Investigation of cardiovascular risk factors in type 2 diabetes in a rural Australian Division of General Practice

The Southern Highlands Division of General Practice (SHDGP) is a rural Division, established in 1994, located 130 km south-west of Sydney. In 1995, the SHDGP began collecting diabetes clinical data in CARDIAB (a Division-based electronic diabetes register system12) from local general practices, and employed a diabetes educator. SHDGP provided various services, including identifying patients at risk and those who should be on insulin, one-to-one and group education for patients with diabetes, an exercise and lifestyle program, ongoing professional development for GPs, supporting practices with care planning, and electronic messaging of diabetes data from practice software to SHDGP. We found previously that GPs who use Division-based electronic diabetes registers are more likely to provide better-quality patient care than those who do not.13

The targets for diabetes care in this study (levels of HbA1c ≤ 7%, total cholesterol [TC] < 4.0 mmol/L, low-density lipoprotein cholesterol [LDL-C] < 2.5 mmol/L, total triglycerides [TG] < 2.0 mmol/L, high-density lipoprotein cholesterol [HDL-C] ≥ 1.0 mmol/L, and blood pressure [BP] < 130/85 mmHg) were based on the Australian guidelines for diabetes management in general practice in 2002. The only change between 2002 and 2005 was a small reduction in the BP target from 130/85 mmHg to 130/80 mmHg.14

Based on population estimates using age and sex distributions from AusDiab (Australian Diabetes, Obesity and Lifestyle Study), an average of 85.8% of people with diagnosed diabetes in the SHDGP area were on the SHDGP register each year during the study period.15,16 The remainder were assumed to have diabetes but not attend general practice, or to attend general practice but not consent to have their data registered with SHDGP. A cohort of patients with type 2 diabetes attending the annual evaluation during the study period was identified.

Sample size calculations on HbA1c level confirmed that, after adjustment for clustering, 628 patients would have sufficient power (1 − β = 0.8 and α = 0.05) to detect an effect size of 0.13 between 2002 and 2005.

De-identified patient information was extracted from CARDIAB. Criteria for data cleaning and management were based on the procedures used in our previous study, the Divisions Diabetes and CVD Quality Improvement Project (DDCQIP).16

Ethics approval for this study was obtained from the Human Research Ethics Committee of the University of New South Wales.

Study variables included age, sex, duration of diabetes, BP, HbA1c level, serum lipid levels, and smoking. Study outcomes were changes in these variables over time.

SPSS version 14.0 (SPSS Inc, Chicago, Ill, USA) and MLwiN17 (Centre for Multilevel Modelling, University of Bristol, Bristol, UK) were used to analyse the data. Comparisons were conducted by multilevel multivariate linear regression adjusting for age, sex and the duration of diabetes; this took into account the multiple observations (GPs and evaluation time) on the same patients and clustering.18 To achieve the most parsimonious multivariate model, non-significant factors were excluded by backward elimination. Therefore, only the factors significantly associated with the outcomes (P < 0.05) were included in the final multilevel model.

At baseline, systolic BP, TC, LDL-C and TG were all higher than the targets for general practice diabetes management (Box 1).14 By Year 4 of the study, BP, HbA1c level and smoking tended to be worse compared with baseline values, but BMI and lipid profile tended to improve. After adjusting for age, sex, duration, and clustering at patient and practice levels (Box 2), only TC and LDL-C significantly improved over the 4 years.

In the multilevel analysis, older patients had worse systolic BP, but better BMI and lipid profile than younger patients (Box 2). Compared with men, women had worse BMI and TC, but better HDL-C and HbA1c level. Longer duration of diabetes was associated with worse systolic BP and HBA1c level, but better HDL-C. People with higher BMI were likely to have worse systolic BP, HDL-C and HbA1c level, but better TC than those with lower BMI.

Cohorts of patients from primary care disease registers can provide useful information on trends in quality of care for diabetes. However, this depends on the coverage of GPs and the population with diabetes. There was a high engagement of GPs (more than 90%) and a high proportion of patients with diabetes enrolled with the register (85.8%) in this study. Our data were retrospectively extracted from the practices without the selection bias that may occur in patients selected for a trial. Finally, we used multilevel analyses to adjust for clustering of patients recruited from general practices. Like all cohort studies and in the absence of a control group, it is not possible to attribute changes to particular interventions or policy. Furthermore, the conclusions may be generalised to other areas only with caution — the strong engagement of GPs with SHDGP may not be typical of other Divisions.

In Australia, the management of hyperlipidaemia is mostly carried out by GPs.6 In the SHDGP cohort, both TC and LDL-C decreased over the study period. This achievement may be attributed to greater attention by GPs to lipid pharmacotherapy and dietary education in patients with type 2 diabetes. Lipid levels tended to be better among older patients. This suggests that GPs may be providing more active treatment, there may be better compliance with treatment among older patients, or younger age may be associated with other worse risk factors (eg, BMI). There were no significant changes to the Pharmaceutical Benefits Scheme guidelines for prescribing lipid therapy over the study period. However, more recently there have been changes that will make it easier for GPs to offer pharmacotherapy to patients with diabetes.19

Compared with lipid management, control of BP and HbA1c level is a more complex task. After adjusting for clustering, there were no significant changes over the study period in systolic BP, HbA1c level or BMI (Box 2). Trend data from the United Kingdom Prospective Diabetes Study Group suggest that these physiological markers tend to worsen with time.20,21 Thus, the absence of a negative secular trend for these indicators in our study is encouraging. It suggests the diabetes care in SHDGP may have prevented what might otherwise have been worse results. However, the low performance against targets for those indicators (Box 1) indicates a need for more comprehensive and intensive pharmacological and behavioural interventions to reduce these risk factors.

BMI did not improve over time and was associated with worse BP and HDL-C in our study. Currently, excess weight has replaced smoking as the most common risk factor for CVD in the Australian population.22 The control of weight is even more challenging than control of BP and hyperlipidaemia, as it requires very active patient self-management.

Box 3 compares our study with two other studies from Australian general practice: the national DDCQIP cohort,8 and a smaller regional cohort from rural south-east Queensland (SEQ).23 The significant changes in systolic BP and HbA1c level in DDCQIP were not seen in the SHDGP study. In the SEQ study, BP improved, but HbA1c level did not. An important reason for these differences may be the selection of practices and patients. Both the DDCQIP and SEQ studies involved a more selective group of volunteer practices and patients than the SHDGP cohort.

Internationally, our study can be compared with the Steno-2 cohort study24 (Box 3) and the National Health and Nutrition Examination Survey (NHANES) cross-sectional surveys in the United States.25 Steno-2 showed better improvements in BP, lipids and HbA1c level. As a clinical trial, Steno-2 had a more selective study group with worse risk factors at baseline than in our study. Both NHANES and our study showed significant reductions in TC. However, NHANES also showed a significant decrease in systolic BP and an increase in BMI (whereas we found no change). This may be due to less aggressive management of BP in Australia than in the US, or may be an artefact of differences in the nature of the studies (NHANES is a repeated cross-sectional study rather than a cohort study).

Our research provides evidence of the quality of preventive care in cardiovascular risk for people with type 2 diabetes in one Division of General Practice. Despite national initiatives and an active Division program, after adjusting for covariates and clustering, only lipid levels improved over the 4 years. HbA1c level, BP and BMI did not change significantly. This indicates the continuing challenge of managing these risk factors comprehensively in primary care, and the importance of very active treatment to targets identified in evidence-based guidelines.