運用大數據技術於醫療資訊的分析探討

大數據（Big data）是近代最炙手可熱的技術之一，其核心價值在於挖掘數據中的隱藏資訊。以幫助企業進行預測，並制定合宜的策略。本研究觀察到臺灣目前面臨社會高齡化的危機，失智症的議題逐漸受到重視。失智症越早接受治療，控制病情的效果越佳。然而失智症初期的症狀不明顯，導致容易被忽略而錯過最佳診療時期。若能找出有哪些因素和失智症相關，並進行預測，就能使失智症受到更有效的管理。於是本研究將大數據技術，應用於合作醫院的失智症醫療數據。並使用IBM的SPSS統計軟體來進行大數據分析。利用卡方檢定（Chi-squared test），找出失智症的相關變數。發現高血脂和腦中風，和失智症的種類有相關性。有高血脂和腦中風的患者，罹患可逆性失智症的比例較高。也發現教育程度和糖尿病，和失智症的嚴重度有相關性。教育程度較低、有糖尿病的患者，失智嚴重度為中重度的比例較高。再運用邏輯迴歸（Logistic regression），進一步挖掘不同教育程度的患者，以及有無罹患糖尿病的患者，在預測失智嚴重度上的年齡分界點，以利於警示失智症的高危險群。最後用判別分析（Discriminant analysis），找出預測患者有無罹患失智症的判別函數。發現患者的年齡、教育程度、血糖、總膽固醇（Total cholesterol）、三酸甘油酯（Triglyceride）、高密度膽固醇（High-density lipoprotein）和血紅素（Hemoglobin），是較具影響力的判別指標。其中又以年齡和教育程度的影響力最大。推論患者可以透過控制這些判別指標的數值，達到降低失智症的罹患風險的目的。上述分析結果能提供醫護人員或患者一個失智症的評估參考，提前找出罹患失智症的高危險群，幫助患者提早就醫，提高失智症的治療效果。也可以提高患者的警戒心，提前做好防範措施，預防失智症的發生。也改善了失智症被忽略的社會問題。

關鍵字

醫療大數據；失智症；卡方檢定；邏輯迴歸；判別分析； SPSS統計軟體

並列摘要

As technology advances and the Internet matures, the Big Data becomes more popular and important. There is useful information hidden in the big data that requires proper means to dig it out for prediction and decision making. Taiwan faces an aging and the elderly care issue which becomes critical and gets people’s attention. Among aging diseases, dementia is the most popular one. If dementia can be detected earlier, the patient has more chances to be healed. Yet the initial symptom of dementia is not obvious, the patient is easily neglected and lost the best chance to cure the disease. On the other hand, if we can find factors related to dementia, then the disease might be controlled efficiently. We collect the medical data of dementia patients and use IBM SPSS Statistics software to analyze the data. This research finds out that hyperlipidemia and stroke are related to the type of dementia by the chi-squared test. Patients with hyperlipidemia and stroke are more likely to suffer from reversible dementia. The logistic regression is used to find that the patient’s education degree and diabetes are significant factors for predicting the severity of dementia. Patients with lower education and diabetes have a higher risk of dementia. Then, the discriminant analysis is used to predict whether the patient has dementia. The results reveal that age, education, blood sugar, total cholesterol, triglyceride, high-density lipoprotein, and hemoglobin are vital factors to discrete whether patients may get dementia. This research aims to provide the reference criteria for dementia so that physicians can detect the disease accurately and provide early help to reduce the burden on patients’ families. It might also help to alert the patients to take early precautions. Eventually, we hope this research can be used to enhance patients’ health and reduce the medical cost of aging patients with dementia.