OBJECTIVE

Cardiovascular diseases (CVD) are a long-term sequela of diabetes. Better individual-based continuity of care has been reported to reduce the risk of chronic complications among patients with diabetes. Maintaining a one-to-one patient–physician relationship is often challenging, especially in public health care settings. This study aimed to evaluate the relationship between higher team-based continuity of care, defined as consultations provided by the same physician team, and CVD risks in patients with diabetes from public primary care clinics.

RESEARCH DESIGN AND METHODS

This was a retrospective cohort study in Hong Kong of 312,068 patients with type 2 diabetes and without any history of CVD at baseline (defined as the earliest attendance at a doctor’s consultation in a public-sector clinic between 2008 and 2018). Team-based continuity of care was measured using the usual provider continuity index (UPCI), calculated by the proportion of consultations provided by the most visited physician team in the 2 years before baseline. Patients were divided into quartiles based on their UPCI, and the characteristics of the quartiles were balanced using propensity score fine stratification weights. Multivariable Cox regression was applied to assess the effect of team-based continuity of care on CVD incidence. Patient demographics, smoking status, physiological measurements, number of attendances, comorbidities, and medications were adjusted for in the propensity weightings and regression analyses.

RESULTS

After an average follow-up of 6.5 years, the total number of new CVD events was 52,428. Compared with patients in the 1st quartile, patients in the 2nd, 3rd, and 4th quartiles of the UCPI had a CVD hazard ratio (95% CI) of 0.95 (0.92–0.97), 0.92 (0.89–0.94), and 0.87 (0.84–0.89), respectively, indicating that higher continuity of care was associated with lower CVD risks. The subtypes of CVD, including coronary heart disease and stroke, also showed a similar pattern. Subgroup analyses suggested that patients <65 years of age had greater benefits from higher team-based continuity of care.

CONCLUSIONS

Team-based continuity of care was associated with lower CVD risk among individuals with type 2 diabetes, especially those who were younger. This suggests a potential flexible alternative implementation of continuity of care in public clinics.

Diabetes is one of the most prevalent noncommunicable diseases and leading causes of mortality worldwide. In 2017, there were ∼5 million deaths globally attributed to diabetes (1). The number of patients with diabetes is projected to increase from 451 million in 2017 to 693 million by 2045 (1). Patients with diabetes were 2.3 times more likely to develop cardiovascular diseases (CVD) than those without, and ∼30% die of CVD-related causes (2). With these rising trends, there is a need for more effective preventive measures for CVD development among patients with diabetes.

Evidence suggests that better continuity of care, specifically relational continuity on the individual level, reduces CVD risk among patients with diabetes (35). Continuity of care is frequently assessed as “the extent to which patients are cared for by the same physician.” However, in many public-sector health care settings, due to workforce constraints, temporary absence or retirement of physicians, and frequent rotation of physicians-in-training, maintaining a long-term one-to-one patient–physician relationship may not be feasible (68). In larger health care settings, delivering care on a team-based framework permits greater flexibility in workforce mobilization, ensuring that the patients will be cared for by the same team of physicians even when their personal physician is not available.

High team-based continuity of care should theoretically achieve similar benefits with individual-based continuity of care through improved patient–physician relationships and information transmission (9). To date, there have been few studies examining the effects of team-based continuity of care on patients with diabetes. This study aimed to investigate the association between team-based continuity of care and the incidence of CVD in patients with diabetes.

Study Design

This is a retrospective cohort study conducted in patients with diabetes and managed under the Hospital Authority (HA), the statutory body for public health care services in Hong Kong. The HA manages ∼90% of the known cases of diabetes in Hong Kong (10). Eligible subjects of this study were patients with type 2 diabetes—identified with the International Classification of Primary Care, 2nd edition, code T90—and had attended at least one physician consultation during 2008 to 2018. Patient demographics, physiological measurements, disease and medication records, and the number of attendances were extracted from the HA Clinical Management System. This database is the centralized electronic health care system for public hospitals and clinics in Hong Kong. It has previously been used in studies on the health service and epidemiology in Hong Kong (1113). The earliest attendance date within the period above was established as the baseline. Patients who were <18 years old, diagnosed with type 1 diabetes, had a CVD event on or before baseline, or had fewer than three attendances in the 2 years before baseline were excluded. Patients were followed until a CVD event, death, or 31 December 2019, whichever occurred first. The detailed study design is illustrated in Fig. 1.

Figure 1

Flowchart of the study design. Team-based UPCI was calculated by dividing the number of attendances to the most visited physician team with the total number of attendances within the 2-year measurement period.

Figure 1

Flowchart of the study design. Team-based UPCI was calculated by dividing the number of attendances to the most visited physician team with the total number of attendances within the 2-year measurement period.

