發展急診病人抗藥性金黃色葡萄球菌與多重抗藥性格蘭氏陰性菌預測模式以預防抗藥性細菌於醫院內傳播

Abstract

多重抗藥性細菌在社區與醫院內同時的快速增加已經在近年變成重要的公共衛生與民眾健康議題，抗藥性細菌，例如近年來出現的抗藥性金黃色葡萄球菌(methicillin-resistant Staphylococcus aureus，MRSA)、抗萬古黴素腸球菌(vancomycin-resistant enterococci，VRE)、製造超廣譜乙內醯胺酶格蘭氏陰性菌(extended-spectrum β-lactamases producing gram-negative bacteria，ESBL producing GNB)或其他多重抗藥性陰性菌(multi-drug resistant gram-negative bacteria，MDR-GNB)等，均會降低抗生素治療的效果進而威脅到全世界成千上萬病人的生命，這在新一代抗生素的研發愈加困難的今日更顯出其嚴重性。雖然抗藥性細菌的出現主要源自於醫院內醫療抗生素的過度使用與不適當使用，但是近年的研究也顯示越來越多的社區民眾帶有或出現各種抗藥性細菌的移生或感染，特別是高抗藥性金黃色葡萄球菌(MRSA)及各種多重抗藥性格蘭氏染色陰性菌(MDR-GNB)，而這些身上移生抗藥性細菌的病人一旦入住到醫院內，往往因為醫師臨床警覺的缺乏或實驗室微生物學診斷的不足，導致經驗性抗生素的不適當選用以及感控隔離措施的延遲介入，不僅影響病人的治療預後，更助長抗藥性細菌在醫院內傳播的風險，因此在社區抗藥性細菌盛行率大幅升高的今日，以具實證基礎的抗藥性細菌預測模式(prediction model)，來協助臨床醫師及感控專家在病人治療及感控防治的正確醫療決策，更具有其重要性及實用性。 由於抗藥性機轉的不同，以往的研究已經證實病人出現或移生格蘭氏染色陽性菌或革蘭氏染色陰性菌抗藥性的流行病學危險因子，其實有相當大的差異，病人移生抗藥性格蘭氏染色陽性細菌例如MRSA的危險因子，可能並不是病人移生抗藥性格蘭氏染色陰性細菌的危險因子，因此在預測模式的建立上，就必須分別建立不同的抗藥性細菌預測模式才能具有最佳的預測能力，進而有效的作為臨床決策的參考依據。因此，為了建構完整且實用的社區病人抗藥性細菌移生或感染的預測模式，我們嘗試就抗藥性金黃色葡萄球菌(社區間最主要的抗藥性格蘭氏陽性菌)及抗藥性格蘭氏染色陰性菌，以急診病人作為社區感染症病人的替代研究對象，分別建立不同的預測模式。 在抗藥性金黃色葡萄球菌MRSA預測模式上，我們利用前後長達11年的急診金黃色葡萄球菌菌血症的縱貫性世代資料，分別利用兩個不同的研究，嘗試建立前後兩套的預測模式。第一套預測模式，是找出社區民眾當臨床上懷疑或證實發生金黃色葡萄球菌感染時，病人出現抗藥性菌株(methicillin resistance isolate)感染的危險因子，並據以建立有效的模式預測病人出現MRSA感染風險的高低，進而協助臨床醫師及感染管制人員作為是否給予病人進行篩檢性微生物學培養、早期接觸隔離、以及經驗性使用對抗藥性金黃色葡萄球菌有效的抗生素如萬古黴素(vancomycin)等之決策依據。 其次，由於近年來在國外及我們團隊的研究顯示，同樣在一般藥物敏感度測試結果為具萬古黴素抗生素具藥物感受性的MRSA菌株中，如果該MRSA菌株對萬古黴素有較高的最低抑菌濃度(minimal inhibitory concentration, MIC)，雖然仍屬對萬古黴素具藥物感受性之菌株，但即使是以萬古黴素治療，仍然會對感染該MRSA菌株的病人造成較差的預後，而需考慮萬古黴素以外的抗生素以改善病人的預後。因此我們接著在已知有MRSA感染的社區病人身上，更進一步的嘗試建立第二套預測模式，協助臨床醫師判斷病人所感染的MRSA菌株，是否可能為高萬古黴素最低抑菌濃度菌株(high vancomycin MIC MRSA isolate)，以決定是否需要選擇萬古黴素以外的抗生素以治療病人的MRSA感染。 而在預測抗藥性格蘭氏陰性菌移生或感染的第三個研究方面，我們在2009年間對995位急診後續住院的病人進行主動式微生物篩檢並合併早期臨床培養結果以偵測病人身上是否移生有任何一種的MDR-GNB菌株，然後配合臨床資料的蒐集，建立了針對社區病人移生MDR-GNB的預測模式。為了更進一步的了解我們所發展的這個多重抗藥性格蘭氏染色陰性菌移生預測模式，對於預測病人實際發生多重抗藥性格蘭氏染色陰性菌臨床感染的效果，我們利用2015年另外998位急診住院病人做另外的效度檢驗，以評估其區分急診病人後續住院過程中發生多重抗藥性格蘭氏染色陰性菌感染風險高低的能力。 在第一篇研究中，我們建立了兩種方便使用的預測模式，用以預測社區病人一旦懷疑或證實為金黃色葡萄球菌感染時，出現抗藥性MRSA菌株感染的風險，以提供臨床醫師決定是否需要經驗性的開立anti-MRSA藥物或進行篩檢性培養的必要性。而第二個研究則提供進一步的預測模式，針對確定感染MRSA的病人，評估預測其感染較高萬古黴素最低抑菌濃度菌株的風險，以協助臨床醫師決定是否應該經驗性採用萬古黴素以外的後線藥物，以降低high vancomycin MIC MRSA感染所造成的病人死亡率。而最後一個預測MDR-GNB移生的預測模式，則是協助臨床醫師鑑別病人移生或感染MDR-GNB風險的高低，進而協助經驗性抗生素及感控措施的介入決策。 總結而言，本論文這一系列的三個研究結果所提供的預測模式，我們完整的建構了一套針對社區病人評估移生或感染抗藥性細菌風險高低的預測模式，不管是抗藥性金黃色葡萄球菌(MRSA)還是多重抗藥性格蘭氏染色陰性菌(MDR-GNB)。而這些預測模式有效地提供臨床醫師及感染控制人員重要的決策依據，針對來自於急診或門診的社區住院病人，第一時間提供臨床醫師更精準的經驗性抗生素決策以避免藥物濫用，同時也提供感控人員在有限醫療資源的考量下決定是否採取篩檢性培養或經驗性接觸隔離的決策參考，以達到最有效率的防治效果。

