以ICD-10疾病診斷系統建立之多重共病衰弱指數與其於心衰竭病患之臨床應用

Abstract

研究背景 衰弱是心臟衰竭病人中重要的議題，由於心臟衰竭被診斷出來多屬於老年時期，因此心臟衰竭的病人也有一半的人伴隨有衰弱的表徵。而衰弱與心臟衰竭並存不僅會對個人造成不良影響，也可能增加醫療利用負擔。如何評估並量化衰弱之於心臟衰竭病人的影響，例如死亡、再住院與處方型態，在現今的高齡化社會格外重要。 研究目的 本研究擬利用ICD-10版本診斷碼建立多重共病衰弱指數(mFI-v10)，並應用於老年人和因心臟衰竭而首次住院的病人，同時也比較本次建立之mFI-v10與過去本研究室以ICD-9版本診斷碼建立的多重共病衰弱指數(mFI-v9)，對於因心臟衰竭而首次住院的病人之不良結果、醫療利用、處方型態和衰弱狀態變化之間的關聯。 研究方法 本研究主要分成兩個部分，第一部分會利用健保資料庫建立mFI-v10，，並分析其於老年族群之應用，包含衰弱程度分群及其與不良結果(全因性死亡、非計劃性住院和ICU住院)的關聯性。第二部分則是將mFI-v10，進一步應用到心臟衰竭的病人，由於研究目的之一是希望比較不同診斷碼系統下所建立的多重共病衰弱指數的臨床應用性，因此本部分會分成兩個不同世代，分別是2013年世代(ICD-9)與2018年世代(ICD-10)，分別利用mFI-v9與mFI-v10，區分研究對象為健康組(fit)、輕度衰弱(mild frailty)、中度衰弱(moderate frailty)與嚴重衰弱(severe frailty)，並使用Cox風險比例模型探討不同衰弱程度與結果之相關性，包含全因性死亡、全因性再住院、心衰竭再住院及綜合指標（全因性死亡或心衰竭再住院）。處方型態則是使用間斷時間序列分析(interrupted time series analysis)衡量臨床指引建議之心衰竭相關藥物的處方用量變化，並比較發生指標住院事件前後用量的差異。由於衰弱狀態可能隨著時間變動，因此使用桑基圖(sankey plot)來描述追蹤期間衰弱狀態的變化。 研究結果 第一部分研究利用144,567位65-100歲的老年研究對象，所建立之多重共病衰弱指數(mFI-v10)涵蓋38個不同的共病缺陷(deficits)，其中循環系統(circulatory system)相關的缺陷數佔將近1/4。研究對象平均年齡為73.97歲，平均多重共病衰弱指數(mFI-v10)為0.051(標準差為0.048)，有52%的研究對象為輕微衰弱以上，且比率隨著年齡上升提高。 在一年的追蹤期之中，與健康組相比，在校正性別與年齡之後，都可以觀察到，隨著衰弱程度上升，則全因性死亡、非計劃性住院還是ICU住院的風險也跟著提高。相比於健康組，嚴重衰弱組的一年死亡風險為3.86(95% 信賴區間為3.54-4.20)，一年非計畫性住院風險則為3.71(95% 信賴區間為3.49-3.94)，而一年ICU住院風險則為3.98(95% 信賴區間為3.64-4.35)，以上皆達到統計上顯著差異(P<0.0001) 第二部分研究的2013年世代之中，首次因為心臟衰竭而住院的共計44044人，而在2018年世代之中則是50474人，2013年世代平均年齡為74.9歲，而2018年世代為74.7歲。兩個世代中隨著衰弱程度的提高，年齡的平均數與中位數皆越高(P<0.0001)，與2013年世代相比，2018年世代的平均多重共病衰弱指數較高(0.142 vs 0.12)，輕微衰弱以上的比例上升 (78.28% vs 72.45%)。 在本研究中發現，不論是2013年世代或是2018年世代，衰弱會顯著增加全因性死亡、全因性住院、或是心衰竭住院的風險。此外，心臟衰竭合併衰弱的患者在5年的觀察期間，都具有高度醫療資源利用的特性。 發生指標住院事件前兩個月，不論世代皆觀察到不論是用藥比例上或是人數上都有一個較為明顯的攀升，並且在住院當下達到最高峰，雖然出院之後平均心臟衰竭用藥使用種類上升，但是隨著時間不論是用藥人數的比例，還是平均使用的數量都有下降的趨勢，同時也發現不同衰弱程度的病患處方型態有所差異。 心衰竭病人的衰弱狀態會隨著時間變化而有所改變甚至回復，在2013年世代基線原有28%是健康組，到了出院後兩年後約有25%的人處於健康狀態，2018年世代則是基線原有22%，而到了出院後兩年則是34%。 結論: 心臟衰竭住院的病人衰弱盛行率高，且隨著衰弱程度上升，不良結果和資源使用也隨之提高。衰弱狀態會隨著時間變動，甚至也可能回復，故心衰竭合併衰弱的病人應接受更為全面性的照護計畫。藥品處方型態在不同衰弱程度病人有所不同，也使得心衰竭的用藥策略需要朝向更為個人化的方向前進，而mFI-v9和mFI-v10則可以作為臨床結果和醫療資源利用的風險分層工具，應用於一般老年人及心臟衰竭的病人。

