Designing safer use of antipsychotics in youth: A human-centered approach to an algorithm-based solution

Abstract

Objective:

Guidelines recommend against antipsychotics without indication, yet prescription rates remain undesirably high for youth. Information technology can facilitate guideline-based prescribing, but little is known about providers’ needs. The Safer Use of Antipsychotics in Youth project is implementing an algorithm-based workflow for peer consultation, care navigation, and expedited psychotherapy access. To optimize workflow for a multi-site trial, we engaged providers for input.

Methods:

Guided by human-centered design (HCD), we interviewed 15 providers from Kaiser Permanente Washington and Nationwide Children’s Hospital about their prescribing barriers and workflow preferences. We identified qualitative themes on barriers impacting implementation and design opportunities to optimize workflow.

Results:

Providers expressed two major barriers: potential disruptions to clinical practice and threats to professional autonomy. Three design opportunities emerged: a la carte orders, passive review of orders, and consultation self-acknowledgement.

Conclusion:

HCD offers an innovative approach to improve guideline-based prescribing with optimizations that are grounded in providers’ needs.

INTRODUCTION

Rates of prescribing antipsychotic medications to children and adolescents remain high^1,2 despite guidelines against their use for behavioral symptoms of childhood mental disorders in the absence of approved indications.³ Most antipsychotic use in youth is for non-psychotic conditions (e.g., attention deficit hyperactivity disorder, disruptive behavior).^2,4 Side effects can be severe and impact patients differently.^5,6 Lack of timely access to evidence-based psychotherapies leads to substitution with medication. For example, less than half of youth initiating an antipsychotic medication received psychosocial treatment in the previous 90 days.⁷ We need to improve the quality of prescribing and care youth receive.

To enhance guideline-based prescribing for youth, the Safer Use of Antipsychotics in Youth (SUAY) project is testing an algorithm-based workflow in a multi-site pragmatic effectiveness trial. Developed by a national consensus panel of child and adolescent psychiatrists and developmental pediatricians, the proposed SUAY workflow (Figure 1) triggers peer consultation and care navigation with expedited psychotherapy access for youth with antipsychotic medication orders for non-psychotic disorders. A related program in Washington requiring mandatory prescription review reduced antipsychotic use in half.⁸ However, the program’s “hard stops” on pharmacy refills is cumbersome for clinicians. SUAY seeks to streamline this workflow.

Figure 1. Proposed SUAY Workflow.

New antipsychotic medication orders (i.e., excludes refills) or medication changes for youth aged 3-17 with disorders other than psychosis, mania, tics, autism spectrum, or moderate to severe intellectual disability trigger the alert, which recommends real time phone consultation with an expert child and adolescent psychiatrist, a one-time 72-hour medication supply, and referral to a care navigator to coordinate follow-up with expedited access to evidence-based psychotherapy through a registry system.

Although technology can guide evidence-based prescribing,⁹ best practices for designing effective decision support are limited.¹⁰ Without meeting the needs of users, technology risks poor adoption, limited clinical integration and effectiveness, and unintended healthcare consequences.¹¹ To optimize the proposed workflow, we engaged providers in human-centered design (HCD) for input on guideline-based prescribing. HCD is an informatics framework for optimizing information systems and services with formative input from healthcare stakeholders to produce useful technology that engenders adoption.¹² SUAY “users” are prescribing providers who will incorporate the proposed workflow in their future practice.

METHODS

We interviewed pediatric and adolescent providers to assess fit of the proposed workflow with their prescribing needs and practices. We recruited primary care providers, developmental behavioral pediatricians, advanced registered nurse practitioners, and psychiatrists with experience prescribing antipsychotics from Kaiser Permanente Washington and Nationwide Children’s Hospital. Kaiser Permanente Washington’s integrated healthcare system treats approximately 56,000 youth with behavioral health services. Nationwide Children’s Hospital treats approximately 31,000 youth annually with inpatient and outpatient behavioral health services. Both sites use the EpicCare electronic health record.

We conducted 1-hour audio-recorded interviews following a 3-part guide: prescribing needs and barriers elicited through critical incidents¹³ by recounting prior prescribing experiences; design preferences for the proposed workflow visually depicted with a storyboard¹⁴ that we walked through using vignettes of fictitious patients for feedback across different clinical scenarios; and demographic survey (Appendix 1, Online supplement). IRB approval deemed study procedures exempt. Participants received $100.

We analyzed survey data with descriptive statistics and used thematic analysis¹⁵ to qualitatively code interview recordings for themes about workflow barriers and design opportunities. Following Braun and Clarke¹⁵ we first explored patterns across interviews. Subsequent coding focused on workflow reactions to alert-guided prescribing, considerations for peer consultation (e.g., timing), and characteristics of care navigation, from which themes emerged.

RESULTS

Fifteen participants completed interviews (P1-P15), including 6 providers from Kaiser Permanente Washington and 9 providers from Nationwide Children’s Hospital. Appendix 2 (online supplement) summarizes participant characteristics.

Participants found value in standardizing antipsychotic prescribing. Although several have unfettered access to expert consultants through call services, curb-side meetings, preceptors, and other informal channels, the proposed workflow was seen as potentially duplicative but beneficial for formalizing consultative practices in the electronic health record. Most participants recognized added benefits of care navigation but felt it was not appropriate for every case. Despite these positive reactions, providers expressed two major barriers of the proposed workflow: potential disruptions to clinical practice and threats to professional autonomy.

