Exploring factors associated with pressure ulcers: A data mining approach

Abstract

Background

Pressure ulcers are associated with a nearly three-fold increase in in-hospital mortality. It is essential to investigate how other factors besides the Braden scale could enhance the prediction of pressure ulcers. Data mining modeling techniques can be beneficial to conduct this type of analysis. Data mining techniques have been applied extensively in health care, but are not widely used in nursing research.

Purpose

To remedy this methodological gap, this paper will review, explain, and compare several data mining models to examine patient level factors associated with pressure ulcers based on a four year study from military hospitals in the United States.

Methods

The variables included in the analysis are easily accessible demographic information and medical measurements. Logistic regression, decision trees, random forests, and multivariate adaptive regression splines were compared based on their performance and interpretability.

Results

The random forests model had the highest accuracy (C-statistic) with the following variables, in order of importance, ranked highest in predicting pressure ulcers: days in the hospital, serum albumin, age, blood urea nitrogen, and total Braden score.

Conclusion

Data mining, particularly, random forests are useful in predictive modeling. It is important for hospitals and health care systems to use their own data over time for pressure ulcer risk prediction, to develop risk models based upon more than the total Braden score, and specific to their patient population.

Introduction

Pressure ulcers (PU) are a substantial burden for patients and for the health care system in general. The National Patient Care Safety Monitoring Study (Lyder et al., 2012) of over 51,000 patients found that 4.5% of Medicare beneficiaries developed a pressure ulcer during their hospital stay and 5.8% had a pressure ulcer on admission. Pressure ulcers regardless of whether they were present on admission were associated with a nearly three-fold increase in in-hospital mortality, 69% increase in 30-day mortality, and an increased length of stay of 6.4 days (Lyder et al., 2012). As of 2008, hospital acquired stage III and IV pressure ulcers are no longer reimbursed by the U.S. Centers for Medicare and Medicaid Services (CMS), leaving the hospitals themselves to absorb the cost of care for patients, which is estimated at $43,180 per patient (Armstrong et al., 2008). Thus it is imperative to discover factors associated with both community and hospital acquired pressure ulcers and institute additional care measures to prevent their occurrence.

Because the causative factors for pressure ulcers are “multifactorial and not well understood” (Benoit and Mion, 2012, p. 341), it is critical for hospitals, nursing homes, and home care agencies to systematically monitor patients for pressure ulcer rates, assess risk, and enhance prevention efforts. Although not all pressure ulcers are avoidable (Black et al., 2011), frequent monitoring may lead to better risk predictions and more thoughtful application of resources (i.e., evidence-based nursing preventive interventions such as turning and repositioning) to those who need it most. As more hospitals adopt electronic medical records, the large clinical data repositories could help improve clinical care through the study of their own best practices and lessons learned directly from their patients. Analyzing clinical data collected from discharge abstracts or directly from clinical records and comparing those who developed or did not develop a pressure ulcer can inform problem identification in quality improvement. Data mining modeling techniques can be beneficial to conduct this analysis. The purpose of this paper is to build and compare data mining models for pressure ulcer prevalence (both community and hospital acquired) and determine the variables that are associated with pressure ulcers based on a four year study database collected from 12 military hospitals. The variables included in the analysis are easily accessible demographical information and medical measurements. We carefully selected a group of data mining techniques that not only supply high predictive accuracy but also allow for meaningful interpretations. The Braden scale developed by Bergstrom et al. (1987) is one currently available tool for assessing pressure ulcer risk. Given the multifaceted nature of pressure ulcers, it is of keen interest to see whether and how other factors could enhance the performance of predicting pressure ulcers when combined with the Braden scale. The eventual wide scale use of electronic medical records will enable hospitals to apply these data mining techniques to their own patient level data to determine factors associated with pressure ulcers.