一致性指標（concordance index， C-index）為評估存活模型預測能力的重要工具，主要用於衡量模型能否正確區分事件較早發生之高風險個體與事件較晚發生之低風險個體。現有方法多以右設限資料為主要研究對象，然而在區間設限情境下，由於事件時間僅能觀測於區間內，使一致性評估更加困難。 此外，在實際醫學存活資料中，常伴隨長期追蹤之時間相依共變數，且此類資料通常存在測量誤差（measurement error）與資訊性退出（informative drop-out）問題，使得傳統方法在估計上可能產生偏誤。特別是在區間設限與時間相依共變數同時存在時，若未適當處理測量誤差與潛在長期追蹤過程，將進一步影響模型預測能力之評估。 為克服上述問題，本研究將傳統以右設限為基礎之 一致性指標 推廣至區間設限資料，並進一步延伸至時間相依情境，結合聯合模型（Joint Model），同時納入長期追蹤與存活資訊，以刻畫個體潛在風險隨時間之變化。透過聯合模型，可有效處理測量誤差與資訊性退出所帶來之偏誤，進而提升一致性指標之估計品質。 模擬結果顯示，所提出方法在不同設限情境下具有較佳之一致性評估表現。最後，將此方法應用於 HIV 實際資料分析，以驗證其實用性與穩健性。;The concordance index (C-index) is a widely used measure for evaluating the discriminative ability of survival models， indicating whether a model can correctly distinguish individuals at higher risk of earlier event occurrence from those at lower risk. Although most existing methods are developed for right-censored data， concordance assessment becomes more challenging under interval censoring because the exact event time is not directly observed. In medical studies， survival outcomes are often accompanied by longitudinally measured time-dependent covariates. In such settings， conventional methods may yield biased results due to measurement error and informative drop-out， especially when interval censoring and time-dependent covariates are present simultaneously. To address these challenges， this study extends the traditional C-index to interval-censored data and further develops a time-dependent C-index through a joint modeling framework. By jointly modeling the longitudinal and survival processes， the proposed approach captures the dynamic evolution of individual risk over time and improves the estimation of concordance. Simulation results show that the proposed method provides better concordance assessment under different censoring scenarios. The method is further applied to real HIV data to demonstrate its practical usefulness and robustness.