Close modal

In Hong Kong, there are 73 public general outpatient clinics organized in 7 geographic clusters in the whole territory. In 2008, ∼500 physicians worked in public primary care clinics across 7 HA clusters (14). Each HA cluster sets its own policy on how to organize the clinic services, including whether to adopt team-based care or allow patients to book appointments with the same team in one or more of its clinics. The physician teams are clinic-based instead of cluster-based. A typical physician team consists of around three physicians at any one time, and a physician is usually posted to a clinic for 2–5 years. For quality assurance, the HA has established standardized guidelines, protocols, and key performance indicators for diabetes management, which all physicians would need to follow.

Measurement of Continuity of Care

To calculate team-based continuity of care, we included all physician attendances in 2 years before baseline. The usual provider continuity index (UPCI), defined as the proportion of visits to the most frequently attended physician team over the total number of attendances for a patient, was used. As the focus was on the association between having a usual physician team and complications development, the UPCI is a more suitable assessment of continuity of care than the Continuity of Care Index (COCI) or the Modified Modified Continuity Index, which measure the dispersion across teams, or the Sequential Continuity Index, which measures the frequency of revisiting the same physician team as the prior consultation (calculations of the measures are shown in Supplementary Table 2) (15,16). The UPCI scores range from 0 to 1, with 1 indicating that the same physician team managed the patient in every visit. A UPCI of 0.50 indicates that the patient visited the same team 5 out of every 10 attendances.

Outcomes

The primary outcome of this study was CVD incidence, including coronary heart disease (CHD), stroke, and heart failure. Secondary outcomes included the incidence of peripheral vascular disease and all-cause mortality. Disease outcomes were defined by the International Classification of Primary Care, 2nd edition, or ICD-9-CM codes (Supplementary Table 1). Patient mortality data were obtained from the Hong Kong Death Registry.

Baseline Characteristics

Baseline characteristics included age, sex, smoking status, the Charlson Comorbidity Index (CCI), hemoglobin A1c (HbA1c), systolic blood pressure (SBP), diastolic blood pressure (DBP), LDL cholesterol (LDL-C), BMI, estimated glomerular filtration rate (eGFR), the total number of attendances within 2 years before the baseline, uses of antidiabetic drugs, antihypertensive drugs, and lipid-lowering agents. All laboratory assays were carried out in laboratories accredited by the College of American Pathologists, the National Association of Testing Authorities of Australia, or the Hong Kong Accreditation Service.

Statistical Analysis

To reduce bias from having missing data, values of missing characteristics were estimated using multiple imputation (17). The estimates and their CIs were computed according to Rubin’s rules from five sets of imputations (18). Patients were divided into quartiles based on their UPCI. However, as many patients had a UPCI of 0.50, 0.75, or 1.0, they could only be divided into approximate quartiles with respective UPCI ranges and number of patients of: 1) <0.50 (N = 63,402); 2) 0.50–0.74 (N = 91,766); 3) 0.75–0.91 (N = 76,597); and 4) 0.92–1.0 (N = 80,303). The 1st quartile was defined as the reference group. An assumption of linearity of the exposure-response relationship between the UPCI and outcomes was verified with restricted cubic splines using three knots (19). Fine stratification weights were applied to improve balance among groups, in which strata were created based on propensity scores before calculating the weights among the stratum members (20). This method reduces loss by matching for subjects that have extreme propensity scores. A propensity score was obtained from multivariable logistic regression between the patient groups and baseline characteristics. Standardized mean difference (SMD) was computed to evaluate the dissimilarity in characteristics among quartiles before and after the matching. An SMD <0.2 indicated that a characteristic was sufficiently balanced between groups (21).

Multivariable Cox regression adjusted with the baseline characteristics was applied to estimate the hazard ratios (HRs) and 95% CIs. A Kaplan-Meier curve was generated to compare the survival functions among the quartiles. Subgroup analyses was performed based on sex, age (<65 vs. ≥65 years), smoking status (smoker vs. nonsmoker), the CCI (<4 vs. ≥4), HbA1c (<7%/53 mmol/mol vs. ≥7%/≥53 mmol/mol), blood pressure (SBP <130 mmHg with DBP <80 mmHg vs. SBP ≥130 mmHg or DBP ≥80 mmHg), BMI (<25 kg/m2 vs. ≥25 kg/m2), and eGFR (≥60 mL/min/1.73 m2 vs. <60 mL/min/1.73 m2). The cutoff age of 65 years was adapted based on the World Health Organization definition of elderly (22), while those for HbA1c and blood pressure were based on their respective recommended levels for patients with diabetes according to the International Diabetes Federation (23). The cutoff value for the CCI was determined based on the median of the study population. The standard definitions for obesity and chronic kidney disease in Hong Kong were used for BMI and eGFR. Bonferroni method was applied for multiplicity adjustment (24), with an adjusted significance level of 0.05/8 = 0.00625. A P value of <0.00625 in the heterogeneity test would indicate significant differences within subgroups.

