Primary care practices throughout the Netherlands that were not involved in other diabetes care improvement programs were block randomized to intervention (26 practices) or the control group (29 practices). The number of PCPs working in each practice and the presence of a practice nurse before intervention were taken into account before randomization. The intervention, also described elsewhere (5
), consisted of 1
) diabetes consultation hour run by a practice nurse, 2
) a CDSS that contained a diagnostic and treatment algorithm based on the Dutch type 2 diabetes guidelines (6
) and provided patient-specific treatment advice, 3
) a recall system, and 4
) feedback every 3 months regarding the percentage of patients meeting the treatment targets (cessation of smoking, A1C <7%, systolic blood pressure <140 mmHg, total cholesterol <4.5 mmol/l, LDL cholesterol <2.5 mmol/l, and BMI <27 kg/m2
) on both the practice and the patient levels (6
). The PCPs were advised that they should prescribe new medication and refer patients if necessary. The control group continued with the same diabetes care that they had received before entering the study, which means that diabetes care was provided by the PCP or by a practice nurse under PCP responsibility. The University Medical Center Utrecht ethics committee approved the study, and patients provided written consent.
From the 171,821 registered patients, all type 2 diabetic patients were identified. Patients who had a short life expectancy, were unable to visit the primary care practice, or were receiving diabetes treatment from a medical specialist were excluded. Initially, 3,979 patients were eligible (2,136 in the control group and 1,843 in the intervention group), but 548 subjects refused to participate (409 control and 139 intervention subjects), and an additional 40 (35 control and 5 intervention subjects) failed to participate for unknown reasons (for both groups, P < 0.05). The final, mainly Caucasian, study population consisted of 3,391 patients (1,692 control and 1,699 intervention). After 1 year, 2,841 patients (1,389 control and 1,452 intervention) completed a follow-up examination; 187 patients (115 control and 72 intervention) refused to participate in the final measurements, and 13 others (12 control and 1 intervention) failed to show for unknown reasons (for both groups, P < 0.05). The groups did not differ with regard to the number of patients who died, moved, became terminally ill, or were referred to a specialist.
Between March 2005 and August 2007, patients were each seen twice for annual diabetes checkups. Patients who did not show received one reminder. In the CDSS, age, sex, ethnicity, duration of diabetes, and smoking habits were registered. A1C, total cholesterol, and HDL cholesterol were measured in local laboratories. LDL cholesterol was calculated. Blood pressure was measured according to a standard operating procedure.
The 10-year coronary heart disease (CHD) risk estimate, as established by the UK Prospective Diabetes Study (UKPDS) (7
), was calculated using the above-mentioned variables, excluding LDL cholesterol.
The primary outcome was the 1-year difference in A1C. Secondary outcomes were the 1-year difference in the 10-year UKPDS CHD risk estimate and the percentage of patients that reached A1C ≤7%, systolic blood pressure ≤140 mmHg, total cholesterol ≤4.5 mmol/l, and LDL cholesterol ≤2.5 mmol/l (6
We performed intention-to-treat analyses with baseline values carried forward in the case of missing values. To correct for clustering at the practice level, generalized linear models were used, and after clustering had been taken into account, a 0.3% difference in A1C and a 2% difference in UKPDS CHD risk could be detected with 90% power (α = 0.05), with at least 1,080 patients in each treatment arm.