OBJECTIVE

To evaluate the 5-year effectiveness of a multidisciplinary Risk Assessment and Management Programme–Diabetes Mellitus (RAMP-DM) in primary care patients with type 2 diabetes.

RESEARCH DESIGN AND METHODS

A 5-year prospective cohort study was conducted with 121,584 Chinese primary care patients with type 2 DM who were recruited between August 2009 and June 2011. Missing data were dealt with multiple imputations. After excluding patients with prior diabetes mellitus (DM)-related complications and one-to-one propensity score matching on all patient characteristics, 26,718 RAMP-DM participants and 26,718 matched usual care patients were followed up for a median time of 4.5 years. The effect of RAMP-DM on nine DM-related complications and all-cause mortality were evaluated using Cox regressions. The first incidence for each event was used for all models. Health service use was analyzed using negative binomial regressions. Subgroup analyses on different patient characteristics were performed.

RESULTS

The cumulative incidence of all events (DM-related complications and all-cause mortality) was 23.2% in the RAMP-DM group and 43.6% in the usual care group. RAMP-DM led to significantly greater reductions in cardiovascular disease (CVD) risk by 56.6% (95% CI 54.5, 58.6), microvascular complications by 11.9% (95% CI 7.0, 16.6), mortality by 66.1% (95% CI 64.3, 67.9), specialist attendance by 35.0% (95% CI 33.6, 36.4), emergency attendance by 41.2% (95% CI 39.8, 42.5), and hospitalizations by 58.5% (95% CI 57.2, 59.7). Patients with low baseline CVD risks benefitted the most from RAMP-DM, which decreased CVD and mortality risk by 60.4% (95% CI 51.8, 67.5) and 83.6% (95% CI 79.3, 87.0), respectively.

CONCLUSIONS

This naturalistic study highlighted the importance of early optimal DM control and risk factor management by risk stratification and multidisciplinary, protocol-driven, chronic disease model care to delay disease progression and prevent complications.

With population aging and increasing prevalence of obesity, the number of patients with diabetes mellitus (DM) and global health expenditure related to DM are forecast to grow substantially, totaling an estimated U.S. $642 million and U.S. $802 billion, respectively, by 2040 (1). Increasing demand for health care services coupled with limitations in resources have led to several international guidelines, including those by the American Diabetes Association and the National Institute for Health and Care Excellence, recommending regular risk assessment and multidisciplinary management (13). A chronic disease service delivery model that incorporates risk-stratified care planning, a multidisciplinary team of health professionals to provide ongoing treatments, patient education, and scheduled health assessments for monitoring of disease control and complications has been promoted globally as a more holistic and cost-effective way to manage patients with diabetes (48).

Several studies (921) have demonstrated that treatments targeting blood glucose, blood pressure (BP), and cholesterol control improve surrogate outcomes and delay macrovascular and/or microvascular complications. However, only two randomized controlled trials (RCTs) and one observational study have examined the impact of implementing a chronic disease service delivery model using multidisciplinary management and risk-stratified protocol-based treatments. These studies have found that the chronic disease model of care is associated with reductions in hemoglobin A1c (HbA1c), BP, and predicted 10-year cardiovascular disease (CVD) risks (1416), but their results have been based on surrogate CVD markers rather than actual clinical events such as CVD and mortality and are not conclusive about the effectiveness in preventing DM-related complications. One RCT with a short 6-month follow-up period conducted in the U.S. demonstrated the effectiveness of multidisciplinary management on health service use such as hospital admissions (22).

To date, most studies examining the effectiveness of therapeutic interventions for diabetes have been performed under relatively artificial conditions with strict inclusion criteria or have used study populations from hospital-based settings. The findings of such studies may not be reflective of the outcomes of care for patients with diabetes who are managed in real-world primary care settings. As the trend for the management of type 2 DM has shifted from hospitals to primary care (23), naturalistic population-based studies evaluating the effectiveness of chronic disease interventions for patients with diabetes are needed to confirm the benefits of this model of care.

The Risk Assessment and Management Programme–Diabetes Mellitus (RAMP-DM) was introduced to complement the usual care into the public primary care general outpatient clinics (GOPCs) of Hong Kong in 2009 (24). RAMP-DM is based on a chronic disease model of care and uses risk-stratified care planning, multidisciplinary care (coordinated by a nurse manager), and scheduled monitoring of complications. Our preliminary analyses found an improvement in surrogate outcomes and cardiovascular complications over 3 years (25,26). This study evaluated the long-term effectiveness of the RAMP-DM regarding all DM-related complications and health service uses over 5 years to determine the characteristics of patients receiving the greatest health benefits from the program.

Study Design

This was a territory-wide prospective cohort study to compare the risks of CVD, microvascular complications, and all-cause mortality, and the frequencies of health service uses over 5 years between RAMP-DM participants and patients receiving usual primary care (i.e., in GOPCs).

Setting of RAMP-DM

The Hong Kong Hospital Authority (HA), an organization governing all public sector hospitals and primary care clinics in Hong Kong, launched RAMP-DM in August 2009 as a territory-wide program to improve the quality of care for primary care patients with DM. The details of RAMP-DM have been reported in the previous study protocol (24). All patients with DM attending the HA GOPCs were eligible to be enrolled into the RAMP-DM. All clinics used the same RAMP-DM protocol. Given that Hong Kong has a substantial subsidized public health care system, the HA provides care for at least 90% of the patients with diagnosed diabetes in Hong Kong (27).