First, real time consultation, hard stops and other features of the proposed workflow raised concern about disrupting clinical practice by reducing care quality and causing delays and safety risks. Most participants were concerned about limited time to “squeeze” (P6) real time consultation into packed schedules. Real-time consultation was not found practical nor necessary. Some felt that rushing the consultation could provide a “disservice” (P1) by limiting time for information gathering and thoughtful discussion. A couple of participants described alternative emergency services already in place for real-time consultation. There was consensus that a thorough and thoughtful consultation, which might require time outside of patient visits, was more important than efficiency.

Enforcing a hard stop for every antipsychotic order until documentation of the consultation also raised concern about care delays and safety, such as for immediate or emergency access to antipsychotics: “So my concern would be, here I am I have done a thorough evaluation and this is well within my area of expertise and my only choice is an emergency prescription and I can’t reorder it until I consult with someone and document that was done – it’s extra steps, it’s extra time and it could be a delay in treatment”(P14). Others described needing to start or wean patients off antipsychotics to “bridge care” (P12) while waiting for therapy or specialty care. Although participants agreed emergency prescriptions help, they found 72 hours insufficient without override options when wait times are weeks to months. Consultation with a 14-day emergency prescription was seen as reasonable, depending on the urgency of the case and consultant availability.

Threats to professional autonomy was the second major barrier we identified. Behavioral health providers with extensive prescribing experience shared concern about a blanket requirement for peer consultation described as “aggravating” (P1), “scrutinizing” (P2), or even “offensive” (P6). They saw little value in consultation for what was perceived as reasonable prescribing by experts with training in psychiatry and pediatrics. In contrast, primary care providers considered peer consultation “reassuring” (P3), a “safety net” for inexperienced prescribers (P4), and a resource rather than a barrier: “Nobody likes to be told what to do, but everyone likes to be offered help!” (P1). A couple of primary care providers felt the workflow implicitly encouraged them to manage complex cases they would otherwise “punt” (P2) to psychiatry. Some felt the proposed workflow should provide a “happy medium” (P4) between monitoring and guiding practice given diversity in provider experience and patient cases. Rather than treating all providers the same (i.e., alert fires for all medication orders irrespective of clinical context), some participants suggested targeting peer consultation to those who most need it.

To address these barriers, participants suggested improvements, including training, information flows, and checklists to guide prescribing. P1 suggested design improvements to reduce disruptions: “What if the workflow helped guide providers to consultation by making it easy to do the right thing, such as embedding the [consultation] referral in the alert and having a place for me to enter my phone number and best times for the consultant to call me?” P13 recommended pre-review to only flag orders outside of guideline that require consultation: “Have the prescriber write up the plan. The consultant reviews and ok’s without needing a phone discussion except for the questionable cases.” P14 suggested encouraging autonomy: “[Design it] On the honor system, for people who don’t have that expertise, to encourage them to seek a consultation, especially if it is something they have not dealt with before and particularly if it is a behavioral health issue and they are not a behavioral health provider.“ Documentation from the consultant was valued for more than a “Yes” or “No” judgment, but a plan for management, expectations, next steps, and patient education. Given participants’ input, we identified three consistent design opportunities to optimize SUAY workflow: provide a la carte order sets in the alert to tailor available resources to varied needs; facilitate passive review of orders to minimize disruptions and target behavior change; and encourage self-acknowledgement of completed consultation to preserve professional autonomy (Appendix 3, Online supplement).

DISCUSSION

Using HCD, we solicited provider input to optimize SUAY workflow. Interviews identified two major barriers – clinical practice disruptions and threats to professional autonomy—that can be addressed through design improvements providers recommended: a la carte orders, passive review, and self-acknowledgement of completed consultations. Because they are grounded in the needs and practices of future intervention users, these optimizations could improve adoption and effective use. We implemented these improvements for our multi-site pragmatic trial.

Findings, based on interviews with a small number of providers focused on a specific prescribing scenario, have limited generalizability and may not fully translate to other clinical settings. Although we engaged prescribers from two healthcare systems, findings could reflect selection bias. Future work should further compare needs of behavioral health and primary care providers. Successful adoption requires additional research to align system design with input from a larger sample of providers, staff, and patients. Despite these limitations, findings provide insight into complex clinical workflows and professional autonomy, which are valuable for implementing SUAY and guiding development of similar interventions.

HCD offers an innovative approach to improve guideline-based prescribing. Providers’ input, which reflects the needs and preferences of future users, helped us arrive at a more acceptable solution with increased likelihood of adoption. Our work illustrates a structured process for negotiating difficult design choices in which system-wide charges and individual priorities may conflict. HCD helped us engage providers in formative design of our future system that will positively impact their work.

HIGHLIGHTS.

• Guidelines recommend against antipsychotics without indication, yet prescription rates remain undesirably high for youth

• Information technology interventions designed to meet providers’ needs have the potential to facilitate guideline-based prescribing

• Human-centered design provides an innovative approach to identify and address barriers through workflow optimizations that inform intervention development