| Summary: | 碩士 === 國立陽明大學 === 衛生資訊與決策研究所 === 94 === Hospital information system (HIS) is very popular in Taiwan after National Health Insurance Program launched on March 1, 1994. Its main reason is the needs of reimbursement in convenient way. Though most health providers have designed or bought different type of computer program for management of their patient’s visits, its major goal is still the reimbursement. From the viewpoint of better health care and patient safety, we may provide better patient care by way of reusing and analyzing of those data already stored in HIS with modern data mining tools.
Electronic medical record (EMR) is a provider-based medical record that includes all health documentation for one person covering all services provided within an enterprise. The data stored in HIS may be in different data types or come from different sources. Longitudinal patient record (LPR) is an EMR that includes all healthcare information from all sources. They include the patient’s characteristics, their ambulatory visit, hospital stays, laboratory or radiological examinations, etc.
It is always a dilemma in correct diagnosis, treatment planning or prognosis prediction of disease. The most important reason is not only due to the high variability of disease, but also the different degree of presentation of symptoms and signs. Most clinicians make decision according to their experience only.
Evidence-based medicine/healthcare is looked upon as a new paradigm, replacing the traditional medical paradigm which is based on authority. It is dependent on the use of randomized controlled trials, as well as systematic reviews (of a series of trials) and meta-analysis, although it is not restricted to these. There is also an emphasis on the dissemination of information, as well as its collection, so that the evidence can reach clinical practice. It therefore has commonality with the idea of research-based practice.
Traditional clinical studies use data from collection of paper-based chart record or single database of HIS. Some important information cannot be collected and analyzed in real time. And most doctors in Taiwan have more than 60 visits in 4-hour ambulatory clinic. They have not enough time to give their patient adequate explanation, not to say, the evidence-based practice. In order to improve the quality of patient care in such a busy clinics, we try to evaluate the possibility and difficulty in developing the value-added EBM HIS via the implement of real-time analysis of longitudinal patient records.
We use all data of patients receiving extracorporeal shock wave lithotripsy (ESWL) from HIS of two branches of Buddhist Tzu Chi General Hospital (BTCGH) – Dalin and Taipei. Their unscheduled return visits were analyzed as outcome. Those input data were reconstructed into LPR and fed into data-mining tools for analysis. And the artificial neural network-based approaches obtained a highest area under receiver operative characteristics (AUROC=0.87) for those data from Dalin branch. But AUROC decreased to 0.66 after applying the model to Taipei branch. We also find out some important prognosis predictive variable, the stone long-to-short axis ratio and the pretreatment emergency room visits, which were not mentioned in literature.
We concluded that it is beneficial and possible to implement data-mining tool in LPR of HIS. But further effort is needed to translate the idea into a practical software toolbox for actual clinical usage.
|
|---|