All eligible GOPC patients were invited to enroll into the RAMP-DM randomly during their regular follow-up consultations with the GOPC doctor. The workflow of the RAMP-DM is shown in Supplementary Fig. 1. Patients enrolled in the RAMP-DM initially undergo an intake risk assessment, which includes a physical examination, laboratory testing, eye and foot assessment, drug adherence and lifestyle assessment, and screening for existing diabetic complications. The screening results were reviewed by trained registered nurses who are engaged as RAMP-DM care managers, who stratify participants into “very high-risk,” “high-risk,” “medium-risk,” or “low-risk” groups according to the classification rules developed by the Joint Asia Diabetes Evaluation (JADE) study (28). The nurse care manager also provides individualized diabetes education, lifestyle advice (exercise, diet, smoking, and drinking), and an explanation of the CVD risk levels. The patient’s disease profile is recorded on an electronic clinical management system platform, which is used for the sharing of information between the multidisciplinary health care team, including doctors, nurses, and other allied health professionals (optometrists, dietitians, podiatrists, and physiotherapists), and for referrals. Care plans are developed based on individual’s risk factors according to a standardized risk-stratified guide.

Patients in the usual care group continued to be managed by their GOPC doctors based on the Hong Kong reference framework for diabetes care in primary care (29) without, however, performance of any risk assessment and stratification. Care was coordinated by the GOPC doctor who would arrange for tests and referrals to allied health as deemed necessary. Routine follow-up visits were scheduled every 3 months for review by the GOPC doctor and the dispensing of medications. Usual care patients had access to the same drug formulary and could still be referred for physical examination, laboratory testing, and various allied health services at their doctor’s discretion. Although the RAMP-DM is a territory-wide program intended for all DM patients managed in the GOPC, the rollout of the program could only be performed in stages because of the enormous number of patients with diabetes being managed and limitations of resources and manpower. This saturation of health service delivery provided a small window of opportunity to conduct a comparison of the effect of RAMP-DM and usual care in a real-world primary care setting. Patient selection for earlier enrollment into the RAMP-DM was by random allocation.

Subjects

Subject inclusion criteria were as follows: 1) age at least 18 years, 2) clinical diagnosis of type 2 DM (identified by the International Classification of Primary Care-2 [ICPC-2] code “T90” before the baseline date), and 3) no prior CVD or microvascular complications. Clinical data were extracted from the clinical management system of the HA. The RAMP-DM group was composed of subjects with DM who were enrolled into the RAMP-DM between 1 August 2009 and 30 June 2011 with the corresponding baseline date defined as the first date of the nurse intake assessment.

The usual care group included GOPC patients with DM who had not yet been enrolled into the RAMP-DM by 30 November 2015 and had at least one GOPC attendance between 1 August 2009 and 30 June 2011. The corresponding baseline date for the usual care cohort was defined as the first attendance date of GOPC within the period. For each clinical outcome, each patient was observed from their baseline date to the date of incidence of an outcome event, all-cause mortality, or last follow-up as censoring until 30 November 2015, whichever came first.

Propensity Score Matching

To reduce the selection bias, all RAMP-DM participants and usual care patients were well matched using a propensity score-matching technique. Propensity score matching aims to create similar comparison groups by using a logistic regression model that summarizes all relevant baseline covariates for each patient and generates an index score (known as the propensity score) and then matches the two groups by that score (3032). A propensity score is the conditional probability of participating in the program given the observed covariates (33). For every patient with DM, a propensity score was generated using all baseline covariates (including sociodemographic data, laboratory results, and clinical characteristics) modeled as independent variables and the RAMP-DM intervention as the dependent variable. The propensity score mapping was performed using the psmatch2 command with one-to-one matching without replacement method and with a caliper measurement of 0.001 in Stata (34). Unmatched subjects were excluded from the analysis.

Outcome Measures

The primary outcome in this study was the incidence of all-cause mortality. Mortality data were extracted from the Hong Kong Death Registry, a population-based official government registry covering all registered deaths for the residents of Hong Kong. Other outcome measures included the incidences of CVD events (consisting of coronary heart disease [CHD], heart failure, or stroke), microvascular complications (consisting of retinopathy, nephropathy, neuropathy, end-stage renal disease [ESRD], and sight-threatening diabetic retinopathy [STDR]), and service use rates. The diagnosis coding systems of ICPC-2 and ICD-9-CM from the clinical management system of the HA were used to identify events (shown in Supplementary Table 1). Service use data included the number of (overnight) hospitalizations, accident and emergency (A&E) attendances, specialist outpatient clinic (SOPC) attendances, and GOPC attendances.

Baseline Covariates

Baseline covariates consisted of a patient’s sociodemographic data, laboratory results, and clinical characteristics. Sociodemographic data included sex, age, and smoking status. Laboratory results included HbA1c, systolic BP (SBP), diastolic BP (DBP), lipid profile (LDL cholesterol [LDL-C], total cholesterol [TC] to HDL cholesterol [HDL-C] ratio, and triglyceride levels), BMI, and estimated glomerular filtration rate (eGFR). Clinical characteristics included self-reported duration of DM; diagnosed hypertension; usage of insulin, oral antidiabetic drugs, antihypertension drugs, and lipid-lowering agents; and Charlson index. Hypertension was defined by the ICPC-2 code “K86” or “K87.” All laboratory assays were performed in laboratories accredited by the College of American Pathologists, the Hong Kong Accreditation Service, or the National Association of Testing Authorities, Australia.

