探討癌症病人五年內罹患焦慮症與憂鬱症之預測模式

Prediction Model of Anxiety Disorders and Depression Disorders among Cancer Patients after Diagnosis in Five Years

石鈺齊(Yu-chi Shih)

指導教授 ： 許弘毅

高雄醫學大學/健康科學院/醫務管理暨醫療資訊學系碩士班/碩士(2019年)

摘要

研究目的
根據統計資料，罹患癌症的人數有逐年上升的趨勢，這些病人常常擔心隨時會復發，導致焦慮與憂鬱的狀況，因此，本研究乃針對癌症病人確診之後，利用類神經網路（ANN）、鄰近節點演算法(KNN)、支持向量機(SVM)及單純貝氏分類(NBC)預測模式評估罹癌病人五年內罹患憂鬱症與焦慮症之準確性，進而評估影響罹癌病人五年內罹患憂鬱症與焦慮症之重要影響因子。
研究方法
本研究架構以「全民健康保險學術資料庫」為研究樣本，採回溯性研究設計，研究期間為1997至2013年，研究對象為年齡≧18歲之所有癌症病人，樣本數共54697人。首先，利用病人特性、臨床特性、醫療機構特性來探討病人罹癌後五年內罹患焦慮症與憂鬱症之重要影響因子，另外利用類神經網路、鄰近節點演算法、支持向量機及單純貝氏分類，比較罹癌後五年內焦慮症與憂鬱症之準確性。
研究結果
本研究發現，癌症病人之病人特性、臨床特性及醫療機構特性與罹癌後五年內焦慮症與憂鬱症具有顯著性相關(P<0.05)，本研究利用類神經網路、鄰近節點演算法、支持向量機及單純貝氏分類作為預測模型之建構，結果顯示類神經網路在敏感性、特異性、陽性預測值、陰性預測值、準確性及曲線下面積等績效評估指標皆較鄰近節點演算法、支持向量機及單純貝氏分類等預測模式為佳。經由全域敏感度分析，類神經預測模式顯示醫院服務量是病人罹癌後五年內焦慮症與憂鬱症之最重要影響因子(VSR=9.15)，其次為醫院層級及醫師服務量。
結論與建議
從本研究結果可以證明類神經網路預測模式優於鄰近節點演算法、支持向量機及單純貝氏分類等預測模式，故醫學相關研究可多運用此模式做為臨床的評估及決策參考，找出最重要的預測因子，以提升醫療品質。

關鍵詞：癌症、焦慮症、憂鬱症、類神經網路

關鍵字

癌症 ； 焦慮症 ； 憂鬱症 ； 類神經網路

並列摘要

Purpose :
Anxiety and depression are associated condition due to fear of disease recurrence among cancer patients. Therefore, this study will identify the best prediction models and the significant predictors of postoperative 5-year anxiety disorder and depression disorder among cancer patients by using the artificial neural network (ANN), k-nearest neighbor (KNN), support vector machine (SVM) and naive bayes classifier (NBC) prediction models.

Methods :
This is a population-based retrospective longitudinal study using the data from National Health Insurance Research Database (NHIRD) during 1997 to 2013. The 54,697 cancer patients aged more than 18 were included into the study. The patient attributes, clinical attributes and institutional attributes were used to evaluate important predictors of postoperative 5-year anxiety disorder and depression disorder among cancer patients. The ANN, kNN, SVM and NBC prediction models were employed to evaluate the accuracy.

Results:
It showed that patient attributes, clinical attributes and institutional attributes were significantly associated with postoperative 5-year anxiety disorder and depression disorder in cancer patients (P<0.05). It also showed that the ANN model is better than the KNN, SVM and NBC models in prediction of postoperative 5-year anxiety disorder and depression disorder among cancer patients in sensitivity, specificity, positive predictive values, negative predictive values, and receiver operator characteristic curves. Additionally, through global sensitivity analysis, the ANN predictive model showed that the best important predictors for postoperative 5-year anxiety disorder and depression disorder among cancer patients is hospital volume, following by hospital level and surgeon volume.

Conclusions and suggestions:
It found that the ANN predictive model is better than the KNN, SVM and NBC models in prediction of post-operative 5-year anxiety and depression. Therefore, this predictive model is recommended to use in future related medical research. It will be helpful that clinical evaluation and policy making to find the important predictors and to improve the medical quality.

Key Words: cancer; anxiety disorder; depression disorder; Artificial Neural Network

並列關鍵字

cancer ； anxiety disorder ； depression disorder ； Artificial Neural Network

參考文獻

英文文獻

1. Amato, F., López, A., Peña-Méndez, E. M., Vaňhara, P., Hampl, A., & Havel, J. (2013). Artificial neural networks in medical diagnosis.47-58.

2. Andersen, B. L., Rowland, J. H., & Somerfield, M. R. (2014). Screening, assessment, and care of anxiety and depressive symptoms in adults with cancer: an American Society of Clinical Oncology guideline adaptation. Journal of oncology practice, 11(2), 133-134.

3. Bray, F., Ferlay, J., Soerjomataram, I., Siegel, R. L., Torre, L. A., & Jemal, A. (2018). Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: a cancer journal for clinicians, 68(6), 394-424.

4. Beura, S., Majhi, B., & Dash, R. (2015). Mammogram classification using two dimensional discrete wavelet transform and gray-level co-occurrence matrix for detection of breast cancer. Neurocomputing, 154, 1-14.