Characteristics of Diabetics with Poor Glycemic Control Who Achieve Good Control

Population

The study population included all diabetic patients, older than 40 years of age with at least one HbA1c measure >9.5 mg% during 2001, who were still alive and insured by CHS in 2003.

This group was divided into 3 categories according to their last HbA1c measure in 2003. Patients with HbA1c <7.5 mg% were considered to be well controlled. Patients with HbA1c >9.5 mg% were considered to be poorly controlled. Patients with HbA1c between 7.5 and 9.5 mg% were excluded from further analysis.

The value of HbA1c >9.5 mg% was chosen to make sure that only diabetic patients with very poorly controlled diabetes would be included in this study. HbA1c of <7.5 mg% was chosen to enlarge the well-controlled group. The intermediate group was excluded from the study in advance to sharpen the differences.

Patient Demographics

The following data were collected for each patient: age (in 2001), gender, place of birth, and socioeconomic status. Patients exempt from paying the monthly National Insurance contribution were defined as being of low socioeconomic status. This exemption from payments is given by the National Insurance according to income. The information about the exemption is retrieved directly from the National Insurance database. This marker of low socioeconomic status was used by the quality indicators national program.¹⁷

Administrative Data

Data about the primary care physicians was collected, including demographics (including age, gender, place of birth, medical school, years since graduation from medical school, years of working at the HMO, and the physicians’ managerial responsibilities); changing the primary care physician during the study period; and primary care clinic size (small, <5,000 patients; intermediate, 5,000 to 10,000 patients; large, >10,000 patients).

Medication costs and total patient costs in 2001 and in 2004 (1 year after diabetes control had been achieved/not achieved) were calculated as US$ (fixed prices). Costs were measured as the costs to the HMO. A log of the costs was used in an attempt to normalize the distribution of the costs.

Health Determinants

Health determinants include the number of chronic disease diagnoses other than diabetes, according to the chronic diseases register, the number and duration of hospitalizations from any cause in 2001 and in 2004, and mortality after 2003.

Diabetes Follow-up

Follow-up included low-density lipoprotein (LDL) cholesterol examination and calculated LDL cholesterol values for 2001 and 2004; ophthalmologist visits in 2001 and 2004 (fundus examination as part of diabetes follow-up is done in Israel routinely by ophthalmologists); and microalbumin examination in 2004 (because microalbumin examinations were evaluated using more than one method in 2001 the information was not included). All laboratory tests were performed in the districts’ central laboratory and results were sent directly to the patient's electronic medical files.

Statistics

χ² and t tests were used for the comparisons of each variant. Logistic regression was used to evaluate the relationship between predictors. The dependent variable was the glycemic control in 2003. STATA software, version 8.0 (Stata Corporation, College Station, TX) was used for all statistical analysis.

Patient Characteristics and Good Glycemic Control (Table 1)

Men were more likely than women to achieve well-controlled diabetes (52% vs 43.6%; P < .001). The average age of those who achieved well-controlled diabetes was slightly younger than that of those who maintained poorly controlled diabetes (63.2% vs 64.5%; P < .001). There were fewer patients from a low socioeconomic status in the well-controlled group compared with those patients in the poorly controlled group. There was no difference between the groups relating to origin, the number of other chronic diseases (besides diabetes), or mortality rate.

Physician and Clinic Characteristics and Good Glycemic Control (Table 2)

The patients’ primary care physician was highly correlated with good glycemic control in any combination of predictors in the logistic regression (physicians were entered as a catregorical variable in the analysis). There was a full spectrum of results related to physicians’ outcomes. Although some physicians had all of their patients with poorly controlled diabetes in 2001 become well controlled in 2003, others had none. None of the primary care physician characteristics were related to glycemic control except being a clinic manager.

The primary care clinic to which the patient belonged was highly correlated with good glycemic control (P < .001). Smaller (small and intermediate size) clinics had a slightly higher percentage of patients who achieved well-controlled diabetes compared with large clinics.

Patients from the group with poorly controlled diabetes were more likely to change their primary care physician (because of patients’ choice or physician replacement) than patients from the group with well-controlled diabetes (31.1% vs 27.7%; P = .044).

Diabetes Follow-up (Table 3)

Patients who had poorly controlled diabetes in 2003 were more likely to have a better diabetes follow-up in 2004 compared with those who were well controlled in 2003. Among the patients with poorly controlled diabetes in 2003, 59.6% visited the ophthalmologist vs 55.5% of patients in the group with well-controlled diabetes (P = .025); microalbumin was checked in 72.6% vs 68.2% (P = .009). LDL cholesterol was checked in 87.3% vs 82.8% (P < .001), but average LDL cholesterol was significantly lower in the well-controlled group (103.1 vs 109.8; P < .001) using both t test and a pair-wise comparison.

Patient Costs (Table 4)

There were no differences in patient hospitalizations, average drug costs, or average total cost between the 2 groups in 2001. A higher percentage of patients in the group with poorly controlled diabetes were hospitalized during 2004 (34.9% vs 31.1%; P = .033). The average patient cost was 8% lower in 2004 among patients in the well-controlled group compared with the poorly controlled group, although the average drug costs were higher for those in the well-controlled group compared with those in the poorly controlled group in the basis year 2001. Using a log of costs, only the total patient cost in 2004 had a borderline significance.