Ranjeet Bamber, Brian Sullivan, Léo Gorman, Winnie W Y Lee, Matthew B Avison, Andrew W Dowsey, Philip B Williams
{"title":"基于本地表型耐药性数据的贝叶斯模型,为经验性抗生素升级决策提供依据。","authors":"Ranjeet Bamber, Brian Sullivan, Léo Gorman, Winnie W Y Lee, Matthew B Avison, Andrew W Dowsey, Philip B Williams","doi":"10.1007/s40121-024-01011-3","DOIUrl":null,"url":null,"abstract":"<p><strong>Introduction: </strong>Clinicians commonly escalate empiric antibiotic therapy due to poor clinical progress without microbiology guidance. When escalating, they should take account of how resistance to an initial antibiotic affects the probability of resistance to subsequent options. The term \"escalation antibiogram\" (EA) has been coined to describe this concept. One difficulty when applying the EA concept to clinical practice is understanding the uncertainty in results and how this changes for specific patient subgroups.</p><p><strong>Methods: </strong>A Bayesian model was developed to estimate antibiotic resistance rates in Gram-negative bloodstream infections based on phenotypic resistance data. The model generates a series of \"credible\" curves to fit the resistance data, each with the same probability of representing the true rate given the inherent uncertainty. To avoid overfitting, an integrated penalisation term adaptively smooths the curves given the level of evidence.</p><p><strong>Results: </strong>Rates of resistance to empiric first-choice and potential escalation antibiotics were calculated for the whole hospitalised population based on 10,486 individual bloodstream infections, and for a range of specific patient groups, including ICU (intensive care unit), haematolo-oncology, and paediatric patients. The model generated an expected value (posterior mean) with 95% credible interval to illustrate uncertainty, based on the size of the patient subgroup. For example, the posterior means of piperacillin/tazobactam resistance rates in Gram-negative bloodstream infection are different between patients on ICU and the general hospital population: 27.3% (95% CI 18.1-37.2 vs. 13.4% 95% CI 11.0-16.1) respectively. The model can also estimate the probability of inferiority between two antibiotics for a specific patient population. Differences in optimal escalation antibiotic options between specific patient groups were noted.</p><p><strong>Conclusions: </strong>EA analysis informed by our Bayesian model is a useful tool to support empiric antibiotic switches, providing an estimate of local resistance rates, and a comparison of antibiotic options with a measure of the uncertainty in the data. We demonstrate that EAs calculated for the whole hospital population cannot be assumed to apply to specific patient group.</p>","PeriodicalId":13592,"journal":{"name":"Infectious Diseases and Therapy","volume":" ","pages":"1963-1981"},"PeriodicalIF":4.7000,"publicationDate":"2024-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11343932/pdf/","citationCount":"0","resultStr":"{\"title\":\"A Bayesian Model Based on Local Phenotypic Resistance Data to Inform Empiric Antibiotic Escalation Decisions.\",\"authors\":\"Ranjeet Bamber, Brian Sullivan, Léo Gorman, Winnie W Y Lee, Matthew B Avison, Andrew W Dowsey, Philip B Williams\",\"doi\":\"10.1007/s40121-024-01011-3\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Introduction: </strong>Clinicians commonly escalate empiric antibiotic therapy due to poor clinical progress without microbiology guidance. When escalating, they should take account of how resistance to an initial antibiotic affects the probability of resistance to subsequent options. The term \\\"escalation antibiogram\\\" (EA) has been coined to describe this concept. One difficulty when applying the EA concept to clinical practice is understanding the uncertainty in results and how this changes for specific patient subgroups.</p><p><strong>Methods: </strong>A Bayesian model was developed to estimate antibiotic resistance rates in Gram-negative bloodstream infections based on phenotypic resistance data. The model generates a series of \\\"credible\\\" curves to fit the resistance data, each with the same probability of representing the true rate given the inherent uncertainty. To avoid overfitting, an integrated penalisation term adaptively smooths the curves given the level of evidence.</p><p><strong>Results: </strong>Rates of resistance to empiric first-choice and potential escalation antibiotics were calculated for the whole hospitalised population based on 10,486 individual bloodstream infections, and for a range of specific patient groups, including ICU (intensive care unit), haematolo-oncology, and paediatric patients. The model generated an expected value (posterior mean) with 95% credible interval to illustrate uncertainty, based on the size of the patient subgroup. For example, the posterior means of piperacillin/tazobactam resistance rates in Gram-negative bloodstream infection are different between patients on ICU and the general hospital population: 27.3% (95% CI 18.1-37.2 vs. 13.4% 95% CI 11.0-16.1) respectively. The model can also estimate the probability of inferiority between two antibiotics for a specific patient population. Differences in optimal escalation antibiotic options between specific patient groups were noted.