The Role of Clinical Setting and Management Approach in Metabolic Testing Among Youths and Adults Treated With Antipsychotics

This naturalistic retrospective cohort study evaluated individuals enrolled in the fee-for-service Medicaid program in the state of Missouri from August 2008 to April 2011. Administrative health care claims data were obtained for individuals receiving an oral antipsychotic during this time frame (N=110,406). All medical and pharmacy claims during the study period were identified by using a single unique identifier for each participant. The Colorado Multiple Institution Review Board and Washington University Institutional Review Board approved this study.

A new-user cohort (N=20,980) of patients with an index prescription from August 2009 thru April 2010 was identified. New use was defined as having not received oral antipsychotic medication in the year before the index prescription. Antipsychotics included were aripiprazole, asenapine, clozapine, iloperidone, lurasidone, olanzapine, paliperidone, quetiapine, risperidone, and ziprasidone. Patients who were not Medicaid eligible for 12 months before and after their index prescription (N=5,281) or were Medicare dual eligible (N=6,226) were excluded, leaving a total sample of 9,473 patients. Patients were divided into two cohorts for analysis on the basis of their age at the time of the index prescription: youths (ages 0–18 years, N=4,271, 45%) and adults (ages 19 and older, N=5,202, 55%).

Metabolic testing was defined as any glucose or lipid test, including nonfasting tests, in the 11 months following the month of the index prescription (31 days to 365 days from baseline). Testing was indicated by a code from the Current Procedural Terminology, fourth revision, or the International Classification of Diseases, ninth revision, clinical modification.

The primary independent variable of interest, having received any care at a CMHC during the study period, was defined by claims data indicating location of provider. To evaluate whether type of care (case management or not) or setting of care (CMHC or not) affected testing rates, we created a three-level “care environment” variable. Demographic variables included age, sex, and race. Medical comorbidity, health care utilization, and medication use were ascertained from the medical and pharmacy claims data for the 12 months preceding the index prescription. Days supplied for all prescriptions for oral antipsychotics were calculated for each patient. [Additional information about the coding for type of care and other variables of interest is available as an online supplement to this report.]

Descriptive statistics were computed for each cohort overall as well as for patients with a glucose test and patients with a lipid test. Multivariable logistic regression was used to determine characteristics that were independently associated with metabolic testing. Variables with sparse distributions or those that were highly correlated with key variables were not entered into the model. Testing rates were adjusted for care environment, sex, age, race, risk conditions for cardiovascular disease (diabetes, dyslipidemia, and hypertension or heart disease for adults and only hypertension and heart disease for youths, given that diabetes or dyslipidemia affected <2% of youths), psychiatric diagnoses, concurrent psychotropic drug use, length of antipsychotic treatment (<120 days, 120–239 days, and ≥240 days), and health care utilization, as described elsewhere by our group (12). [A list of the unique mental health categories defined by the Clinical Classifications Software can be found in the online supplement.]

Analyses were run with and without the individuals who did not have a claim with a primary psychiatric diagnosis during the study period (youths, N=151, 4%; adults, N=299, 6%).

Table 1 summarizes the characteristics of the study cohorts (youths and adults) and reports glucose and lipid testing by age group, care environment, and other variables. CMHC users made up 1,539 (36.0%) of the youth sample and 1,894 (36.4%) of the adult sample. Testing rates were lower among youths than among adults. Youths and adults who received care within a CMHC setting were more likely than those who did not receive care at a CMHC to receive glucose and lipid testing. Case management did not appear to have an impact on testing rates, with the exception of lipid testing for adults—the percentage of adults who received lipid testing in a CMHC setting was eight percentage points higher among those who received case management. These results did not notably change when individuals without a primary psychiatric diagnosis were removed from the analysis.

Because the composition of the patient populations in CMHC versus non-CMHC settings may differ, we adjusted for differences in patient demographic characteristics, clinical conditions, and overall health care utilization. After this adjustment, the odds of receiving testing were higher among youths who received case management in CMHC settings (glucose, adjusted odds ratio [AOR]=1.68; lipids, AOR=2.40) and among youths who received care at a CMHC but did not receive case management (glucose, AOR=1.89; lipids, AOR=2.35), compared with youths who received no care at a CMHC. The adjusted odds of receiving testing were higher among adults who received case management in CMHC settings (glucose, AOR=1.43; lipids, AOR=1.97) and for adults who received care in a CMHC setting but did not receive case management (glucose, AOR=1.44; lipids, AOR=1.48), compared with adults who received no care at a CMHC.

Among both youths and adults, the odds of testing increased with increasing age, although the increase was not always statistically significant. The odds of testing also increased with increasing length of antipsychotic treatment.

We found that receiving care in a CMHC setting was associated with increased odds of metabolic testing among both youths and adults. There is good news overall, too, in that rates of testing were higher than those reported for an earlier time period (5,13). Nonetheless, it is important to note that significant undertesting remains, particularly among youths, even though a decade has passed since the first drug warnings and consensus recommendations were published. These results should be interpreted with caution. Specifically, adjustment for care setting and type of care received cannot fully eliminate the bias that individuals who receive care in a CMHC setting may be more likely to participate in follow-up care, including testing. To fully address the question of whether care setting and type of care affect testing rates, further randomized controlled study is necessary.

The increased odds of metabolic testing observed within the CMHC setting suggest an increased awareness of the need to test, which may be associated with the fact that Missouri DMH and Medicaid programs place greater emphasis on screening compared with other organizations. In 2012 (after the study period), the state implemented a “Health Homes” initiative for Medicaid-eligible residents with chronic diseases, including mental illness. Care managers use data analytic tracking to find and address care gaps, such as gaps in metabolic testing. Preliminary data suggest that testing rates have improved; in addition, clinical indicators and laboratory values suggest that there have also been improvements in the chronic conditions (14). Outcomes related to these specific initiatives are the subject of further investigation.

These results are subject to limitations. We defined follow-up testing as testing performed within the 11 months following the month of antipsychotic treatment initiation. This could have included testing done for reasons other than for screening related to antipsychotic treatment screening. Increased testing rates for glucose, in particular, could have been affected by the recommendation to use Hgb A1c as a diabetes-screening tool during the period of study. Fasting or random plasma blood glucose values are considered diabetes screening tests; because finger-stick glucose values could have been included in this sample, the results here do not represent true diabetes screening. Medicaid claims data can miss testing that is part of contracted bundled services, as well as finger-stick testing done by handheld devices in the office setting; in addition, results may be missed because of human error. Finally, the cohorts studied were limited to individuals with 12 months of Medicaid eligibility before and after their index antipsychotic prescription, and individuals with dual Medicaid and Medicare eligibility were excluded from the analysis, which could limit the generalizability of results.

The authors thank Michael Yingling, M.S., and Vincent Huang, B.A., for administrative assistance in development of the manuscript.