Ten-Year Effectiveness of the Multidisciplinary Risk Assessment and Management Programme–Diabetes Mellitus (RAMP-DM) on Macrovascular and Microvascular Complications and All-Cause Mortality: A Population-Based Cohort Study

Diabetes is a serious public health problem associated with significant increases in the development of comorbidities and risks of premature mortality. According to the International Diabetes Federation estimates, the number of people with diabetes worldwide will increase from 536.6 million in 2021 to 783.7 million by 2045 (1).

The chronic nature of diabetes necessitates lifelong management processes that require patients’ continuous engagement with the health care system and are best achieved at the primary care level (2). Interventions designed based on the chronic care model aim to deliver multidimensional and evidence-based health care interventions in primary care (3–5) and have been recommended for the management of diabetes by the American Diabetes Association (6), the European Society of Cardiology, and the European Association for the Study of Diabetes (7). These management programs integrate clinical care and diabetes self-management support through team-based care coordination, including primary care providers, other health care practitioners, and administrative staff. Multiple randomized clinical trials have demonstrated the effectiveness of disease management programs that used the chronic care model with team-based care on reducing the incidence or delaying the development of diabetes-related complications (8–15) or mortality (9,10,12,14–16). However, studies investigating the long-term effectiveness of a multidisciplinary program for diabetes care implemented in the real-world primary care setting remain scarce.

Hong Kong has a heavily subsidized health care system where public medical care is provided at little to no cost. The majority (90%) of patients with diabetes are treated in public hospitals and clinics operated by the Hospital Authority, which is the statutory body responsible for all public hospitals and clinics in Hong Kong (17). To enhance the quality of care for patients with diabetes, the Hospital Authority established the Risk Assessment and Management Programme–Diabetes Mellitus (RAMP-DM) in 2009. The RAMP-DM is a territory-wide chronic disease management program incorporated into the public primary care system for all patients with type 2 diabetes and is offered in addition to usual care. RAMP-DM participants receive a comprehensive assessment of existing diabetes complications, disease control indicators, use of medications, and lifestyle habits to guide risk stratification and individualized management. We have previously shown that RAMP-DM is associated with improvements in intermediate clinical outcomes, including hemoglobin A_1c (HbA_1c), LDL cholesterol (LDL-C), and blood pressure (BP) in patients with diabetes after 36 months (18,19). Significant reductions in the risks of macrovascular (57%) and microvascular (12%) complications and mortality (66%) were also observed after 5 years (20) compared with patients receiving usual care only. However, it is uncertain whether the benefits of RAMP-DM could be sustained beyond 5 years, given that a longer disease duration is associated with increased risks of coronary heart disease (21) and mortality (22), potentially diluting the beneficial effects of RAMP-DM. This study aimed to examine the long-term effects of RAMP-DM on the risks of macrovascular and microvascular complications and all-cause mortality in patients with type 2 diabetes over a 10-year observation period.