Several sensitivity analyses on CVD risk were conducted, including an analysis without weighting, an analysis with complete cases, and an analysis including patients with at least 3 years of follow-up. Sensitivity analyses were performed to assess the robustness of our findings by dividing the patients into quartiles defined by the COCI, the Modified Modified Continuity Index, and the Sequential Continuity Index, respectively. To affirm that the inclusion criterion of at least three attendances had a negligible impact on the result, the associations were examined again by including patients with at least five attendances and at least eight attendances. The relationship between team-based continuity of care and CVD/all-cause mortality risk after accounting the effect of control subjects with diabetes and medications, by applying HbA1c as a time-varying covariate, and adding the use of two relatively newer classes of antidiabetic drugs—sodium–glucose cotransport 2 inhibitors or glucagon-like peptide 1 receptor agonists—as confounders was also investigated. Another sensitivity analysis was performed by adjusting for the yearly average attendances during the follow-up period as a confounder.

All analyses were performed using Stata 16.1. Statistical significance was defined as a two-tailed P value <0.05 except in the subgroup analyses, where the Bonferroni method was applied.

Ethics Approval and Consent to Participate

This study was approved by the Institutional Review Board of The University of Hong Kong—the HA Hong Kong West Cluster (reference number UW 19–329). As all data used in this study have been anonymized and retrieved from the Clinical Management System of the Hong Kong HA, no consent to participate from the patient was required.

Data and Resource Availability

Due to the local regulations on the distribution of personal data and the privacy policy from the data holder, the data used in this study cannot be released through public data repositories. However, access to the data can be applied through the Data Sharing Portal of the Hong Kong HA (https://www3.ha.org.hk/data/DCL/Index/).

All baseline characteristics had a data completion rate of ≥84% (Supplementary Table 3). Based on the SMD values before weighting, patients with higher continuity of care tended to be older, had fewer attendances, had a higher CCI, and were more likely to be taking antihypertensive drugs (Table 1). After multiple imputation and weighting, 99 participants were excluded due to the lack of an available match. As a result, a total of 312,068 eligible patients were included after weighting, among which 146,932 (47.1%) were male, and the average age was 63.4 years. All SMD values were <0.2 after weighting, indicating sufficient balance between groups. The cohort for this study had an average UPCI (SD) of 0.71 (0.24) and an average number of attendances (SD) of 12 (4.6) in the 2 years before baseline.

Table 1

Baseline characteristics by team-based UPCI quartiles after multiple imputation with and without weighting adjustments

