探討台灣醫院照護流程組合分數和病患預後之關係-以急性心肌梗塞為例

Abstract

前言：為了更廣範及便利地呈現照護流程測量，以及機構間比較的需要，組合測量是一種新的評估照護品質的方法，正大量的被發展及應用。然而，臨床上應用時卻出現了部分缺點。一、不同的組合方式會導致不同的機構排名。二、組合分數和其他構面指標(例如結果)間的關係仍不清楚。 目的：探討病人層級及醫院層級急性心肌梗塞照護流程組合分數和病人短期預後間的關係，並比較不同方法醫院層級組合分數間的關係。最終希望能找出一種最好的組合分數方法能代表醫院在急性心肌梗塞的照護品質。 方法：資料來源為台灣國家健康保險研究資料庫2005年1月1日到2009年12月31日第一次急性心肌梗塞(國際疾病分類代碼，第9版，ICD-9-CM 410.xx，排院410.x2)的病人。組合流程分數是依據六個照護流程指標(包含阿斯匹靈使用，乙型阻斷劑使用，Statin類降血脂藥物使用，左心室功能評估，左心室心輸功能不良病人使用ACEI/ARB類藥物，純藥物治療病人使用Clopidogrel)所完成並分別計算出三種病人層級組合分數(包含病人平均patient average，70%標準，及全有全無All-or-none) 及四種醫院層級組合分數（包含全平均overall average，簡單權重平均simple weighted average，全有全無All-or-none，及因素分析Factor analysis）。預後則選用住院後30天內死亡及出院後30天內再住院。多因子羅吉斯迴歸用來檢驗病人層級組合分數和病人預後間的關係，而醫院層級組合分數和病人預後間的關係則利用多層次階層羅吉斯迴歸完成。 結果：於2006年1月1日至2009年9月30日，分別有31899及25119名病人資料分別用於住院後30天內死亡及出院後30天內再住院分析。全部的病人分佈於19間醫學中心、71間區域醫院及34間地區醫院。三種病人層級組合分數的表現均和病人住院後30天內死亡及出院後30天內再住院有關。全體醫院在四種醫院層級組合分數的表現分別為全平均0.57±0.17，簡單權重平均0.62±0.16，全有全無0.18±0.15，及因素分析0.61±0.17（平均±標準差）。醫院中心在四種組合分數的表現均明顯高於區域及地區醫院。四種組合分數間不論是原始分數或醫院排名上，均高度相關，但不同方法會導致醫院落入不同的等級分類。在多層次分析方面，當病人層級照護品質當成控制變項時，只有區域醫院間，在全平均、簡單權重平均、及因素分析這三種醫院層級組合分數表現較高的醫院，其病人的住院後30天內死亡機率較低。當病人層級組合分數或醫院評鑑層級被當成控制變項時，四種醫院層級組合分數均和病人出院後30天內再住院無關。 結論：醫療服務在價值競爭的觀點上，病人預後是個重要的決定因素。三種病人層級組合分數均能和病人短期預後有關。70%標準在病人層級對和病人預後的關係並不差於全有全無。然而，不同的醫院層級組合分數會導致機構不同的排名，若要在急性心肌梗塞疾病上要使用組合流程分數來當成公開資訊或是論質計酬基礎必須要相當謹慎。進一步發展新的醫院層級組合分數是必須的。

Background: Composite quality-of-care measures have been increasingly developed and applied for public reporting and pay-for-performance initiative. However, some defects have been noted when composite measures were applied in a clinical setting. First, different methods of computing composite scores can lead to unfair ranking. Second, the relationship between the composite score and other indicators, such as patient outcomes, remains unclear. Objective: We explored the relationship between composite process scores, at the patient and hospital levels, and short-term patient outcomes and compared the differences types of hospital-level composite score. Ultimately, we intended to determine the composite score that most accurately represents the quality of hospital care for patients with acute myocardial infarction (AMI). Methods: All patients who were admitted for AMI (International Classification of Diseases, Ninth Revision, Clinical Modification 410.xx [excluding 410.x2]) for the first time in Taiwan between January 1, 2005 and December 31, 2009 were identified in the National Health Insurance Research Database. Six process indicators (ie, Aspirin usage, Beta-blocker usage, Statin usage, Left ventricle function evaluation, ACEI/ARB usage on left ventricular systolic dysfunction, and Clopidogrel usage on medical treatment) were used to assemble the composite process scores. Three methods for calculating patient-level composite scores (ie, patient average, 70% standard, and all or none) and 4 methods for calculating hospital-level composite scores (ie, overall average, simple weighted average, all or none, and factor analysis) were employed. The outcomes were 30-day mortality after hospitalization and 30-day all-cause readmission after discharge. Multivariate logistic regression was applied to examine the relationship of the patient outcomes to the patient-level composite process measures. A multilevel hierarchical logistic regression model was applied to examine the relationships between the patient outcomes and factors at the two levels. Results: Between January 1, 2006 and September 31, 2009, we identified 31 899 patients with 30-day mortality and 25 119 patients with all-cause readmission. The patients were distributed among 19 medical centers, 71 regional hospitals, and 34 district hospitals. All 3 patient-level composite scores were inversely related to 30-day mortality and all-cause readmission in the multivariate logistic regression. The mean ± standard deviation was calculated for each method for calculating hospital-level composite scores: overall average (0.57 ± 0.17), simple weighted average (0.62 ± 0.16), all or none (0.18 ± 0.15), and factor analysis (0.61 ± 0.17). The 4 hospital-level composite scores of medical centers were significantly higher than those of the regional and district hospitals. The 4 hospital-level composite scores, including both the raw scores and those determined according to hospital rankings, were highly correlated to each other, but using different methods caused the hospitals to be categorized into different categories. In the multilevel analysis, only the hospital-level composite overall average, simple average, and factor analysis scores for the regional hospitals were inversely associated with patient 30-day mortality when the patient-level quality of care was controlled for. The 4 hospital-level composite scores were not correlated to patient 30-day all-cause readmission when the hospital accreditation level or patient-level quality of care were controlled for. Conclusion: Because of value competition among medical services, patient outcomes are a critical factor. All 3 patient-level composite scores can be used to represent patient-level quality of care related to short-term outcomes for patients with AMI. 70% standard is not inferior to all-or-none in patient-level. The hospital-level composite scores were related to patient outcomes only under certain conditions, and using different composite scoring methods might lead to different rankings. Selecting methods for public reporting or pay-for-performance initiatives for patients with AMI should be considered carefully. Further research on developing new hospital-level composite quality scoring is warranted.