Setting
MHS, the second largest health plan in Israel, with 1,807,521 members when the study was performed.
Study Period
Data were extracted on November 15, 2008 (Due Day) and reflects care provided during the previous 12 months (December 2007-November 2008)
Study Subjects
Our reference population for the assessment of diabetes prevalence among different groups consisted of all MHS members, aged 18-80, who had visited a general practitioner (GP) at least once during the previous two years. The study population (diabetic patients) for the assessment of diabetes care and control were all MHS members, aged 18-80, who were recorded in the computerized Diabetes Registry (DR) on the due day and visited a GP at least once during the previous two years. We used the latter criterion as a proxy for a routine patient-physician relationship in all measures that reflect care at the practice level.
Diabetes Registry (DR)
Diabetic patients were defined according to the criteria suggested by the American Diabetes Association [20]. Symptoms were diagnosed as diabetes with fasting plasma glucose > 125 mg/dL or casual (namely, any time of day without regard to time since last meal) plasma glucose concentration ≥200 mg/dL, in addition to a diabetes diagnosis documented in the medical records. Patients were included also if they (1) purchased at least two hypoglycemia medications or a single insulin dose during any six-month period or (2) had HbA1C results of at least 7.25 gr%. HbA1C of 6.5 gr% was used as an entry criterion only if a diabetes diagnosis had been documented in the electronic medical records [21]. Women of child-bearing age who entered the DR due to one purchase of insulin but without any other diabetic medications or no purchase of insulin for the next three months were considered Gestational Diabetics and subsequently excluded.
Data sources
Data were extracted from: (1) the MHS computerized DR and billing system, (2) the MHS computerized PMS, (3) Israel's Census Registry for data on Socio-Economic Rank (SER) and ethnicity.
Performance measures
In order to assess the effectiveness of management of diabetes care and its potential complications, we analyzed the following quality indicator process measures: HbA1C and LDL cholesterol testing, a pre-defined optimal follow-up, and intermediate outcome measures: HbA1C control and LDL cholesterol control.
Definitions
HbA1C testing
Proportion of patients registered in the DR who performed at least one HbA1C test during the previous year.
LDL cholesterol test
Proportion of patients registered in the DR who performed at least one LDL cholesterol test during the previous year.
Optimal Follow-up
This is an "all or none measure" [22], which includes performance of all the following tests: HbA1C testing, LDL Cholesterol testing, nephropathy monitoring (urine micro-albumin), eye and foot exams, blood pressure and Body Mass Index recording.
Intermediate outcomes
(1) Poor HbA1C control: Proportion of DR patients who performed an HbA1C test during the previous year and had demonstrated HbA1C ≥9 gr%; (2) Appropriate LDL control: Proportion of DR patients who performed an LDL test in the previous year and had achieved LDL cholesterol ≤ 100 mg%.
Socio Economic Rank (SER)
Israel is divided into geographic sub-districts, each with a population of approximately 3,000. Israel's Central Bureau of Statistics characterizes each sub-district according to a socio-economic status ranking system based on the 1995 national census, updated in 2003. Sub-districts are ranked on a scale of 1-20, with 1 being the lowest and 20 the highest. The following variables are among those included in the SER equation: housing density, household employment, income, and education [23]. In our analysis, subjects were assigned to a sub-district according to his or her last recorded address and then ranked for socio-economic status by that sub-district.
Supplementary Voluntary Health Insurance (SVHI)
As mentioned, Israel has a universal health care system with a unified benefits package of services provided by four health plans. All health plans offer their members the opportunity to purchase SVHI for services that are not included in the basic package, such as certain types of medical devices or greater choice among the services included in the package. Every member, irrespective of health status, can purchase his or her plan's SVHI; the tariff is age-based and not related to health status. Relevant benefits that diabetic patients can utilize include, for example, reduced prices when purchasing devices for self-monitoring of blood glucose levels, reimbursement for consultations with private experts; and reduced co-payment for sophisticated new drugs. 87% of DR MHS members and 88% of all MHS members have SVHI.
Ethnicity (Arabs vs. Non Arabs)
Since personal data on ethnicity are not available to HMOs, Israeli Arabs were identified by the ethnicity of their residence in a settlement or (in large mixed cities such as Nazareth) by their sub-district, as recorded in the National Census. By using this method, we estimate that 90% of MHS's Arab members were allocated.
Immigration
Starting in 1990, Israel became the objective of a wave of immigration (about 1 million people) from countries belonging to the Former Soviet Union (FSU). These immigrants contributed to 15% growth in total population. The MHS data set contains immigration status for this group of immigrants only. Thus, under the term "Immigration," we considered only these immigrants. We should note that other immigration groups arriving during this period are relatively small in number.
Data set construction
We combined the MHS DR, Billing System, PMS, and Israeli Census into one data set using the AS400 query system.
Statistical Methods
(A) We performed an analysis of variance for prevalence as well as for process and intermediate outcome measures, between sub-groups (Gender, SERs, Ethnicity, Immigration and SVHI) using Chi-square techniques; (B) Logistic Regression Models were estimated for three dependent variables: (1) Optimal follow-up; (2) recorded HbA1C > 9 gr% and (3) achieving LDL < 100 gr.%. The independent variables were included in the models according to whether they reached the significance level (p ≤ 0.05) in the univariant analysis. For SER, we included SER 1-10 and 11-20 as two separate dummy variables in the final models although we compared four ranks in the univariant analysis. Our decision was based on the fact that in the univariant analysis, the SER categories 1-5 and 6-10 differed from the two others (11-15 and 16-20), and the preliminary models indicated robust results only for two rather than four dummies. Odds Ratios (OR) and 95% Confidence Intervals (95%CI) were calculated.