Without weightingAdjusted with weighting
1st quartile2nd quartile3rd quartile4th quartileSMD1st quartile2nd quartile3rd quartile4th quartileSMD
UPCI <0.50 (N = 63,473)UPCI 0.50–0.74 (N = 91,777)UPCI 0.75–0.91 (N = 76,608)UPCI 0.92–1.0 (N = 80,309)UPCI <0.50 (N = 63,402)UPCI 0.50–0.74 (N = 91,766)UPCI 0.75–0.91 (N = 76,597)UPCI 0.92–1.0 (N = 80,303)
Male, n (%) 32,594 (51.4) 43,416 (47.3) 34,386 (44.9) 35,635 (44.4) 0.14 30,108 (47.5) 43,033 (46.9) 35,969 (47.0) 37,821 (47.1) 0.01 
Age, years 61.1 (11.4) 63.3 (11.4) 64.2 (11.3) 64.9 (11.2) 0.34 63.1 (11.5) 63.4 (11.5) 63.4 (11.5) 63.6 (11.3) 0.04 
Smoker, n (%) 5,195 (8.2) 6,267 (6.8) 4,446 (5.8) 4,056 (5.1) 0.13 4,124 (6.5) 5,789 (6.3) 5,038 (6.6) 5,468 (6.8) 0.02 
SBP, mmHg 133.1 (16.8) 133.5 (16.7) 133.5 (16.5) 134.0 (16.4) 0.06 133.6 (17.0) 133.5 (16.7) 133.5 (16.5) 133.7 (16.4) 0.01 
DBP, mmHg 75.7 (10.3) 74.7 (10.2) 74.3 (10.0) 74.0 (9.9) 0.17 74.9 (10.4) 74.7 (10.3) 74.7 (10.0) 74.6 (9.9) 0.03 
HbA1c, % 7.2 (1.3) 7.1 (1.2) 7.0 (1.1) 7.1 (1.1) 0.11 7.1 (1.2) 7.1 (1.2) 7.1 (1.2) 7.1 (1.2) 0.02 
HbA1c, mmol/mol 54.9 (14.3) 53.9 (12.9) 53.4 (12.2) 53.6 (12.2) 0.11 54.1 (13.6) 53.9 (13.0) 54.0 (12.8) 54.2 (12.7) 0.02 
BMI, kg/m2 25.9 (4.2) 25.8 (4.1) 25.7 (4.1) 25.8 (4.0) 0.03 25.8 (4.1) 25.8 (4.1) 25.8 (4.1) 25.8 (4.0) 0.01 
LDL-C, mmol/L 2.8 (0.8) 2.8 (0.8) 2.8 (0.8) 2.8 (0.9) 0.04 2.8 (0.9) 2.8 (0.8) 2.8 (0.8) 2.8 (0.9) 0.02 
eGFR, mL/min/1.73 m2 107.6 (30.6) 105.1 (89.3) 103.6 (42.6) 102.0 (39.9) 0.16 105.1 (29.9) 104.4 (40.0) 104.2 (29.6) 104.4 (44.9) 0.03 
CCI 3.7 (1.3) 3.9 (1.2) 4.0 (1.2) 4.0 (1.2) 0.26 3.9 (1.2) 3.9 (1.2) 3.9 (1.2) 3.9 (1.2) 0.02 
Use of antidiabetic drugs, n (%) 50,990 (80.3) 75,108 (81.8) 62,848 (82.0) 65,226 (81.2) 0.04 51,595 (81.4) 74,522 (81.2) 62,463 (81.5) 65,755 (81.9) 0.02 
Use of antihypertensive drugs, n (%) 43,739 (68.9) 67,999 (74.1) 59,111 (77.2) 64,413 (80.2) 0.26 46,945 (74.0) 69,163 (75.4) 57,320 (74.8) 59,502 (74.1) 0.03 
Use of lipid-lowering agents, n (%) 21,410 (33.7) 31,789 (34.6) 27,952 (36.5) 28,639 (35.7) 0.06 21,804 (34.4) 32,354 (35.3) 26,728 (34.9) 27,435 (34.2) 0.02 
Number of attendances 14 (7.8) 12 (4.7) 12 (3.5) 11 (3.4) 0.52 12 (5.2) 12 (4.6) 12 (4.2) 12 (4.5) 0.03 
UPCI 0.36 (0.08) 0.61 (0.07) 0.84 (0.06) 0.99 (0.02) NA 0.37 (0.08) 0.61 (0.07) 0.84 (0.06) 0.99 (0.02) NA 
Without weightingAdjusted with weighting
1st quartile2nd quartile3rd quartile4th quartileSMD1st quartile2nd quartile3rd quartile4th quartileSMD
UPCI <0.50 (N = 63,473)UPCI 0.50–0.74 (N = 91,777)UPCI 0.75–0.91 (N = 76,608)UPCI 0.92–1.0 (N = 80,309)UPCI <0.50 (N = 63,402)UPCI 0.50–0.74 (N = 91,766)UPCI 0.75–0.91 (N = 76,597)UPCI 0.92–1.0 (N = 80,303)
Male, n (%) 32,594 (51.4) 43,416 (47.3) 34,386 (44.9) 35,635 (44.4) 0.14 30,108 (47.5) 43,033 (46.9) 35,969 (47.0) 37,821 (47.1) 0.01 
Age, years 61.1 (11.4) 63.3 (11.4) 64.2 (11.3) 64.9 (11.2) 0.34 63.1 (11.5) 63.4 (11.5) 63.4 (11.5) 63.6 (11.3) 0.04 
Smoker, n (%) 5,195 (8.2) 6,267 (6.8) 4,446 (5.8) 4,056 (5.1) 0.13 4,124 (6.5) 5,789 (6.3) 5,038 (6.6) 5,468 (6.8) 0.02 
SBP, mmHg 133.1 (16.8) 133.5 (16.7) 133.5 (16.5) 134.0 (16.4) 0.06 133.6 (17.0) 133.5 (16.7) 133.5 (16.5) 133.7 (16.4) 0.01 
DBP, mmHg 75.7 (10.3) 74.7 (10.2) 74.3 (10.0) 74.0 (9.9) 0.17 74.9 (10.4) 74.7 (10.3) 74.7 (10.0) 74.6 (9.9) 0.03 
HbA1c, % 7.2 (1.3) 7.1 (1.2) 7.0 (1.1) 7.1 (1.1) 0.11 7.1 (1.2) 7.1 (1.2) 7.1 (1.2) 7.1 (1.2) 0.02 
HbA1c, mmol/mol 54.9 (14.3) 53.9 (12.9) 53.4 (12.2) 53.6 (12.2) 0.11 54.1 (13.6) 53.9 (13.0) 54.0 (12.8) 54.2 (12.7) 0.02 
BMI, kg/m2 25.9 (4.2) 25.8 (4.1) 25.7 (4.1) 25.8 (4.0) 0.03 25.8 (4.1) 25.8 (4.1) 25.8 (4.1) 25.8 (4.0) 0.01 
LDL-C, mmol/L 2.8 (0.8) 2.8 (0.8) 2.8 (0.8) 2.8 (0.9) 0.04 2.8 (0.9) 2.8 (0.8) 2.8 (0.8) 2.8 (0.9) 0.02 
eGFR, mL/min/1.73 m2 107.6 (30.6) 105.1 (89.3) 103.6 (42.6) 102.0 (39.9) 0.16 105.1 (29.9) 104.4 (40.0) 104.2 (29.6) 104.4 (44.9) 0.03 
CCI 3.7 (1.3) 3.9 (1.2) 4.0 (1.2) 4.0 (1.2) 0.26 3.9 (1.2) 3.9 (1.2) 3.9 (1.2) 3.9 (1.2) 0.02 
Use of antidiabetic drugs, n (%) 50,990 (80.3) 75,108 (81.8) 62,848 (82.0) 65,226 (81.2) 0.04 51,595 (81.4) 74,522 (81.2) 62,463 (81.5) 65,755 (81.9) 0.02 
Use of antihypertensive drugs, n (%) 43,739 (68.9) 67,999 (74.1) 59,111 (77.2) 64,413 (80.2) 0.26 46,945 (74.0) 69,163 (75.4) 57,320 (74.8) 59,502 (74.1) 0.03 
Use of lipid-lowering agents, n (%) 21,410 (33.7) 31,789 (34.6) 27,952 (36.5) 28,639 (35.7) 0.06 21,804 (34.4) 32,354 (35.3) 26,728 (34.9) 27,435 (34.2) 0.02 
Number of attendances 14 (7.8) 12 (4.7) 12 (3.5) 11 (3.4) 0.52 12 (5.2) 12 (4.6) 12 (4.2) 12 (4.5) 0.03 
UPCI 0.36 (0.08) 0.61 (0.07) 0.84 (0.06) 0.99 (0.02) NA 0.37 (0.08) 0.61 (0.07) 0.84 (0.06) 0.99 (0.02) NA 