Data Analysis

Missing baseline covariates were handled by multiple imputation (35). This method can effectively reduce unnecessary biases (35,36), raise the power of the analysis, and produce more reliable and applicable models within clinical practice (3739). Each missing value was imputed five times by the chained equation method. For each of the five imputed data sets, the same analysis was performed and the five sets of results were combined using the rules of Rubin (40). Multiple imputation was performed using the “mi impute” command in Stata.

The RAMP-DM group subjects were first matched with usual care group subjects by propensity score matching. Descriptive statistics were displayed for both groups after matching and differences in the baseline characteristics between both groups were compared using χ2 tests for categorical variables or independent t tests for continuous variables. Differences in clinical characteristics between baseline and 60-month follow-up were assessed by paired t tests for each of the RAMP-DM and usual care groups and the difference-in-difference between the two groups was assessed by independent t tests.

Incidence rates for CVD events, microvascular complications, and all-cause mortality were estimated by an exact 95% CI based on a Poisson distribution (41). The cumulative number of incidences and incidence rates of outcome events with 95% CIs were reported. Multivariable Cox proportional hazards regression models were conducted to estimate the effect of RAMP-DM on the incidences of CVD, microvascular complications, and all-cause mortality, which accounted for all baseline characteristics of subjects. Proportional hazards assumption was also checked by examining the plots of the scaled Schoenfeld residuals against time for the covariates, and the presence of multicollinearity was also assessed by the variance inflation factor. Analysis of the data showed that all models fulfilled the proportional hazards assumption and no multicollinearity existed.

Frequencies and event rates for service uses in both groups were compared, and negative binomial regression models with the adjustment of baseline covariates were used to evaluate the effect of RAMP-DM on the frequencies of episode events as the count outcomes were overdispersed. A sensitivity analysis was conducted using a complete case cohort, in which missing values of baseline covariates were not imputed.

To explore the benefit of RAMP-DM among different patient subsets, subgroup analyses were performed on the effect of RAMP-DM versus usual care for all outcomes stratified by sex (male, female), age (<65, ≥65 years), smoking status (smoker, nonsmoker), duration of DM (<2, ≥2 years), eGFR (<60, ≥60 mL/min/1.73 m2), HbA1c (<7%, ≥7%), BMI (<27.5, ≥27.5 kg/m2), BP (<130/80, ≥130/80 mmHg), LDL-C (<2.6, ≥2.6 mmol/L), and CVD risk (low, medium, high).

All significance tests were two tailed, and those with a P value <0.05 were considered to be statistically significant. Statistical analysis was performed using Stata version 13.0.

Ethics Approval

Ethics approval of this study was granted by all local institutional review boards.

Supplementary Fig. 2 shows the subject recruitment flow. In total, 121,584 Chinese subjects with DM (RAMP-DM group 73,366 subjects; usual care group 48,218 subjects) received care for their DM in primary care clinics of the HA between 1 August 2009 and 30 June 2011. After excluding 30,418 patients (RAMP-DM group 11,596 patients; usual care group 18,822 patients) without fulfilling the subject inclusion criteria, 91,166 patients (RAMP-DM group 61,770 patients; usual care group 29,396 patients) remained. Supplementary Table 2 shows the completion rates for all baseline covariates, which ranged from 50.2% to 100%. In total, 26,718 subjects in each group were included for analysis after multiple imputation and propensity score matching. Supplementary Table 3 demonstrated that the completion rates in most laboratory and clinical assessments were nearly 100% in the baseline nurse assessment among individuals in the RAMP-DM group but were substantially lower before enrollment in the program and in the usual care group.

Table 1 displays the baseline and 60-month characteristics between the RAMP-DM and usual care groups. As expected, there were no significant differences between the two groups after propensity score matching. A total of 97.8% and 89.8% of patients, respectively, were followed up until 2015 in the RAMP-DM and usual care groups. After 60 months, the RAMP-DM group showed significant reductions in all clinical parameters, indicating that the RAMP-DM group had greater improvements than the usual care group. Compared with usual care subjects, significantly more RAMP-DM participants were diagnosed with hypertension and used oral antidiabetic drugs, antihypertensive drugs, and lipid-lowering agents. The cumulative incidence of all events (DM-related complications and all-cause mortality) was 23.2% in the RAMP-DM group and 43.6% in the usual care group. The cumulative incidences of each outcome event and comparisons of service uses are shown in Table 2. In general, fewer cases of observed events were observed in the RAMP-DM group. For instance, there were 4.34 cases of CVD/microvascular complications per 100 person-years for RAMP-DM participants, whereas the incident rate of CVD/microvascular complications in the usual care group was 7.73 per 100 person-years during a median follow-up period of 57.5 months. For all-cause mortality, the incident rates were 1.68 for RAMP-DM and 5.07 for the usual care groups. Fewer numbers of hospitalizations, A&E attendances and SOPC attendances were observed in the RAMP-DM group. Conversely, the event rates of GOPC attendance for the RAMP-DM and usual care groups were 457.0 and 354.3, respectively, indicating that the RAMP-DM group experienced more primary care attendances than the usual care group.