The rapid emergence of multi-drug resistant bacteria is occurring simultaneously in both community and hospital and has become a major public health issue worldwide. Antimicrobial-resistant of bacteria, including methicillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant enterococci (VRE), extended-spectrum β-lactamases (ESBL) producing and other multi-drug resistant gram-negative bacteria (MDR-GNB), endanger the efficacy of antibiotics and threatens millions of lives of the world. It is especially alarming that the development of new antibiotics effective for these highly drug-resistant bacteria has been more difficult than decades before. Despite the drug resistance has been attributed to the overuse and misuse of antibiotics in hospital settings, recent researches also showed an increasing prevalence of antimicrobial-resistant bacteria colonization or infection among patients coming from the community. Because hospitalized patients with antimicrobial-resistant bacteria colonization or infection are usually undetected or delayed diagnosed, the risk of inappropriate empirical antibiotics and nosocomial spread of these drug-resistant bacteria have become a serious concern in making therapeutic and infection control decision. In facing the rising influx of multi-drug resistant bacteria from the community to hospital, especially the MRSA and MDR-GNB, useful prediction models and comprehensive strategy that helps clinician and infections control specialists in their clinical treatment and infection control policy is urgently required. Previous studies have shown the significant differences between risk factors for antimicrobial-resistant gram-positive bacteria and gram-negative bacterial colonization or infection among community patients. It might be due to the different mechanisms for the emergence of antimicrobial resistance of a bacterium. Therefore, it is necessary to develop different prediction models specifically for antimicrobial-resistant gram-positive and antimicrobial-resistant gram-negative bacteria for best clinical utility. Therefore, we were dedicated to develop different prediction models for MRSA and MDR-GNB colonization or infection for emergency department (ED) physicians for risk stratification among community patients. To develop a MRSA infection prediction model, we used a longitudinal S. aureus bacteremia cohort dataset that crossed an 11-year study period to develop two prediction models. The first prediction model identified the risk factors for contracting MRSA infection among community patients when S. aureus infection was suspected or documented. We aimed to develop a useful prediction model for first-line physicians in their decisions of cost-effective surveillance culture, early contact isolation and empirical anti-MRSA therapy. Because recent studies demonstrated that patients infected with MRSA of higher vancomycin minimal inhibitory concentration (MIC) within the susceptible range had worse outcomes than patients infected with MRSA of lower vancomycin MIC value if these patients received vancomycin treatment. To early identify community patients potentially infected with high vancomycin MIC MRSA isolate, we therefore used the same cohort dataset to develop another prediction model to identify risk of high vancomycin MIC isolate infection among community patients with MRSA infection. It was important for first-line physicians to consider empirical anti-MRSA antibiotics other than vancomycin once high vancomycin MIC MRSA isolate infection was highly suspected. The final prediction model was developed for the risk stratification of MDR-GNB colonization or infection among ED patients with suspected or confirmed gram-negative bacterial infection. We used microbiological and clinical data of 995 ED patients collected in 2009 to develop and validate a MDR-GNB colonization prediction model (COP model). Then we tested the COP model in another 998 ED patients in 2015 to evaluate the usefulness of the model in stratifying the risk of subsequent MDR-GNB infection after ED admission. In summary, we proposed three different prediction models for clinicians in stratifying risk of MRSA or MDR-GNB colonization or infection. The first MRSA prediction model help clinicians in stratifying risk of MRSA colonization or infection among community patients when S. aureus infection was impressed. This first model helps physicians with their decisions on active microbiological surveillance, early contract precaution and empirical anti-MRSA antibiotics. The second model, high vancomycin MIC MRSA prediction model, helps clinicians in stratifying risk of contracting high vancomycin MIC isolate infection among community patients with MRSA infection. This model helps physicians in selecting anti-MRSA antibiotics other than commonly used vancomycin as empirical antibiotics for patients with confirmed MRSA infection. The final model helps clinicians in differentiating risk of MDR-GNB colonization or infection among community patients and therefore their decision of empirical antibiotics and infection control interventions. The results from our serial studies provide scientific evidences for the development of a comprehensive strategy in mitigating the adverse impact on patient outcomes and also the preventing the nosocomial spread of these highly antimicrobial-resistant microorganisms early at patients’ admission.