Data are mean (SD) unless otherwise indicated.

In the adjusted groups, fine stratification weights were applied. The propensity score was generated by multivariable logistic regression adjusted with the listed parameters, except UPCI. The SMD listed is the largest SMD between any pairs of the quartiles. A SMD <0.2 indicates sufficient balance for the characteristic.

NA, not applicable.

Over a follow-up of 2,111,580 person-years, the total number of new CVD events was 52,428. Overall, CVD, CHD, stroke, and peripheral vascular disease displayed a linear relationship with UPCI (Supplementary Fig. 1). The median CVD follow-up period for the quartiles was: 1) 77.5 months; 2) 77.5 months; 3) 77.5 months; and 4) 78.5 months (Fig. 2). As the UCPI increased, both the incidence rate and HRs for CVD decreased. Incidence rates (cases per 1,000 person-years) for the four groups were: 1) 26.1; 2) 25.4; 3) 24.6; and 4) 23.6. The CVD HRs (95% CI) of patients in quartiles 2nd–4th, relative to the 1st, were: 2) 0.95 (0.92–0.97); 3) 0.92 (0.89–0.94); and 4) 0.87 (0.84–0.89), respectively. Similar trends were observed in CHDs, stroke, and peripheral vascular disease. Patients with UPCI of ≥0.50 had lower risks of all-cause mortality than those with UPCI <0.50, but the risk did not reduce further upon further increase of UPCI. There were no significant differences in the risk for heart failure. The baseline characteristics for each of the sensitivity analyses are presented in Supplementary Table 4AG. Similar associations between team-based continuity of care and CVD/all-cause mortality risks were observed in the sensitivity analyses compared with the main analysis (Supplementary Fig. 2). The Kaplan-Meier curve for CVD showed that groups with higher UPCI were significantly less likely to develop CVD throughout the follow-up period (Supplementary Fig. 3).

Figure 2

Associations of team-based continuity of care with CVD and all-cause mortality among patients with diabetes. UPCI cutoffs of the quartiles were: <0.5, 0.5–0.74, 0.75–0.91, and 0.92–1.0. HR was adjusted by age, sex, smoking status, SBP and DBP, BMI, HbA1c, LDL-C, eGFR, CCI, number of attendances, use of antihypertensive drug, use of lipid-lowering drug, and use of antidiabetic drug at baseline. CVD includes CHD, heart failure, and stroke.

Figure 2

Associations of team-based continuity of care with CVD and all-cause mortality among patients with diabetes. UPCI cutoffs of the quartiles were: <0.5, 0.5–0.74, 0.75–0.91, and 0.92–1.0. HR was adjusted by age, sex, smoking status, SBP and DBP, BMI, HbA1c, LDL-C, eGFR, CCI, number of attendances, use of antihypertensive drug, use of lipid-lowering drug, and use of antidiabetic drug at baseline. CVD includes CHD, heart failure, and stroke.

Close modal

The association between team-based continuity of care and reduction in CVD risk was significantly stronger in patients who were <65 years old than those aged ≥65 years (Fig. 3). Relative to patients in the 1st quartile, the CVD HRs (99.4% CI) for those in the three higher quartiles were: 2) 0.93 (0.87–0.98), 3) 0.88 (0.83–0.94), and 4) 0.82 (0.77–0.87) for those <65 years old; and 2) 0.97 (0.93–1.02), 3) 0.96 (0.92–1.01), and 4) 0.93 (0.89–0.97) for those ≥65 years.