The RAMP-DM has been implemented in all Hospital Authority public general outpatient clinics (GOPCs) with the intention of including all patients with type 2 diabetes who have been followed regularly in the clinics since 2009. All patients with type 2 diabetes were invited by primary care doctors and nurses to enroll in RAMP-DM when they attended GOPCs. The design and components of the intervention have been described in detail previously (18–20,23,24). In brief, a comprehensive assessment was performed by a nurse case manager when patients first enrolled in RAMP-DM. Patients were then categorized into different risk groups according to existing diabetes-related complications (cardiovascular diseases [CVD] and end-stage renal disease [ESRD]) and risk factors based on the definition of the Joint Asia Diabetes Evaluation Program (24,25). Risk factors screened included 1) smoking status, 2) obesity (BMI ≥27.5 kg/m² or waist circumference ≥90 cm [male] or 80 cm [female]), 3) dyslipidemia (LDL-C ≥2.6 mmol/L, triglycerides ≥1.7 mmol/L, HDL cholesterol [HDL-C] <1 mmol/L [male] or 1.3 mmol/L [female], or on lipid-lowering drug treatment), 4) hypertension (BP ≥130/80 mmHg or on antihypertensive drugs), 5) presence of diabetes retinopathy, 6) albuminuria (albumin-to-creatinine ratio >2.5 mg/mmol [male] or 3.5 mg/mmol [female]), 7) foot problems (deformities, skin abnormalities, or neuropathy), 8) HbA_1c ≥8%, and 9) estimated glomerular filtration rate (eGFR) (≥90, 60–89, or <60 mL/min/1.73 m²). After risk stratification, the nurse case manager would explain the risk level and provide appropriate management advice to patients. Additional consultations with RAMP-DM nurses and doctors were scheduled for patients categorized as high risk and with poor HbA_1c control. Patients may also be referred by the nurse to allied health professionals or other health services provided by the Hospital Authority, including the Patient Empowerment Programme, Patient Support Call Centre, and Smoking Counseling and Cessation Centre, based on needs. RAMP-DM participants were booked to repeat the RAMP-DM assessment once every 1–3 years on the basis of their estimated risk level categorized at the first assessment session of RAMP-DM and diabetes (HbA_1c) control. In the meantime, they were followed by their doctors as usual in the same manner as non–RAMP-DM participants.

The usual care of patients with type 2 diabetes in public GOPCs includes doctor-led follow-up consultations once every 2–4 months guided by the Hong Kong Reference Framework for Diabetes Care for Adults in Primary Care Settings (26). Primary care doctors would review disease control parameters and adjust antidiabetic medications if needed. Doctors would also perform physical examinations and laboratory investigations and make referrals to specialists or allied health professionals, as indicated.

Data were sourced from the Clinical Management System of the Hospital Authority, which is a centralized database that stores all patient information in the public health care system. The data include sociodemographics, disease diagnoses, laboratory results, medical procedures performed, medication prescription records, and public health care services’ attendance records.

Electronic medical records of adult patients with doctor-diagnosed type 2 diabetes, defined by the International Classification of Primary Care, 2nd Edition (ICPC-2), code T90, and managed in GOPCs or family medicine specialist clinics under the Hospital Authority between 1 August 2009 and 30 September 2010, were extracted from the Clinical Management System. Patients were categorized into two groups: 1) RAMP-DM participants and 2) patients who received usual care only. RAMP-DM participants were defined as patients who attended at least one RAMP-DM intake assessment session between 1 August 2009 and 30 September 2010 and continued to be followed with usual care, while usual care–only patients were those who had not received any RAMP-DM services by 31 December 2019 (the end of the observation period). The index date for follow-up was defined as the date of first RAMP-DM intake assessment attendance for RAMP-DM participants or the date of the first GOPC attendance between 1 August 2009 and 30 September 2010 for patients who received usual care only. Patients with type 1 diabetes, gestational diabetes mellitus, or a history of diabetes-related complications, including CVD (a composite of coronary heart disease, heart failure, and stroke), peripheral vascular disease (PVD), nonproliferative diabetic retinopathy, sight-threatening diabetic retinopathy (STDR), diabetic neuropathy, or ESRD on or before the index date, with a recorded eGFR <45 mL/min/1.73 m² on or before the index date or without any follow-up after the index date (i.e., no records of attendance at GOPCs, specialist outpatient clinics, accident and emergency departments, hospitals, and allied health services under the Hospital Authority), were excluded from the study. Patients were observed from their index date until one of the following: 1) date of any occurrence of diabetes-related complications, 2) date of mortality, or 3) the last follow-up date under the Hospital Authority.

The primary outcome was the incidence of diabetes-related complications and all-cause mortality. Diabetes-related complications were defined by ICPC-2 or ICD-9-CM diagnosis codes (Supplementary Table 1). The date and primary cause of death of patients were extracted from the Hospital Authority database, which is linked to the Hong Kong Death Registry. Secondary outcomes included the pattern of changes in disease control parameters, including blood glucose levels, BP, lipid profile, and eGFR.

