Design, setting and participants
Cross-sectional data from electronic medical records (EMRs) from 11 practices (49 GPs’ with 58857 listed patients) from multiethnic parts of Eastern Oslo were used . A specially designed data programme was used to identify patients with diabetes and to capture predefined data from the EMR from the years 2003–2005. Data regarding HbA1c, SBP, DBP, microalbuminuria, body weight and foot examinations were from 2005, for eye examinations from 2004 or 2005, and for s-cholesterol and smoking habits from 2003 to 2005. A total of 2064 patients with a diagnosis of diabetes were identified. As our aim was to explore GPs’ adherence to guidelines for T2DM, we excluded T1DM patients (n = 103), T2DM patients with two or more diabetes related hospital visits the previous 12 months (n = 178) as they also had incomplete information of current medication, those with less than six months of follow-up or who had moved or were deceased (n = 128), or who had incomplete information about the country of birth (n = 2), leaving 1653 T2DM patients cared for by their GPs to be included in the present study. The study was approved by the Regional Ethics Committee West, the Directorate for Health and the Data Inspectorate.
Patients who according to the EMR had a diagnosis of angina pectoris, previous myocardial infarction, stroke or transitory ischemic attack, or intermittent claudication, were categorized as having CVD and thus requiring secondary prevention (SP) to reduce their risk of new events. All other patients were considered to require primary prevention (PP) of CVD.
According to the Norwegian national general practice diabetes guidelines from 2005 [33, 34], diabetes patients without CVD should have anti-hypertensive therapy if their BP > 140/90 mmHg, and lipid lowering therapy if their total cholesterol/HDL-cholesterol ratio > 5.0 and they had at least one additional risk factor (current anti-hypertensive therapy, smoking, microalbumiuria or a family history of premature CVD). For diabetes patients with CVD, the corresponding treatment thresholds were BP > 140/90 mmHg and total cholesterol/HDL-cholesterol ratio ≥ 4.0 irrespective of other risk factors. Treatment targets were: HbA1c ≤ 7.5%, SBP ≤ 140 mmHg, diastolic blood pressure (DBP) ≤ 85 mmHg, and total cholesterol/HDL-cholesterol ratio < 4.0 [33, 34].
Age, gender, recorded measurements of CVD risk factors, patient’s age when diabetes was diagnosed, disease duration, and prescription data were captured . For HbA1c, SBP, DBP and lipids, the most recent results were selected.
Prescription of glucose-lowering therapy (anti-diabetic agents, insulin or any combinations), anti-hypertensive (angiotensin converting enzyme inhibitors, calcium channel blockers, alfa blockers, beta blockers, angiotensin II receptor antagonists, diuretics, or any combinations), and lipid-lowering therapy (statins) were dichotomized as “yes” versus “no” for all three therapy groups. The intensity of treatment was categorized by numbers of agents used in combination to target elevated BP, and to lower blood glucose. Because too intensive treatment may put patients at increased risk for side effects like hypoglycaemia and hypotension, we identified patients on glucose-lowering therapy with HbA1c < 6.0% [27, 35] and those on anti-hypertensive therapy with SBP < 130 mmHg or DBP < 65 mmHg .
Ethnicity was based on self-reported country of birth as recorded in the EMR, and categorized as: Norwegians (including about 2% from other Scandinavian countries or Western Europe/North America), South Asians (Pakistanis, Sri Lankans and Indians), others (from other low- and middle income countries).
To estimate individual 10-year absolute risk for CHD for patients without prior CVD in the two largest ethnic groups (Norwegians and South Asians), we used the UKPDS Risk Engine version 2 , which includes age, gender, diabetes duration, HbA1c, SBP, total cholesterol, HDL-cholesterol, smoking status (never, past or current smoker).
Chi- squared, one-way ANOVAs, T-tests and Wald tests were used to test differences between proportions and means in the groups. Because HbA1c values were highly skewed, they were log-transformed and presented as geometric means (estimates with 95% confidence intervals, transformed back to the original scale). Generalized linear models were applied to estimate means for the CVD risk factors, the proportions receiving pharmacological therapy, and the proportions not reaching treatment targets, all adjusted for age and gender. The geometric mean values for HbA1c were also adjusted for diabetes duration.
HbA1c was recorded in 95%, BP in 91%, total cholesterol in 94% and smoking habits in 59% of all patients, but the complete set of variables to be used in the UKPDS risk engine was available for only 54% of Norwegian and 43% of South Asian patients. We therefore used multiple imputation techniques for individuals with incomplete data as this is the recommended procedure to limit bias due to missing data when values are missed at random . Age, gender, ethnicity, BMI, glucose-lowering therapy, anti-hypertensive therapy and lipid-lowering therapy were used as predictors for the imputed values and five imputed datasets were created. Multiple regression models were applied to estimate age- and gender adjusted 10-year mean risk for cases with complete data and for all cases after imputation and pooling of the original and imputed data sets.
Two-sided tests were used and p-values ≤ 0.05 were considered statistically significant. The analyses were performed with SPSS 19.0 for Windows and Stata 12.