Adherence to Oral Diabetes Medications Among Users and Nonusers of Antipsychotic Medication

Individuals with serious mental illnesses represent a high-risk subgroup of patients who have complex treatment requirements. Several studies have shown that patients with mental illnesses also experience cardiometabolic conditions, such as diabetes, hypertension, and hyperlipidemia (1–3). Patients with schizophrenia or other mental illnesses have a higher incidence of diabetes compared with the general population (4–6). A population-based longitudinal study found that the age- and sex-adjusted point prevalence of diabetes was 9.3% among patients with mental illnesses and 5.1% among those without mental illnesses (7). Citrome and colleagues (8) found that the prevalence of diabetes among inpatients at a large state psychiatric hospital system increased from 6.9% in 1997 to 14.5% in 2004.

Diabetes is a chronic medical condition with a prevalence of 8.3% in the United States; in 2007, the estimated annual total cost (direct medical costs and indirect costs) was $174 billion (9). Patients diagnosed as having diabetes are typically prescribed insulin, oral diabetes medications, or both and are instructed to modify their diet and lifestyle to control their blood glucose levels. Optimal diabetes medication adherence is associated with positive outcomes, including improved blood glucose control, fewer hospitalizations, and lower costs (10–17). Patients with mental illnesses have an increased risk of negative health outcomes and mortality. This may be due to several predisposing risk factors, such as obesity and smoking, but may also include poor medication adherence (18). Antipsychotic medications are the cornerstone of treatment for schizophrenia and other psychotic illness.

Several studies have demonstrated poor adherence to antipsychotic medications among patients with schizophrenia; however, there is limited information on adherence to medications for other comorbid conditions among this group of patients (19–21). Adherence to both antipsychotics and medications for non-mental health–related conditions among patients with mental illnesses should be monitored in order to improve outcomes (22).

Few studies have examined adherence to medications for comorbid conditions among patients taking antipsychotics. Nelson and others (23) evaluated adherence to antihyperlipidemic medication and lipid control among diabetic Veterans Affairs (VA) patients with and without psychiatric illnesses. Although there was no difference between the two groups in total mean±SD cholesterol (167.1±42.8 and 167.6±39.4 for patients with and without psychiatric illnesses, respectively), patients who had psychotic illnesses were without antihyperlipidemic therapy for fewer days than patients without psychotic diseases (44 and 62 days, respectively). In other words, medication adherence was significantly higher (p<.05) among patients with psychotic conditions. A study by Dolder and others (24) evaluated antihypertensive medication adherence among patients with and without psychotic illnesses. Adherence to antihypertensive medication was similar between the two groups of patients. However, blood pressure was less likely to be under control among patients with psychotic illnesses.

Some studies have specifically evaluated adherence to diabetes medications among patients using antipsychotics. In a study of VA outpatients, Dolder and colleagues (25) observed that prescriptions for oral antidiabetics among patients who had psychotic conditions and chronic medical conditions were filled on schedule only 52% of the time. Piette and others (26) found that VA patients with schizophrenia and comorbid hypertension and diabetes were more likely to be nonadherent (medication possession ratio [MPR] <.80) to antipsychotic medications than to antihypertensive and hypoglycemic medications (35% versus 26% and 29% of patients, respectively).

To our knowledge, only one published study has compared adherence to oral antidiabetics among patients with and without psychiatric diseases. Kreyenbuhl and others (27) examined adherence to oral antidiabetics for type 2 diabetes among VA patients with and without schizophrenia. Patients with schizophrenia were more likely than those without schizophrenia to be adherent.

The objective of the study was to compare adherence to oral antidiabetics among users and nonusers of antipsychotic medications in the Texas Medicaid population. In addition, adherence to oral antidiabetics and antipsychotics among users of antipsychotics was compared.

Methods

This study was a retrospective analysis of prescription claims data from the Texas Medicaid data set between July 1, 2008, and December 31, 2011. It was approved by the Institutional Review Board of the University of Texas at Austin. The index date was defined as the date of a first prescription claim for an oral diabetes medication. The preindex period was defined as the six-month period before the index claim, and patients were followed for 12 months after the index date.

Prescription claims data were extracted and analyzed for patients who met the following criteria: age between 18 and 64 years on the index date, at least two claims for oral antidiabetics within the index period (January 1, 2009, to December 31, 2010), no claims for oral antidiabetics during the preindex period, and continuous enrollment during the preindex and postindex periods. Antipsychotic medication users were identified as patients who had at least two claims for oral antipsychotic medication in both the preindex and postindex periods.

Because there may have been differences in severity of illness among the antipsychotic users and nonusers, we used a caliper-matching algorithm to match patients in the two groups on the basis of their chronic disease score (CDS) in the postindex period (28). The CDS is a measure of disease severity that was developed by Von Korff and others (29). It uses pharmacy claims data to determine the number of selected prescription medications used by the patient. Because diabetes medications may be dispensed differently among antipsychotic users and nonusers, claims data were used to categorize patients in each group by mean days' supply of diabetes medication (1–30 days, 31–60 days, and >60 days).

