A Medicaid Quality Improvement Collaborative on Psychotropic Medication Prescribing for Children

An advisory panel of behavioral health experts identified the objectives described above and developed evidence-supported antipsychotic treatment algorithms and online modules, fact sheets, and shared decision-making tools for prescribers, school and agency personnel, parents, and youths. These resources were pilot-tested in three regional community collaboratives (including 13 counties, 25 stakeholder organizations, and nearly 120 prescribers) identified on the basis of high-volume antipsychotic prescribing patterns and maximal opportunity for population impact for youths in Medicaid and foster care. In each region, two distinct groups were identified. In group 1 (N=8,222 youths with a filled prescription for a psychotropic medication), a high volume of youths were served by general pediatricians in large-volume practices and tertiary children’s hospitals and a few high-volume prescribers from community behavioral health agencies. Group 2 (N=3,412 youths with a filled prescription for a psychotropic medication) had a high volume of psychotropic medication use prescribed mostly by psychiatrists in community mental health and residential treatment centers (and a few general pediatric providers) working with youths with very complex conditions. Each regional collaborative was organized as a learning community, with multidisciplinary representation from clinical, educational, child welfare, and other social service entities.

After consensus standards were disseminated, participating clinicians agreed to receive monthly feedback on individual prescriber claims data compared with recommended prescribing practice. Pharmacy reports included the child’s age, gender, and length of antipsychotic exposure. Agency medical directors and individual prescribers were asked to review these reports, explain variances from standard, and document an action plan to align with prescribing guidelines. Aggregate data feedback was discussed during monthly technical assistance calls with participating practices in conjunction with consultation and training from academic mentors. Within the collaborative learning community, these mentors also served as referral leads to address adverse effects of medication adjustments. Notably, minimal adverse effects were reported during monthly calls.

The target population included Medicaid-enrolled children ages two through 17 who received psychotropic medication during October 2013 through February 2015. Children with at least one prescriber in group 1 or group 2 were compared with the other group and with a state comparison group of all Ohio children enrolled in Medicaid without an OMM participating prescriber (N=143,057). Compared with group 1 and the state group, group 2 had a significantly larger proportion of males and of disabled, low-income, and nonwhite youths. [Tables in an online supplement present additional information.]

Measure 1, antipsychotic use rate among young children, was defined as the number of young children (ages two through five) receiving antipsychotic medication divided by the number of young children receiving any psychotropic medications. Measure 2, antipsychotic polypharmacy, was defined as children ages six through 17 years receiving two or more different antipsychotics for a minimum of two consecutive months (to account for acceptable cross-titration). Rates were calculated for this age group by dividing the number with antipsychotic polypharmacy by the number receiving any antipsychotic. Measure 3, psychotropic polypharmacy, was defined as receipt of one or more antipsychotics and three or more other psychotropics with different generic names for at least two consecutive months. Rates were calculated by dividing the number of children (ages two through 17) receiving psychotropic polypharmacy by the number receiving any antipsychotic. Follow-up rates were measured at 17 months for group 1 and 10 months for group 2. Measures reflect the proportion of children who filled a prescription (either a refill or a new start) that met the targeted prescribing pattern during a given month.

Significant progress occurred on each measure, although not uniformly across groups. Because of OMM’s structure, it was not possible to discern which specific intervention elements resulted in the changes described below.

For measure 1, antipsychotics prescribed to children under age six, group 2 had a significantly lower baseline rate, compared with the state population and group 1. Furthermore, the prevalence decreased significantly in group 2 (−49.1%, p<.05) and the state comparison group (−11.2%, p<.001), with no significant change in group 1.

For measure 2, prescription of two or more concomitant antipsychotics for longer than two months, group 1 had lower rates—and group 2 higher rates—at both baseline and follow-up. The prevalence of antipsychotic polypharmacy decreased significantly in group 1 (−23.8%, p<.05) and group 2 (−29.8%, p<.001), with no significant change in the state group (−2.6%).

For measure 3, psychotropic polypharmacy, compared with the state group, both groups 1 and 2 had higher rates at baseline and follow-up. Psychotropic polypharmacy decreased significantly in group 2 (−18.5%, p<.01) and the state group (−5.0%, p<.001), whereas no significant change was observed in group 1 (−5.3%).

OMM demonstrated that a state-level QI collaborative can be a successful strategy in reducing antipsychotic overprescribing for youths. The initial effort to engage an expert advisory panel to review the complex medical literature and develop clear guidelines on safe prescribing paid off in increased guideline buy-in from multiple stakeholders. Furthermore, we learned that by educating prescribers and providing rapid feedback, in the context of a learning community involving agency medical directors and academic mentors, we could achieve significant improvements in overprescribing, and associated savings, without adverse clinical impact.