Table 3 shows the adjusted effect of RAMP-DM on DM-related complications and all-cause mortality and the usage frequencies of different public health services. After adjusting for all baseline covariates, the RAMP-DM group was associated with a 40.6% (hazard ratio [HR] 0.594, P < 0.001) and 66.1% (HR = 0.339, P < 0.001) greater reduction in the risks of CVD/microvascular complications and all-cause mortality than the usual care group. Significantly lower incidences of individual CVD and microvascular complications (HR 0.383–0.742, P < 0.001), except retinopathy (HR 1.256, P < 0.001), were observed among individuals in the RAMP-DM group. The incidence rate ratios (IRRs) for service use using negative binomial regression models found that the RAMP-DM group had significantly fewer numbers of hospitalizations (IRR 0.415, P < 0.001), A&E attendances (IRR 0.588, P < 0.001), and SOPC attendances (IRR 0.650, P < 0.001) but a higher number of GOPC attendances (HR 1.326, P < 0.001) compared with usual care subjects. The results of the sensitivity analysis found that the complete case cohort was similar to that of the main analysis (Supplementary Tables 4 and 5).

Subgroup analyses were conducted on the outcomes of CVD, microvascular complications, all-cause mortality. and service uses, as shown in Fig. 1A and B. In general, RAMP-DM participants in all subgroups observed a 40% greater risk reduction in each CVD/microvascular complications and a 55–85% risk reduction in all-cause mortality over that of usual care subjects. In addition, RAMP-DM participants in all subgroups had significantly fewer hospitalizations, A&E attendances, and SOPC attendances but more GOPC attendances than usual care patients. Among these subgroups, RAMP-DM participants <65 years of age with a DM duration of <2 years or with low/medium CVD risks received the greatest benefits from the RAMP-DM.

This was a large population-based study to investigate the effectiveness of a RAMP involving multidisciplinary interventions based on a chronic disease model of care for patients with diabetes in a real-world primary care setting. Our findings demonstrated that the RAMP-DM led to significantly greater reductions in CVD/microvascular complications and secondary/tertiary care service uses compared with usual care. The number needed to treat to prevent one CVD event was 8 and the number needed to treat for all-cause mortality was 6. Our results also showed that patients with lower CVD risks received the greatest benefits from the RAMP-DM.

This study found that a chronic disease model of care was effective in reducing DM-related complications and all-cause mortality when compared with usual care. It was startling to find that the risk reductions in DM-related complications and mortality were much higher than the observed improvements in the disease parameters including HbA1c, BP, and LDL-C between groups. This discrepancy suggests that the benefits of a RAMP-DM extends beyond simply improving conventional disease parameters. Several previous studies (1113,42) have found that aggressive treatments targeting a single clinical parameter did not reduce the risks of CVD and mortality compared with standard treatments. However, there have been other studies that have had findings similar to ours. The results from a post hoc analysis of the Steno-2 RCT found that their multifactorial intervention resulted in only small improvements in disease parameters but remarkable relative risk reduction in CVD (59%), diabetic nephropathy (56%), and all-cause mortality (46%) (9). Similarly, the post hoc analysis of the UK Prospective Diabetes Study (UKPDS) found that intensive glucose treatment had relative risk reductions in myocardial infarction and all-cause mortality, even though the intervention and control groups had similar HbA1c, BP, and body weight at the end of follow-up (10). Nevertheless, these studies required a much longer follow-up period of at least 10 years to obtain the differences in outcomes (9,10). Since the effectiveness of RAMP-DM within different subgroups (including smokers and patients with suboptimal HbA1c, BP, and LDL-C values) were similar, we can only conclude that there must be other factors aside from conventional disease parameters that have not been measured, which contribute to the clinical outcomes of RAMP-DM.

The large beneficial effect of the RAMP-DM on the DM-related complications and mortality might be attributed to the implementation of the following quality improvement strategies, which were also recommended to optimize the chronic care model for the management of diabetes (7,43). Through observations of a significantly higher data completion and detection rate during comprehensive screening and drug prescriptions after the RAMP-DM compared with the usual care group, the structured and systematic protocol-driven risk assessment including retinopathy screening and annual blood and urine tests enabled the identification of the patient’s risk and reversible factors and complications such as microalbuminuria and preproliferative diabetic retinopathy. Therefore, timely detection and subsequent treatments could be provided to prevent further deterioration. In the usual care group, opportunistic care was provided by GOPC doctors who were responsible for the coordination of care including the arrangement of assessments and referrals to allied health as deemed necessary. Unfortunately, the doctors in the Hong Kong public health system are under tremendous workload pressures, with an average consultation length of 6 min (44). The quality of care provided can be suboptimal as it is very challenging to deal with multifactorial disease problems and complicated lifestyle or compliance issues within such a short consultation time. Because of the skewed distribution of the GOPC attendance in the usual care group, some of the beneficial effect of the entire RAMP-DM program may be due to the fact that some patients in the usual care group had very few visits, but higher complication and mortality rates. A further analysis showed that patients in the usual care group who had fewer visits, particularly below the median of 13 GOPC visits, had disproportionally higher rates of cardiovascular events (HR 2.52, P < 0.001) or mortality (HR 3.71, P < 0.001) compared with the usual care patients with a higher median number of GOPC visits. However, after adjusting all baseline covariates including patient sociodemographic data, laboratory results, and clinical parameters, the RAMP-DM group was associated with lower risks of CVD and all-cause mortality than the usual care group, irrespective of the frequency of GOPC attendance (Supplementary Table 6).