Figure 3

Subgroup analyses on the association between team-based continuity of care and CVD among patients with diabetes. UPCI cutoffs of the quartiles were: <0.5, 0.5–0.74, 0.75–0.91, and 0.92–1.0. HR was adjusted by age, sex, smoking status, SBP and DBP, BMI, HbA1c, LDL-C, eGFR, CCI, number of attendances, use of antihypertensive drug, use of lipid-lowering drug and use of antidiabetic drug at baseline. Bonferroni method was applied using an adjusted significant level of 0.00625. CVD includes CHD, heart failure, and stroke.

Figure 3

Subgroup analyses on the association between team-based continuity of care and CVD among patients with diabetes. UPCI cutoffs of the quartiles were: <0.5, 0.5–0.74, 0.75–0.91, and 0.92–1.0. HR was adjusted by age, sex, smoking status, SBP and DBP, BMI, HbA1c, LDL-C, eGFR, CCI, number of attendances, use of antihypertensive drug, use of lipid-lowering drug and use of antidiabetic drug at baseline. Bonferroni method was applied using an adjusted significant level of 0.00625. CVD includes CHD, heart failure, and stroke.

Close modal

The findings of this study suggest that higher team-based continuity of care was associated with lower risks of CVD among patients with type 2 diabetes. Higher team-based continuity of care correlated with lower risks of overall CVD, CHD, stroke and peripheral vascular diseases. While the risk for all-cause mortality was lower in patients with a team-based UPCI of ≥0.50 than those with UPCI <0.50, there was no apparent improvement upon further increases in team-based continuity of care. Noticeable differences in the magnitude of associations between continuity of care and CVD risk were also observed depending on the patients’ age.

Our findings correspond to previous studies that a higher continuity of care was associated with a reduction in the incidence rates of complications in patients with diabetes (35). Hussey et al. (3) showed that for each 0.1-unit increase in individual-based COCI, there was a corresponding reduction in the odds for acute myocardial infarction or coronary thrombolysis by 13%. Similarly, compared with patients with low individual-based continuity of care (UPCI <0.7), patients who received care from a single physician were 35% less likely to be diagnosed with diabetes-related complications (including CVD, renal diseases, neural diseases, and others) (4). The effect observed in our study was smaller than those in the previous literature. However, previous studies either have not differentiated the impact of continuity of care by individual complications (4,5), had a relatively smaller sample size (5), or had a short follow-up period (1 year) (3). Having addressed these limitations, our findings provided more reliable and precise evidence of the association between continuity of care and CVD risk among patients with diabetes. Notably, our results revealed that visiting the same physician team more frequently was generally correlated with reduced CVD risk. This could suggest a certain degree of flexibility is feasible for implementing continuity of care within a public health care setting. However, as this study was conducted in the Hong Kong government–subsidized public-sector health setting, our findings, therefore, might not be generalizable to other settings. As the adoption of team-based models of care becomes more popular worldwide, further studies conducted in different locations will be needed to confirm and build upon our findings.

Managing patients using a few physicians working as a team could improve interpersonal relationships between the patients and the physicians. Higher continuity of care helps improve patient satisfaction (25), develops patients’ trust in their physicians (26), and facilitates treatment compliance. Physicians could better understand the patient’s health conditions through more frequent monitoring of HbA1c, blood pressure, and lipid levels (27), thus making more informed decisions and providing timely and appropriate treatment. Studies have shown that a well-controlled HbA1c could mitigate the risk of CVD development (28,29), and higher continuity of care could promote glycemic control (3032). We also found that patients with a UPCI of ≥0.50 had lower risks for all-cause mortality than those with a UPCI <0.50, but the benefits did not increase proportionately with a higher UPCI. This might be because patients who had died were more likely to have more severe conditions and comorbidities, which modulated the effect of continuity of care.

The weaker associations between team-based continuity of care and reduction in CVD events observed in older patients could be because these patients were likely to have poorer health, which contributes to the increased CVD risk. Their conditions could be less treatable, and the benefits from a more enduring and consonant relationship with their physicians may not be as apparent.

A key strength of our study was the large sample size and robust statistical methods used. We minimized bias using multiple imputations and improved the sample balance using fine stratification weights (33). Compared with previous similar studies, the relatively long follow-up period and assessment by individual complications provided a more comprehensive overview of the association between continuity of care and CVD development than previous studies (35). By examining the differential associations among patients with different characteristics, we provided a more nuanced picture and additional insights for better policy implementation for continuity of care. Sensitivity analyses, including different levels and measures of continuity of care, and inclusion of only patients with at least 8 visits in 2 years showed a consistent effect of continuity of care, indicating the robustness of the findings.

