Using Apriori Algorithm to Determine Investigation Templates Specific to Intensive Care Units

Abstract

Objective: The development of new technologies, the improvements in the quality of health care provision and competition among health institutions are constantly increasing the expenditures in the field of health. It has become compulsory for health institutions, with their existing financial resources, to take serious measures to meet this increasing spending in order to sustain their existence. The aim of this study is to propose test templates suitable for the patient's profile in intensive units, to help physicians in diagnosis and treatment. This will help increasing service speed as well as quality and also reducing the laboratory expenditures, an important component of health expenditures. Method: Laboratory test results, which health institutions refer to for diagnosis, are an intense and important data source of hospital information management systems. These data, when analyzed by data mining methods, can provide information to support healthcare professionals in their decisions and help the effective management of the health institutions. For this study, a data set containing examinations during the diagnosis/treatment processes and the demographic information of patients hospitalized in intensive care units in the year 2018 was created. The data set was initially analyzed with the Apriori Algorithm included in the open source WEKA program. Afterwards, laboratory tests and their rules of association suggested by the Apriori algorithm were determined. Laboratory test templates were then shaped according to the common characteristics of the medical and demographic information of the patients. Results: First of all, data including 63686 records laboratory examinations of patients hospitalized in intensive care in 2018 were prepared. Using the Apriori Algorithm, the laboratory tests for which the confidence value was greater than or equal to 0.95 and the lift values indicating the importance of the rule were large were determined. From these tests, 17 rules that match the conditions of coexistence were formed, and a new data set containing 15785 records providing the rules was created. Templates specific to intensive care units were then created according to the results of the K-mean clustering algorithm applied on the data set. Conclusion: Decision support systems, which are designed to help physicians to make more effective and correct decisions in their health care services are a part of hospital information management systems. Test templates that can be a module for these systems are defined in this study. Providing physicians with an intensive care unit specific template determined according to the patient's department and demographic characteristics will support the effective health service delivery. These templates will at the same time support the demand for rational orders be made by clinicians. This overall can reduce the laboratory test and treatment costs.

Keywords: Apriori algorithm, intensive care units, laboratory tests, templates

Special Issue of Health Sciences

DOI: 10.7176/JSTR/6-03-08