Management of New Hyperglycemia in Patients Prescribed Antipsychotics

This retrospective cohort analysis included veterans who were receiving antipsychotic treatment in 32 VA facilities in the western states, Alaska, and Hawaii. Pharmacy, laboratory, service use, and diagnostic data were extracted from Veterans Health Administration Medical SAS Data Sets, and vital signs were extracted from the regional data warehouse. This project was approved by the Central Arkansas Veterans Healthcare System Institutional Review Board and Research and Development Committee.

Patients were included if they had a new antipsychotic prescription (index prescription) that was prescribed between October 1, 2005, and September 30, 2011, and that had not been prescribed in the preceding 180 days. If more than one new antipsychotic prescription was identified in the study period, we selected the most recent occurrence. We used the methods of Mittal and colleagues (7), including patients with a stable antipsychotic regimen before the index prescription, at least a 60-day supply in the subsequent 90 days, and no recent hospital, nursing home, or residential program stays. We also required at least two clinic visits (one primary care) in the year preceding the index date. We excluded patients with a diagnosis of diabetes or prediabetes (ICD-9 codes 249.xx, 250.xx, and 790.29), prescription of a medication for diabetes (insulin or oral hypoglycemic agent), or impaired glucose tolerance (plasma glucose ≥140 mg/dL or hemoglobin A1c ≥5.7%) (10) in the year before the index date to allow for sufficient time to identify preexisting hyperglycemia. Finally, we selected veterans who had evidence of clinically significant hyperglycemia (plasma glucose ≥200 mg/dL or hemoglobin A1c ≥6.5%) (10,11) between day 1 and day 180 after the index prescription date and examined metabolic management practices after the first occurrence of hyperglycemia. The threshold for clinically significant hyperglycemia in need of intervention was set at 200 mg/dL because VA laboratory data do not reliably indicate whether glucose tests were drawn in the fasting state. [A patient selection flow chart is available in an online data supplement to this report.]

Of 54,908 veterans with a new antipsychotic prescription, 33,403 did not have evidence of impaired glucose tolerance in the prior year. Of these, 19,290 had glucose or hemoglobin A1c tests in the 180 days following the index prescription. The final sample included 403 veterans whose testing indicated new clinically significant hyperglycemia after beginning an antipsychotic. For 294 veterans (73%), the index prescription was the first antipsychotic prescription in the previous 365 days and likely represented a new start of antipsychotic treatment; for the 109 veterans (27%) who had an antipsychotic prescription in the previous year, the index prescription may indicate a switch to, or addition of, a new antipsychotic.

Management was defined as any of the following actions occurring within 30 days after the abnormal result: a primary care encounter; a MOVE! behavioral weight management program encounter; a nutrition encounter; a diabetes clinic encounter; a new prescription for an oral hypoglycemic agent or insulin; or the addition of an antipsychotic with a lower risk of weight gain. The addition of a lower-risk antipsychotic likely indicates the provider’s intent to decrease metabolic risk after the abnormal test. Medications were classified according to risk on the basis of the joint statement of the American Diabetes Association and others (12); comprehensive review articles (13,14); and results of the Clinical Antipsychotic Trials of Intervention Effectiveness study (15): high (clozapine and olanzapine), medium (quetiapine, risperidone, chlorpromazine, thioridazine, loxapine, perphenazine, paliperidone, thiothixene, and trifluoperazine), or low (aripiprazole, ziprasidone, haloperidol, fluphenazine, molindone, pimozide, and mesoridazine). In addition, we created variables to indicate whether the management actions occurred within a 60-day window.

Frequencies of each type of action were calculated. Multivariable logistic regression models were used to examine the association of patient characteristics with the likelihood of receiving recommended management within 30 days. Patient characteristics examined include age, gender, preexisting comorbidities (obesity, dyslipidemia, hypertension, and heart diseases) within 360 days before or on the index prescription date and any psychiatric diagnosis during the 360 days before or 180 days after the index prescription date. Adjusted odds ratios and 95% confidence intervals are reported. A sensitivity analysis was conducted in which we defined management as any of the actions occurring within 60 days after the abnormal result.

All analyses were conducted using SAS, version 9.1, and p values of ≤.05 were considered statistically significant.

Of the 403 veterans in the final sample, 378 (94%) were male, and 268 (67%) were white. The mean±SD age of veterans who received any management (N=254) was 56.3±12.0; for those who did not receive management (N=149), it was 60.1±13.8. The two groups did not differ significantly by race, gender, or marital status. Diagnoses in the full sample were as follows: psychiatric diagnosis other than schizophrenia, bipolar disorder, or other psychotic disorder, 50% (N=203); bipolar disorder, 19% (N=78); schizophrenia, 17% (N=68); and other psychotic disorder, 12% (N=49). Five veterans (1%) had no psychiatric diagnosis. The prevalence of metabolic comorbidities included obesity (N=168, 42%), dyslipidemia (N=192, 48%), hypertension (N=202, 50%), and heart disease (N=42, 10%). Most veterans were prescribed antipsychotics with a medium risk of weight gain (N=275, 68%). Among veterans who did not receive management, 9% (N=14) were prescribed high-risk medications, 71% (N=106) medium-risk medications, and 19% (N=29) low-risk medications. Among those who received management, 7% (N=18) were prescribed high-risk medications, 67% (N=169) medium-risk medications, and 26% (N=67) low-risk medications. [A table presenting these and other results is included in the online data supplement].