Nevertheless, there were several limitations that might require further exploration. First, this study only examined team-based continuity of care at public primary care clinics. Hong Kong implements a dual health care system consisting of both public and private services. Patients with higher socioeconomic status may be more likely to visit private clinics and, therefore, be underrepresented in our study. Nevertheless, ∼90% of patients with diabetes were managed under the public sector (10); thus, most medical attendances for diabetes consultations should have been included in our data set. Second, some important confounders such as socioeconomic status, alcohol consumption, duration of diabetes, or details of the physician teams were not routinely recorded and unavailable in the study data set. In addition, patterns on health-seeking behavior in patients are not readily available from electronic record databases; it remains uncertain if patients who are more conscious about health are more likely to continue with follow-up medical appointments. Third, our study design precludes any causal relationship to be drawn between higher team-based continuity of care and better health outcomes in patients with diabetes. Therefore, more research should be performed, ideally with randomized controlled trials, to verify the effect of team-based continuity of care on patients with diabetes.

In conclusion, higher team-based continuity of care was associated with a reduction in CVD risk and mortality in primary care patients with diabetes, especially among patients <65 years old. Maintaining a one-to-one patient–physician relationship is challenging in public primary care clinics with many physicians. Continuity of care from the same physician team may serve as a more flexible alternative to achieve better health outcomes for patients with diabetes.

This article contains supplementary material online at https://doi.org/10.2337/figshare.19165346.

Acknowledgments. The authors thank the HA head office, staff in primary care clinics in all seven clusters, and the Statistics and Workforce Planning Department at the Hong Kong HA for the contributions. Multiple imputation was performed using research computing facilities offered by Information Technology Services, The University of Hong Kong.

Funding. This study is supported by the Health and Medical Research Fund, Food and Health Bureau, the Government of Hong Kong Special Administrative Region (SAR; project no. CFS-HKU4).

No funding organization has any role in the design and conduct of the study, collection, management, analysis, or interpretation of the data, or preparation of the manuscript.

