Process quality of type 2 diabetes mellitus care and association with patient perceived attributes of family doctor service in urban general practices, Beijing, China

We carried out a cross-sectional study from May to December 2019 in community health service centers in Beijing China. Non-probability sampling method was adopted as this was a preliminary study in the exploration of process quality of T2DM in primary care [23]. To ensure a stable and adequate amount of contracted T2DM patients, general practice experts and researchers were consulted before sampling. The sampling plan had a three-stage design. First, from fourteen CHSCs that were teaching sites of Capital Medical University, five CHSCs were chosen by convenience sampling considering the following criteria: (1) stable amounts (> 1000) of contracted T2DM patients; (2) had five or more GPs; (3) available for the study. Second, there were 118 GPs in total in the five CHSCs and we had chosen three GPs per CHSC by convenience sampling considering the following criteria: (1) worked in the community for over 5 years; (2) ranked above intermediate grade professional title; (3) had > 50 contracted T2DM patients; (4) available for the study. Third, consecutive patients visiting the selected GPs were invited and informed about the study. All those who were eligible and consented were recruited in the study as “consecutive consenting sample.” [24]. The inclusion criteria of patients were as follows: (1) diagnosed with T2DM for at least 1 year before the study; (2) signed a contract with the selected GPs; (3) being able to answer the questions clearly and logically. Total 400 patients (80 patients per CHSCs) were recruited in this study. These patients were from fifteen GPs in five CHSCs located in two of the six urban districts of Beijing. Please see Fig. 1. The sample size was chosen based on the available funding granted.

A pilot study had been conducted at two CHSCs in Beijing in April 2019 to test the feasibility of monitoring and health counselling indicators. We found that the information of partial indicators was not required by the FDS model (e.g., records of foot and retinopathy monitoring), which was not available or missing in the general practice EHR system. Indicators in the two domains were further modified (i.e., modification of data collection method and indicator content) based on general practice expert consultation and research team meeting to guarantee data credibility and availability (see Supplementary Table 2, Additional file 1). Revised process indicators were shown in Table 1. Combination of patient questionnaire and review of EHR was recommended to collect data of the monitoring and health counseling indicators. The questionnaire and EHR extraction form were revised through multiple rounds of research team meetings.

The questionnaire consisted of three sections. The first section contained questions regarding patient socio-demographic variables (i.e., gender, age, marital status and education level) and health statues (i.e., coexisting health problems, number of years since diagnosis, T2DM associated therapy, T2DM related visits and self-reported health status). Number of years since diagnosis was collected by patients’ self-reported information. Patients reported their overall health on the Visual Analogue Scale (EQ-VAS) based on their health on the day of survey, with anchor points 0 (‘worst imaginable health state’) and 100 (‘best imaginable health state’). In the second section, we used three adopted items to learn about patient perceived attributes of FDS (i.e., accessibility, continuity and team-based care) based on literature review and FDS related policies in China [3, 25]. The three items included: (1) is it accessible to get health care from general practice when needed? (2) is there a personal GP providing continuous care for you? (3) could you perceive a functioning GP team (including the GP, nurse, and preventive care physician)? The items were with ternary results ‘Yes’, ‘No’ or ‘Not sure’, and only if the answer is ‘Yes’ will the attribute be considered as perceived by patient. The third section investigated the process quality of T2DM care using 7 monitoring and 6 health counseling indicators (see Table 1). In order to reduce recall bias, patients just answered whether relevant quality indicators had been performed, not the frequency. The items of process quality indicators in patient questionnaire were with ternary results ‘Yes’, ‘No’ or ‘Not sure’ and only if the answer is ‘Yes’ will the service be considered as received by patients. In order to ensure good response rates to all questions and the exchange of information to reduce misunderstanding, we conducted the questionnaire survey through one-on-one, face-to-face interviews by five investigators (i.e., two PhD. candidate, two master candidates and one lecturer of general practice) in the waiting room before or after consultation. The investigators were mainly responsible for explaining the items to the patients and recording patients’ answers. Before the investigation, the investigators had received unified training by JGH (lecturer of general practice) on data collection, and potential problems that may be encountered during the investigation were also discussed.

