Special Section on Implications of CATIE: Implications of CATIE for Mental Health Services Researchers

Mental health services research provides a critical bridge between efficacy demonstrated in randomized controlled trials and treatment effectiveness in real-world settings. It does so by taking into account multiple influences—for example, characteristics of service systems, consumers, and providers—that influence the effectiveness of treatments in service delivery settings ( 1 ).

One way to understand the impact of these multiple influences is by developing a program theory ( 2 , 3 ) or heuristic model ( 1 ) to guide an effectiveness evaluation from inception to interpretation. The heuristic model proposed by Shern and Evans ( 1 ) provides a framework for guiding mental health services research by identifying five domains that affect both consumer and service outcomes: consumer characteristics, such as demographic characteristics, clinical status, and functional status; organizational characteristics, such as rules that govern behavior in an organization; practitioner behavior, such as prescribing, and practitioner characteristics, including training, values, beliefs, and therapeutic orientation; services environment, such as attitudes, values, and beliefs of staff; and service system characteristics, such as external laws and regulations influencing service delivery and reimbursement.

From the perspective of this heuristic model, we discuss the implications that the findings of the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) have for monitoring and influencing the behavior of prescribers through changes in the service system that should have an impact on both consumer outcomes (for example, change in clinical status or side effect burden) and system outcomes (for example, changes in resources devoted to different prescribing patterns).

Effectiveness of antipsychotic medications

The CATIE results underscore the complexities of peoples' responses to antipsychotic medications. The effectiveness of each of the study medications differed depending upon the study phase and the participant population. Specifically, when discontinuation of assigned treatment for any reason was examined, olanzapine was superior to quetiapine and risperidone but was not superior to ziprasidone or perphenazine in phase 1 ( 4 ); however, olanzapine was equally as effective as risperidone and superior to quetiapine and ziprasidone among participants who had discontinued a second-generation antipsychotic ( 5 ) and equally as effective as quetiapine and superior to risperidone among participants who had discontinued perphenazine ( 6 ). Differences between agents in phase 1 were attenuated among those who used illicit substances ( 7 ). In addition, when the analysis excluded data from people who were randomly assigned to the same medication that they were already taking at study entry, although the pattern of findings did not change, the groups were no longer significantly different ( 8 ). When clozapine was a treatment option, it was superior to olanzapine, risperidone, and quetiapine for participants who failed to improve on a second-generation antipsychotic in phase 1 ( 9 ).

This variety in outcomes makes clear the need to match prescribing decisions with clinical and medical history ( 10 , 11 ) and a person's preference regarding the relative importance of side effects versus symptom control. A prescriber using such an approach would naturally prescribe a variety of agents across his or her caseload, including first-generation antipsychotic medications, such as perphenazine, that may be equally effective for some people and significantly cheaper because of their lower cost ( 12 ). However, what the distribution of various antipsychotics should be within a prescriber's caseload or within a system or whether some common practices, such as prescribing two or more antipsychotic medications, are reasonable are issues that are not informed by CATIE.

Given the complexity and variability of response to antipsychotic medications in CATIE ( 4 , 5 , 6 , 7 , 8 , 9 ), collaborative efforts between mental health services researchers and service systems to change prescriber behavior should not result in restricted formularies ( 13 ). Instead, staff time that would be consumed by monitoring and enforcing such "one size fits all" service system rules could be reallocated to support analysis of secondary data to identify questionable practices, such as underuse of clozapine at a clinic, where a clinical review and possibly targeted interventions might support practice change. Using such a process in lieu of restricting formularies has been suggested by others as well ( 14 ), with the most effective prescriber interventions focusing on training needs, areas of strength, and problematic organizational support processes rather than on individual punishments or awards ( 3 ).

