Patient characteristics correlated with quality indicator outcomes in diabetes care

INTRODUCTION

Organisations such as the Organisation for Economic Cooperation and Development (OECD)1 and National Diabetes Quality Improvement Alliance (NDQIA)2 have recently adopted quality indicator (QI) programmes to compare quality of care across health systems. In Israel, a national QI programme in primary care was introduced in 2001,3 integrating data from all the health maintenance organisations (HMOs) in Israel. Although the long-term outcomes of this programme are still unclear, in the short term, improvements in QI scores have been achieved.3 Quality assessment has highlighted disparities in medical care attributed to factors such as insurance programme,4 ethnicity,5,6 and socioeconomic status.7

Patients' sociodemographic characteristics explain a substantial proportion of racial disparities in diabetes outcomes,8 while use of a uniform treatment algorithm was found to abolish racial disparities in glycaemic control within 2 years.9 A study on the UK GP Quality and Outcomes Framework data found practices located in deprived areas performed less well on quality measures than those based in affluent areas.10

Israel has a national health insurance with universal access to medical care. Health care is delivered by four HMOs through the country. Every Israeli resident receives medical coverage from the HMO of their choice. Clalit Health Services is the largest HMO in Israel. It serves more than 50% of the population and more than 70% of the older patients (≥65 years). Patient records in Clalit Health Service have been completely computerised for over a decade, and an extensive healthcare database has been created. Patient records include demographic data, working diagnoses, medications, lab results, hospitalisations, referrals, and administrative data including patients' costs to the HMO. The demographic data are updated directly from the population registry of the Interior Ministry. All laboratory tests are sent to a central lab, and the results are sent directly to the primary care physician and to the patient's file. The working diagnosis of ‘diabetes mellitus’ detects over 90% of the known patients with diabetes.11

Clalit Health Service clinics are spread throughout the country; most of the primary care clinics serve their local neighbourhoods, which usually represent a relatively homogenous population. Clalit Health Service uses a passive capitation system, and all patients in Clalit Health Service have to choose a primary care physician who is responsible for their care. There is no co-payment for visiting the primary care physician or for having laboratory tests. There is a small co-payment for specialist visits (equivalent to $US 5).

The primary care physicians are responsible for diabetes follow-up for all patients on their list, including those seen in specialist diabetes clinics. QIs in diabetes care were introduced in Clalit Health Service as part of the national QI programme. All the physicians in Clalit Health Service are updated routinely about their QI scores, and get lists of patients who need improvement in their care. This has created a unique opportunity to evaluate factors influencing QI outcomes in the primary care setting.

METHOD

The study was conducted in the Central District of Clalit Health Service in Israel, and received local ethics committee approval.

All QIs related to diabetes mellitus in the national QI programme were examined. Five performance measures related to diabetes follow-up: the percentage of patients with diabetes for whom glycosylated haemoglobin (HbA_1c), microalbumin, low-density lipoprotein (LDL)-cholesterol, and blood pressure were measured at least once during 2005, and the percentage of patients with diabetes who were seen by an ophthalmologist for retinal examination during 2005. Three outcome measures related to diabetes control: the percentage of patients with diabetes and HbA_1c <7 mg%, the percentage of patients with diabetes and blood pressure <130/80 mmHg, and the percentage of patients with diabetes and LDL-cholesterol levels <100 mg/dl according to their last result in 2005.

To avoid inconsistent results due to a change in the treating physician, data were gathered for patients of all the primary care physicians who started working in Central District of Clalit Health Service before 2004 and who were working in the same clinic until the end of the study period. Physicians were included if they treated at least 10 patients with diabetes.

All patients with diabetes who were treated by the same physician for over 2 years were included, and sociodemographic data including: age, sex, year of immigration to Israel (a patient was defined as a new immigrant if he or she had immigrated to Israel after 1990), and socioeconomic level (low socioeconomic status as defined by the National Insurance Institute of Israel according to patients' income) were recorded. Clinical information was retrieved, including HbA_1c, LDL-cholesterol, microalbumin, body mass index (BMI), blood pressure, diabetes clinic attendance, insulin use, and oral medication for diabetes use.

For each physician list, information was retrieved about the total number of patients and total number of patients with diabetes. To evaluate patient characteristics, a profile of patients with diabetes was created for every physician in the study. The following values were calculated for the physician's patients with diabetes: the percentage of females, percentage of new immigrants, and percentage of patients of low socioeconomic level from the total number of patients with diabetes; the mean age, mean BMI, and mean systolic blood pressure of the patients with diabetes; and the percentage of patients attending specialist diabetes clinics, percentage taking oral medications, and percentage treated with insulin from the total number of patients with diabetes. Since patients with diabetes represent a significant proportion of the workload for the primary care physician, the percentage of these patients out of the total number of patients treated by each physician was calculated.

Linear regression was used to identify correlated variables. A stepwise method was used to identify correlated variables. The primary model included all available variables including demographics, treatment modalities, endocrinologist consultations, and workload of patients with diabetes. The final regression model included mean age and mean BMI as continuous variables, and sex (female versus male as baseline), low socioeconomic status (low versus non-low socioeconomic status), and immigrant status (new immigrants versus non-new immigrants) as dichotomous variables. Robust regression models with the same variables were used for QIs that were not normally distributed (microalbumin and blood pressure measure and blood pressure <130/80 mmHg). QIs of each physician were considered as dependent variables. A P-value of 0.01 was considered as significant. STATA (version 8.0) was used for statistical analysis.

RESULTS

One hundred and seventy primary care physicians were responsible for 301 625 patients (range 253–3296 per physician); 18 316 (6.1%) were patients with diabetes. The mean number of patients with diabetes per physician was 107 (range 10–203). Table 1 describes the cohort as a whole and the variability among the different physicians' practices.

Patients of low socioeconomic status with diabetes were older (63.9 versus 67.2 years; P<0.001) with slightly higher BMI (30.8 versus 29.7 kg/m²; P<0.001), were more likely to be referred to an endocrinologist (14.1% versus 12.9% P = 0.012), and were more likely to be treated with insulin (21.3% versus 15.3% P<0.001).

New-immigrant patients with diabetes were younger (58.4 versus 68.7 years; P<0.001), were less likely to receive oral medications (74.7% versus 81.0%; P<0.001), were more likely to receive insulin (20.3% versus 16.0%; P<0.001), and more likely to be referred to an endocrinologist (15.7% versus 12.0%; P<0.001).

Patients who were referred to an endocrinologist were younger compared to those who were not referred (60.0 versus 65.9 years; P<0.001). Patients who were treated with insulin were younger than those not receiving insulin (60.3 versus 66.2 years; P<0.001). Table 2 presents the quality indicators of the cohort and the different practices.

DISCUSSION