We also learned that the same intervention for different prescriber groups had different levels of impact. Group 1 prescribers were mostly general pediatricians who significantly reduced the prescription of two or more antipsychotics, but who showed little improvement on the other two measures. Group 2 prescribers, on the other hand, were primarily from community behavioral health, residential treatment, and other settings serving children with the greatest needs. Group 2 achieved significant reductions on all three measures. We planned to provide the same intervention to each group, not anticipating that, in fact, interventions should be tailored to each group’s specific barriers and challenges. Both prescriber groups identified timely access to evidence-based psychosocial treatment and community support as a significant issue, and thus it is not surprising that the group with the least behavioral health expertise and with more limited access to outside resources (group 1) showed the lowest impact on the three measures collectively. Furthermore, the single measure on which group 1 successfully improved, reduction of antipsychotic polypharmacy, is easiest to modify in general practice without outside access to nonpsychopharmacology resources, whereas reducing the total number of medications or finding alternatives to medication for younger children is more likely when evidence-based nonpsychopharmacologic interventions are more readily available. It is likely that group 2 prescribers, in more organized treatment settings, were better able to combine medication changes with initiation of nonpsychopharmacologic interventions.

These findings reinforce what may be the most important lesson: successful long-term impact on antipsychotic overprescribing requires more than one intervention targeted at one aspect of the issue—specifically prescriber knowledge and feedback about best practices. As noted, prescribing decisions do not result solely from the prescriber’s knowledge of best practice but rather reflect multiple factors, including access to nonpharmacologic interventions and pressures to maintain medications from schools, parents, courts, and others.

Another issue in implementing improvement in a large system is related to the capacity of the provider organizations to manage change in their own settings and structures. Communities and provider organizations in the collaborative varied in their ability to provide leadership, infrastructure, internal collaborations, and performance data management to drive specific improvement outcomes. For example, clinicians did not consistently participate in the collaborative; many were busy medical directors and supervisors. Many technical assistance calls focused on strategies that medical directors and supervisors could use to review data feedback with their supervisees as part of an internal QI process. However, only a limited number of clinicians made full use of the feedback. It might have been helpful to focus more on developing the internal QI structures at the provider level, including helping providers develop the processes needed to support QI and maintenance-of-certification competencies, before launching the project.

This complexity is a major reason why continuous QI strategies, such as plan-do-study-act (PDSA) cycles of change, are so valuable. As demonstrated in the IHI Breakthrough Series model, the first intervention plan (for example, provider education and feedback) is studied and then modified, according to the data and to identification of additional factors that might contribute to outcome, to generate a next-step intervention, with a further PDSA cycle, continuing incremental improvements over time.

Long-term success requires a sustainable change process. Unfortunately, OMM was a time-limited intervention because of funding. Potential savings from antipsychotic prescribing reductions were not redirected to continue the effort, suggesting the importance of building sustainable QI processes not as distinct projects but as core elements of system infrastructure. A compendium of examples and resources from 16 state Medicaid agencies is available (4).

As state Medicaid agencies move to managed care to increase value and quality, creating mechanisms for continuous QI collaborations to address National Committee for Quality Assurance (NCQA) quality measures (such as “use of multiple concurrent antipsychotics in children and adolescents”) can help uphold standards of care for vulnerable populations. Managed care plans offer loci of accountability and feedback for states and providers, as well as flexibility to drive continued improvements through contracting incentives, including for safer antipsychotic prescribing in general medical and behavioral health settings. To sustain this work in Ohio, the NCQA antipsychotic measures (developed after OMM initiation) have been incorporated into patient-centered medical home standards.

OMM demonstrated that a Medicaid-led QI collaborative could significantly improve pediatric antipsychotic overprescribing and polypharmacy. A multistrategy approach incorporating data-driven feedback and dissemination of evidence-based recommendations was effective in achieving a nearly 25% reduction in antipsychotic exposure. Development of an infrastructure for continuing QI efforts through multiple PDSA cycles may have led to further progress. Ohio has thus moved toward embedding QI in value-based purchasing as a strategy to sustain and spread best practices, including in antipsychotic prescribing. Further studies will be useful in analyzing the impacts of specific interventions across provider groups.

The authors acknowledge the Ohio Minds Matter Clinical Advisory Panel under the leadership of John Campo, M.D., and Eric Benjamin, M.D., and the members of the regional pilot sites for their input and guidance.