Quality of Diabetes Care Among Adult Medicaid Enrollees With Mental Disorders

A retrospective cross-sectional and descriptive analysis of quality of care for diabetes in the population of adult Medicaid enrollees in a large mid-Atlantic urban area during a one-year period (July 2009–June 2010) was conducted. Using Healthcare Effectiveness Data and Information Set (HEDIS) criteria for defining the Medicaid diabetes population, we identified individuals with diabetes who were between the ages of 18 and 75 and who had no more than one gap in Medicaid enrollment of up to 45 days during the study year and a diagnosis of diabetes alone or of diabetes and a comorbid mental disorder (13,14). Enrollees in long-term nursing home care were excluded from the analysis.

Administrative encounter data from a Medicaid managed behavioral health care database were used to identify patients with a mental disorder. Medicaid medical claims from the state Department of Human Services (DHS) were used to identify individuals in the study population with diabetes and to determine prevalence rates for treated diabetes and receipt of HEDIS measures of quality of diabetes care. Treated prevalence and receipt of quality measures were constructed with variables contained in the claims records, including diagnosis codes, type of service, and Current Procedural Terminology code for procedure type. Medicaid enrollment files from DHS were used to provide demographic characteristics, category of Medicaid assistance, and patterns of enrollment during the study year.

The sample consisted of 355,856 adult Medicaid enrollees, of whom 20% (N=72,857) were treated for a mental disorder and 80% (N=282,999) had no diagnosis of a mental disorder. The group with mental disorders was further divided into three subgroups. Enrollees with a diagnosis of schizophrenia (ICD-9 code 295) or affective disorder (code 296) were grouped as having serious mental illness; those with one or more claims with a primary diagnosis of a substance use disorder (codes 291, 292, and 303–305) coupled with one or more claims with a diagnosis of any mental disorder were placed in the group with co-occurring disorders. Enrollees with other psychiatric diagnoses were assigned to the other mental disorders group. In order to minimize the potential for misclassification, individuals with more than one primary diagnosis for a mental disorder other than a substance use disorder were classified by the ICD-9 code associated with the primary diagnosis that they had received most frequently during the study year. Enrollees with no mental health claim during the study year (comparison group) were classified as having no psychiatric diagnosis. Groups were mutually exclusive.

Diagnosis of diabetes was identified by one inpatient or one emergency claim or two face-to-face outpatient claims with a primary or secondary diagnosis of diabetes (ICD-9 codes 250, 357.2, 362.0, and 366.41) during the year prior to and during the study year. When these criteria were used, 5.9% (N=21,015) of adult Medicaid enrollees had a diagnosis of diabetes. Two study groups were created, one for patients with both a mental disorder and a diabetes diagnosis (N=7,061), and one for patients with a diabetes diagnoses and no mental disorder (N=13,954). The treated prevalence for diabetes was 9.7% among enrollees with a mental disorder compared with 4.9% among enrollees without a mental disorder (data not shown).

HEDIS quality measures were calculated for four indicators of diabetes treatment: HbA1c testing, low-density lipoprotein cholesterol (LDL-C) screening, nephropathy screening, and retinal eye exam. The percentage of patients in both groups who received each of the four HEDIS-recommended diabetes tests was calculated. A cumulative variable ranging from 0 to 4 was constructed to reflect the number of quality indicators associated with the treatment of diabetes. The median number of quality indicators for the sample (N=3) was used to represent the expected outcome.

The cross-sectional analysis compared persons with a mental disorder and diabetes with persons with diabetes only for the following variables: sociodemographic characteristics, Medicaid program category, site of diabetes treatment, number of outpatient visits, and diabetes quality measures. Chi square tests were used to compare categorical variables, and t tests were used to compare continuous variables. Tests of significance were two-tailed and were adjusted for multiple comparisons by using the Bonferroni step-down method. To determine the likelihood of having at least three HEDIS quality-of-care measures, two logistic regression models were constructed. The first model compared having any mental disorder versus no mental disorder. The second model compared having each type of mental disorder (serious mental illness, co-occurring substance use disorders, and all other mental disorders) with no mental disorder. Adjusted models controlled for age, gender, race-ethnicity, disability status, and number of outpatient visits for diabetes. All comparisons were performed with SAS, version 9.3. The study was approved by the Institutional Review Board of the University of Pennsylvania.