Baseline covariates included sociodemographics, laboratory results, and clinical characteristics. Sociodemographics comprised sex, age, current smoking status, and whether patients had received public assistance. Laboratory results included fasting glucose (FG), HbA_1c, systolic BP (SBP) and diastolic BP (DBP), LDL-C, total cholesterol (TC)/HDL-C ratio, triglycerides, BMI, eGFR, and urinary albumin-to-creatinine ratio (ACR). Clinical characteristics included diabetes duration, hypertension diagnosis, the Charlson Comorbidity Index, and use of insulin, oral antidiabetic drugs, antihypertensive drugs, and lipid-lowering agents.

Descriptive statistics for baseline characteristics of patients are presented as mean ± SD for continuous variables or frequency and proportion for categorical variables. Multiple imputation by chained equation was adopted to account for missing baseline characteristics (27,28). Each missing datum was imputed five times on the basis of other known baseline variables. Five complete data sets were then generated and analyzed separately. The overall estimates were combined based on Rubin’s rule (29).

To reduce selection bias due to differences in characteristics of patients between the two groups, a one-to-one propensity score matching was applied. Logistic regression was fitted by the binary variable, which indicates the comparison group against all baseline covariates. The predicted probability of each patient (i.e., the propensity score) was then generated from the regression results. Patients with similar propensity scores (characteristics) were matched between the two groups. A caliper of 0.001 was applied to ensure that patients with similar propensity scores were matched. All steps of the propensity score matching were done using the Stata package psmatch2 (30). After matching, the balance of baseline characteristics between comparison groups was assessed by absolute standardized mean difference (ASMD). An ASMD >0.1 implies an imbalance of a covariate between groups (31).

Cumulative incidence and incidence rates (cases per 100 person-years) of outcome events were calculated. The absolute risk reduction (ARR) was calculated as the difference in the cumulative incidence between the RAMP-DM and usual care–only groups, and the number needed to treat (NNT) to prevent one additional event was obtained as the reciprocal of the ARR. Annualized incidence rates of outcome events for both groups were compared. Cox proportional hazards regression models were used to assess the risk of outcome events between the groups. The proportional hazards assumption was checked by the plots of the scaled Schoenfeld residuals against time. The trend of events over time was shown using Kaplan-Meier survival curves. A log-rank test was used to assess the equality of survival curves between groups. Changes in clinical parameters within and between treatment groups were compared by linear mixed-effects and generalized linear models, respectively.

The cohort was further stratified into subgroups, and the analyses were repeated for any differences in treatment effect by patient characteristics. The subgroups include sex (male, female), age (<65, ≥65 years), smoking status (current smoker, nonsmoker), received public assistance (yes, no), HbA_1c (<7, ≥7%), FG (<6.5, ≥6.5 mmol/L), BP (<130/80, ≥130/80 mmHg), LDL-C (<2.6, ≥2.6 mmol/L), TC/HDL-C ratio (<4.5, ≥4.5), triglycerides (<1.7, ≥1.7 mmol/L), BMI (<23, 23–24.9, ≥25 kg/m²), eGFR (<60, ≥60 mL/min/1.73m²), diabetes duration (<2, ≥2 years), hypertension (yes, no), and estimated CVD risk (low, medium, high) at baseline.

Four scenarios were considered in the sensitivity analyses, including 1) an intention-to-treat approach that included patients who were invited to join RAMP-DM but did not enroll in the program in the RAMP-DM group (contributing 4.4% of all invited participants), 2) multiple imputation of baseline covariates without propensity score matching, 3) complete case analysis with propensity score matching, and 4) complete case analysis without propensity score matching. Two additional analyses were conducted to explore the possible impact of unmeasured confounders on the outcomes in this cohort study. First, the risk of sleep disturbance was compared between groups as a negative control outcome analysis to assess whether the intervention had any significant effect on an irrelevant outcome due to the presence of uncontrolled confounders (32). Second, E-value represents the minimum strength of association, on the risk ratio scale, that an unmeasured confounder would need to have with both the treatment and outcome to fully explain away a specific treatment–outcome association, conditional on the measured covariates (33).

