Glucose Testing for Adults Receiving Medicaid and Antipsychotics: A Population-Based Prescriber Survey on Behaviors, Attitudes, and Barriers

Improvements in prevention efforts in primary care offer the largest potential for reducing cardiovascular disease mortality among individuals with severe mental illness (1). Prevalence rates for obesity, hypertension, type 2 diabetes, and dyslipidemia are one-and-a-half to two times higher among adults with serious mental illness, yet these conditions have been historically underdiagnosed and undertreated. Moreover, the use of antipsychotic medication contributes to increased metabolic risk.

Metabolic risk information on antipsychotic medications has been widely disseminated over the past decade. Several studies of Medicaid and privately insured patients and within integrated systems of care, such as the Veterans Health Administration, have shown low rates of screening after the drug warnings. These findings triggered health system quality improvement initiatives. Diabetes screening is now a Healthcare Effectiveness Data and Information Set (HEDIS) performance measure for adults who are served by Medicaid and who have a serious mental illness and are receiving antipsychotic medication (2).

Taking a population health perspective for this study, we examined knowledge, behaviors, and attitudes regarding diabetes screening among all Missouri Medicaid providers who prescribed antipsychotic medications to adults. Population-based studies have primarily examined patient factors associated with screening; provider and system factors remain understudied. Knowledge is particularly limited about attitudes of primary care and non–behavioral health prescribers who are estimated to start therapy for at least half of all patients initiating antipsychotic treatment (3,4). The purpose of this survey was to understand underlying mechanisms and gaps affecting diabetes screening so performance improvement interventions could be better targeted.

Methods

The survey was fielded among 4,863 providers who prescribed oral second-generation antipsychotic medication (hereto referred to as “antipsychotic medication”) to patients served by Missouri Medicaid in 2011. Provider identification and addresses were obtained from Missouri Medicaid administrative data and supplemented with publicly available physician market data. Surveys were mailed according to established protocols (5) in two waves: community mental health center (CMHC) providers (late 2011–2012) and all providers (2013). CMHCs were resurveyed with a supplemental survey in 2013. For each wave, up to three survey attempts were made to each provider over the initial six-week recruitment period. A final attempt to reach nonresponders was made via fax or phone.

The survey instrument assessed a range of physician, practice, and patient factors hypothesized to independently affect metabolic screening. Attitudinal questions included screening intention, responsibility, knowledge, beliefs that screening will reduce risk (response efficacy), confidence in ordering and interpreting results (self-efficacy), and barriers to screening. Screening advocacy was assessed with the Net Promoter Score, an index used in the consumer industry to measure advocacy (6). Promoters are defined as providers who responded 9 or 10 (on a 10-point scale, with 10 being extremely likely) to the question, “How likely are you to recommend glucose testing for adults taking antipsychotics to a colleague?”

Providers were categorized into four specialty-setting groups: behavioral health in a CMHC, behavioral health (non-CMHC), primary care providers, or all others. Prescriber demographic characteristics, practice and setting characteristics, antipsychotic prescribing practices, and diabetes screening intent and advocacy were compared between specialty groups (the primary independent variable of interest) by using Pearson’s chi square test of association and adjusting for multiple comparisons. Missing values were excluded from the denominator when calculating the descriptive statistics if they represented <10% of survey responses.

The principal outcome measure was metabolic screening intent. We asked in regard to when the provider first prescribed an antipsychotic and at annual follow-up, “How likely would you be to order a blood glucose test?” Prescriber factors associated with screening intent were assessed with log-binomial regression adjusted for differences in provider, practice, and prescribing characteristics; screening attitudes; and perceived barriers to screening. To identify providers with the strongest beliefs, we dichotomized responses to survey questions as “definitely,” “agree strongly,” or “very confident” versus other responses. Secondary modeling examined factors associated with specific attitudes as the outcome measure. To evaluate survey response bias, we obtained provider and practice demographic data from ProviderPRO, a publicly available health care provider database from Healthcare Data Solutions, and compared data between survey respondents and nonrespondents.

