Multiplex PCR point of care testing versus routine, laboratory-based testing in the treatment of adults with respiratory tract infections: a quasi-randomised study assessing impact on length of stay and antimicrobial use

Andrews, Denise; Chetty, Yumela; Cooper, Ben S.; Virk, Manjinder; Glass, Stephen K; Letters, Andrew; Kelly, Philip A.; Sudhanva, Malur; Jeyaratnam, Dakshika

The primary outcome was analysed with a linear regression model (after log-transformation of length of stay data) according to a pre-specified analysis plan with an individual patient taken as the unit of analysis. Secondary outcomes were analysed using linear regression models for continuous outcome data, logistic regression for binary outcomes, and negative binomial regression for count outcome data. For all these patient-related outcomes we adjusted for multiple pre-specified potential confounders (age, sex, Charlson score, EWS, WCC, CRP). Pearson's Chi-squared test was used to test for differences between the arms in categorical antibiotic prescribing decisions with p-values calculated by 10,000 Monte Carlo replicates (to avoid problems associated with small cell counts associated with the usual asymptotic p-values). A t-test was used to compare the time to test between the two arms. A planned subgroup analysis was performed as above for primary and secondary outcomes excluding patients who had infection proven elsewhere after enrolment as it is plausible that a respiratory pathogen POC result would not alter LOS or antibiotic use when the patient had another infective diagnosis made. Analysis was conducted in R [18] . Multiple imputation was used to account for missing data using the package mice [19] .