All statistical analyses were performed using Stata 14.0 software (StataCorp LP, College Station, TX). All tests were two-tailed; P < 0.05 was considered statistically significant.

Ethics approval was granted by the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (reference no. UW 19-329). Informed consent from patients was not required for the use of deidentified data in this prospective cohort analysis.

The data used in this study will not be publicly available. Restrictions are placed on data access and the license to use the data.

A total of 23,496 RAMP-DM participants and 37,342 usual care patients were identified in the cohort after excluding patients with a history of complications and without follow-up. The data completion rate for baseline characteristics of eligible patients is presented in Supplementary Table 2. After propensity score matching, a total of 36,746 patients (18,373 patients in each group) were included in this study. The detailed flowchart for patient inclusion is presented in Supplementary Fig. 1. Baseline characteristics of patients are shown in Table 1. Overall, 52.7% of patients were female, and the mean (SD) age was 64.7 (12.2) years. The mean (SD) duration of diabetes was 7.4 (6.9) years, and HbA_1c at baseline was 7.3% (1.4) (56.1 [15.0] mmol/mol). All ASMDs were <0.1, indicating that a balance of baseline characteristics between the two groups was achieved after propensity score matching.

Over a median follow-up of 9.5 years (or 306,802 person-years), the cumulative incidence and incidence rate of all complications and all-cause mortality were reduced in the RAMP-DM group compared with patients in the usual care group (Table 2). The ARRs of CVD, PVD, STDR, neuropathy, ESRD, and all-cause mortality were 11.6%, 0.9%, 1.0%, 0.7%, 3.2%, and 16.2%, respectively. The ARRs of macrovascular, microvascular, and all complications were 11.9%, 3.3%, and 12.1%, respectively. The corresponding NNT ranged from 6 for all-cause mortality to 148 for neuropathy. The NNT of any macrovascular, microvascular, and all complications was 8, 30, and 8, respectively. The risk of diabetes-related complications, including CVD, PVD, STDR, neuropathy, ESRD, and all-cause mortality, was 42–55% lower (HRs ranging from 0.45 for all-cause mortality to 0.58 for stroke) in the RAMP-DM group, while the risk of nonproliferative diabetic retinopathy (HR 0.92 [95% CI 0.84–1.01]) was not statistically different between groups (Table 3). The HRs of macrovascular, microvascular, and all complications were 0.52 (95% CI 0.50–0.54), 0.68 (95% CI 0.64–0.72), and 0.57 (95% CI 0.55–0.59), respectively. The Kaplan-Meier survival curves of all outcomes are shown in Fig. 1 and Supplementary Fig. 2. The distribution of the survival curves for all outcomes except nonproliferative diabetic retinopathy was significantly different between groups (all P < 0.05 by log-rank test). The reduction in risk of mortality in the RAMP-DM group persisted over the 10-year period, whereas that for macrovascular and microvascular complications attenuated at the 9th and 8th year of follow-up (Supplementary Table 3), respectively.

The subgroup analyses showed largely consistent results compared with the primary analysis (Supplementary Fig. 3 and Supplementary Table 4). The relative beneficial effect of RAMP-DM on diabetes-related complications and all-cause mortality was smaller among patients with an eGFR of <60 mL/min/1.73 m². Conversely, patients with a diabetes duration of <2 years, aged <65 years, and with low or medium baseline CVD risk saw greater relative benefits from RAMP-DM. Among the RAMP-DM participants, patients in the low CVD risk group had a lower risk of macrovascular complications but similar risk of microvascular complications and all-cause mortality compared with patients categorized as having medium CVD risk (Supplementary Fig. 4). Patients classified as having a high CVD risk had the highest incidence of complications and all-cause mortality.

