To the best of our knowledge, this is the first study to evaluate the effectiveness of a risk-stratification management approach involving multidisciplinary interventions for people with diabetes in the real-world primary care setting. This study found that the RAMP-DM intervention led to lower incidence of cardiovascular events and significant improvements in HbA1c and predicted cardiovascular risks compared to the usual care. Further investigation of the effects of each RAMP-DM intervention component found that, after adjusting for potential confounders, associate consultant intervention was associated with improvements in lipid control and predicted CHD risk, while PEP was related to decreases in BMI.
Compared to the previous study on risk stratification and intervention in staff-model primary care clinics by Clark et al. , the magnitude of improvement in HbA1c in our study was smaller. In our study, there was a 0.2% net decrease of HbA1c in RAMP-DM group, while Clark’s study showed about 0.35% between group differences in the changes of HbA1c after 12-month follow-up. Our subjects were much less severe at baseline with an average HbA1c of 7.2%, whereas the mean HbA1c in Clark’s study was above 8.5%. The RAMP-DM is designed to cover all people with diabetes in the primary care in Hong Kong; therefore, we randomly selected subjects to assess the effectiveness of RAMP-DM in general instead of selecting severer cases deliberately. We observed 5.40% more subjects under RAMP-DM reaching treatment target (HbA1c < 7.0%), which addressed the clinical benefits of RAMP-DM.
RAMP group had a greater increase in the proportions of reaching HbA1c < 7%, and SBP/DBP < 130/80 mmHg compared to usual care group, but after adjusting for the baseline parameters and drug treatment, the differences became insignificant, although the results were still favoring RAMP-DM participants. Male sex and no history of myocardial infarction were found to be associated with uncontrolled blood pressure . As all of our study subjects were without cardiovascular complications at baseline and gender was well matched between groups, it was likely that the baseline parameters and drug treatment affected the outcomes. Drug treatment and duration of disease were indicators of diabetes severity. Moe et.al found that compared to people with diabetes and without medication, these on medication subjects had higher risk of cardiovascular death . For severer subjects at baseline, doctors might provide them with more intensive care, no matter they were enrolled in RAMP-DM or not, leading to bigger improvement. Also, regression to the mean might lead to bigger reduction for those with higher baseline HbA1c and SBP levels. In addition, subjects in control group were also eligible to be referred to some of the services in the RAMP-DM intervention package (allied health professionals and PEP) if necessary, which might bias the effects of RAMP-DM towards null.
The RAMP-DM group observed fewer coronary heart disease and total cardiovascular events compared to the control group during 12 months follow-up. This is consistent with the findings of the improvement of HbA1c and predicted cardiovascular risks in RAMP-DM group. A recent study shows that the increase in HbA1c level is significantly associated with the incidence of coronary heart disease during 6 years follow-up . To validate the association in our study, longer follow-up period is needed.
We employed the Framingham cardiovascular risk function developed for primary care  to assess the longer term effects in total CVD risk, and applied the UKPDS risk engines , to predict the changes in CHD and stroke risks. Although we found that RAMP-DM participants showed significantly greater improvement in the total CVD risk, the differences were not significant after adjusting for drug treatment, while the differences in CHD and stroke risk predicted by the UKPDS risk engines remained significant. Our previous study found that the UKPDS risk engine is more sensitive to detect differences in CHD risk in Chinese people with diabetes , as it was developed for people with diabetes specifically. Moreover, the previous study showed that the CHD risk predicted by the UKPDS risk function showed excellent convergent validity with the JADE risk function that was developed in Chinese people with diabetes ,. We could not use the JADE risk function for the estimation of CVD risk in this study because many required parameters such as estimated glomerular filtration rate and urine albumin:creatinine ratio were missing in many subjects.
Very few studies on DM management measured cardiovascular risk reduction as an outcome. Most studies only reported changes in blood pressure and lipid profiles in addition to HbA1c. Comprehensive cardiovascular risk management is getting increasing attention in diabetes care ,. The RAMP-DM provided personalized risk-stratification based care to people with diabetes by multidisciplinary health care professionals, which promoted the concept of cardiovascular risk management and facilitated the optimization of medical resources.
The RAMP-DM addressed four interralated components of the Chronic Care Model , with multidisciplinary management involving doctors, nurses and allied health professionals. By exploration on individual intervention components among RAMP-DM participants, we found that nurse intervention alone was not associated with any improvements in biomedical outcomes and cardiovascular risks.
Previous trials on nurses led interventions resulted in inconsistent findings in changes of HbA1c. The PEACH study in the Australia primary care setting delivering telephone coaching on medication goals by practice nurses failed to achieve improvement in HbA1c and other relevant biomedical measures . A Spain based standardized language in nursing care plans  and a U.S. based nurse care management also showed no improvements in HbA1c . However, a nurse-led telephone coaching intervention in the U.S. found significant reductions in HbA1c . Lacking of prescribing rights is likely to limit the role of nurses in diabetes management, thus affect the benefits of sole nurses intervention .
On the other hand, most multidisciplinary interventions involving at least nurses and physicians found favoring results on blood glucose control. A multidisciplinary intervention for patients with HbA1c higher than 10% in Israel found the intervention group had significant decrease in HbA1c after six months. This multidisciplinary team contained diabetologist, the dietician and the diabetes nurse educator . Positive results were also found in similar multidisciplinary interventions in Taiwan , the U.S.  and France . Most of these multidisciplinary interventions included diabetic education sessions. A symposium convened by American Association of Diabetes Educators acknowledged that the most effective education programs occurred within multidisciplinary teams .
The PEP was associated with improvement in BMI, and had favoring effects on HbA1c, lipid profiles and SBP. The small number of subjects enrolled in PEP (7.2%) might be insufficient to detect significant changes. An independent study on the effectiveness of PEP confirmed these favoring results . A RCT was designed to compare the effects of long-term (2 years) education program with initial education only in France. The results are yet to report .
There are several limitations of this study. First, since it is not a RCT, some unknown potential confounders might affect the results. However, a study found that the positive effects of interventions in controlled trial settings could not be replicated in real-world primary care settings . Second, the lipid profiles and blood pressure between the RAMP-DM and usual care groups were not well matched at baseline, which might affect the changes over 12 months. Third, the lack of blinding of clinicians and patients is the inherent limitation of population based clinical interventions. Fourth, the Framingham risk function and the UKPDS risk engines were developed in western population, and the UKPDS risk engines were developed in subjects with type 2 diabetes. We had less than 1% subjects with type 1 diabetes, which might affect the accuracy of the predicted CHD and stroke risks. As cardiovascular events need time to develop, the follow-up period of one year was not long enough for us to validate the predicted CVD risk with observed cardiovascular events.