One of the benefits of the RAMP-DM is that some of the care, particularly that involving individual counseling (diabetes management, drug adherence, weight management, and lifestyle modification), and coordination of care is shifted to the RAMP-DM nurse. Furthermore, the health care professionals involved in the RAMP-DM program are experienced in the management of diabetes. For patients who are stratified as medium to very high risk, additional consultations are provided by advanced practice nurses and/or family medicine specialists who are allowed longer consultation times. The feedback and reminder system incorporated the clinical management system provided automatic alerts for scheduled assessments, follow-ups, and abnormal results, which could have led to closer monitoring of the disease, prescription of drugs, and treatments. Given a substantially higher data completion rate for laboratory results and clinical parameters in the RAMP-DM group, a higher incidence of new complications uncovered at the first RAMP-DM assessment of reversible complications like diabetic retinopathy (26.7%) and nephropathy (17.9%) or comorbidities like hypertension (20.5%) and hypercholesterolemia (20.5%) during comprehensive nurse assessment were observed. As a result, significantly higher proportions of RAMP-DM participants were prescribed glucose-lowering drugs, antihypertensive drugs, ACE inhibitors/angiotensin receptor blockers, and lipid-lowering agents after the program. Another key aspect of the RAMP-DM is the risk stratification process, which might help to motivate patients to adopt a healthier lifestyle to prevent potential DM complications. Patients also had access to group classes and a patient support call center designed to help facilitate better self-care behaviors and empower the patient to take care of themselves better. The benefits of the structure of diabetes empowerment programs have been confirmed in meta-analyses (4547). Multiple contacts with different health professionals helped to consolidate knowledge, adherence, and self-management (4850).

We found that the RAMP-DM group had a higher risk of diabetic retinopathy but a lower risk of STDR compared with the usual care group. A possible explanation was that all RAMP-DM participants underwent a formal systematic retinopathy screening through the use of a fundi camera by an optometrist during risk assessment, so an early stage of diabetic retinopathy such as background and preproliferative diabetic retinopathy could be identified early in order to prevent further progression to STDR by timely interventions. Several studies (51,52) suggested that the early stage of diabetic retinopathy might be reversible by improved control of DM. On the other hand, patients in usual care may receive informal retinopathy assessment using indirect ophthalmoscopy by the consulting doctor, and such screening was totally dependent on their doctors’ discretion. The early stage of diabetic retinopathy may progress to advanced stages without producing any immediate symptoms in the patients. As a consequence, usual care patients had a higher risk of STDR than RAMP-DM participants.

Our results from the subgroup analysis showed that patients in different risk groups all benefited from the RAMP-DM with a reduction in the risks of CVD and all-cause mortality. We observed that patients with lower baseline CVD risks actually had a higher relative risk reduction from RAMP-DM than those in higher-risk groups. One possible explanation is that many patients with high CVD risk might already have irreversible complications or atherosclerosis, making it harder for any intervention to prevent the development of CVD. Previous observational studies (28,53) also concluded that interventions for patients with diabetes are more effective when commenced at an earlier stage of the disease progression. Our findings highlight the importance of early optimal DM control and risk factor management in order to delay disease progression and prevent the development of complications. A further study should be required to confirm the health benefit of early optimal DM control.

It was encouraging to find lower rates of use of secondary or tertiary care services (SOPC, A&E, and hospitalizations) in the RAMP-DM group compared with usual care group. Despite no direct comparison with the existing literature, previous studies (22,54,55) showed the multidisciplinary management, computer-supported care management, and improvement of surrogate outcomes such as HbA1c decreased the health service use. One reasons for the reduction of service uses was that patients receiving RAMP-DM care had fewer complications and therefore needed less secondary/tertiary care. The reduced risk of CVD enabled patients receiving RAMP-DM care to continue to be followed up in primary care. Supporting this, higher rates of primary care service use (GOPC) was observed in the RAMP-DM groups, which could be attributed to a higher acceptance of the program.

There were several strengths to this study. First, a comprehensive evaluation on the effectiveness of RAMP-DM included not only surrogate markers but also actual observed events and service uses. Second, the propensity score matching and subgroup analyses increased the reliability of the findings because of the large population-based sample. Third, all data were extracted from a computerized administrative database, which helps to assure data accuracy.

