Exploring Factors Associated with Pressure Ulcers: A Data Mining Approach
Raju, D, Su, X, Patrician, P, Loan, L, McCarthy, M
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.
Using Item Response Theory Models to Evaluate the Practice Environment Scale
Raju, Dheeraj; Su, Xiaogang; Patrician, Patricia A.
Background and Purpose: The purpose of this article is to introduce different types of item response theory models and to demonstrate their usefulness by evaluating the Practice Environment Scale. Methods: Item response theory models such as constrained and unconstrained graded response model, partial credit model, Rasch model, and one-parameter logistic model are demonstrated. The Akaike information criterion (AIC) and Bayesian information criterion (BIC) indices are used as model selection criterion. Results: The unconstrained graded response and partial credit models indicated the best fit for the data. Almost all items in the instrument performed well. Conclusions: Although most of the items strongly measure the construct, there are a few items that could be eliminated without substantially altering the instrument. The analysis revealed that the instrument may function differently when administered to different unit types.
The role of physical distribution services as determinants of product returns in Internet retailing
Rao, S, E, Rabinovich, & Raju, D.
Pressure continues to build on Internet retailers to squeeze out inefficiencies from their day-to-day operations. One major source of such inefficiencies is product returns. Indeed, product returns in Internet retailing have been shown to be, on average, as high as 22% of sales. Yet, most retailers accept them as a necessary cost of doing business. This is not surprising since many retailers do not have a clear understanding of the causes of product returns. While it is known that return policies of retailers, along with product attributes, are two important factors related to product return incidents, little is known about which aspects of the online retail transaction make such a purchase more return-prone. In the current study, we seek to address this issue. We use a large data set of customer purchases and returns to identify how process attributes in physical distribution service (PDS) influence product returns. The first attribute involves perceptions of scarcity conditions in inventory availability among consumers when retailers reveal to consumers information on inventory levels for the products that they intend to buy. Our results show that orders in which items are sold when these conditions are revealed to shoppers have a higher likelihood of being returned than orders in which these conditions are not revealed. While prior research has argued that inventory scarcity perceptions have an effect on purchases, our findings suggest that they are also related to the likelihood of these purchases being returned. The second attribute involves the reliability in the delivery of orders to consumers. We find that the likelihood of orders being returned depends on the consistency between retailer promises of timeliness in the delivery of orders and the actual delivery performance of the orders. Moreover, we find that the effect that consistency in the delivery has in the likelihood of returns, is stronger for orders that involve promises for expedited delivery than for orders with less expeditious promises. That is, although the occurrence of returns depends on the delays in the delivery of orders to consumers relative to the initial promises made by the retailers, this effect is more notable for orders that involve promises of fast delivery.
Never stop improving