建構乳癌治療處置與存活分析於視覺化儀表板工具：以台灣單一醫療機構為例

Abstract

背景：乳癌為台灣女性發生率最高之癌症，乳癌防治為一重要公共衛生議題，針對不同的分子亞型族群提供適切的治療處置，可以改善乳癌病人的預後，存活率為癌症治療的常見指標。病人的預後結果為治療品質的指標，傳統靜態報表難以即時呈現不同乳癌病人族群之預後資訊，建構互動式視覺化分析工具，有助於提升臨床醫療團隊資料查找之效率與便利性。 目的：本研究以台灣某醫療機構乳癌登記資料庫為基礎，分析分子亞型與治療處置對存活預後之影響，建構Power BI互動式視覺化儀表板工具，提供描述性及分析結果，供臨床團隊動態查找乳癌個案存活率等重要醫療品質指標。 方法：納入2011–2022年間診斷之乳癌個案，進行描述性統計，並以Kaplan-Meier法估計存活率，同時依分子亞型分別建立Cox比例風險模型，分析各分子亞型之各治療處置存活分析結果。將上述分析結果匯入Power BI，建構包含人口學分布、存活率分析及校正後存活曲線等功能區塊之互動式視覺化儀表板。 結果：本研究共篩選納入7,686名乳癌病人，整體5年、10年存活率分別為90.0%、81.8%。依分子亞型分層之Cox模型顯示，HR+/HER2-亞型（n=4,861）荷爾蒙治療（HR=0.40, 95% CI: 0.28–0.59, p<0.001）與放射治療（HR=0.57, 95% CI: 0.46–0.70, p<0.001）具顯著保護效果；HR+/HER2+亞型（n=830）化學治療（HR=0.45, 95% CI: 0.28–0.73, p=0.001）及荷爾蒙治療（HR=0.55, 95% CI: 0.34–0.89, p=0.015）達統計顯著保護效果；HR-/HER2-亞型（n=717）化學治療（HR=0.57, 95% CI: 0.36–0.89, p=0.015）與放射治療（HR=0.59, 95% CI: 0.40–0.87, p=0.008）具顯著保護效果；HR-/HER2+亞型（n=681）各項治療處置之風險比點估計方向雖多呈保護趨勢，但均未達統計顯著。 互動式儀表板整合上述分析結果，使臨床團隊可依分子亞型、病理分期及治療處置等任一維度，動態查找病人之描述性統計結果、查詢不同子族群之存活率，並可查詢不同分子亞型中各項治療處置之校正後存活分析結果。 結論：本研究建構之乳癌治療處置與存活率分析視覺化儀表板工具，透過臨床問卷回饋顯示使用者對儀表板易用性與整體滿意度評價良好，滿意度問卷調查平均4.7–4.9分(滿分5分），可以互動方式查找各項重要乳癌品質指標，有助於提升臨床團隊查找資訊之便利性與時效性，可作為醫療品質監測之實務工具。

Background: Breast cancer is the most common cancer among Taiwanese women, and breast cancer control is a key public health issue. Tailoring treatment to molecular subtype can improve prognosis, and survival rate is a common outcome indicator reflecting care quality. Static reports cannot provide real-time prognosis information across patient groups; an interactive visualization tool can improve the efficiency and convenience of clinical data retrieval. Objective: Using a breast cancer registry from a medical institution in Taiwan, this study examined the effects of molecular subtype and treatment on overall survival, and developed an interactive Power BI dashboard providing descriptive and analytic results for clinical teams to dynamically retrieve key quality indicators such as survival rates. Methods: Patients diagnosed between 2011 and 2022 were included. Descriptive statistics were performed, and overall survival was estimated using the Kaplan-Meier method. Subtype-specific Cox proportional hazards models were separately built for each molecular subtype to analyze treatment-specific survival outcomes. Results were imported into Power BI to build an interactive dashboard comprising demographic distribution, survival analysis, and adjusted survival curve modules. Results: A total of 7,686 patients were included after screening; the overall 5-year and 10-year survival rates were 90.0% and 81.8%. Subtype-stratified Cox models showed significant protective effects for hormone therapy (HR = 0.40, 95% CI: 0.28–0.59, p < 0.001) and radiotherapy (HR = 0.57, 95% CI: 0.46–0.70, p < 0.001) in HR+/HER2- (n = 4,861); for chemotherapy (HR = 0.45, 95% CI: 0.28–0.73, p = 0.001) and hormone therapy (HR = 0.55, 95% CI: 0.34–0.89, p = 0.015) in HR+/HER2+ (n = 830); and for chemotherapy (HR = 0.57, 95% CI: 0.36–0.89, p = 0.015) and radiotherapy (HR = 0.59, 95% CI: 0.40–0.87, p = 0.008) in HR-/HER2- (n = 717). No treatment reached statistical significance in HR-/HER2+ (n = 681), though protective trends were observed across treatments. The interactive dashboard integrates these results, allowing clinical teams to dynamically retrieve descriptive statistics, subgroup survival rates, and treatment-specific adjusted survival results by molecular subtype, pathological stage, or treatment. Conclusion: A usability survey showed high user satisfaction with the dashboard's ease of use and overall performance (mean scores of 4.7–4.9 out of 5). This dashboard enables interactive lookup of key breast cancer quality indicators, improving convenience and timeliness for clinical teams, and serves as a practical tool for quality-of-care monitoring.