There were also several limitations to this study. First, this was a prospective cohort study and not an RCT. Hence, unobserved potential confounders might influence the conclusion. Nonetheless, it may be infeasible to carry out such a high-evidence trial in primary care settings. High attrition rates, a low number of incident events, short follow-up times, and strict subject inclusion criteria are some known drawbacks of the RCT that reduce the applicability to patients with diabetes in clinical practices (56,57). To overcome some of the limitations of cohort studies, the comparison group was selected by propensity score matching with no significant differences in patient characteristics between groups. Second, ICPC-2 and ICD-9-CM diagnosis coding were used to identify complications. There was no validation study performed to assess the accuracy and completeness of the coding, and thus the data may be susceptible to misclassification bias. Third, data on the patient’s financial burden, compliance, drug adherence, mental health, and lifestyle behaviors were not collected and may have contributed to CVD and mortality risk. A further study including these factors may help to confirm the current findings. Last, because of the limited number of usual care patients, not all RAMP-DM participants could be matched with usual care patients, which may cause potential biases on the results.

In conclusion, this large territory-wide naturalistic study in a real-world primary care setting showed that RAMP-DM, irrespective of any patient characteristics, led to significantly greater reduction in any CVD or microvascular complications and secondary or tertiary service use for patients with diabetes. Among all subgroups, patients with lower CVD risk may have received the most benefit from the RAMP-DM. Our findings highlight the importance of early optimal DM control and risk factor management by risk assessment and stratification and multidisciplinary management in order to delay disease progression and prevent the development of complications. Further studies to evaluate the cost-effectiveness of RAMP-DM from health service provider and societal perspectives should be conducted to confirm whether the RAMP-DM is cost-effective.

Acknowledgments. The authors thank the multidisciplinary risk-stratification–based DM management program team at the HA head office, the chief of service in primary care, the program coordinator in each cluster, and the Statistics and Workforce Planning Department at the HA for their contributions to this work.

Funding. This study was funded by the HA (Reference #8011014157) and the Health and Health Services Research Fund, Hong Kong Special Administrative Region Food and Health Bureau, and the Health and Medical Research Fund Commissioned Study on Enhanced Primary Care (Reference #EPC-HKU-2).

