This timely supplement would support our goals for the current award: An Evaluation of the National Zero Suicide Model Across Learning Healthcare Systems (U01MH114087) by capitalizing on a natural experiment, the planned the national roll-out of safety planning templates in behavioral health departments across five Kaiser Permanente regions and Henry Ford Health System in 2019. Safety planning is a highly recommended practice within the Zero Suicide (ZS) framework, but little is known about the effectiveness of the individual elements that can make up a safety plan, such as lethal means assessment, identification of supportive contacts, coping skills, warning signs, and sources of distraction. The current Zero Suicide award proposes to examine the impact of safety planning and lethal means assessment using a stepped-wedged interrupted time- series (ITS) approach, measuring each as a binary variable (e.g. safety planning did or did not occur). The ITS approach requires that some sites implement safety planning (intervention sites for safety planning), while others do not (control sites for safety planning). The proposed ITS approach is now problematic without further work for two reasons: 1) All Kaiser Permanente sites and Henry Ford have decided to uniformly implement safety planning around the same time, therefore there are no control sites 2) Without control sites, metrics that can accurately measure variation in safety planning/lethal means assessment at baseline and then longitudinally thereafter would enable our evaluation to take place, but all of the documentation lives in text- based clinical narratives. In working with our health system leads on the development of Zero Suicide metrics, we have been informed that the rate for safety planning and lethal means assessment at baseline is not zero, but the actual rate is unknown. This supplement will support development of new metrics using Natural Language Processing to determine baseline rates, from which, we can quantify the change in safety planning and lethal means assessment practice longitudinally after implementation of new safety planning templates using our Zero Suicide main award. Furthermore, we propose to take advantage of the newly implemented templates to address an important mediator of the effect of safety planning on suicide outcomes, the impact of fidelity to the new templates, which we define as quality, completeness, and level of integration with ongoing care. We propose the following three specific aims for this supplemental work: 1) Identify key terms for safety planning and lethal means assessment 1.) Develop Natural Language Processing (NLP) metrics to assess the occurrence of safety planning and lethal means assessment at three Zero Suicide sites 2) Implement NLP queries for identification of safety planning and lethal means assessment and measure baseline rates 3) Upon implementation of electronic safety planning templates in medical records, develop and implement metrics using NLP for assessing fidelity (completeness, quality, integration with care) to safety planning templates.
This supplement to the current award: An Evaluation of the National Zero Suicide Model Across Learning Healthcare Systems (U01MH114087) will take advantage of a national roll-out for safety planning templates in the electronic medical record across 5 Kaiser Permanente regions and Henry Ford Health System. It will support the development of innovative measures using natural language processing to efficiently quantify the baseline rates of safety planning and lethal means assessment across multiple sites, which will enable a rigorous evaluation to take place upon implementation of the new electronic templates. Furthermore, we propose to take advantage of the safety planning template roll-out by measuring fidelity (quality, completeness, and level of integration with care), because it may be an important mediator of the relationship between safety planning and suicide outcomes.