The study received approval from the Colorado Multiple Institutional Review Board and adhered to data use agreements with the State of Missouri.

All statistical analyses used SAS software version 9.4. [The survey instrument and conceptual framework, information on respondent characteristics, analysis of respondents versus nonrespondents, and model results for attitudinal measures are available in an online supplement to this report.]

Results

The effective survey response rate was 25% (N=1,041 respondents). All survey respondents who prescribed antipsychotics to adults (N=924) were included in the descriptive analysis. The subset of respondents with complete survey responses (N=669) were included in the multivariable analysis.

Most respondents were primary care providers (499 of 924, 54%), followed by CMHC providers (156 of 924, 17%), psychiatrists in non-CMHC settings (136 of 924, 15%), and all others (133 of 924, 14%). A subset of respondents (13%, N=118) treating Missouri Medicaid patients practiced in the states bordering Missouri. A majority of respondents (74%, N=654) used an electronic medical record system, and a minority (24%, N=217) practiced in facilities offering shared mental health and general medical care.

CMHC providers were more likely to report that the majority of their patients were served by Medicaid (68%, N=106) compared with behavioral health (non-CMHC, 36%, N=49), primary care (15%, N=75), and other (25%, N=33) providers (p<.001). CMHC providers (71%, N=111) were also more likely to report that most of the patients they treated in a typical week were receiving antipsychotic medication compared with behavioral health (non-CMHC, 36%, N=49), primary care (4%, N=20), and other (11%, N=15) providers (p<.001).

With regard to screening intent, CMHC providers were more likely to report they would “definitely” order a glucose test when initiating antipsychotic therapy with an adult patient (57%, N=86) compared with primary care (39%, N=193) or other (24%, N=31) providers (p<.001). At patients’ one-year follow-up visits, CMHC providers were more likely to report that they would “definitely” order a glucose test (78%, N=119) compared with behavioral health (non-CMHC, 61%, N=83), primary care (60%, N=297), or other (31%, N=40) providers (p<.001). Most CMHC (76%, N=80) and behavioral health (non-CMHC, 62%, N=84) providers reported that they were “extremely likely” to recommend to colleagues screening for adults receiving antipsychotic medication compared with 49% (N=243) of primary care providers and 33% (N=42) of other providers who also prescribed antipsychotics (p<.001).

After controlling for provider, practice, and prescribing characteristics, analyses showed that behavioral health (non-CMHC), primary care, and other providers were less likely to “strongly agree” that all adults starting antipsychotic medication should receive baseline glucose and lipid testing compared to behavioral health (CMHC) providers (relative risk [RR]=.62, 95% confidence interval [CI]=.40–.92; RR=.58, CI=.47–74; and RR=.35, CI=.20–.59, respectively). Behavioral health (non-CMHC), primary care, and other providers were also less likely to “strongly disagree” that metabolic screening was not a priority for them or their organization compared to behavioral health (CMHC) providers (RR=.49, CI=.31–.74; RR=.67, CI=.55–85; and RR=.24, CI=.13–.42, respectively).

Table 1 reports the adjusted likelihood of diabetes screening intent. Providers who disagreed “strongly” that metabolic screening was not a priority for them or their organization were more likely than those responding otherwise (RR =1.94, p<.001) to report that they “definitely” would order a blood glucose test when prescribing an antipsychotic medication to a new or restarting patient and at a one-year follow-up visit (RR=1.74, p<.001).

After adjustment for differences in attitudes and practice characteristics, differences in the likelihood of glucose testing intent between CMHC providers and non-CMHC providers did not achieve statistical significance.