The funders had no role in study design, data collection and analysis, decision to publish, or the preparation of the manuscript.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. E.Y.F.W., C.S.C.F., and C.L.K.L. contributed to the study design and acquisition of data, researched the data, contributed to the statistical analysis and interpretation of the results, and wrote the manuscript. F.F.J., E.Y.T.Y., D.Y.T.F., C.K.H.W., K.H.Y.C., and R.L.P.K. contributed to the interpretation of the results. All authors reviewed and edited the manuscript. W.Y.C. and A.K.C.C. contributed to the statistical analysis and the interpretation of the results and wrote the manuscript. E.Y.F.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
International Diabetes Federation
.
IDF Diabetes Atlas
. 7th ed.
Brussels, Belgium
,
International Diabetes Federation
,
2015
2.
American Diabetes Association
.
Children and adolescents
. Sec. 11. In Standards of Medical Care in Diabetes—2015.
Diabetes Care
2015
;
38
(
Suppl. 1
):
S70
S76
[PubMed]
3.
National Institute for Health and Care Excellence. Type 2 diabetes in adults: management [Internet], 2015. Available from https://www.nice.org.uk/guidance/ng28. Accessed 15 February 2017
4.
Bodenheimer
T
,
Wagner
EH
,
Grumbach
K
.
Improving primary care for patients with chronic illness: the chronic care model, part 2
.
JAMA
2002
;
288
:
1909
1914
[PubMed]
5.
Bodenheimer
T
,
Wagner
EH
,
Grumbach
K
.
Improving primary care for patients with chronic illness
.
JAMA
2002
;
288
:
1775
1779
[PubMed]
6.
Wagner
EH
,
Grothaus
LC
,
Sandhu
N
, et al
.
Chronic care clinics for diabetes in primary care: a system-wide randomized trial
.
Diabetes Care
2001
;
24
:
695
700
[PubMed]
7.
Wagner
EH
,
Austin
BT
,
Davis
C
,
Hindmarsh
M
,
Schaefer
J
,
Bonomi
A
.
Improving chronic illness care: translating evidence into action
.
Health Aff (Millwood)
2001
;
20
:
64
78
[PubMed]
8.
Larme
AC
,
Pugh
JA
.
Attitudes of primary care providers toward diabetes: barriers to guideline implementation
.
Diabetes Care
1998
;
21
:
1391
1396
[PubMed]
9.
Gaede
P
,
Lund-Andersen
H
,
Parving
H-H
,
Pedersen
O
.
Effect of a multifactorial intervention on mortality in type 2 diabetes
.
N Engl J Med
2008
;
358
:
580
591
[PubMed]
10.
Holman
RR
,
Paul
SK
,
Bethel
MA
,
Matthews
DR
,
Neil
HAW
.
10-year follow-up of intensive glucose control in type 2 diabetes
.
N Engl J Med
2008
;
359
:
1577
1589
[PubMed]
11.
Gerstein
HC
,
Miller
ME
,
Byington
RP
, et al.;
Action to Control Cardiovascular Risk in Diabetes Study Group
.
Effects of intensive glucose lowering in type 2 diabetes
.
N Engl J Med
2008
;
358
:
2545
2559
12.
Patel
A
,
MacMahon
S
,
Chalmers
J
, et al.;
ADVANCE Collaborative Group
.
Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes
.
N Engl J Med
2008
;
358
:
2560
2572
13.
Duckworth
W
,
Abraira
C
,
Moritz
T
, et al.;
VADT Investigators
.
Glucose control and vascular complications in veterans with type 2 diabetes
.
N Engl J Med
2009
;
360
:
129
139
[PubMed]
14.
Vargas
RB
,
Mangione
CM
,
Asch
S
, et al
.
Can a chronic care model collaborative reduce heart disease risk in patients with diabetes?
J Gen Intern Med
2007
;
22
:
215
222
[PubMed]
15.
Clark
CM
 Jr
,
Snyder
JW
,
Meek
RL
,
Stutz
LM
,
Parkin
CG
.
A systematic approach to risk stratification and intervention within a managed care environment improves diabetes outcomes and patient satisfaction
.
Diabetes Care
2001
;
24
:
1079
1086
[PubMed]
16.
Yu
GC
,
Beresford
R
.
Implementation of a chronic illness model for diabetes care in a family medicine residency program
.
J Gen Intern Med
2010
;
25
(
Suppl. 4
):
S615
S619
[PubMed]
17.
Lee
JD
,
Saravanan
P
,
Patel
V
.
Alphabet strategy for diabetes care: a multi-professional, evidence-based, outcome-directed approach to management
.
World J Diabetes
2015
;
6
:
874
879
[PubMed]
18.
Hull
S
,
Chowdhury
TA
,
Mathur
R
,
Robson
J
.
Improving outcomes for patients with type 2 diabetes using general practice networks: a quality improvement project in east London
.
BMJ Qual Saf
2014
;
23
:
171
176
19.
Weber
V
,
Bloom
F
,
Pierdon
S
,
Wood
C
.
Employing the electronic health record to improve diabetes care: a multifaceted intervention in an integrated delivery system
.
J Gen Intern Med
2008
;
23
:
379
382
[PubMed]
20.
Stock
S
,
Drabik
A
,
Büscher
G
, et al
.
German diabetes management programs improve quality of care and curb costs
.
Health Aff (Millwood)
2010
;
29
:
2197
2205
[PubMed]
21.
Olivarius
NF
,
Beck-Nielsen
H
,
Andreasen
AH
,
Hørder
M
,
Pedersen
PA
.
Randomised controlled trial of structured personal care of type 2 diabetes mellitus
.
BMJ
2001
;
323
:
970
975
[PubMed]
22.
Sadur
CN
,
Moline
N
,
Costa
M
, et al
.
Diabetes management in a health maintenance organization. Efficacy of care management using cluster visits
.
Diabetes Care
1999
;
22
:
2011
2017
[PubMed]
23.
Gelding
SV
.
Improving cardiovascular risk in type 2 diabetes: time to get personal. Eur Heart J Qual Care Clin Outcomes
.
2016
;
2
:
233
234
24.
Fung
CS
,
Chin
WY
,
Dai
DS
, et al
.
Evaluation of the quality of care of a multi-disciplinary Risk factor Assessment and Management Programme (RAMP) for diabetic patients
.
BMC Fam Pract
2012
;
13
:
116
[PubMed]
25.
Jiao
F
,
Fung
CS
,
Wan
YF
, et al
.
Long-term effects of the multidisciplinary Risk Assessment and Management Program for Patients with Diabetes Mellitus (RAMP-DM): a population-based cohort study
.
Cardiovasc Diabetol
2015
;
14
:
105
[PubMed]
26.
Jiao
FF
,
Fung
CS
,
Wong
CK
, et al
.
Effects of the Multidisciplinary Risk Assessment and Management Program for Patients with Diabetes Mellitus (RAMP-DM) on biomedical outcomes, observed cardiovascular events and cardiovascular risks in primary care: a longitudinal comparative study
