高風險藥物與急性住院：病例─時間─對照研究

Abstract

研究背景 藥物不良事件會惡化病人的健康情況，也會為有限的醫療資源帶來沉重的負擔，若能有效率地找出可能導致嚴重藥物不良事件(如：藥物不良事件引起之住院)的高風險藥物，將有助臨床工作者與政策制定者針對此議題進行改善。過去研究探討不同藥物相關之住院風險此一議題時，多採用分析自動通報紀錄、病歷回顧、以及從病歷資料庫中篩選藥物不良反應相關之診斷碼等方式進行，然而這些方式普遍存在通報不足的問題且各個藥物不良反應之特性會影響通報率；此外，多數研究僅呈現藥物引起之住院事件數而未考慮藥物在整體族群之使用盛行率。 研究目的 為解決上述的研究限制，本研究使用臺灣全民健康保險研究資料庫為資料來源，以病例─時間─對照設計 (case–time–control design) 探討高風險藥物和急性住院事件之間的相關性。 研究方法 由2000年承保抽樣歸人檔中選取年滿20歲且2000年1月1日至2011年12月31日至少有一筆門診處方紀錄者為研究族群。於研究族群中篩選出急性住院事件為index visits而門診就醫事件為reference visits，並將index visits與reference visits以性別、年齡、index date (定義為index 及reference visits 的首日)、Charlson Comorbidity Index、門診次數進行1:1配對。Index visits與reference visits之病例期定為index date 前1-14天，對照期則定為index date 前366-379天，再比較高風險藥物分別在病例期與對照期之暴露情形，本研究選定的高風險藥物為diabetic agents、diuretics、nonsteroidal anti-inflammatory drugs (NSAIDs)、anticoagulants、antiplatelets、antihypertensives、 antiarrhythmics、anticonvulsants、antipsychotics、antidepressants、benzodiazepine (BZD)/Z-hypnotics以及narcotics。統計分析使用條件式邏輯迴歸模型進行odds ratio (OR)的預測，並校正隨時間改變之干擾因子 (包含病例期和對照期之Charlson Comorbidity Index、門診就診次數、急診就診次數、慢性用藥總數)，另計算急性住院事件中與高風險藥物相關之可歸因風險 (attributable fraction (AF))。 研究結果 於12種本研究選定的高風險藥物中，急性住院風險顯著增加者有7種，風險高至低依序為antipsychotics (adjusted OR: 1.54, 95% confidence interval [1.37-1.73], AF: 35.0%)、NSAIDs (1.50, [1.44-1.56], 33.3%)、anticonvulsants (1.34, [1.10-1.64], 25.6%)、diuretics (1.24, [1.15-1.33], 19.1%)、BZD/Z-hypnotics (1.23, [1.16-1.31], 18.8%)、antidepressants (1.17, [1.05-1.31], 14.7%)及antiplatelets (1.16, [1.07-1.26], 14.0%)。anticoagulants、antiarrhythmics、antihypertensives、narcotics與急性住院無顯著之相關性；diabetic agents則顯示保護作用 (0.86, [0.77-0.97])。 以全部急性住院事件進行分析時，使用narcotics之風險增加未達統計顯著，但僅分析住院天數≥10天之急性住院事件時，風險顯著增加。將各個高風險藥物細分為不同子類別進行分析時急性住院風險不同，急性住院風險最高之antipsychotics及NSAIDs中，又以typical antipsychotics及non-selective NSAIDs之風險較高。依年齡分層分析時，antipsychotics、NSAIDs、diuretics、BZD/Z-hypnotics及antiplatelets在<65歲者和≥65歲者之急性住院風險皆顯著增加，anticonvulsants和antidepressants則只在≥65歲者和急性住院有顯著相關性。 研究結論 本研究提供一種全新的「主動監測藥物不良事件」之方式，與現有之被動監測機制互補可提供更完整的資訊，本研究中急性住院風險顯著增加之藥物antipsychotics、NSAIDs、anticonvulsants、diuretics、BZD/Z-hypnotics、antidepressants及antiplatelets等，可作為未來臨床研究與政策執行之目標。