Duality of Interest. E.Y.F.W. has received research grants from the Food and Health Bureau of the Government of the Hong Kong SAR and the Hong Kong Research Grant Council, outside the submitted work. E.Y.T.Y. has received research grants from the Food and Health Bureau of the Government of the Hong Kong SAR, outside the submitted work. C.L.K.L. has received research grants from the Food and Health Bureau of the Government of the Hong Kong SAR, the Hong Kong Research Grant Council, the Hong Kong College of Family Physicians, and Kerry Group Kuok Foundation, outside the submitted work. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. E.Y.F.W. conceived of the presented idea. E.Y.F.W. and C.L.K.L. contributed to the acquisition of data. K.S.C., E.Y.F.W., W.Y.C., E.Y.T.Y., and C.L.K.L. contributed to the study design, statistical analysis, and interpretation of the results. K.S.C., E.Y.F.W., I.L.M., W.H.G.C., and M.K.H. contributed to manuscript draft. All authors contributed to the review and edited the manuscript. K.S.C. and E.Y.F.W. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Cho
NH
,
Shaw
JE
,
Karuranga
S
, et al
.
IDF Diabetes Atlas: global estimates of diabetes prevalence for 2017 and projections for 2045
.
Diabetes Res Clin Pract
2018
;
138
:
271
281
2.
Chan
JCN
,
Lim
L-L
,
Wareham
NJ
, et al
.
The Lancet Commission on diabetes: using data to transform diabetes care and patient lives
.
Lancet
2021
;
396
:
2019
2082
3.
Hussey
PS
,
Schneider
EC
,
Rudin
RS
,
Fox
DS
,
Lai
J
,
Pollack
CE
.
Continuity and the costs of care for chronic disease
.
JAMA Intern Med
2014
;
174
:
742
748
4.
Liao
P-J
,
Lin
Z-Y
,
Huang
J-C
,
Hsu
K-H
.
The relationship between type 2 diabetic patients’ early medical care-seeking consistency to the same clinician and health care system and their clinical outcomes
.
Medicine (Baltimore)
2015
;
94
:
e554
5.
Nam
JH
,
Lee
C
,
Kim
N
, et al
.
Impact of continuous care on health outcomes and cost for type 2 diabetes mellitus: analysis using National Health Insurance Cohort database
.
Diabetes Metab J
2019
;
43
:
776
784
6.
Raivio
R
,
Jääskeläinen
J
,
Holmberg-Marttila
D
,
Mattila
KJ
.
Decreasing trends in patient satisfaction, accessibility and continuity of care in Finnish primary health care - a 14-year follow-up questionnaire study
.
BMC Fam Pract
2014
;
15
:
98
7.
Hellesø
R
,
Fagermoen
MS
.
Cultural diversity between hospital and community nurses: implications for continuity of care
.
Int J Integr Care
2010
;
10
:
e036
8.
Majeed
A
.
Shortage of general practitioners in the NHS
.
BMJ
2017
;
358
:
j3191
9.
Reddy
A
,
Wong
E
,
Canamucio
A
, et al
.
Association between continuity and team-based care and health care utilization: An observational study of Medicare-eligible veterans in VA patient aligned care team
.
Health Serv Res
2018
;
53
(
Suppl. 3
):
5201
5218
10.
Lau
IT
.
A clinical practice guideline to guide a system approach to diabetes care in Hong Kong
.
Diabetes Metab J
2017
;
41
:
81
88
11.
Wan
EYF
,
Chin
WY
,
Yu
EYT
, et al
.
The impact of cardiovascular disease and chronic kidney disease on life expectancy and direct medical cost in a 10-year diabetes cohort study
.
Diabetes Care
2020
;
43
:
1750
1758
12.
Wan
EYF
,
Chui
CSL
,
Lai
FTT
, et al
.
Bell’s palsy following vaccination with mRNA (BNT162b2) and inactivated (CoronaVac) SARS-CoV-2 vaccines: a case series and nested case-control study
.
Lancet Infect Dis
2022
;
22
:
64
72
13.
Wan
EYF
,
Fung
CSC
,
Jiao
FF
, et al
.
Five-year effectiveness of the multidisciplinary Risk Assessment and Management Programme–Diabetes Mellitus (RAMP-DM) on diabetes-related complications and health service uses—a population-based and propensity-matched cohort study
.
Diabetes Care
2018
;
41
:
49
59
14.
Government of the Hong Kong Special Administrative Region
.
LCQ12: Doctors working in general outpatient clinics [press release], 2008
.
15.
Saultz
JW
.
Defining and measuring interpersonal continuity of care
.
Ann Fam Med
2003
;
1
:
134
143
16.
Dreiher
J
,
Comaneshter
DS
,
Rosenbluth
Y
,
Battat
E
,
Bitterman
H
,
Cohen
AD
.
The association between continuity of care in the community and health outcomes: a population-based study
.
Isr J Health Policy Res
2012
;
1
:
21
17.
Royston
P
.
Multiple imputation of missing values
.
Stata J
2004
;
4
:
227
241
18.
Rubin
DB.
Multiple Imputation for Nonresponse in Surveys
.
New York
,
John Wiley & Sons
,
2004
19.
Desquilbet
L
,
Mariotti
F
.
Dose-response analyses using restricted cubic spline functions in public health research
.
Stat Med
2010
;
29
:
1037
1057
20.
Desai
RJ
,
Franklin
JM
.
Alternative approaches for confounding adjustment in observational studies using weighting based on the propensity score: a primer for practitioners
.
BMJ
2019
;
367
:
I5657
21.
Austin
PC
.
Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples
.
Stat Med
2009
;
28
:
3083
3107
22.
World Health Organization
.
Definition of an older or elderly person, 2013
.
Available from https://www.who.int/healthinfo/survey/ageingdefnolder/en/. Accessed 30 June 2015
23.
International Diabetes Federation Guideline Development Group
.
Global guideline for type 2 diabetes
.
Diabetes Res Clin Pract
2014
;
104
:
1
52
24.
Bland
JM
,
Altman
DG
.
Multiple significance tests: the Bonferroni method
.
BMJ
1995
;
310
:
170
25.
Hjortdahl
P
,
Laerum
E
.
Continuity of care in general practice: effect on patient satisfaction
.
BMJ
1992
;
304
:
1287
1290
26.
Parchman
ML
,
Burge
SK
.
The patient-physician relationship, primary care attributes, and preventive services
.
Fam Med
2004
;
36
:
22
27
27.
Parchman
ML
;
Residency Research Network of South Texas Investigators
.
Continuity and quality of care in type 2 diabetes: a Residency Research Network of South Texas study
.
J Fam Pract
2002
;
51
:
619
624
28.
Singer
DE
,
Nathan
DM
,
Anderson
KM
,
Wilson
PW
,
Evans
JC
.
Association of HbA1c with prevalent cardiovascular disease in the original cohort of the Framingham Heart Study
.
Diabetes
1992
;
41
:
202
208
29.
Kelly
TN
,
Bazzano
LA
,
Fonseca
VA
,
Thethi
TK
,
Reynolds
K
,
He
J
.
Systematic review: glucose control and cardiovascular disease in type 2 diabetes
.
Ann Intern Med
2009
;
151
:
394
403
30.
Mainous
AG
3rd
,
Koopman
RJ
,
Gill
JM
,
Baker
R
,
Pearson
WS
.
Relationship between continuity of care and diabetes control: evidence from the Third National Health and Nutrition Examination Survey
.
Am J Public Health
2004
;
94
:
66
70
31.
Dearinger
AT
,
Wilson
JF
,
Griffith
CH
,
Scutchfield
FD
.
The effect of physician continuity on diabetic outcomes in a resident continuity clinic
.
J Gen Intern Med
2008
;
23
:
937
941
32.
Lustman
A
,
Comaneshter
D
,
Vinker
S
.
Interpersonal continuity of care and type two diabetes
.
Prim Care Diabetes
2016
;
10
:
165
170
33.
Dehejia
RH
,
Wahba
S
.
Propensity score-matching methods for nonexperimental causal studies
.
Rev Econ Stat
2002
;
84
:
151
161
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.