Outcomes

Adherence to oral diabetes medication was calculated for all patients during the 12-month postindex period. Adherence to antipsychotic medications among antipsychotic users in the 12-month postindex period was also evaluated. Proportion of days covered (PDC), the adherence measurement for oral antidiabetics and antipsychotics, was calculated by dividing the number of days in which at least one study medication (oral antidiabetic or oral antipsychotic) was available by the number of days in the study period (N=365). Persistence was defined as the number of days of continuous access to oral antidiabetics, allowing for a 30-day gap period.

Statistical analyses

Means±SDs, frequencies, and percentages were used to describe the sample. Because antipsychotic users and nonusers were matched by using CDS, a paired-sample t test was used to analyze the relationship between adherence to oral antidiabetics and use of antipsychotics. A paired-sample t test was also used to analyze the relationship between adherence to antipsychotics and oral antidiabetics among antipsychotic users. Comparison of the two groups on categorical variables was done by using McNemar’s tests. Logistic regression procedures were used to find predictors of adherence (PDC ≥.80) to oral antidiabetics. For all statistical tests, assumptions were tested and met before analysis. SAS, version 9.2, was used for analyses, and the alpha level was set at <.05.

Results

A total of 18,999 patients met the study’s inclusion criteria. [A table describing the execution of the inclusion criteria is available online as a data supplement to this article.] The mean age was 46.4±12.5 years, and 66% (N=12,539) were female. Of the 18,999 patients, 1,956 (10.3%) were antipsychotic users. Antipsychotic users and nonusers were matched by the caliper matching algorithm, resulting in two groups of 1,821 patients each. The mean CDS for both groups was 5.3±2.8. For oral antidiabetics, the overall mean PDC was .59±.29, and mean persistence was 163.4±121.0 days.

Table 1 shows the characteristics of the CDS-matched groups of antipsychotic users and nonusers. The proportion of women was higher than that of men in each group, and the gender proportions between patient cohorts were significantly different (p<.001). Furthermore, the mean PDC for oral antidiabetics was significantly higher among antipsychotic users than nonusers (.63±.29 versus .55±.28, p<.001). A similar trend was observed for mean persistence (180.8±124.8 days for users versus 146.0±114.5 days for nonusers, p<.001). Adherence to oral antidiabetics was significantly greater among antipsychotic users (37.2%) versus nonusers (24.0%).

Logistic regression analysis showed that being prescribed antipsychotic medications was significantly and positively related to adherence to oral antidiabetics (p<.001), after holding other variables constant (Table 2). Antipsychotic users were 2.10 times more likely than nonusers to be adherent to oral antidiabetics (χ²=97.63, df=1, p<.001). Females were 23% less likely than males to be adherent to oral antidiabetics. In addition, older age (odds ratio [OR]=1.02) and a higher CDS (OR=1.09) were associated with adherence to oral antidiabetics (p<.001). Patients with a mean days’ supply per prescription fill between 1 and 30 days or between 31 and 60 days were less likely—by 50% and 32%, respectively—to be adherent to oral antidiabetics than patients with a mean days’ supply greater than 60 days.

Among antipsychotic users, the mean PDC was higher for antipsychotics than for oral antidiabetics (.78±.25 versus .63±.29, t=−20.76, df=1,820, p<.001). About 62% (N=1,132) of these patients were considered adherent to their antipsychotic medications.

In the primary analyses, antipsychotic users were defined as patients who had at least two prescription claims for antipsychotics in the preindex and postindex periods. This may lead to the selection of patients who are highly adherent. Thus, as a sensitivity analysis, antipsychotic users were defined as patients who had at least one claim for an antipsychotic during both the preindex and postindex periods. We identified 2,610 patients who were users and 17,043 patients who were nonusers of antipsychotics. After matching, there were 2,430 patients in each group. [A table comparing the characteristics of the two groups is available online in the data supplement.] The users of antipsychotics had higher mean PDC (.61±.29 versus .54±.28) and mean persistence (172.1±123.6 versus 144.8±114.4 days) for antidiabetics than nonusers. The logistic regression procedure demonstrated that after analyses adjusted for all other factors, antipsychotic users were 1.82 times more likely than nonusers to be adherent to antidiabetics. [The results of the logistic regression analysis are available online in the data supplement.] Among antipsychotic users, the mean PDC for antipsychotics was higher than for oral antidiabetics (.72±.75 versus .60±.62, t=−18.89, df=2,429, p<.001). About 54% (N=1,314) of antipsychotic users were adherent to their antipsychotic medications.