Background: Adverse drug events (ADEs) could threaten the health of the patients and imposed a substantial economic burden on the healthcare system. Therefore, efficient identification of severe ADEs, such as ADE-related hospitalizations, can help highlight area in which improvement of treatment practices clinicians and policy-makers can put efforts in. Previous studies have adopted several approaches to identify ADE-related hospitalizations, such as analysis of spontaneous reporting data, medical chart review, and screening diagnostic codes from electronic medical charts databases. However, these methods were limited by under-reporting and reporting bias. Furthermore, most of the studies only examined the number of ADE-related hospitalizations, and did not consider the prevalence of medication exposure within the population. Objectives: To address the limitations of current approaches, we used a case–time–control design to evaluate the association between high-risk medications and emergency hospitalizations from the National Health Insurance Research Database (NHIRD) in Taiwan. Methods: Using data from the Longitudinal Health Insurance Database (LHID) 2000, patients aged 20 years and over and had received at least one outpatient prescription during 2000 to 2011 were included as our study cohort. Among them, emergency hospitalizations and outpatient visits were identified as index visits and reference visits, respectively. The first date of index and reference visit was defined as the index date. Each index visit was then matched to a randomly selected reference visits by age, gender, index date, Charlson Comorbidity Index and number of outpatient visits. For both index and reference visits, the period of 1-14 days and 366-379 days prior to the index date were defined as case period and control period, respectively. The exposure of high-risk medications during the case period and the control period were then compared. High-risk medications included in our study were diabetic agents, diuretics, nonsteroidal anti-inflammatory drugs (NSAIDs), anticoagulants, antiplatelets, antihypertensives, antiarrhythmics, anticonvulsants, antipsychotics, antidepressants, benzodiazepine (BZD)/Z-hypnotics and narcotics. Conditional logistic regression models were used to estimate odds ratios (ORs) with adjustment for time-varying variables, including Charlson Comorbidity Index, number of outpatient visits, number of emergency visits and number of chronic drugs used. Attributable fractions (AFs) were also calculated to determine the proportion of emergency hospitalizations attributable to exposure of high-risk medications. Results: Overall, antipsychotics (adjusted OR: 1.54, 95% confidence interval [1.37-1.73], AF: 35.0%), NSAIDs (1.50, [1.44-1.56], 33.3%), anticonvulsants (1.34, [1.10-1.64], 25.6%), diuretics (1.24, [1.15-1.33], 19.1%), BZD/Z-hypnotics (1.23, [1.16-1.31], 18.8%), antidepressants (1.17, [1.05-1.31], 14.7%) and antiplatelets (1.16, [1.07-1.26], 14.0%) were significantly associated with increased risks of emergency hospitalizations. Such associations were not found in anticoagulants, narcotics, antiarrhythmics and antihypertensives. Diabetic agents showed protective effect (0.86, [0.77-0.97]). However, when we limited our analyses in emergency hospitalizations with a length of stay ≥10 days, the narcotics significantly associated with an increased risk of emergency hospitalizations. In subgroup analysis by pharmacological properties, typical antipsychotics were associated with higher risks of emergency hospitalizations than atypical antipsychotics. Non-selective NSAIDs was linked with higher risk of emergency hospitalizations than COX-2 selective NSAIDs. Age-stratified analyses showed that antipsychotics, NSAIDs, diuretics, BZD/Z-hypnotics, and antiplatelets were significantly associated with emergency hospitalizations in aged<65 years and aged ≥65 years, but anticonvulsants and antidepressants were associated with increased risk of emergency hospitalizations in the elderly (age ≥65 years) only. Conclusion: This study provides an alternative approach for active pharmacovigilance, which may complement the present passive pharmacovigilance system in Taiwan. Future research can focus on medications significantly associated with increased risks of emergency hospitalizations, including antipsychotics, NSAIDs, anticonvulsants, diuretics, BZD/Z-hypnotics, antidepressants and antiplatelets.