Furthermore, the CATIE results suggest that some existing quality measures for antipsychotic medications need to be revisited. For example, all states are currently participating in the Uniform Reporting System of the Substance Abuse and Mental Health Services Administration. The system includes only one measure related to psychotropic prescribing—the proportion of a population prescribed a second-generation antipsychotic medication. A larger proportion is considered to reflect better quality of care. However, given that perphenazine performed well in phase 1 of CATIE ( 4 ), a large proportion of people taking second-generation antipsychotics may no longer be, and perhaps never was, a useful proxy for better-quality care. Measures need to evolve to keep pace with new findings, and before such measures are adopted as new quality indicators, mental health services researchers can help by testing the ability of the measures to monitor change in one or more of the five domains described above and to improve consumer outcomes. The step linking improvements in consumer outcomes to quality indicators is particularly important given that some studies, such as an examination of adolescents receiving mental health outpatient services ( 15 ), have demonstrated weak and inconsistent relationships between quality indicators and client outcomes.

Cardiac and metabolic impact of antipsychotics

The CATIE findings have also expanded our understanding of the prevalence of cardiac and metabolic disorders and the potential impact of antipsychotic medications on these conditions. As in previous studies, significant numbers of CATIE participants with schizophrenia had abnormal lipid levels (48%), hypertension (33%), and diabetes (10%) at study entry ( 16 ). In addition, compared with a matched sample in the National Health and Nutrition Examination Survey (NHANES), men in the CATIE study were 85% more likely and women were 140% more likely to have metabolic syndrome, even after the analysis controlled for demographic characteristics and body mass index (BMI) ( 17 ). Similarly, in another study that controlled for BMI and compared participants with a matched NHANES sample, the ten-year risk for coronary heart disease was significantly elevated among male participants (9.4% versus 7.0%) and among female participants (6.3% versus 4.2%) ( 18 ). Despite the high prevalence of serious medical conditions among CATIE participants with schizophrenia, 88% of those with abnormal lipid levels, 62% of those with hypertension, and 30% of those with diabetes were not receiving treatment for their illness at study entry ( 16 ).

The same secondary data approach used to identify questionable prescribing could also help identify whether prescribers are providing appropriate screening (such as regular glucose and lipid checks) and treatment (such as medications for diabetes or hypertension) for cardiac and metabolic conditions in this vulnerable cohort. In fact, one recent example using secondary data to monitor quality in a Veterans Affairs facility found that the frequency of follow-up lipid monitoring for people with schizophrenia who were receiving a second-generation antipsychotic increased after an abnormal total cholesterol level, although the median time to follow-up (about ten months) was less than desirable for clinical care ( 19 ). Mental health services researchers can collaborate with service systems to help identify practitioners who are not providing appropriate screening or treatment and to develop appropriate interventions to increase good clinical care.

As part of these possible interventions, the effect that different antipsychotic medications have on weight, blood glucose, and lipid measures should be considered when various treatment options are weighed ( 20 ). Complementing a growing body of evidence and as summarized by others ( 20 ), the CATIE results showed that people assigned to olanzapine had greater weight gain and increases in measures of glucose and lipid metabolism in both phase 1 ( 4 ) and phase 2 ( 5 ). Over time, the prevalence of metabolic syndrome changed significantly among those taking olanzapine (an increase from 35% to 44%) compared with those taking ziprasidone (a decrease from 38% to 30%), whereas the rates remained relatively level for risperidone, quetiapine, and perphenazine ( 21 ). Newcomer ( 20 ) provided a summary of the CATIE evidence coupled with results from other studies, and he concluded that olanzapine and clozapine can produce substantial increases in weight and other associated metabolic disturbances. In comparison, risperidone and quetiapine have an intermediate effect on weight and a less clear impact on other metabolic disturbances, whereas aripiprazole and ziprasidone are considered low risk of weight gain and other adverse metabolic changes ( 20 ).

Given this mounting evidence, mental health services researchers can help stakeholders develop approaches to support the use of effective ( 4 , 5 , 6 , 9 ) but higher-risk ( 4 , 5 , 20 ) antipsychotic medications. Prescriber and consumer education, support for metabolic monitoring, and shared decision making are critical components in the selection of an antipsychotic medication. In the face of new findings, prescribers and consumers need to be made aware of the relative risks of different antipsychotic medications and strategies for managing metabolic side effects, including the consideration of a lower-risk medication when behavioral or medical interventions have failed to reduce metabolic disturbances. Systems for identifying individuals at risk and encouraging behavior change among prescribers need to be developed. For example, timely secondary data can be used to track or prompt appropriate laboratory monitoring or to identify individuals in need of a clinical review. For individuals at risk of developing the metabolic syndrome, e-prescribing systems can identify concerns about quality of care before initiation of treatment.

