以申報資料進行中風嚴重度指標 的開發與驗證之研究

使用申報資料進行中風後果研究時，風險校正是相當困難的，本研究目的為開發並驗證中風嚴重度指標，以供申報資料研究之用。本研究自某醫院中風登錄取得缺血性中風病人同時探勘病人之高維度申報資料，病人實際中風嚴重度是以美國國家衛生研究院中風量表來測量，共使用四種資料探勘之迴歸方法及傳統之多重線性迴歸來建立中風嚴重度預測模式，模式之優劣是以中風量表分數與中風嚴重度指標間之皮爾森相關係數來比較。外部驗證方面，是以其他兩家醫院之中風登錄資料連結健保研究資料庫來取得研究資料，並使用邏輯斯迴歸來探討納入中風嚴重度指標對於死亡之預測能否提高預測效能。本研究共辨識出七項預測特徵，同時發展了五種預測模式，其中k最近鄰居法於衍生組病人上表現最優，至於在驗證組病人上，則以支援向量迴歸法最佳。中風嚴重度指標與入院時之中風量表分數呈高度相關，與中風後三個月、六個月、及一年之日常功能指數亦呈相關。納入中風嚴重度指標於死亡預測模式中，可明顯提高預測效能。

關鍵字

資料探勘；申報資料；急性缺血性中風；疾病嚴重度

並列摘要

Case-mix adjustment is difficult for stroke outcome studies using claims data. This study aimed to develop and validate a claims-based stroke severity index (SSI). We analyzed patients with acute ischemic stroke from a hospital-based registry and explored claims data with high-dimensional features. Stroke severity was measured using the National Institutes of Health Stroke Scale (NIHSS). We used four data mining methods and conventional multiple linear regression to develop prediction models, comparing the model performance according to the Pearson correlation coefficient between the SSI and the NIHSS. We externally validated these models in patients recruited from hospital-based registries linked with the National Health Institute Research Database. To investigate the added predictive ability of the SSI for mortality, we fitted separate logistic regression models with or without the SSI. We identified 7 predictive features and developed 5 models. The k-nearest neighbor model outperformed other models in the derivation cohort and the support vector regression model performed best in the validation cohort. The SSI correlated with admission NIHSS and functional outcomes at 3 months, 6 months, and 1 year after stroke. Mortality models with the SSI demonstrated superior discrimination than those without.

並列關鍵字

data mining ； claims data ； acute ischemic stroke ； disease severity

參考文獻

Liao, C. C., Chang, P. Y., Yeh, C. C., Hu, C. J., Wu, C. H., & Chen, T. L. (2014). Outcomes after surgery in patients with previous stroke. The British journal of surgery.

Aha, D. W., Kibler, D., & Albert, M. K. (1991). Instance-based learning algorithms. Machine Learning, 6(1), 37–66.

Allen, N. B., Holford, T. R., Bracken, M. B., Goldstein, L. B., Howard, G., Wang, Y., & Lichtman, J. H. (2010). Trends in one-year recurrent ischemic stroke among the elderly in the USA: 1994-2002. Cerebrovascular diseases, 30(5), 525–532.

Ananthakrishnan, A. N., McGinley, E. L., Binion, D. G., & Saeian, K. (2010). A novel risk score to stratify severity of Crohn's disease hospitalizations. The American journal of gastroenterology, 105(8), 1799–1807.

Balami, J. S., Chen, R.-L., Grunwald, I. Q., & Buchan, A. M. (2011). Neurological complications of acute ischaemic stroke. Lancet neurology, 10(4), 357–371.

國際替代計量

以申報資料進行中風嚴重度指標的開發與驗證之研究

未授權

主題瀏覽