The results of the sensitivity analyses are shown in Supplementary Table 5. Regardless of the analytic approach used, the results were consistent with the primary analysis, where RAMP-DM was associated with a lower risk of all diabetes-related complications and all-cause mortality among patients with diabetes. For the analysis of the negative control outcome, the risk of sleep disturbance was comparable between RAMP-DM participants and usual care patients (HR 1.08, 95% CI 0.96–1.22, P = 0.109), indicating that there was no significant effect of RAMP-DM on an irrelevant outcome. This suggested that the bias due to unmeasured confounders on the study results was small. In addition, the E-values of study outcomes based on the HRs ranged from 2.32 to 3.56 (Supplementary Table 6), implying that an unobserved confounder has to have at least a 2.32–3.56-fold stronger association with the outcomes to explain away the effect of RAMP-DM on the study outcomes, which is unlikely.

The changes in means of disease control parameters over the 10-year follow-up are shown in Supplementary Fig. 5 and Table 4. FG values showed a gradual decrease during the first 3 years (from 7.54 to 7.28 mmol/L in the RAMP-DM group and 7.56 to 7.44 mmol/L in the usual care group) but increased in the subsequent years for both groups (from 7.28 to 7.66 mmol/L in the RAMP-DM group and 7.44 to 7.60 mmol/L in the usual care group). HbA_1c increased gradually after the 1st year for usual care patients and from the 4th year for RAMP-DM participants. However, HbA_1c remained significantly lower in the RAMP-DM group compared with the usual care group at the end of the observation period (7.40% vs. 7.46%, P = 0.015). RAMP-DM participants had better BP control than usual care patients throughout the entire follow-up period (P < 0.001 across all follow-up years for both SBP and DBP). SBP in both groups decreased initially in the first 5 years (from 136.6 to 129.8 mmHg in the RAMP-DM group and 136.7 to 132.7 mmHg in the usual care–only group) yet increased afterward (from 129.8 to 133.3 mmHg in the RAMP-DM group and 132.7 to 136.8 mmHg in the usual care–only group). Nevertheless, RAMP-DM participants had significantly lower SBP compared with usual care patients after 10 years (133.3 vs. 136.8 mmHg, P < 0.001). DBP decreased gradually for both groups, with a greater reduction over 10 years in the RAMP-DM group (from 75.2 to 69.9 mmHg in usual care plus RAMP-DM group and 75.3 to 71.8 mmHg in the usual care–only group). Both groups showed similar decreasing trends in LDL-C and TC/HDL-C ratio. In addition, RAMP-DM participants showed significantly lower values for triglycerides at the end of the observation period (1.45 vs. 1.52 mmol/L, P < 0.001). There was a significant reduction for average BMI from 25.8 to 25.0 kg/m² in RAMP-DM participants but not in usual care patients. The decline in renal function evaluated by eGFR level was slower in the RAMP-DM group, but both groups had similar eGFR levels by the 10th year (88.1 vs. 87.7 mL/min/1.73 m², P = 0.587).

For the paired change of metabolic factors from baseline to the 10th year between treatment groups, the RAMP-DM group had significant improvements in SBP (difference in differences [DID] −5.1 mmHg, P < 0.001), DBP (DID −2.4 mmHg, P < 0.001), TC/HDL-C ratio (DID −0.08, P = 0.005), triglycerides (DID −0.05 mmol/L, P = 0.039), BMI (DID −0.3 kg/m², P < 0.001), and eGFR (DID 3.3 mL/min/1.73 m², P < 0.001) over the usual care group but not for FG (DID 0.06 mmol/L, P = 0.376), HbA_1c (DID −0.006%, P = 0.829), and LDL-C (DID −0.002 mmol/L, P = 0.920). In addition, the RAMP-DM group had significantly higher odds of achieving disease control targets for metabolic risk factors, including HbA_1c <7%, BP <130/80 mmHg, LDL-C <2.6 mmol/L, and eGFR ≥60 mL/min/1.73 m² throughout the follow-up period compared with the usual care group (Supplementary Table 7 and Supplementary Fig. 6).

