以類馬可夫模式評估疾病風險

Abstract

本研究以部分設限資料之類馬可夫模型為基礎，並根據健保資料庫蒐集之數據直接驗證此模型之可行性。\n\n根據研究顯示，類馬可夫模型可以在醫學研究中分析患者狀態轉移的過程，且醫學研究中時常存在病程不完整的情況，所以我們認為部分設限資料之類馬可夫模型非常適合被應用在醫學研究中。希望透過現有的健保資料庫數據，將模型與實際數據作結合，活化醫療數據和使用。\n\n驗證結果顯示，此估計模型在資料完整情況下是相當好的估計方法; 而在設限情況下透過模型估計出來的轉移機率與實際轉移機率無太大差異，所以此估計模型確實可以用來估計我們所感興趣的轉移機率。並且，雖然資料完整未必在所有疾病都可準確估計，但可以看出整體趨勢往實際數值靠近。

This thesis is based on the semi-Markov models for partially censored data. Data from the National Health Insurance Research Database are used to evaluate the feasibility of the model.\n\nAccording to the research, semi-Markov models can be used to analyze the process of state transitions of the patient. However, the patient history in medical research is sometimes incomplete. We evaluate the semi-Markov models for partially censored data and find it can greatly fit for medical research. Patient data is extracted from National Health Insurance Research Database, enhancing the sustainability of medical data and application.\n\nThe verification result shows that this model performs well when the data is complete. Meanwhile, the estimate of transition probability under the censored situation is nonsignificantly different compared to the case with complete information. We can conclude that this model is suitable to estimate the transition probability that we are interested in. Still, although the completeness of information may not always induce precise prediction of all risks, but the approximation by the model correctly reflects the trend.