Discussion

The results of this survey indicate that Missouri Medicaid providers who prescribe antipsychotics in a CMHC setting reported greater diabetes screening intent and advocacy for their adult patients taking antipsychotic medication compared with other prescriber specialties and settings. The survey findings are consistent with analysis of Missouri Medicaid claims data which found that receiving care at a CMHC was associated with higher rates of glucose and lipid laboratory testing versus other settings, even after adjustment for differences in patient mix and receipt of care management (7).

The perceived greater organizational priority for metabolic screening observed among CMHC providers can be hypothesized to be the culmination of a series of targeted efforts by the Missouri Department of Mental Health. Missouri was the first state to take advantage of the resources and tools in the Affordable Care Act in order to set up a health home model (8). Missouri’s CMHCs serve as the health home and central source of general medical and mental health care for clients. As a result, the message that diabetes screening is necessary for adults with mental illness who are receiving antipsychotic medication has been reinforced in CMHC settings. In addition, the Missouri Department of Mental Health invested in the CMHC health home infrastructure, including the institution of audit-and-feedback systems to oversee metabolic monitoring and ensure delivery of continuing education on cardiometabolic screening to both clinicians and clinic staff, which may have further reinforced the organizational priority of screening.

The attenuation of the adjusted association between provider specialty setting and intentions to order glucose testing may be due to the inclusion of attitudes and organizational priorities related to screening that were targeted within CMHCs and in the model specification. This result suggests that the differential across provider types and setting may be due to modifiable factors affected by organizational priority setting rather than to idiosyncratic characteristics of CMHCs. Thus these results support the potential for increasing diabetes screening of mental health patients through purposeful priority and performance improvement initiatives at the Medicaid system level, not just in the CMHC setting. The recent Medicaid-specific HEDIS performance measures for diabetes screening and management for adults with serious mental illness receiving antipsychotic medication should reinforce the priority of screening regardless of health care setting.

Missouri was effectively able to target its organizational messaging, prioritization, and health home infrastructure investment toward a concentrated “market segment” of 200–250 providers operating in a finite number of mental health service areas. However, from a population health perspective, 70% of adults receiving antipsychotic medication in Missouri Medicaid do not initiate antipsychotic therapy within the CMHC health home infrastructure, according to administrative claims records. Because the challenge will be to scale up to practice settings that are somewhat dissimilar to the CMHCs studied in this initial work, gathering formative data about how the model is perceived in terms of cost, compatibility, and ease of implementation will be important before dissemination of the approach (9).

Another challenge in directing state-based performance improvement within Medicaid is that many states have reciprocating arrangements when it comes to reimbursed care. In this study, one in ten adults received at least one of their antipsychotic prescriptions from providers practicing in one of the seven bordering states. Missouri agencies have less influence on how care is prioritized, organized, and delivered outside its borders. This has implications for Medicaid performance improvement initiatives as states consider strategies to improve health metrics for all of their citizens.

One limitation of survey studies is nonresponse selection bias (5). Physician response to unsolicited surveys is known to be low and declining. A systematic review of physician survey response bias indicates that the amount of bias may be minimal (10). In this study, the effective response rate was consistent with rates observed in unsolicited physician surveys (11). Demographic differences between survey responders and nonresponders were generally small, with the exception that responders treated more patients receiving antipsychotic medication than nonresponders did. However, quality improvement efforts would likely target heavier prescribers, and so the findings provide valuable insights.

Antipsychotic prescribers within Missouri Medicaid, and the state health care system in which they practice, may not be nationally representative but can provide a model for others. The 2014 Excellence in Mental Health Act, which established criteria for certified community behavioral health clinics, was enacted to meet the needs of all Americans with serious mental illnesses and promote whole-person medical care. The survey instrument used in this study can be used by medical directors and policy makers to assess attitudes and barriers toward diabetes testing within their own states as certified clinics are implemented.

Conclusions

Significant disparities in attitudes toward diabetes screening and intention to screen were found among various prescriber specialty settings. Establishing organizational priority across all treatment settings will be important for achieving screening goals in Medicaid and reducing population-based risk.