</p><p><strong>Conclusions: </strong>EA analysis informed by our Bayesian model is a useful tool to support empiric antibiotic switches, providing an estimate of local resistance rates, and a comparison of antibiotic options with a measure of the uncertainty in the data. We demonstrate that EAs calculated for the whole hospital population cannot be assumed to apply to specific patient group.</p>\",\"PeriodicalId\":13592,\"journal\":{\"name\":\"Infectious Diseases and Therapy\",\"volume\":\" \",\"pages\":\"1963-1981\"},\"PeriodicalIF\":4.7000,\"publicationDate\":\"2024-09-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11343932/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Infectious Diseases and Therapy\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.1007/s40121-024-01011-3\",\"RegionNum\":3,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"2024/7/18 0:00:00\",\"PubModel\":\"Epub\",\"JCR\":\"Q1\",\"JCRName\":\"INFECTIOUS DISEASES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Infectious Diseases and Therapy","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1007/s40121-024-01011-3","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2024/7/18 0:00:00","PubModel":"Epub","JCR":"Q1","JCRName":"INFECTIOUS DISEASES","Score":null,"Total":0}
引用次数: 0
摘要
导言:临床医生通常会在没有微生物学指导的情况下,因临床进展不佳而升级经验性抗生素治疗。在升级治疗时,他们应考虑到对初始抗生素的耐药性如何影响对后续方案的耐药性概率。人们创造了 "抗生素升级图"(EA)一词来描述这一概念。将 EA 概念应用于临床实践的一个困难是理解结果的不确定性,以及这种不确定性在特定患者亚群中的变化情况:方法:根据表型耐药性数据,开发了一个贝叶斯模型来估计革兰氏阴性血流感染的抗生素耐药率。该模型生成一系列 "可信 "曲线来拟合耐药性数据,鉴于固有的不确定性,每条曲线代表真实耐药率的概率相同。为避免过度拟合,一个综合惩罚项根据证据水平自适应地平滑曲线:根据 10,486 例血液感染病例计算出了住院病人对经验性首选抗生素和潜在升级抗生素的耐药率,并计算出了一系列特定病人群体的耐药率,包括 ICU(重症监护室)、血液肿瘤科和儿科病人。该模型根据患者亚组的规模生成了预期值(后验平均值)和 95% 可信区间,以说明不确定性。例如,在革兰氏阴性血流感染中,ICU 患者和普通医院患者的哌拉西林/他唑巴坦耐药率后验均值不同:分别为 27.3% (95% CI 18.1-37.2 vs. 13.4% 95% CI 11.0-16.1)。该模型还能估算出两种抗生素在特定患者群体中的劣效概率。我们注意到了特定患者群体在最佳升级抗生素选择上的差异:由我们的贝叶斯模型提供的 EA 分析是支持经验性抗生素转换的有用工具,它提供了局部耐药率的估计值,并通过数据的不确定性对抗生素选择进行了比较。我们证明,不能假定为整个医院人群计算的 EAs 适用于特定的患者群体。
A Bayesian Model Based on Local Phenotypic Resistance Data to Inform Empiric Antibiotic Escalation Decisions.
Introduction: Clinicians commonly escalate empiric antibiotic therapy due to poor clinical progress without microbiology guidance. When escalating, they should take account of how resistance to an initial antibiotic affects the probability of resistance to subsequent options. The term "escalation antibiogram" (EA) has been coined to describe this concept. One difficulty when applying the EA concept to clinical practice is understanding the uncertainty in results and how this changes for specific patient subgroups.
Methods: A Bayesian model was developed to estimate antibiotic resistance rates in Gram-negative bloodstream infections based on phenotypic resistance data. The model generates a series of "credible" curves to fit the resistance data, each with the same probability of representing the true rate given the inherent uncertainty. To avoid overfitting, an integrated penalisation term adaptively smooths the curves given the level of evidence.
Results: Rates of resistance to empiric first-choice and potential escalation antibiotics were calculated for the whole hospitalised population based on 10,486 individual bloodstream infections, and for a range of specific patient groups, including ICU (intensive care unit), haematolo-oncology, and paediatric patients. The model generated an expected value (posterior mean) with 95% credible interval to illustrate uncertainty, based on the size of the patient subgroup. For example, the posterior means of piperacillin/tazobactam resistance rates in Gram-negative bloodstream infection are different between patients on ICU and the general hospital population: 27.3% (95% CI 18.1-37.2 vs. 13.4% 95% CI 11.0-16.1) respectively. The model can also estimate the probability of inferiority between two antibiotics for a specific patient population. Differences in optimal escalation antibiotic options between specific patient groups were noted.
Conclusions: EA analysis informed by our Bayesian model is a useful tool to support empiric antibiotic switches, providing an estimate of local resistance rates, and a comparison of antibiotic options with a measure of the uncertainty in the data. We demonstrate that EAs calculated for the whole hospital population cannot be assumed to apply to specific patient group.
期刊介绍:
Infectious Diseases and Therapy is an international, open access, peer-reviewed, rapid publication journal dedicated to the publication of high-quality clinical (all phases), observational, real-world, and health outcomes research around the discovery, development, and use of infectious disease therapies and interventions, including vaccines and devices. Studies relating to diagnostic products and diagnosis, pharmacoeconomics, public health, epidemiology, quality of life, and patient care, management, and education are also encouraged.
Areas of focus include, but are not limited to, bacterial and fungal infections, viral infections (including HIV/AIDS and hepatitis), parasitological diseases, tuberculosis and other mycobacterial diseases, vaccinations and other interventions, and drug-resistance, chronic infections, epidemiology and tropical, emergent, pediatric, dermal and sexually-transmitted diseases.