Nurse call data analysis using Bayesian network modeling

Abstract

This paper considers and suggests efficient patrol of nurses from analysis result of nurse calls log. As the method of analysis, we consider applying Bayesian network. This is one of probability model that is available for prospects of a phenomenon, rational mind decision and so on. In conventional studies, correlation coefficient was used to examine relation between phenomena. However, we could analyze in detail by using Bayesian network because we didn't overlook the information that may overlook by using correlation coefficient. Applying the proposed method to total 5,478 nurse call data, we have obtained following knowledge: (a) some reasons of nurse call are inferred by situation of patient and demand level of nurse call, (b) an efficient patrol of nurse can be devised, and (c) Bayesian network didn't overlook the information compared with conventional method.We believe that nurse calls can play a role not merely as a means to convey patients' needs, but also as a medium to analyze and predict patients' needs using nurse calls log. In conventional study, the factor that influences the patient satisfaction often includes "Speed of correspondence to the nurse call" [1, 2]. In order to improve patient's satisfaction, the medic should adequately catch the desire of the patient. The real of reasons for calling was demonstrated.This paper considers and suggests efficient patrol of nurses from analysis result of nurse calls log. As the method of analysis, we consider applying Bayesian network (BN) [9,10]. This is one of probability model that is available for prospects of a phenomenon, rational mind decision and so on. In conventional studies, correlation coefficient was used to examine relation between phenomena. However, we could analyze in detail by using BN because we didn't overlook the information that may overlook by using correlation coefficient.In the following, overview of BN is introduced first. Then we will describe the feature of proposed BN modeling. Showing the numerical study, we will summarize what we did and potential future works. 978-1-4244-8314-3/10/$26.00 ©201O IEEE 272 II. BAYESIAN NETWORK A. Overview of Bayesian Network (BN)BN is a graphical modeling method where the quantitative dependence between nodes expressed as a random variable was shown at the conditional probability [4,5,7]. The conditional probability is calculated by the event (parameter) that happens at the node. Each node is connected with the direct link, and the link from node X to nodes Y indicates that Y has received the direct influence from X. In a word, it means there is a causal relation in the direction of the link. Figure I shows an example of BN which is consisted of three nodes "Gender", "Time Zone", and "Reason of Nurse Call". Each node has some parameters. Once a set of nodes have their own probability (it is called evidence), the conditional probabilities are calculated. It should be something like the posterior probability. To compare the prior probability with the posterior probability, called as "probabilistic ...

An Overview of Bayesian Network Applications in Uncertain Domains

“…Nevertheless, effectiveness cannot be judged without validated outcome. Aoki et al [26] developed a BN-based methodology for efficient patrolling of nurses. The purpose of nurse patrolling is to improve the dependability of calling a nurse according to patient's needs by analyzing data based on demand level of nurse calls (time zone of previous nurse call, interval of previous nurse call, reason of previous call, and degrees of freedom in life) and patient's condition (sex, age, hospital department, and degrees of freedom in life).…”

“…Therefore, BN techniques have been applied effectively in diagnosis like examining dental pain [29], BPD [28], risk analysis in alternative medical diagnosis [31] and efficient patrolling of nurses [26]. The advantage of BN in such domains is the handling of incomplete or missing data for prediction with precision.…”