Signal Detection to Measure Allostatic Load

Abstract

Purpose: To illustrate a means to calculate allostatic load in hospitalized patients using big data from the electronic medical record (EMR).

Organizing construct: To describe the development of the Troubled Outcome Risk (TOR) scale using signal detection in big data.

Methods: Using both retrospective and prospective observational studies, I describe a mechanism to determine meaning from retrospective data then use the results to improve nursing surveillance to reduce length of stay (LOS) and nursing sensitive indicators on an inpatient medical surgical unit.

Findings: Results from the retrospective study contained over 290,000 individual data points and established an interpretation standard for the TOR score using an algorithm to detect signals. The prospective observational study used the TOR scale and developed an interpretation standard to assist unit charge nurses in assigning staff to patients based on a fully objective measure of patient allostatic load.

Conclusions: The TOR scale in conjunction with existing nurse staffing methodology reduced inpatient LOS by 0.3 days, reduced allostatic load as measured by the TOR scale, and changed staffing patterns from purely geographic to patient-need based.

Clinical relevance: The TOR scale demonstrates that careful evidence-based criteria can be easily gathered from the EMR and used to positively impact nursing practice and patient outcomes.

Keywords: Allostatic load; nursing surveillance; nursing-sensitive indicators; signal detection.