Generating Treatment Plan in Medicine: A Data Mining Approach

Abstract

This study reports on a research effort on generating treatment plan to handle the error and complexity of treatment process for healthcare providers. Focus has been given for outpatient and was based on data collected from various health centers throughout Malaysia. These clinical data were recorded using SOAP (Subjective, Objective, Assessment and Plan) format approach as being practiced in medicine and were recorded electronically via Percuro Clinical Information System (Percuro). Cross-Industry Standard Process for Data Mining (CRISP-DM) model has been utilized for the entire research. We used data mining analysis through decision trees technique with C5 algorithm. The scopes that have been set are patient's complaint, gender, age, race, type of plan and detailed item given to patient. Acute upper respiratory infection disease or identified as J06.9 in International Classification of Diseases 10 by World Health Organization has been selected as it was the most common problem encountered. The model created for J06.9 disease is that type of plan recommended through giving drug to patients without the need to consider patient's complaint, gender, age and race, with accuracy obtained for the model is 94.73%. Inspite of that, we also identified detailed items that have been given to J06.9 patients and the occurancy of them. This can be as a guideline for future treatment with item recommendation is less than 0.078% compared to item inventory in Percuro database. The research is expected to aid healthcare provider as well as to minimize error during treatment process while benefited from technology information to increase the health care delivery.

Discrete derivative: a data slicing algorithm for exploration of sharing biological networks between rheumatoid arthritis and coronary heart disease

“…For data mining, data preparation, pretreatment, and treatment is fundamental and necessary [ 30 - 33 ] for the final result.…”

A Systematic Review on Healthcare Analytics: Application and Theoretical Perspective of Data Mining

“…If a clinician has knowledge about this relation, he/she can prescribe similar medications when faced with a similar set of problems. One group of researchers [ 61 ] developed an approach which achieved 90% accuracy in finding association between medications and problems, and 55% accuracy between laboratory tests and problems. Among outpatients diagnosed with respiratory infection, 92.79% were treated with drugs.…”

“…Authors often tested their models with multiple algorithms; SVM was at the top of that list and often outperformed other algorithms. However, 15 [ 38 , 40 , 42 , 45 , 47 , 51 , 54 , 56 , 58 , 60 , 61 , 66 , 73 , 74 , 76 ] of the studies did not incorporate expert opinion from doctors, clinician, or appropriate healthcare personals in building models and interpreting results (see the study characteristics in Supplementary Materials Table S3 ). We also noted that there is an absence of follow-up studies on the predictive models, and specifically, how the models performed in dynamic decision-making situations, if doctors and healthcare professionals comfortable in using these predictive models, and what are the challenges in implementing the models if any exist?…”

Clustering of Drug Sampling Data to Determine Drug Distribution Patterns with K-Means Method : Study on Central Kalimantan Province, Indonesia

“…Many studies have been conducted in health sector. Data mining has been able to provide many contributions to the decision-making system for health services [8] [19], to find out the pattern of a disease [11], to find the cause of the spread of a disease [14], to predict or early diagnose of a disease [6] [12], to be used in finding great information on the costs of care and treatment of patients who have certain diseases [9] [16] and can provide alternative recommendations in treatment [7][10] [17]. However, study on knowledge discovery in drug surveillance data has never been done.…”