Pressure Injury Prediction Modeling (PIPM) Project
Table of contents
This SOCR project is focused on a clinical collaboration to study pressure injury in bed-confined hospitalized patients.
Overview
Pressure injuries (PIs) are caused by stress on the skin (the largest organ in the human body) that compromise its integrity. It is commonly observed when extremities are restrained (e.g., shoes, straps, etc.). PIs may also be acquired during patient hospitalization, which leads to substantial burden, patient suffering, increased medical costs, and co-morbidities. This project utilizes advanced data science analytics and machine learning techniques to interrogate large, incongruent, incomplete, heterogeneous, and time-varying data of hospital-acquired PIs. Using model-based and model-free techniques, we examine PI outcomes and determine the salient features that are highly predictive of the severity of the outcomes.
Goals
- Develop model-based statistical methods and model-free AI/ML predictive techniques to identify the critical factors in severity of pressure injury (PI) hospitalized patients (e.g., post-operative).
- Identify salient features and patient clusters associated with PI risk and severity.
- Validate the forecasting models and report the precision, consistency, and reliability of the predictions.
- Determine the clinical implications and suggest implementation strategies and interventions to alleviate the PI impact on surgical patients in hospital settings.
Team
Zerihun Bekele, Christine Anderson, Dana Tschannen, Ivo Dinov, and the SOCR Team.
Code
The complete code and end-to-end pipeline workflow for this project is available in the code folder.
Application
The pressure injury prediction model (PIPM) live application (app) is available on shinyapps.io and Michigan Medicine server for interactive testing by researchers, clinicians, developers, and healthcare providers.
Acknowledgments
The PIPM project supported in part by the University of Michigan School of Nursing; NSF grants 1916425, 1734853, 1636840, 1416953, 0716055 and 1023115; and NIH grants P20 NR015331, U54 EB020406, P50 NS091856, P30 DK089503, UL1TR002240, R01CA233487, R01MH121079.
References
Anderson, C, Bekele, Z, Qiu, Y, Tschannen, D, and Dinov, ID. (2021) Modeling and prediction of pressure injury in hospitalized patients using artificial Intelligence, BMC Med. Inform. Decis. Mak., 21:253, DOI: 10.1186/s12911-021-01608-5.