Descriptive statistics for the 21,015 enrollees treated for diabetes, classified by mental disorder, are shown in Table 1. Having any mental disorder was associated with higher use of primary care services for diabetes (70% compared with 55%), lower rates of hospitalization (20% compared with 24%), and a greater mean number of outpatient visits (5.5 compared with 4.4). Similar patterns were found in the quality of diabetes care, such that enrollees with mental disorders performed better on HEDIS criteria compared with enrollees without a mental disorder. These differences were especially notable for enrollees with serious mental illness. Conversely, a greater proportion of patients with a mental disorder utilized emergency services for diabetes compared with patients with diabetes only (42% compared with 30%). Use of emergency services for diabetes was highest among patients with co-occurring substance use disorders (60%) compared with patients with no mental disorders (30%), serious mental illness (36%), and other mental disorders (43%).

In the first adjusted regression model, the odds of having three or more HEDIS quality indicators for diabetes care were higher for patients with any mental disorder than for patients with no mental disorder (odds ratio [OR]=1.71, 95% confidence interval [CI]=1.59–1.85). In the second adjusted model, the odds of having three or more HEDIS quality indicators were higher for patients with serious mental illness (OR=1.78, CI=1.62–1.96), co-occurring substance use disorders (OR=1.69, CI=1.41–1.97), and other mental disorders (OR=1.64, CI=1.46–1.83), compared with patients with no mental disorder.

Among diabetes patients, the odds of receiving quality diabetes care were higher among non-Hispanic blacks (OR=1.16, CI=1.06–1.28), Hispanics (OR=1.50, CI=1.34–1.68), and patients who identified as other race-ethnicity (OR=1.67, CI=1.45–1.92), compared with non-Hispanic whites. Being disabled (OR=1.32, CI=1.23–1.41), being female (OR=1.22, CI=1.15–1.31), and having more outpatient contacts for diabetes (OR=1.32, CI=1.30–1.34) were also associated with higher odds of high-quality diabetes care.

Quality measures used in this analysis were not available on inpatient claims. Thus we may not have accounted for all quality indicators for patients in either group who were hospitalized for diabetes. To address this possibility, we removed any patient hospitalized for diabetes (N=4,732) and reran the logistic regression. Results were comparable to the initial analysis, with the group with a mental disorder having significantly higher odds of receiving three or more HEDIS measures compared with the comparison group (OR=1.58, CI=1.45–1.72). The same patterns held when we repeated the regression analysis for specific groups of mental disorders.

This study found that among Medicaid enrollees with diabetes, those with mental disorders were more likely to meet the criteria for quality of care for the treatment of diabetes compared with enrollees with no mental disorder. The concordance with diabetes care HEDIS measures was strongest for individuals with serious mental illness. Persons with mental disorders also used more outpatient and primary care services for diabetes and had fewer diabetes-related hospitalizations compared with persons with no mental disorder.

There are several possible explanations for this study’s finding of higher performance on HEDIS measures among patients with diabetes and serious mental illness, which was contrary to previous findings. One possibility is the high density of health care providers available for our study participants in the urban geographic region. Also, our sample participated in a managed medical care plan that had a designated primary care provider for medical care. A separate carve-out program for behavioral health care had a case manager for individuals with serious mental disorders. Access to a physician and coordination of care by the case manager may have enabled better opportunity for delivery of care.

The use of administrative data limited our analysis to process measures of the quality of care for diabetes. Studies utilizing outcome measures acquired from electronic health records may enhance our understanding of diabetes control among persons with serious mental illness who receive the recommended tests examined in this study.

Counter to the belief that integrating mental health and general medical care within a single organization such as a PCMH provides a vast improvement over separate carve-out systems of care for persons with comorbid mental disorders and chronic health conditions, our findings suggest that alternative models of care coordination can provide high-quality medical care, especially for patients with serious mental disorders. Thus managed care behavioral health carve-out systems may be as effective in coordinating general medical and mental health care for persons with serious mental disorders as a primary care medical care home that focuses primarily on the general medical problem. Future studies should investigate the quality of care for patients with mental disorders and diabetes, as well as other chronic medical problems, in the Medicaid sector to determine whether managed care mental health carve-out programs in other areas of the United States replicate these outcomes. Additional findings related to better HEDIS outcome measures for nonwhite and disabled individuals warrant further study as well.

The authors thank the Pennsylvania Department of Human Services Office of Medical Assistance Programs and the Philadelphia Department of Behavioral Health and Intellectual disAbility Services for facilitating access to the data used in the study.