To our knowledge, this study is the first to evaluate the real-world effectiveness of a multidisciplinary program for diabetes care implemented in the primary care setting over 10 years. The results suggest that RAMP-DM was associated with significant reductions in diabetes-related macrovascular and microvascular complications and all-cause mortality over 10 years compared with usual care. Additionally, RAMP-DM participants showed greater improvements in disease control, including blood glucose, BP, and lipid profile.

RAMP-DM participants had 46–55% relative reduction rates in the risks of diabetes-related complications, including CVD, PVD, STDR, diabetic neuropathy, ESRD, and all-cause mortality. These observations are largely consistent with our previous analysis examining the effectiveness of RAMP-DM after 5 years (20); however, the magnitude of relative risk reduction was slightly larger in the 5-year analysis. Consistently, the differences in the annual incidence for complications between RAMP-DM participants and usual care patients were greatest in the first 5 years but gradually diminished after the 8th year of follow-up. These observations suggest that while RAMP-DM is effective for preventing the development of complications and mortality, benefits of the program may slowly attenuate over longer follow-up periods. Additional treatments and care may therefore be required for RAMP-DM participants after the 8th year of enrollment for the effectiveness of the program to sustain. Alternatively, the reduction in all-cause mortality in RAMP-DM participants persisted for 10 years (P < 0.05 for all annual incidence rates), possibly suggesting that the initial reduction in complication rates had a sustained benefit on decreasing mortality rates over the long term. It is also notable that RAMP-DM participants had 26% greater relative risks for developing nonproliferative diabetic retinopathy after 5 years (20) but not by the 10th year (HR 0.918, P = 0.069). This may be attributed to the availability of retinopathy assessments provided as part of the RAMP-DM, which screened out patients with a higher risk for nonproliferative diabetic retinopathy at an earlier time. Therefore, a relatively greater number of patients were identified during the program’s initial phase, but the risks for developing nonproliferative diabetic retinopathy was eventually lowered by routine follow-up and management.

Prior large-scale clinical trials examining the effectiveness of multifactorial management strategies for preventing adverse clinical outcomes among patients with diabetes reported mixed results. The Danish Steno-2 trial showed that patients who underwent behavioral modification and team-based treatment had a 59% (HR 0.41, 95% CI 0.25–0.67), 43% (relative risk 0.57, 95% CI 0.37–0.88), 47% (relative risk 0.53, 95% CI 0.34–0.81), and 46% (HR 0.54, 95% CI 0.32–0.89) relative risk reduction in CVD, retinopathy, autonomic neuropathy, and all-cause mortality, respectively, compared with patients who received conventional therapy over a mean follow-up of 13.3 years (16). The Japan Diabetes Optimal Integrated Treatment Study for 3 Major Risk Factors of Cardiovascular Diseases (J-DOIT3) trial, which compared an intensive therapy group with additional lifestyle modification, self-care management, and review of medication use with conventional care, reported a significant reduction in the risks of cerebrovascular disease (HR 0.42, 95% CI 0.24–0.74) but not for coronary events or all-cause mortality (12). Alternatively, other trials, including the UK Prospective Diabetes Study (UKPDS) (15) (a median follow-up of 10.0 years), the Anglo-Danish-Dutch Study of Intensive Treatment in People With Screen-Detected Diabetes in Primary Care (ADDITION-Europe) (a mean follow-up of 9.61 years) (9), and another study conducted in Denmark (10) reported no significant benefits in multifactorial interventions over conventional care. The current study demonstrates that a significant reduction in the risks of macrovascular and microvascular complications and all-cause mortality were seen among RAMP-DM participants. This beneficial effect may in part be explained by the risk stratification strategy at enrollment before intervention, as patients would subsequently receive individualized interventions based on their initial risk level. In addition, most trials have targeted the control of clinical parameters (e.g., HbA_1c, BP, lipid profile) as the primary goal (9,12,15,16), while RAMP-DM focused on total CVD risk reduction, which may represent a more comprehensive approach for lowering the overall risks of developing diabetes-related complications and, subsequently, mortality.

