS, concern about one’s own health; the degree of family doctors’ protection of patients/consumers’ privacy, cost and process of signing a contract, satisfaction with the community, and advocacy of contracting services, etc. The keywords extracted from the interviews were counted and we found that the factors affecting CFDS from the perspective of consumers were as follows: national policy factors, contracted team factors, and consumer-related factors.
Demographic characteristics of community medical staff
Of the 320 respondents,72.2% were women, 56.6% had received an undergraduate education, and 43.4% had intermediate titles. Among respondents, general practitioners were the main occupations, with 57.5% of respondents working longer than 10 years. The demographic characteristics of the medical personnel are shown in Table 1.
Analysis of the factors influencing CFDS
Exploratory factor analysis
The Kaiser-Meyer-Olkin value was 0.836, which showed that the data could be used for factor analysis (if the Kaiser-Meyer-Olkin value is close to 1, the variable group is suitable for factor analysis).
After finishing the orthogonal rotation of the factor loading matrix, the remaining 25 items had eigenvalues >1, orthogonal rotation explained the maximum amount of variance, seven factors were extracted from the system, and the cumulative variance contribution rate of 67.613% is shown in the Supplementary form. Table 2 shows that 25 observational variables were clearly classified into seven common factors. Based on the results of the group discussion among project team members, seven factors were named according to the characteristics of the variables observed. F1 was ‘national policy factor’, the combination of F2 and F3 was ‘resident factors’, the combination of F4 and F5 was ‘contract doctor factors’, and the combination of F6 and F7 was ‘community factors’.
Confirmatory factor analysis
The results of the confirmatory factor analysis were as follows: the RMSEA was 0.059, and thus was less than the 0.08 cutoff that indicates a good fit; the RMR was 0.05; The TLI, NFI, GFI, and CFI were 0.913, 0.902, 0.905, and 0.917, respectively.
Results of expert consultations
The health and family planning commission, contracted services researchers, and administrators were selected to carry out an expert consultation. The above results were modified according to their inputs, and the final versions of the predisposing factors are listed below.
In the first round of consultation, the name of each dimension in the model was modified: ‘national policy factors’ was revised to ‘national government factors’, ‘resident factors’ was revised to ‘consumers-related factors’, ‘contracted doctor factors’ was revised to ‘contracted doctor-related factors’, and ‘community factors’ was revised to ‘community health service agency factors’.
The second round of consultation integrated the dimensions of the model. The experts deemed that the ‘situation of the first diagnosis of the patients/consumers’ should be incorporated into the resident-related factors dimension rather than contracted doctor-related factors.
The final determinants of the factors of CFDS
The final determinants of the factors of CFDS are shown in Table 3. The factors influencing CFDS from the perspective of medical staff were divided into four dimensions and 25 items. The four dimensions were named national government factors, community health service agency factors, consumer-related factors and contracted doctor-related factors, respectively.
Calculation of the factor weights of factors influencing CFDS
The cumulative variance contribution rate of F1–F7 was 67.613% (shown in the Supplementary form). The weighted mean of the variance contribution rate of each factor was calculated, and the evaluation formula of the comprehensive score was obtained: F = (0.302F1 + 0.107F2 + 0.0701F3 + 0.0571F4 + 0.0531F5 + 0.0458F6 + 0.0406F7)/0.676. Indicator weight = composite score/model coefficient.
Based on the above results, we merged similar factors into four dimensions. Finally, the weight coefficients of the common factors were 0.319, 0.247, 0.226, and 0.208, respectively (see Table 4 for specific details).