EHR information extraction form consisted of patient’s basic information (i.e., name, age, gender) and the frequency of five process quality indicators (see Table 1). After completing the questionnaire survey, patients’ basic information was recorded into the EHR information extraction form. EHR database was reviewed to extract data by two investigators together (one master candidates and one lecturer of general practice) using the EHR information extraction form.

Information on the GPs, including age, gender, highest degree, working years, professional title, GP training experience, total number of contracted T2DM patients were collected. Written informed consent was obtained from each participant in this study.

Of the 400 patients, 58.0% were male, 79.8% were with a spouse and the average age was 68.9 ± 9.16 (ranging from 38 to 92) years. More than half of the patients (54.5%) completed high school or above. There were 90.2% of patients had at least one and 35.2% had three or more coexisting health problems. The median number of years since diagnosis (with 1 missing) was 10 (ranging from 1 to 59) years. The most common therapy was oral medication only (74.2%). The percent of patients who had more than 12 visits at the CHSC in the 12 months preceding investigation was 61.5% (with 3 missing) (see Table 2). All the 15 GPs were female with an average age of 42.5 ± 4.14 (ranging from 37 to 49) years. 10 GPs had a bachelor’s degree. The average working years was 16.3 ± 6.48 (ranging from 8 to 30) years. For each GP, the total number of contracted T2DM patients ranged from 120 to 539. The information of the 15 GPs was shown in Table 2.

The median number of T2DM care received by patients was 9 (ranging from 0 to 12) in the dimension of monitoring, 4 (ranging from 0 to 6) in the dimension of health counseling, and 12 (ranging from 2 to 18) in total. The utilization rate of all the 12 indicators in monitoring, 6 indicators in health counseling and all the 18 process indicators, were 12.8%, 23.8% and 6.0% respectively. The least frequently received care was peripheral neuropathy examination (28.7%) in the monitoring domain and emergency help counseling (42.5%) in the health counseling domain. It was accessible for most patients (87.8%) to get health care from general practice. Most of the patients (86.5%) had a personal GP providing continuous care for them. There were 67.3% of the patients recognized that they were being managed by a GP team (see Table 3). The distribution of all patients was 3.5%, 8.3%, 31.5% and 56.8% for those perceived 0, 1, 2 and 3 attributes of FDS, respectively.

There were statistically significant associations between accessibility of care and lipid (p = 0.008), electrocardiogram (p = 0.016), retinopathy (p = 0.037) and peripheral neuropathy (p = 0.006) monitoring and each of the 6 health counseling indicators (all the p values < 0.05). Regular follow up (p = 0.039), plasma blood glucose (p = 0.020), blood pressure (p = 0.026), body mass index (p = 0.044) and foot (p = 0.005) monitoring as well as each of the 6 health counseling indicators (all the p values < 0.05) were more likely to be received by patients when continuity of care was ensured. Patients who were managed by a GP team had higher utilization rate of glycosylated hemoglobin monitoring (p = 0.026) and each of the 6 health counseling indicators (all the p values < 0.05) (see Table 4).

The AIC and BIC values for the three models were 2067.69/2079.59, 2003.16/2014.93, and 1988.78/2000.48, respectively. The results indicated that the full model (Model-3) with the smallest AIC and BIC values fitted the data better than the other models. The ICC in Model-3 indicated that 2.4% (ICC: 0.024) of the variation in the quality performance was attributable to differences across GPs.