RAMP-DM participants showed more favorable changes in disease control parameters compared with patients receiving usual care. The proportion of patients who had achieved metabolic targets for HbA_1c, BP, LDL-C, and eGFR was greater among RAMP-DM participants during the observation period (Supplementary Table 7 and Supplementary Fig. 6). Although mean FG and HbA_1c levels were similar at the end of observation, the mean reduction in HbA_1c as well as BP levels was significantly greater among RAMP-DM participants over the follow-up period. SBP reduced significantly among RAMP-DM participants, but an increase in SBP was seen among usual care patients. Both groups had a reduction in DBP, yet a larger reduction was observed in RAMP-DM group. For the lipid profile, significant reductions with similar magnitudes were seen in both groups after 10 years. Decreases in BMI and eGFR were observed in both groups, with greater changes found in the RAMP-DM group. This suggested that the reduction in key clinical parameters related to disease control could be sustained, although the beneficial effect on metabolic risk factors became smaller or even insignificant toward the end of the 10-year follow-up period. These results were consistent with the trends in the annual incidence rates of complications, where the protective effect of RAMP-DM on complications was most significant in the first 7 years and attenuated after the 8th year.

Similar results were reported in the J-DOIT3 trial, where the observed reductions for the intervention group compared with the control group were −0.37% (95% CI −0.43 to −0.32) for HbA_1c, −5.3 mmHg (95% CI −6.1 to −4.5) for SBP, −2.9 mmHg (95% CI −3.5 to −2.3) for DBP, and −0.47 mmol/L (95% CI −0.51 to −0.43) for LDL-C but not for BMI (0.05 kg/m², 95% CI −0.05 to 0.15) after 9 years (12). The UKPDS trial also reported a decrease in HbA_1c and FG levels among patients who received an intensive treatment strategy compared with patients who received conventional treatment (15). Conversely, the Steno-2 trial did not observe any significant improvements in most of the metabolic factors in the intervention group (16). The ADDITION-Europe trial observed a reduction in TC (−0.08 mmol/L, 95% CI −0.16 to −0.002) and median eGFR (0.99 mL/min/1.73 m², 95% CI 0.96–1.01) but not for other metabolic factors (9). These results signified that effective control of clinical parameters may accompany the reduction in complication and all-cause mortality rates.

We previously noted that the reported effects of RAMP-DM exceeded the benefit of various diabetes interventions found in other randomized clinical trials (9,16). However, the beneficial effects of RAMP-DM on complication and mortality reduction were at least in part associated with the improvements in metabolic control. We observed similar trends in the changes in metabolic risk factors (Table 4) and the annual incidence of complications and mortality (Supplementary Table 3) during the 10 years of follow-up. For instance, a larger reduction in HbA_1c was recorded (0.10–0.29% on average) in the first 7 years compared with that in 10th year (0.06%), which corresponded to a higher relative risk reduction in the incidence of complications and mortality in the first 7 years than in the 10th year. There could be other mechanisms through which RAMP-DM could reduce complications and mortality. Patient motivation for self-care and additional interventions such as weight reduction classes, smoking cessation counseling, and patient empowerment programs can further reduce CVD and mortality risk beyond the control of metabolic factors.

