Better Health Care With Data Mining
Statistical analysis helps health care professionals turn raw data into valuable information
Source: www.spss.com
Copyright SPSS, Inc. 2004


Introduction
Health care generates mountains of administrative data about patients, hospitals, bed costs, claims, etc. Clinical trials, electronic patient records and computer supported disease management will increasingly produce mountains of clinical data. This data is a strategic resource for health care institutions. 

With the advent of data warehousing techniques, specific areas of interest may be investigated more thoroughly. Products such as INFoCOM from SMS, which is a clinically-based data warehouse product designed for use throughout a hospital, bring the potential for specialized information production to the clinicians and managers desktop through the use of clinical workstations and Executive Information Systems (EIS). 

Data mining products are designed to take this one stage further. It brings the facility to discover patterns and correlation hidden within the data repository and assists professionals to uncover these patterns and put them to work. Therefore, decisions rest with health care professionals, not the information system experts.

The key to successful data mining is to first define the business or clinical problem to be solved. New knowledge is not discovered by the algorithms, but by the user. This paper will prove that knowledge can automatically be obtained by the use of machine learning techniques in the hands of health care decision-makers.

Inpatient Length of Stay
The analysis is to look at the contributory factors influencing length-of-stay (LOS) of a patient during a consultant episode — a major component in the cost of inpatient treatment.

The purpose is to identify patterns affecting LOS that may help in the reduction of cost (and potentially reduce patient trauma), based upon the premise that it is possible to reduce costs by seeking to reduce patient LOS.

Discussion
The findings from any data mining exercise can bring patterns to surface (that might otherwise remain undiscovered), which may suggest alternative ways for treating patients making better use of resources.

In the described scenario, actual interpretation of these results can only be made by speculation. Using an indirect approach to patient costing — in this case by analyzing length of stay — it may be shown that there is an exceptionally low-cost for an emergent technique for surgery, as opposed to the more established techniques for the same primary procedure. However, a hospital is a complex environment and keyhole surgery is likely to extend time spent in the hospital, which would bring further pressures on resource availability.

It would appear from this scenario that there are previously undiscovered patterns which can be induced using data mining techniques. Evidence-based data could stimulate further discussion and further investigation by both managers and clinicians working in partnership for the good of the patient and the hospital. 

The discussion and further analysis would potentially include the balance of increased institution utilization against reduced utilization of wards and associated support services. Further discriminating factors may then be discovered which could hone the clinical decision, in that there may be other factors that influence the decision to opt for open surgery in preference to keyhole surgery techniques.

Conclusion
The use of data mining has focused on evidence-based patterns from previous patient treatment. In all likelihood, the absence of automated discovery of patterns would leave many questions unasked. These questions, if asked, would benefit not only the resource utilization for patient treatment, but also the health of the patient. 

Data mining helps professionals discover these patterns and put them to work. As models are based directly on history, they represent the ultimate in evidence-based care. But technology is no panacea, professional, ethical and practical issues must be addressed. Decisions must rest with the health care professionals, not the information systems experts.