以存活分析法比較疾病之住院日

目的：比較醫院間疾病別住院日數宜採算術平均、幾何平均、存活分析，還是無母數分析，並對傳統醫院管理採算術平均做為時間資料的比較，提出建議。 方法：利用2004年私立大學及財團法人醫學中心健保住院為蜂窩性組織炎（IDC-9碼為6826；DRG碼為278）的1,509件案件，以算術及幾何平均與存活分析比較資料檢定上的差異，期以探討：（1）醫院群組間住院日數比較統計方法的選擇； （2）存活分析設限與未設限條件下對機率（生命）函數的影響。 結果：存活分析是處理設限資料的方法。若設限資料比例過少，分析的結果與其他方法的結果類似，且其結論相同。本文用蜂窩性組織炎住院日數所得之結果，做此應證。 結論：醫院評鑑中規範許多住院日數的分析與比較，若設限比例高及設限機制與住院時間獨立時可考慮採用存活分析方法分析住院時間資料。
Objectives: Length of stay (LOS) is an important indicator in health care and hospital management. The aim of the study is to investigate whether arithmetic mean and geometric mean are suitable for comparing lengths of stay, especially when there is not a normal distribution. Survival analysis was utilized in this study for assessment of LOS for cellulitis.
Methods: A total of 1,509 discharged cases of cellulitis, coded with DRG 278 and IDC-9 6826, from Taiwan's private-run university and non-profit medical centers in 2004 were analyzed. We compared survival analysis with and without data outliers, uncompleted medication and treatment discharge, death and referral with three methods of arithmetic mean, geometric mean and non-parameter approach; and examined differences that existed between the types of medical centers.
Results: Significant difference was exhibited in mean calculations with distinct statistical methods in original, non-normal distribution data; while discharged with uncompleted medical procedures were of concerns. Censored data in survival analysis should be examined thoroughly to obtain informed and interpreted results that are more accurate.
Conclusions: Hospital accreditation stipulated lengths of stay must be analyzed and compared to improvements in medical and service quality. Survival-time data used with the Mantel and Haenszel (Log Rank) test could help hospital staff produce survival curves and covariate coefficient estimates to sufficiently facilitate high quality research studies of lengths of stay for diseases, and in wards or hospitals. Copyright © 2009 臺灣醫務管理學會.