Overall, 63% of patients (N=254) received at least one type of management action within 30 days of receiving an abnormal result. A primary care visit was most common; 55% (N=222) had a visit within 30 days of the abnormal result. In a sensitivity analysis using a 60-day window, 73% (N=294) received at least one management action. [A figure presenting results for the 30- and 60-day windows is included in the online data supplement.]

Table 1 presents the adjusted odds ratios for the association of demographic and clinical variables with management actions. Overall, as age increased, patients were less likely to receive timely management. In addition, females were less likely than males to receive any management. The presence of a preexisting comorbidity inconsistently influenced management practices. Patients with preexisting dyslipidemia were more likely to have a primary care visit, and patients with hypertension were more likely to receive a nutrition or weight management consult. Preexisting obesity or heart disease or having a diagnosis of bipolar disorder did not influence management practices.

To our knowledge, this is the first report of the extent to which individuals who are prescribed a “new” antipsychotic agent and who develop metabolic side effects receive specific management of those effects. Although most patients with evidence of impaired glucose tolerance received some management within 30 days, this consisted mostly of primary care encounters. Further data collection (for example, medical record review) would be necessary to determine whether such visits were scheduled specifically to further assess and manage hyperglycemia and whether the primary care provider was aware of the abnormal result.

Fewer than 8% of patients received a nutrition encounter, and fewer than 2% had a weight management intervention encounter within 30 days. The reasons are unclear; patients or providers could face barriers to these management interventions. Diabetes clinic encounters and antipsychotic medication changes were also infrequent. Again, the reasons for these findings are not clear. Primary care providers may have attempted to provide initial care, such as nutrition and exercise counseling, before sending the patient to a subspecialty clinic. Further, providers have reported a reluctance to change a medication regimen that is effective for psychiatric symptoms, and they perceive that patients are also reluctant (unpublished data, Smith JL, Owen RR, Pope SK, et al., 2014).

In this study, no management action was documented within 30 days for 37% of patients. Further research is needed to determine the extent to which such management actually occurs but is not recorded in administrative data. Nevertheless, these findings suggest that there is room for improvement in management of treatment-emergent metabolic side effects, particularly in terms of referring eligible patients to the MOVE! program and nutrition services. In addition, further research is needed on provider and patient decision-making considerations in regard to changes in antipsychotic medications in the face of significant metabolic side effects.

Of interest, rates of management were not greatly improved when the analysis used a more “lenient” 60-day window for a management action. Although an additional 10.5% of the sample had a primary care encounter in this window, the difference in the proportions of other management actions was 3.5% or less [see online data supplement]).These results suggest that some patients with side effects may experience significant delays from detection of hyperglycemia to initiation of an appropriate management action.

This study examined the extent to which detection of impaired glucose tolerance was followed by certain evidence-based actions. However, as with any analysis using administrative databases, the specific reasons for lack of management could not be determined, nor could we determine whether the actions that did occur were the direct result of the abnormal test. It is possible that some action was taken without consideration of the antipsychotic treatment. In addition, the lack of encounters could reflect either the provider’s inaction or lack of attendance by the patient. Finally, although this study is the largest to date to examine management practices after a test indicating clinically significant hyperglycemia among patients prescribed antipsychotics, the sample size was relatively small.

In the first comprehensive study of management of impaired glucose tolerance during a new antipsychotic treatment episode, most patients received some management action; just over half visited primary care clinics within 30 days. Although these results are encouraging, it is concerning that specific interventions were rarely documented. Further research is needed to investigate reasons for lack of attendance at weight management and nutrition programs, diabetes clinic visits, and changes in antipsychotic medication to reduce metabolic risk. Medical record abstraction may discover additional management actions that were not identified by using administrative data and the extent to which lack of hyperglycemia management is caused by patient, provider, or system factors. Finally, although this analysis focused on management of hyperglycemia among patients prescribed a new antipsychotic, most patients in this data set were not monitored for hyperglycemia. As suggested by Mittal and colleagues (7), continued research into ways to improve monitoring of metabolic side effects of antipsychotics is an important step in improving the management of abnormal results.

This study was supported by grant SHP 08-146 from the Health Services Research and Development Service (HSR&D), Office of Research and Development, U.S. Department of Veterans Affairs (VA), and the VA Mental Health Quality Enhancement Research Initiative. Dr. Viverito is supported by a VA HSR&D fellowship at the Center for Mental Healthcare and Outcomes Research, Central Arkansas Veterans Healthcare System. The authors thank Cynthia Brace, B.S., for assistance. The contents do not necessarily represent the views of the VA or of the U.S. government.

Dr. Li is a consultant for eMaxHealth Systems. The other authors report no competing interests.