Generating treatment plan in medicine: a data mining approach

American Journal of Applied Sciences, Feb, 2009 by Ahmad Mahir Razali, Shahriyah Ali

INTRODUCTION

Accurate and error-free of diagnosis and treatment given to patients has been a major issue highlighted in medical service nowadays. Traditionally, healthcare providers provide services based on their knowledge and experiences whether individually or collectively depending on cases. Research done proved that hospitals do not all provide the same quality of service even though they provide the same type of service (1). To achieve service excellence, hospitals must strive for zero defections (2) that require continuous effort to improve the quality of the service delivery system (3).

Treatment plan on the other hand, refers to management on any interventions consists of treatment and/or examination which will be initiated for each problem based on patient's history, physical examination, provisional diagnosis and differential diagnosis (4). Treatment plan can be generated to provide useful evidence as a basis for future medical practice, by utilizing previous treatment patterns from clinical records database. The amount of collected and stored data in databases has increases dramatically due to advancements in software capabilities and hardware tools, along with decreasing trend of hardware and software cost.

In spite of that, data mining techniques which are part of knowledge discovery in databases (KDD), have become popular research tools for medical researchers who seek to identify and exploit patterns and relationships among large number of variables and be able to predict the outcome of a disease using the historical cases stored within datasets (5,) (6). Applications of data mining have already been proven to provide benefits to many areas of medicine, including diagnosis, prognosis and treatment (7). Data mining techniques have been applied in various medical fields, amongst with are health administration (8), (9) adverse drug reactions (10-12), drug safety (13), (14) predicting breast cancer survivability (15), predicting survival time for kidney dialysis patients (16), knowledge discovery in hypoplastic left heart syndrome (17) and predicting protein function (18).

Because of these issues, we think there is a need of aid for health practitioners during treatment process, as well as consideration of patient's well being. Taking advantage of massive clinical data gathered from information technology and the importance of data mining nowadays in decision making, generating treatment plan seems to encounter the above mentioned problems.

MATERIALS AND METHODS

This research is conducted based on outpatient clinical data gathered from various health centers throughout Malaysia. These data were stored electronically via Percuro Clinical Information System (Percuro), which was provided by RareSpecies Corporation Sdn. Bhd., a medical software development company. Figure 1 shows Percuro main module for consultation and treatment session (19).

[FIGURE 1 OMITTED]

Percuro applies SOAP (Subjective, Objective, Assessment and Plan) format in recording all medical information as being practiced in medicine. Further details on SOAP can be seen in Table 1.

Table 1: SOAP format

SOAP format                         Details

S       Subjective    History: Information requested from patients
        data          on principal symptoms, history of present
                      illness, past history, social history, family
                      history and systems review

O       Objective     Physical examination, provisional diagnosis
        data          and differential diagnosis: Records from
                      physical and laboratory findings that relevant
                      to patient's complaint

A       Assessment    Interpretation of any relevant findings for
                      each problem.

P       Plan          Any interventions that will be initiated for
                      each problem consists of treatment and/or
                      examination. Treatment: drug, procedure
                      Examination: laboratory, imaging

All data related to patients are recorded electronically direct into patients' record and are stored in the Percuro database. Demographic information was recorded by staffs in charge in registration counter during registration while clinical information was recorded by health practitioners during treatment process.

Throughout generating treatment plan, Cross-Industry Standard Process for Data Mining (CRISP-DM) has been used as foundation (Fig. 2). CRISP-DM has a life cycle consisting of six phases. Each phase is followed until research objective is achieved.

[FIGURE 2 OMITTED]

Business understanding was well defined and data understanding was thoroughly observed, before data preparation can be made. After understanding the whole set of data and what can be extracted from it, the objective of the study was determined.

As much as 88,355 clinical data have been gathered for the duration period of 18 months. Acute upper respiratory infection (J06.9 as identified in International Classification of Disease 10 by World Health Organization) was set to be the disease for generating treatment plan as it was the most common problem encountered. Acute upper respiratory infection is a severe adenovirus infection of the respiratory tract characterized by fever, sore throat and cough.