Treatment algorithms and decision support software can be used to support a shared decision-making process between patients and prescribers to help them weigh the risks and benefits of different treatment options. For example, on the basis of the CATIE findings, prescribers may have different views on whether olanzapine should be a first-line agent, because of its relative efficacy, or a second-tier agent, in consideration of its relatively high risk of metabolic abnormalities and high cost. Individual history, status, and preferences should ideally drive these choices. Services researchers have developed shared decision-making programs that increase the concordance between an individual's preference and the treatment that is delivered ( 22 , 23 ).

Secondary data sources

To measure and target change in behavioral care settings, the data collected must involve naturalistic, longitudinal information that is collected as a standard part of the ongoing process of providing mental health care ( 3 ). Claims data, such as Medicaid claims data, have been used successfully to characterize current prescriber behavior, such as prescribing and health monitoring practices, and consumer outcomes, such as medication adherence, service use, and health status, to identify problems in the quality of care and to monitor the impact of interventions. Given that policy governing mental health service delivery has moved largely from single state agencies to public health agencies, social insurance programs, private health insurance programs, and sometimes judicial agencies ( 24 ), it is increasingly important to compile data from multiple sources and agencies to get an accurate picture of all of the services received by individual clients and the quality of the public mental health system.

The challenge for mental health services researchers is to work with mental health delivery settings to develop systems that move beyond simply identifying less than optimal practices by including strategies to modify practice and measure outcomes, continuing the cycle of identification, modification, and monitoring until the desired change is achieved ( 3 ). This necessitates monitoring secondary data in a concurrent and continuing manner so that interventions to improve clinical care can happen on a real-time basis. Secondary data compiled from several agencies and sources have the potential to offer information about both consumer and system outcomes for interventions across many domains, particularly when data about critical outcomes, such as side effects (for example, weight gain) and functional status (for example, employment status), are included. Thus mental health services researchers would be wise to pool their resources, share the burden of compiling these large secondary data sources, and encourage systems to routinely collect information about additional important outcomes in order to create continuous data sources that will inform a variety of interventions and outcomes.

Conclusions

In this commentary we have focused on the implications of the CATIE findings from the perspective of one factor—prescriber behavior—in one heuristic model. The CATIE findings present opportunities for additional mental health services research from other frameworks and within other important domains, including areas such as neurocognition ( 25 ), psychosocial functioning ( 26 ), and vocational outcome ( 27 ), where the assigned treatment in CATIE did not have an impact on outcome. Indeed, some have begun to examine other areas among the CATIE participants, such as the clinical correlates of substance use ( 28 ) and the impact of substance use on psychosocial functioning ( 29 ). CATIE provides a rich data source for mental health services researchers that will continue to inform the field for years to come.

Mental health services researchers should embrace helping mental health delivery systems to incorporate results from trials such as CATIE into practice. Alone, few mental health care delivery systems have the resources to sort through and consider whether or how to incorporate new information into practice and how to balance these decisions among the sometimes competing interests of various stakeholders, including, but not limited to, service system administrators, mental health care providers, consumers, and families of consumers. Yet, objective, multiple-perspective approaches involving concurrent and continuous data collection, such as those proposed above, are critical in promoting the evolution of mental health services ( 30 ). The tradeoff of shifting from an outside expert with a specific research agenda to an involved collaborator in a public-academic collaboration supporting the needs of multiple stakeholders ( 3 ) is well worth the gain in the opportunity to have an impact on policy and improve the quality of care in large systems.

Acknowledgments and disclosures

CATIE was supported by grant N01-MH-90001 from the National Institute of Mental Health (NIMH). A list of study locations and principal investigators can be found at www.catie.unc.edu/schizophrenia/locations.html. Preparation of this article was funded in part by grant R01-MH-59312 from NIMH to Dr. Essock, principal investigator; by grant R03-MH-71663 from NIMH to Dr. Covell, principal investigator; and by the State of New York Office of Mental Health. This article does not express the views of the Connecticut Department of Mental Health and Addiction Services or the State of Connecticut. The views and opinions expressed are those of the authors.

The authors report no competing interests.