The sustained benefits of RAMP-DM may be attributed to the multidisciplinary, team-based approach, incorporation of additional comprehensive assessments, and risk stratification models to screen for risk factors. Patients are prioritized and managed on the basis of their baseline CVD risk and empowered by diabetes education and lifestyle advice provided by RAMP-DM nurses and doctors to manage their own self-care. More importantly, regular follow-ups are scheduled by nurses and GOPC doctors in addition to referrals to family medicine specialists, allied health professionals, or other educational programs, if necessary. In this study, the median follow-up time for a RAMP-DM participant was 7.3 years (range 0–10.3 years). A longer RAMP-DM follow-up period was associated with a lower risk of macrovascular and microvascular complications and all-cause mortality (HRs ranged from 0.65 to 0.71, all P < 0.001) (Supplementary Table 8). This all-round management ensured regular follow-up of the patients’ health conditions and that various underlying risk factors could be thoroughly addressed. Results from this study provide support for integration of RAMP-DM into routine primary care for optimal management of patients with diabetes through improving control of disease parameters and preventing the development of diabetes-related complications. Further studies should assess the long-term cost effectiveness of RAMP-DM to determine the economic impact of operating the program.

There were several strengths of our study. First, the Hospital Authority treats 90% of patients with diabetes in Hong Kong (17); hence, the study participants are highly representative of the population. Second, the data were extracted from the centralized, electronic administrative database, which allowed multiple factors to be considered and adjusted for, and ensured the validity and accuracy of the data. Third, our study observed patients for a relatively long period (the median follow-up period was 9.5 years), which minimized the cash-in effect and may allow for a true measure of the effectiveness of RAMP-DM.

Some limitations should also be acknowledged. First, because this study used electronic health record data, missing and misclassified diseases based on diagnosis codes might be present. Although the validity and completeness of diagnosis codes in patients with diabetes in the Hospital Authority database have not been specifically assessed, previous studies have shown that the database has a high completeness of ICPC-2 (61.0–87.7%) (34) and accuracy of ICD-9-CM codes for identifying CVD outcomes (35). Second, because patients were not randomly assigned in this observational cohort study, selection and confounding bias might exist. We used propensity score matching to select patients with similar characteristics to minimize bias caused by differences in characteristics between groups. However, the use of propensity score matching cannot eliminate possible biases caused by unobserved confounders, such as lifestyle habits (e.g., dietary habits, physical activity level), medication adherence, and patient motivation levels. There was a possibility where patients who agreed to participate in RAMP-DM were more motivated toward self-care and treatment adherence. Nonetheless, results from the negative control analysis (shown in the results section) and E-values (Supplementary Table 6) demonstrated that the impact of unobserved confounders on the study results are likely negligible. Finally, study data were only available until December 2019; hence, complete 10-year data were not available for patients sampled from January to September 2010.

In this population-based study, RAMP-DM, a protocol-driven, risk-stratified multidisciplinary program offered in addition to usual care for patients with diabetes in primary care, might have a long-term benefit in reducing the incidence of diabetes-related complications and all-cause mortality, as well as better control of metabolic risk factors, compared with usual care over a 10-year observational period. The effect of RAMP-DM on diabetes-related complications attenuated at the 8th year of follow-up, while that for all-cause mortality persisted by 10 years. Integration of RAMP-DM into usual primary care could represent an effective strategy for achieving favorable outcomes in patients with diabetes.

Acknowledgments. The authors acknowledge the Central Panel on Administrative Assessment of External Data Requests at the Hospital Authority for providing the data.

Funding. This study was funded by the Health Medical Research Fund (reference no. CFS-HKU4), Food and Health Bureau, the Government of Hong Kong Special Administrative Region.

The funding source did not have any role in the current study, including study design; collection, analysis, and interpretation of data; report writing; and decision to submit the manuscript for publication.

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

Author Contributions. E.H.M.T. and I.L.M. contributed to the literature review, study design, statistical analysis, and writing of the manuscript. E.T.Y.T., E.Y.F.W., E.Y.T.Y., J.Y.C., W.Y.C., C.K.H.W., and C.L.K.L. contributed to the literature review, study design, acquisition of data, statistical analysis, and writing of the manuscript. D.V.K.C., W.W.S.T., and T.K.H.H. provided critical input in clinical aspects. All authors reviewed and edited the manuscript and contributed to the interpretation of the results. C.K.H.W. and C.L.K.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.