Using the National Health Insurance Research Database to Investigate Whether Using Inhaled Corticosteroids Increases the Incidence of Tuberculosis and Pneumonia in Patients with Chronic Obstructive Pulmonary Disease

View Statistics Email Alert RSS Feed

Information

Details

Project title/計畫英文名

Project Number/計畫編號

NSC99-2314-B002-088-MY2

Translated Name/計畫中文名

利用全民健康保險研究資料庫探討慢性阻塞性肺病患者使用吸入型類固醇是否會增加結核病及肺炎的發生率

Project Principal Investigator/計畫主持人

JANN-YUAN WANG

Funding Organization

National Science and Technology Council

Co-Investigator(s)/共同執行人

胡賦強,李枝新

Website

link

Start date/計畫起

01-08-2010

Expected Completion/計畫迄

12-07-2011

Description

Abstract

摘要：背景：最近的研究發現，針對慢性阻塞性肺病的患者，使用高劑量的吸入型類固醇會增加之後發生肺炎的機率。我們針對一家醫學中心的研究中也發現，慢性阻塞性肺病的患者，若使用高劑量的吸入型類固醇，也會增加肺結核的發生率。目的：我們將利用全民健康保險研究資料庫（NHIRD）的資料，著手分析了解慢性阻塞性肺病的患者使用吸入型類固醇之後，未來發生肺結核以及肺炎的危險是否有影響。資料收集：第一步我們將申請國家衛生研究院的2005 年承保抽樣歸人檔（LHID 2005），這個檔案所包括的個案，是以2005 年承保資料檔中「2005 年在保者」隨機取100 萬人，擷取其由1997 年至2008 年共十二年間就醫資料建置而成。由資料庫中，我們將先挑選出具有慢性阻塞性肺病之診斷並同時服用相關藥物的患者，由當中篩選是否有患者在初次診斷慢性阻塞性肺病的三個月之後罹患結核病或嚴重肺炎。藉由評估患者返診的規則性、使用慢性阻塞性肺病相關藥物的總量、以及急診或住院的頻率，我們將試圖推論患者慢性阻塞性肺病的嚴重度以及服藥的順從性。其中慢性阻塞性肺病相關藥物包括長效型或短效型乙型交感神經興奮劑、長效型或短效型乙醯膽鹼受器阻斷劑、吸入型類固醇、全身性類固醇、以及茶鹼類藥物。評估結果：這個研究有兩個評估的結果，分別是”發生結核病的時間”、以及”發生嚴重肺炎的時間”。前者指的是由第一次診斷慢性阻塞性肺病到之後第一次診斷結核病的時間，後者指的是由第一次診斷慢性阻塞性肺病到之後第一次診斷肺炎住院的時間。統計分析：藉由Kaplan‐Meier 法，我們會描繪出發生結核病與發生嚴重肺炎的時間曲線，並且利用log‐rank 法比較。在多變數分析方面，我們利用Cox proportional hazards 的模式，試圖找出發生結核病與嚴重肺炎的重要影響因子。為了確保迴歸統計的品質，在進行迴歸分析當中，我們將利用basic model‐fitting 的技術，進行分析因子的篩選、adjusted generalized R‐square 的計算、以及迴歸診斷。所使用的迴歸診斷方法包括：確立proportional hazards 的假設可以成立、殘差分析、偵測高影響之數據、以及共線性的診斷。在逐步篩選影響因子時，所有可能的臨床因素都包括在內，顯著值若大於0.15會被排除，反之該因子則被納入。所有統計方法當中均以雙尾p 值小於0.05 定義為顯著差異。所有的分析，都使用SAS（SAS Institute Inc., Cary, NC, U.S.A.）。假說：全民健康保險研究資料庫中的慢性阻塞性肺病患者，使用高劑量吸入型類固醇的患者，日後發生結核病或嚴重肺炎的機會較高。研究目標：1. 評估慢性阻塞性肺病患者使用高劑量吸入型類固醇後是否增加結核病的發生率2. 評估慢性阻塞性肺病患者使用高劑量吸入型類固醇後是否增加嚴重肺炎的發生率
Abstract: Background: The use of high‐dose inhaled corticosteroids (ICS) in patients with chronic obstructivepulmonary disease (COPD) has recently been shown to increase the incidence of pneumonia. Our recentstudy conducted in a medical center also demonstrated that use of high‐dose ICS is associated withpulmonary tuberculosis (TB) in COPD patients.Objective: To study the impact of ICS on the incidence of pulmonary TB and pneumonia in COPD patients,we conduct a study using the National Health Insurance Research Database (NHIRD)Data Collection: We will first obtain the Longitudinal Health Insurance Database (LHID) 2005 from theNational Health Research Institute. LHID 2005 contains the entire original claim data of 1,000,000beneficiaries enrolled in year 2005 randomly sampled from the year 2005 Registry for Beneficiaries of theNHIRD. We will then identify patients with COPD and receiving medical treatment in this database. Amongthem, those who had a diagnosis of TB and severe pneumonia requiring treatment more than 3 monthsafter the diagnosis of COPD will be identified. By assessing the regularity of clinic visit, the consumption ofCOPD medications, and the frequency of emergency room (ER) visit and hospitalization, we will try todefine the severity and COPD and compliance for every patient. The COPD medications includelong‐acting/short‐acting beta‐agonists (LABA/SABA), long‐acting or short‐acting cholinergic receptorantagonists, inhaled corticosteroids (ICS), systemic corticosteroids, and aminophyllines.Outcome Measurement: Time to TB (defined as the interval between the initial diagnosis of COPD and TB),and time to severe pneumonia (defined as the interval between the initial diagnosis of COPD andpneumonia requiring hospitalization).Statistic Analysis: Time‐to‐TB curves and time‐to‐severe‐pneumonia for each potential risk factor will begenerated using the Kaplan‐Meier method and compared using log‐rank test. In multivariate analysis fortime to develop TB and severe pneumonia, Cox proportional hazards model will be used. Basicmodel‐fitting techniques for regression analysis, including variable selection, adjusted generalized R‐squareand regression diagnostics (verification of proportional hazards assumption, residual analysis, detection ofinfluential cases, and check for multicollinearity), will be applied to assure the quality of analysis. Instepwise variable selection procedure, all the potential predictors will be included. Significance levels forentry and for stay are set at 0.15. A two‐sided p value < 0.05 is considered significant. All analyses will beperformed using SAS (SAS Institute Inc., Cary, NC, U.S.A.).HYPOTHESIS: In COPD patients enrolled in the National Health Insurance Research Database, use ofhigh‐dose ICS is associated with pulmonary TB and severe pneumonia.SPECIFIC AIMS:1. Using a nationwide database to evaluate if using high‐dose ICS for COPD increases the risk oftuberculosis2. Using a nationwide database to evaluate if using high‐dose ICS for COPD increases the incidence ofpneumonia requiring hospitalization