Background: Frailty often coexists with heart failure (HF) and may have negative impacts on adverse outcomes among HF patients. How to quantify and characterize the impact of frailty on clinical outcomes among HF patients in an aging society is thus critical. Objectives: Owing to the transition of coding system (from ICD-9-CM to ICD-10-CM) in Taiwan since 2016, we aimed to develop a new version of frailty index, and use the multimorbidity frailty index (mFI) developed under the ICD-9-CM or ICD-10-CM codes to estimate the prevalence of frailty and its impact on mortality and healthcare utilization among HF patients. Changes and associated impact on quantification of frailty among HF patients between versions in 2013 (ICD-9 era) and 2018 (ICD-10 era) were also captured. Methods: In the first part, we updated the multimorbidity frailty index using ICD-10 CM codes (mFI-v10) from Taiwan’s National Health Insurance Research Database (NHIRD) and examined the association between frailty and all-cause mortality, unplanned hospitalization, or ICU admission by the updated mFI-v10. In the second part, we identified patients aged over 40 years old and newly admitted for heart failure (index event), who discharged from the index event in 2013 (ICD-9-CM era) and 2018 (ICD-10-CM era). Frailty was measured by the relevant frailty index developed under ICD-9 CM (mFI-v9) or ICD-10 CM codes (mFI-v10) for HF patients identified in the year 2013 and 2018 cohorts. All study subjects were further categorized into: fit, mild frailty, moderate frailty, or severe frailty based on the quartiles of mFI. Outcomes of interest (all-cause mortality, all-cause readmission, readmission due to heart failure and composite endpoint of all-cause mortality or readmission due to heart failure) were reported. Cox regression models were used to estimate the impacts of frailty on outcomes of interest, and interrupted time series analysis was used to evaluate the prescribing trend between pre-index period and post-index period. We also use sankey plot to describe the transitions of frailty status in the following-up period. Results: In the first part, we updated the mFI-v10 incorporated 38 deficits using ICD-10 codes, with mean mFI-v10 score of 0.051 (standard deviation = 0.048) among 144,567 subjects. Compared with the fit group, those with severe frailty were associated with a 4-fold (adjusted HR 3.86, 95% CI 3.54-4.20) higher risk for death at one year. Subjects with moderate frailty or mild frailty were associated with a 2.4-fold (adjusted HR 2.35, 95% CI 2.18–2.55) or 1.6-fold (adjusted HR 1.57, 95% CI 1.47–1.69) higher risk for death at one year than the fit group. Similar risk trends can also be observed in unplanned hospitalization and ICU (intensive care unit) admission among the study population In the second part, we identified 44,044 and 50,474 HF patients in the year 2013 and 2018, respectively. The proportion of frailty measured by mFI-v9 and mFI-v10 among HF patients were 72.5% in 2013 (mild 42.0%; moderate 24.8% and severe frailty 5.7%) and 78.3% (mild 50.0%; moderate 18.3% and severe frailty 10.0%) in 2018. Compared with the 2013 cohort, 2018 cohort had a higher mean mFI (0.142 vs 0.120). Compared with the fit group, those with severe frailty were associated with a 1.2-fold (adjusted HR 1.24, 95% CI 1.211.15-1.381.35) higher risk for 1-year mortality in the 2013 HF cohort and 1.2-fold (adjusted HR 1.18, 95% CI 1.10-1.27) in the 2018 HF cohort, respectively. Similar risk trends can also be observed in all-cause mortality, all-cause readmission, readmission due to heart failure and composite endpoint of all-cause mortality or readmission due to heart failure among the heart failure patients, irrespective of the study cohort. The average use of both outpatients and emergency visits remained high. An increasing trend of average number of HF medication was observed two months prior the index event, with the greatest number during the index hopsitalzation period. After discharged from the index hospitalization, the average number of HF medication gradually decreased as well as the proportion of patients taking any HF medication in both the 2013 cohort and 2018 cohort. The prescribing pattern varies across frailty status. The transition of frailty status was observed, with the majority of transitions either stay at same stauts or became frailer. Conclusions Using the nationwide, claims-based NHIRD in Taiwan, our study reveals that the proportion of frailty measured by mFI increased in newly admitted HF patients as comparing the 2013 and 2018 HF cohorts. In addition, HF patients with frailty were associated with a higher risks of mortality and readmissions compared with the fit group. Considering the prescribing pattern varies across frailty status, a tailored medication strategy may be warranted. HF care programs adopting mFI for risk stratifications are therefore suggested.