Building a Better Computerized Adaptive Test for Suicide Risk Screening and Measurement Through Iterative User Input: An Implementation Study (Preprint)
Authors:
- Edwin D BoudreauxEBEdwin D Boudreaux
- Hyein S LeeHLHyein S Lee
- Rachel E Davis-MartinRDRachel E Davis-Martin
- Celine LarkinCLCeline Larkin
- Cody SchroederCSCody Schroeder
- Robert D GibbonsRGRobert D Gibbons
Preprint
- 5 March 2026
- preprint
- Published by JMIR Publications Inc. in JMIR Preprints
Abstract
BACKGROUND Suicide remains a leading cause of death in the United States, with healthcare settings like emergency departments (EDs) representing key settings for early identification of risk. Existing screening tools are limited by inefficiency, narrow scope, and workflow integration challenges, necessitating the development of solutions that combine accurate risk detection with clinical feasibility. OBJECTIVE This study aimed to develop the Computerized Adaptive Test for Suicide and Expanded (CAT-SSE), an integrated tool that merges computerized adaptive testing with electronic health record (EHR)-based machine learning to improve suicide risk screening and measurement in a large ED. METHODS The development process followed a user-centered, four-phase design involving ED clinicians and patients: 1) initial design, 2) early feedback, 3) build and integration, and 4) final testing and feedback. The CAT-SSE incorporated adaptive tests across multiple psychopathology domains combined with an EHR-based machine learning algorithm. Iterative feedback guided enhancements in usability and workflow integration. Usability, acceptability, appropriateness, and feasibility were evaluated using validated instruments, including the System Usability Scale (SUS), Acceptability of Intervention Measure (AIM), Intervention Appropriateness Measure (IAM), and Feasibility of Intervention Measure (FIM). RESULTS Ten patients and sixteen clinicians participated in early wireframe and prototype testing, and usability improvements were made accordingly. In the final user testing phase, twelve patients and five clinicians who managed nine of the patient encounters assessed the production version. Across earlier and later study phases, patients consistently reported high usability ratings (mean SUS (SD): 87.0 (9.9); 87.7 (12.8)), while clinicians’ usability ratings increased from a mean of 73.4 (16.2) to 79.7 (24.1) after iterative refinements. Both user groups rated the CAT-SSE as highly acceptable and appropriate. Clinicians also endorsed its feasibility for integration into busy ED workflows. CONCLUSIONS The CAT-SSE is one of the first ED-based suicide risk screening tools to effectively combine adaptive tests and EHR-based machine learning algorithms. Iterative development with end-user input yielded a solution rated as usable, acceptable, and feasible by both patients and clinicians. The CAT-SSE shows promise as a scalable strategy to efficiently screen and measure suicide risk and broad psychopathology within ED clinical practice.Keywords
This publication has 1 reference indexed in Scilit: