針對加護病房肺炎病患的抗生素選擇量表之研發

The Development of a Scoring Tool for Accessing the Risk of Multiple Drug Resistant Pathogens in Patients with Pneumonia Presenting To Medical Intensive Center Units

黃聖峰(Sheng-Feng Huang)

指導教授 ： 林英琦

高雄醫學大學/藥學院/藥學系臨床藥學碩士班/碩士(2015年)

摘要

研究背景:
肺炎是發生率高，死亡率亦高的感染症。研究發現使用不適當抗生素會造成死亡率的上升，尤其在重症病人(例如:需要住進加護病房)更是如此，但若為減少不適當抗生素的使用而廣泛給予抗菌範圍更廣的抗生素，可能造成未來的多重抗藥性而惡化治療成效。目前許多文獻指出美國感染症學會(IDSA)臨床指引評估病人感染多重抗藥性肺炎的方式相較於預測量表的評估方式是較不準確的，建議改以量表之方式評估，然而，由於國內外的國情不同，且針對疾病嚴重程度較高的肺炎族群尚未有研究探討，因此在量表的外推性備受質疑下，直接應用國外發展之量表於高死亡率的族群將伴隨著極大的風險。
研究目的與目標:
本研究的目標是針對南台灣某醫學中心之內科加護病房研發一套簡單、精準、方便使用的抗藥性預測量表，希望作為臨床醫師在重症肺炎族群評估是否為高風險多重抗藥性肺炎的參考指標，以提供抗生素經驗療法的選擇依據以增進臨床治療成效。
研究方法:
本研究針對2012至2013年間的高醫內科加護病房肺炎病人執行回溯性世代研究(retrospective cohort study)，藉由回溯高醫之醫院研究資料庫與高醫電子病歷取得研究變項資料，再利用逐步挑選利用羅吉斯回歸(stepwise logistic regression)與重複隨機抽樣驗證法(repeated random sub-sampling validation)找尋病人的多重抗藥性肺炎之危險因子，並探討各危險因子與多重抗藥性肺炎的相關性，再依據迴歸係數評估法(repeated random sub-sampling validation)依其相關性之強弱研發量表並且確認效果。
研究結果:
本研究納入2012至2013年間的300個肺炎住院次之最常見的感染菌種為Klebsiella pneumoniae (19.3%)，其次分別為Pseudomonas aeruginosa (16.3%)與Acinetobacter baumannii (10.7%)。有95個住院次感染多重抗藥性肺炎佔了總住院次的31.7%，多重抗藥性比例最高的菌種為Acinetobacter baumannii (75.0%)，其次為Stenotrophomonas maltophilia (33.3%)、Staphylococcus aureus (38.1%)與Pseudomonas aeruginosa (36.7%)。使用逐步挑選羅吉斯回歸，發現長期入住安養中心(OR=3.764, 95%CI=2.165-6.544, p<.0001)、長期使用呼吸器(OR=2.81, 95%CI=1.223-6.458, p=0.015)和過去90天內曾經住院25天以上(OR=2.768, 95%CI=1.382-5.545, p=0.0041)為顯著的危險因子。在重複隨機抽樣驗證法中，和逐步挑選羅吉斯回歸相同，發現最重要的三個危險因子依序為長期入住安養中心、過去90天內曾經住院25天以上和長期使用呼吸器。此三個變項在訓練組分別被選為顯著重要變項共985(98.5%)、508(50.8%)與558(55.8%)次，之後在驗證組中分別共有788(80.0%)、61(12.0%)與33(5.9%)次，依然維持顯著性而獲得驗證，相對重要程度分別為78.8%、6.1%與3.3%，在最後的模型中勝算比分別為3.49(95%CI=1.897-6.416, p<.0001)、2.47(95%CI=1.141-5.364, p=0.0218)和3.38 (95%CI=1.334-8.541, p=0.0102)。依照危險因子與多重抗藥性菌種感染之相關性強弱給予量表中各項目的分數，長期入住安養中心、過去90天內曾經住院25天與長期使用呼吸器皆分別給予1分，在計算每個住院次的總分後，發現總分與多重抗藥性肺炎發生的比例成正比關係，總分為0分的病人有16.6%感染多重抗藥性肺炎； 總分為1分的病人有50.0%感染多重抗藥性肺炎； 總分為2分的病人有66.7%感染多重抗藥性肺炎。最後利用ROC曲線計算量表的AUC為0.7117 (95%CI=0.6539-0.7694)顯著大於其他研究所研發之量表。
結論與建議:
本研究為單中心回溯性觀察性研究，我們探討在過去研究中曾被提出之危險因子，並依各個獨立危險因子與抗藥性肺炎的相關性強弱研發簡單且精準量表，發現量表分數與多重抗藥性肺炎呈現正相關，且證明本量表之精準度優於其他研究所研發之量表。相信本研究所研發之簡單、精準、方便使用的量表可作為內科加護病房臨床醫師在重症肺炎族群，評估是否為高風險多重抗藥性肺炎的參考指標，以提供抗生素經驗療法的選擇依據以增進臨床治療成效。但本研究依然有些許之研究限制，期待未來可以朝向多中心前瞻性世代研究的方式，改善外推性與樣本數不足的缺憾，並驗證量表的準確性與評估量表對於臨床的影響性。

關鍵字

肺炎 ； 呼吸道感染 ； 抗生素 ； 抗藥性 ； 評分 ； 預測 ； 量表

並列摘要

Background:
Pneumonia is a leading cause of death in intensive care unit. Proper use of antibiotic to cover multiple drug resistant (MDR) pathogens is essential in reducing mortality in these critically ill patients. The classification of pneumonia in the IDSA guideline attempts to predict infection with MDR pathogen, but the precise of such prediction is unclear. Many scoring tools have been developed to predict MDR for patients with pneumonia. However none of them have been validated in Taiwan population as well as focusing on patients in the medical intensive care unit (MICU), who could have the higher mortality than patients in general wards.
Aims:
The object of this study was to identify the risk factors for multiple-drug resistant (MDR) aand to develop a new scoring tool for MICU patients with pneumonia in Taiwan.
Methods:
We conducted a retrospective cohort study on patients admitted to the MICU of Kaohsiung Medical University Chung-Ho Memorial Hospital (KMUH) with pneumonia diagnosis between January 2012 and October 2013. Patient before admission and during this admission was collected from data base of KMUH and medical record review. Stepwise logistic regression and repeated random sub-sampling validation were used to identify the important independent risk factors. To create an easy to use scoring tool, we adapted coefficient-based scoring method to give each predictor an integer point score according to the association between independent risk factors and pneumonia with MDR pathogens. Finally, we calculated the total risk score by adding the individual scores for each predictor together and used receiver operating characteristic (ROC) curve analysis to assess the accuracy of this new scoring tool.
Result:
During the study period, about 300 episodes of pneumonia with positive culture were analyzed. The first three common pathogen were Klebsiella pneumonia (19.3%), Pseudomonas aeruginosa (16.3%) and Acinetobacter baumannii (10.7%). There were 95 pneumonia episode (31.7%) infected by multiple drug resistant (MDR) pathogen. The first four common pathogen with MDR were A. baumannii, of which MDR rate was 75.0%, followed by Stenotrophomonas maltophilia (33.3%), Staphylococcus aureus (38.1%) and P. aeruginosa (36.7%). In the logistic regression, patients from long term care (LTC) centers (OR=3.764, 95%CI=2.165-6.544, p<.0001), hospitalization for 25 days or more during the preceding 90 days (OR=2.768, 95%CI=1.382-5.545, p=0.0041) and patients with long term ventilator (OR=2.81, 95%CI=1.223-6.458, p=0.015) were independent risk factors. In the repeated random sub-sampling validation, these three were still the most important risk factors, which were selected 985(98.5%), 508(50.8%) and 558(55.8%) times as significant variable ,respectively, in training group and were selected 788(80.0%), 61(12.0%) and 33(5.9%) times individually in validation group. The relative importance of those risk factors were 78.8%, 6.1% and 3.3% ,respectively, and the odds ratio in the final model is 3.49 (95%CI=1.897-6.416, p<.0001), 2.47 (95%CI=1.141-5.364, p=0.0218) and 3.38 (95%CI=1.334-8.541, p=0.0102) individually. According to the association between independent risk factors and MDR pathogen, risk score of those three independent risk factors were all assigned 1 point and after calculated the total risk score by adding the individual scores, we found prevalence of MDR of the episodes with score of 0, 1 and 2 were 16.6%, 50.0%, 66.7% respectively. The AUC (area of under curve) of this scoring tool was 0.7117 (95%CI=0.6539-0.7694).
Conclusion:
This single center, retrospective, observational study identified the important risk factors of pneumonia with MDR pathogens in MICU patients. A new scoring tool was develped. The scoring system contains LTC, hospitalization for 25 days or more during the preceding 90 days and long term ventilator. This simple and feasible prediction tool can be used to facilitate appropriate selection of initial antibiotic treatment for patient with pneumonia in MICU.

並列關鍵字

pneumonia ； drug-resistance ； antibacterial agents ； decision support techniques ； respiratory tract infections ； score, prediction

參考文獻

1. Mandell LA, Wunderink RG. Pneumonia. In: Kasper D, Fauci A, Hauser S, Longo D, Jameson JL, Loscalzo J, editors. Harrison's Principles of Internal Medicine, 19e. New York, NY: McGraw-Hill Education; 2015.

2. Organization WH. The global burden of disease: 2004 update (2008), get more information: http://www.who.int/healthinfo/global_burden_disease/2004_report_update/en/. 2008.

3. Marrie TJ, Huang JQ. Epidemiology of community-acquired pneumonia in Edmonton, Alberta: an emergency department-based study. Canadian respiratory journal : journal of the Canadian Thoracic Society. 2005;12(3):139-42.

4. Arnold FW, Wiemken TL, Peyrani P, Ramirez JA, Brock GN. Mortality differences among hospitalized patients with community-acquired pneumonia in three world regions: results from the Community-Acquired Pneumonia Organization (CAPO) International Cohort Study. Respiratory medicine. 2013;107(7):1101-11.

5. File TM, Jr., Marrie TJ. Burden of community-acquired pneumonia in North American adults. Postgraduate medicine. 2010;122(2):130-41.