Probabilistic microsimulation to examine the cost-effectiveness of hospital admission screening strategies for carbapenemase-producing enterobacteriaceae (CPE) in the United Kingdom

Abstract

Listen

Translate article

BackgroundAntimicrobial resistance has been recognised as a global threat with carbapenemase- producing-Enterobacteriaceae (CPE) as a prime example. CPE has similarities to COVID-19 where asymptomatic patients may be colonised representing a source for onward transmission. There are limited treatment options for CPE infection leading to poor outcomes and increased costs. Admission screening can prevent cross-transmission by pre-emptively isolating colonised patients.ObjectiveWe assess the relative cost-effectiveness of screening programmes compared with no- screening.MethodsA microsimulation parameterised with NHS Scotland date was used to model scenarios of the prevalence of CPE colonised patients on admission. Screening strategies were (a) two-step screening involving a clinical risk assessment (CRA) checklist followed by microbiological testing of high-risk patients; and (b) universal screening. Strategies were considered with either culture or polymerase chain reaction (PCR) tests. All costs were reported in 2019 UK pounds with a healthcare system perspective.ResultsIn the low prevalence scenario, no screening had the highest probability of cost-effectiveness. Among screening strategies, the two CRA screening options were the most likely to be cost-effective. Screening was more likely to be cost-effective than no screening in the prevalence of 1 CPE colonised in 500 admitted patients or more. There was substantial uncertainty with the probabilities rarely exceeding 40% and similar results between strategies. Screening reduced non-isolated bed-days and CPE colonisation. The cost of screening was low in relation to total costs.ConclusionThe specificity of the CRA checklist was the parameter with the highest impact on the cost-effectiveness. Further primary data collection is needed to build models with less uncertainty in the parameters.