Discussion

To the investigators’ knowledge, this was the first study that compared adherence to oral diabetes medications among users and nonusers of antipsychotics in a Medicaid population. In the past, studies have compared adherence to oral antidiabetics among patients with and without schizophrenia (27). In this study, we included patients with antipsychotic medication claims, given that some second-generation antipsychotics have been associated with increased risk of diabetes (30). The mean PDC for oral antidiabetics was higher among antipsychotic users (.63±.29) than nonusers (.55±.28). Adherence to oral antidiabetics has been linked with better outcomes for patients with diabetes, including fewer hospitalizations and better glycemic control (10–17). Determining whether better adherence among antipsychotic users was associated with better glycemic control was outside the scope of our study, However, the higher adherence scores for this group of patients adds to the literature by suggesting that the coexistence of a mental illness does not necessarily compromise intermediate outcomes of care associated with better glycemic control. In fact, such outcomes may be better among patients with diabetes and comorbid mental illnesses versus those without mental illnesses (31–34).

The proportion of patients who were nonadherent to oral antidiabetics was 86% (N=1,348) for antipsychotic nonusers and 63% (N=1,143) for antipsychotic users. This was similar to the trend observed by Kreyenbuhl and others (27) with VA data. They found that patients with diabetes and comorbid schizophrenia were significantly less likely to be nonadherent (MPR >.80) than patients with only diabetes (43% versus 52%, respectively). The investigators calculated a weighted MPR for patients who were using more than one oral diabetes medication (27). This study measured adherence by using PDC, which is a more conservative measure of adherence, given that it avoids overestimation. The use of a different measure of adherence and a different population could possibly explain the higher rates of nonadherence reported in our study compared with those reported by Kreyenbuhl and others (27).

Kreyenbuhl and others (35) conducted another study by using self-report data and concluded that significantly more patients with mental illnesses (compared with those without mental illnesses) reported adherence to oral diabetes medications. This trend of better adherence to diabetes treatment among patients with comorbid psychiatric-related diseases has also been observed for other illnesses. A study by Himelhoch and others (36) of HIV patients found that those with serious mental illnesses were less likely than those without serious mental illnesses to discontinue their HIV medications. Better adherence to oral diabetes medications among younger patients (compared with older patients) and among males (compared with females) has also been reported in previous studies (20,37). In this study, patients with shorter days’ supply of oral diabetes prescriptions were less likely to be adherent than patients with longer days’ supply. Although shorter days’ supply promotes frequent assessments and reduces issues associated with undetected adverse events, it increases the burden on the patients to arrange the visit to the clinic to get the prescription, transportation to the pharmacy, and other associated administrative tasks (26). This could be a reason for the observed poor adherence among patients with shorter days’ supply.

Another study reported that patients with psychiatric illness were more likely than their peers without psychosis to adhere to their antihyperlipidemic medications (23). An explanation of these results may lie in the fact that the presence of comorbid diseases is an indication of disease severity; hence, both health care providers and informal care givers, such as family members, pay more attention to the management of the general medical conditions of patients with comorbid conditions. Perhaps the patients with psychiatric conditions themselves play a more active role in medication management than do their counterparts. This improved awareness could lead to better management of other chronic conditions, such as diabetes, which, like mental illnesses, requires self-care to improve short- and long-term outcomes. Although the presence of serious mental illnesses may be detrimental to various aspects of life, the continued self-management of chronic conditions might enhance the management of other co-occurring conditions. Patients with psychiatric conditions may be living in group homes and also could have informal caregivers overseeing their medication management, which could be another reason for higher adherence in this group (38).

Antipsychotic users were significantly more adherent to antipsychotic medications than to oral antidiabetics (.78±.25 versus .63±.29). Nevertheless, adherence to both oral diabetes and antipsychotic medications was suboptimal. About 62% of patients were adherent to their antipsychotic medications, and only about 37% were adherent to diabetes medications. The adherence rate for antipsychotic medications conformed to the rate reported by Piette and others (65%), but adherence to oral antidiabetics was much lower than the rate (71%) reported by Piette et al. (26). A possible reason for our lower estimates is our use of PDC versus MPR as a measure of adherence.

This study had several limitations. Because of the unavailability of International Classification of Diseases (ICD-9) codes, we used the presence of two or more prescription claims for oral antidiabetics and antipsychotics as proxies for the presence of diabetes and psychosis, respectively. We did not have data about several patient characteristics, such as socioeconomic status, literacy level, and psychosocial factors, which might be significant predictors of adherence to either or both classes of medications. This study has limited generalizability beyond a Medicaid population. Finally, we cannot be sure that patients actually took their medications; we only know that they refilled their prescriptions.

Conclusions

Adherence to oral antidiabetics and antipsychotics was generally suboptimal in the Texas Medicaid population. Patients prescribed both oral antidiabetics and antipsychotics were more adherent to oral antidiabetics than patients prescribed only oral antidiabetics. In the subset of people prescribed medications from both classes, adherence was higher for antipsychotics than for oral antidiabetics. The low adherence rates highlight the need for interventions to help improve medication management for patients. Assigning patients with chronic diseases to case managers and highlighting the importance of medication therapy might help improve adherence rates. Future research could focus on conducting similar studies in diverse patient populations and using general medical data, including diagnoses codes, to validate these results.