本文已於2026/4/30更新為更正版。原刊載之部分數值與文字已依勘誤說明更正。詳情請參閱Vol.45 No.2勘誤說明：http://doi.org/10.6288/TJPH.202504_45(1)114074_Corrigendum 或 https://toaj.stpi.niar.org.tw/index/journal/volume/article/4b1141f99df306a6019df6522cde0067

目標：採用健保資料庫與死因檔，建構初發缺血性腦中風的風險系統。方法：採用2011-2019年資料，建立2012-2019年台灣全人口的成人動態世代。涵蓋追蹤期間11個危險因子及缺血性腦中風的每日狀態。執行負二項迴歸，透過點數型風險評分演算法得出風險總分。接著利用分類迴歸樹對風險總分進行分級。結果：總追蹤14,281萬人年，初發缺血性腦中風的發生率為每十萬人年278.58例。風險總分在0至46分間，50-64歲得5分，65-79歲得9分，80歲以上得13分。心房顫動9分、心臟衰竭5分、高血壓5分、糖尿病3分、慢性腎臟疾病3分、周邊動脈阻塞疾病3分、心肌梗塞2分、冠狀動脈疾病2分、男性1分。此模型的解釋力達65.26%。經分類迴歸樹分析後，得3個風險分級：第1級（0-15分）、第2級（16-23分）和第3級（24-46分），此分級模型的解釋力達51.02%。結論：研究結果可協助民眾瞭解個人缺血性腦中風風險，進而採取預防措施。為政府機關與醫療單位提供科學依據，制定更有效的公共衛生策略與優化醫療資源的分配。

This article was corrected on April 30, 2026. Please see the Corrigendum here: http://doi.org/10.6288/TJPH.202602_45(1).114074_Corrigendum or https://toaj.stpi.niar.org.tw/index/journal/volume/article/4b1141f99df306a6019df6522cde0067

Objectives: We developed a risk stratification tool for first-onset ischemic stroke by using data from the National Health Insurance Research Database and Cause of Death Data. Methods: A dynamic cohort of the entire Taiwanese adult population (2012–2019) was established using the aforementioned data (2011–2019). Daily data on 11 risk factors and ischemic stroke were collected. Negative binomial regression was performed to estimate regression coefficients, and a point-based risk-scoring algorithm was used to calculate a total risk score. Classification and regression tree analysis was performed using the total risk score to derive risk strata. Results: During 142.81 million person-years of follow-up, the incidence of first-onset ischemic stroke was 278.58 per 100,000 person-years. Total risk scores ranged from 0 to 46. Key contributors included age (50–64 years, 5 points; 65–79 years, 9 points; ≥80 years, 13 points), atrial fibrillation (9 points), heart failure (5 points), hypertension (5 points), diabetes (3 points), chronic kidney disease (3 points), peripheral arterial obstructive disease (3 points), myocardial infarction (2 points), coronary artery disease (2 points), and male sex (1 point). The model achieved a pseudo R2 value of 65.26%. Classification and regression tree analysis stratified the total risk score into three levels—Level 1 (0–15 points), Level 2 (16–23 points), and Level 3 (24–46 points), achieving a pseudo R2 value of 51.02%. Conclusions: Our findings may help the public accurately understand individual ischemic stroke risk, enabling them to implement preventive measures. Furthermore, the findings may support government agencies and health-care institutions in developing effective public health strategies and optimizing medical resource allocation.

缺血性腦中風；動態危險因子；台灣健保資料研究庫；負二項迴歸；分類迴歸樹

ischemic stroke; dynamic risk factors; National Health Insurance Research Database; negative binomial regression; classification and regression tree