.
Cardiovasc Diabetol
2014
;
13
:
127
[PubMed]
27.
Lau
IT
.
A clinical practice guideline to guide a system approach to diabetes care in Hong Kong
.
Diabetes Metab J
2017
;
41
:
81
88
[PubMed]
28.
Chan
J
,
So
W
,
Ko
G
, et al
.
The Joint Asia Diabetes Evaluation (JADE) Program: a web-based program to translate evidence to clinical practice in type 2 diabetes
.
Diabet Med
2009
;
26
:
693
699
[PubMed]
29.
Department of Health, The Government of the Hong Kong Special Administrative Region. Hong Kong reference framework for diabetes care for adults in primary care settings [Internet], 2017. Available from http://www.pco.gov.hk/english/resource/professionals_diabetes_pdf.html. Accessed 15 February 2017
30.
Rosenbaum
PR
.
Observational Studies
.
New York
,
Springer
,
2002
31.
Rosenbaum
PR
,
Rubin
DB
.
The central role of the propensity score in observational studies for causal effects
.
Biometrika
1983
;
70
:
41
55
32.
Rubin
DB
.
Estimating causal effects of treatments in randomized and nonrandomized studies
.
J Educ Psychol
1974
;
66
:
688
701
33.
D’Agostino
RBJ
 Jr
.
Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group
.
Stat Med
1998
;
17
:
2265
2281
[PubMed]
34.
Leuven E, Sianesi B. PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing [Internet], 2012. Available from https://econpapers.repec.org/software/bocbocode/s432001.htm. Accessed 15 February 2017
35.
Royston
P
.
Multiple imputation of missing values
.
Stata J
2004
;
4
:
227
241
36.
Clark
TG
,
Altman
DG
.
Developing a prognostic model in the presence of missing data: an ovarian cancer case study
.
J Clin Epidemiol
2003
;
56
:
28
37
[PubMed]
37.
Schafer
JL
,
Graham
JW
.
Missing data: our view of the state of the art
.
Psychol Methods
2002
;
7
:
147
177
[PubMed]
38.
Steyerberg
EW
,
van Veen
M
.
Imputation is beneficial for handling missing data in predictive models
.
J Clin Epidemiol
2007
;
60
:
979
[PubMed]
39.
Moons
KG
,
Donders
RA
,
Stijnen
T
,
Harrell
FE
 Jr
.
Using the outcome for imputation of missing predictor values was preferred
.
J Clin Epidemiol
2006
;
59
:
1092
1101
[PubMed]
40.
Rubin
DB
.
Multiple Imputation for Nonresponse in Surveys
.
New York
,
John Wiley & Sons
,
2004
41.
Ulm
K
.
A simple method to calculate the confidence interval of a standardized mortality ratio (SMR)
.
Am J Epidemiol
1990
;
131
:
373
375
[PubMed]
42.
Griffin
SJ
,
Borch-Johnsen
K
,
Davies
MJ
, et al
.
Effect of early intensive multifactorial therapy on 5-year cardiovascular outcomes in individuals with type 2 diabetes detected by screening (ADDITION-Europe): a cluster-randomised trial
.
Lancet
2011
;
378
:
156
167
[PubMed]
43.
Tricco
AC
,
Ivers
NM
,
Grimshaw
JM
, et al
.
Effectiveness of quality improvement strategies on the management of diabetes: a systematic review and meta-analysis
.
Lancet
2012
;
379
:
2252
2261
[PubMed]
44.
Yu
FS
,
Yip
BH
,
Kung
K
, et al
.
The association of types of training and practice settings with doctors’ empathy and patient enablement among patients with chronic illness in Hong Kong
.
PLoS One
2015
;
10
:
e0144492
[PubMed]
45.
Steinsbekk
A
,
Rygg
,
Lisulo
M
,
Rise
MB
,
Fretheim
A
.
Group based diabetes self-management education compared to routine treatment for people with type 2 diabetes mellitus. A systematic review with meta-analysis
.
BMC Health Serv Res
2012
;
12
:
213
[PubMed]
46.
Norris
SL
,
Lau
J
,
Smith
SJ
,
Schmid
CH
,
Engelgau
MM
.
Self-management education for adults with type 2 diabetes: a meta-analysis of the effect on glycemic control
.
Diabetes Care
2002
;
25
:
1159
1171
[PubMed]
47.
Minet
L
,
Møller
S
,
Vach
W
,
Wagner
L
,
Henriksen
JE
.
Mediating the effect of self-care management intervention in type 2 diabetes: a meta-analysis of 47 randomised controlled trials
.
Patient Educ Couns
2010
;
80
:
29
41
[PubMed]
48.
Glasgow
RE
,
Orleans
CT
,
Wagner
EH
.
Does the chronic care model serve also as a template for improving prevention
?
Milbank Q
2001
;
79
:
579
612, iv–v
[PubMed]
49.
Hulscher
M
,
Wensing
M
,
Weijden
T
,
Grol
R
.
Interventions to implement prevention in primary care
.
Cochrane Databse Syst Rev
1997
;
1
:
CD000362
50.
An
LC
,
Foldes
SS
,
Alesci
NL
, et al
.
The impact of smoking-cessation intervention by multiple health professionals
.
Am J Prev Med
2008
;
34
:
54
60
[PubMed]
51.
Stitt
AW
,
Curtis
TM
,
Chen
M
, et al
.
The progress in understanding and treatment of diabetic retinopathy
.
Prog Retin Eye Res
2016
;
51
:
156
186
[PubMed]
52.
Wong
TY
,
Mwamburi
M
,
Klein
R
, et al
.
Rates of progression in diabetic retinopathy during different time periods: a systematic review and meta-analysis
.
Diabetes Care
2009
;
32
:
2307
2313
[PubMed]
53.
Alberti
KGM
,
Zimmet
P
,
Shaw
J
.
International Diabetes Federation: a consensus on Type 2 diabetes prevention
.
Diabet Med
2007
;
24
:
451
463
[PubMed]
54.
Domurat
ES
.
Diabetes managed care and clinical outcomes: the Harbor City, California Kaiser Permanente diabetes care system
.
Am J Manag Care
1999
;
5
:
1299
1307
[PubMed]
55.
Wagner
EH
,
Sandhu
N
,
Newton
KM
,
McCulloch
DK
,
Ramsey
SD
,
Grothaus
LC
.
Effect of improved glycemic control on health care costs and utilization
.
JAMA
2001
;
285
:
182
189
[PubMed]
56.
Benson
K
,
Hartz
AJ
.
A comparison of observational studies and randomized, controlled trials
.
N Engl J Med
2000
;
342
:
1878
1886
[PubMed]
57.
Concato
J
,
Shah
N
,
Horwitz
RI
.
Randomized, controlled trials, observational studies, and the hierarchy of research designs
.
N Engl J Med
2000
;
342
:
1887
1892
[PubMed]
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 http://www.diabetesjournals.org/content/license.

Supplementary data