The fixed effects of Model-3 showed that one more FDS attribute perceived by patients, indicators received by patients significantly increased by 1.50 (coefficient = 1.50, 95%CI 1.08,1.92, p < 0.001) after patient and GP characteristics were adjusted. Several non-modifiable patient and GP factors also had statistically significant influence on the process quality performance. Patients between the age of 65 and 74 years received 1.15 more indicators than those under 65 (coefficient = 1.15, 95%CI 0.40,1.90, p = 0.003). Patients with more than ten years duration of T2DM reported receiving 0.74 more indicators (coefficient = 0.74, 95%CI 0.08,1.41, p = 0.028). Patients who took both insulin and oral medicine received 0.97 more indicators, comparing with those who took oral medication only (coefficient = 0.97, 95%CI 0.11,1.83, p = 0.027). Patients who were managed by GPs with on-job training experience received 1.19 more indicators (coefficient = 1.19, 95%CI 0.06,2.32, p = 0.040). The p-value for the interaction of education level and self-reported health status was 0.023 (see Table 5). Among the patients who had completed junior high school or below, having better self-report health status (≥ 60) received 2.40 less indicators (coefficient = -2.40, 95%CI -4.01,-0.78, p = 0.004). However, among the patients who had completed high school and above, there was not significant association between self-reported health status and quality performance (see Table 6).

Our study represents one of the first attempts to assess the process quality of T2DM care by using detailed indicators based on clinical guidelines and explore the association between clinical quality and patient perception of care in China. The results of our study indicated the significant association between patient perception of FDS attributes and process quality of T2DM care, which might support the importance of patient perceived experience in delivering and measuring the quality of T2DM care. However, there are some limitations. Firstly, the study adopted non-probability sampling with a relatively small sample size, which could affect the precision of statistical analysis. All the fifteen GPs included in this study were female because most GPs who met the inclusion criteria were female in those CHSCs selected. This may be one reason for that only 2.4% (ICC: 0.024) of the variation in the quality performance was attributable to differences across GPs. Due to the explorative nature of this preliminary study, the findings based on non-probability sampling can only be generalized to the population from which the sample is drawn and not to the entire population. Future studies with random sampling frames can enhance the external validity of the findings. Nevertheless, we made attempt to control for the characteristics of participants to provide the clearest possible image of quality of T2DM care in CHSCs in Beijing, China. Secondly, quality indicators used in this study were constructed with only monitoring and health counseling indicators due to the limited availability of more quality-related information in the EHR system. Thus, these indicators may not be an exact representation of all the quality-related aspects. Monitoring and health counseling as indicators for the T2DM care in our study should be interpreted with prudence, since repeating counseling annually may be unnecessary for patients with well-controlled diabetes. And patients near the end of their live span may be less likely to benefit from these services. Thirdly, the information collected through questionnaire survey was self-reported in the past year and patients’ incomprehension of the quality indicators as well as recall biases might induce underreported information of process quality. In order to reduce these biases, the investigator fully explained the meaning of each indicator to the participants and they just answered whether they had received the relevant services, not the frequency. There may be a social desirability bias relying on self-report data, as participants may be inclined to present themselves in a more positive light or appease the course leader. In order to ensure the authenticity of the patient's answer, before the survey, we explained to patients the importance of accurate answers and promised them that data was fully anonymized for the purposes of research access and all identifiable data would be removed. The study variables were reports asking patients whether it happened, not evaluations asking them to rate what happened. What’s more, we set up a "Not sure" option, which is convenient for patients to give a negative answer in a euphemistic way. On the other hand, we conducted the questionnaire survey through interviews by five investigators. Interviewer bias and recording errors could be a possible limitation. The questionnaire was revised through multiple rounds of research team meetings. We also provided careful training for investigators about the principles of questionnaire survey and information recording to ensure the consistency of data collection. Finally, even though we had adapted a two-level GLMM to test the effect sizes of patient and GP-level factors on process quality of T2DM care, there were factors may have been neglected, such as patient’s diabetes knowledge and compliance and CHSCs’ human resources. Future research is needed to identify more variables to explain the quality of T2DM care in general practice in China. As with any observational study causality cannot be determined.