Fall Risk Prediction Models During the Initial COVID-19 Surge: Could Predictive Analytics Be Used in a...

Area of Science:

• Healthcare Management
• Patient Safety
• Epidemiology

Background:

• The COVID-19 pandemic necessitated significant changes in hospital operations, including nurse staffing and workflows.
• These changes occurred in resource-constrained environments, potentially impacting patient care delivery.
• Patient falls are a critical safety concern in hospitals, especially during public health crises.

Purpose of the Study:

• To investigate the rate of inpatient falls during the initial COVID-19 surge compared to the pre-surge period.
• To evaluate the performance of an automated predictive analytic algorithm against the established Johns Hopkins Fall Risk Assessment tool.
• To determine the utility of predictive models for fall risk assessment in resource-limited healthcare settings.

Main Methods:

• A retrospective review of patient falls was conducted, comparing a 3-month period before the COVID-19 surge with the first 3 months of the surge.
• Data collected included the total number of falls, overall fall rates per 1000 patient-days, and fall risk assessment scores.
• The accuracy, sensitivity, and specificity of the Johns Hopkins Fall Risk Assessment and an automated predictive analytic model were compared.

Main Results:

• A statistically significant increase in the inpatient fall rate was observed during the first 3 months of the COVID-19 surge (2.48/1000 patient-days) compared to the preceding 3 months (1.89/1000 patient-days; P = .041).
• The Johns Hopkins instrument demonstrated higher sensitivity (78.9%) than the predictive analytic model (57.0%).
• However, the automated predictive analytic model exhibited superior specificity (71.3%) and accuracy (71.2%) compared to the Johns Hopkins instrument (54.1% and 54.3%, respectively).

Conclusions:

• The COVID-19 surge was associated with a higher rate of inpatient falls, highlighting a critical patient safety issue.
• While the Johns Hopkins Fall Risk Assessment had higher sensitivity, the automated predictive analytic model offered better specificity and overall accuracy.
• The findings suggest that automated predictive analytic models can be effectively utilized for accurate patient fall risk classification, particularly in resource-constrained environments.