An ecological comparison of hospital-level antibiotic use and mortality in 36,124,372 acute/general medicine inpatients in England
Budgell EP., Davies TJ., Donker T., Hopkins S., Wyllie D., Peto TEA., Gill M., Llewelyn MJ., Walker AS.
ABSTRACTObjectivesTo determine the extent to which variation in hospital antibiotic prescribing is associated with mortality risk in acute/general medicine inpatients.DesignEcological analysis, using electronic health records from Hospital Episode Statistics (HES) and antibiotic data from IQVIA.Setting135 acute National Health Service (NHS) hospital Trusts in England.Participants36,124,372 acute/general medicine inpatients (16 years old at admission) admitted between 01/April/2010-31/March/2017 (median age 66 years, 50.4% female, 83.8% white ethnicity).Main outcome measuresRandom-effects meta-regression was used to investigate whether heterogeneity in the adjusted probability of death within 30-days of admission was associated with hospital-level antibiotic use, measured in defined-daily-doses (DDD)/1,000 bed-days. Models also considered DDDs/1,000 admissions and DDDs for selected antibiotics, including narrow-spectrum/broad-spectrum, inpatient/outpatient, parenteral/oral, piperacillin-tazobactam and meropenem, and Public Health England interpretations of World Health Organization Access, Watch, and Reserve antibiotics. Secondary analyses examined 14-day mortality and non-elective re-admission to hospital within 30-days of discharge.ResultsThere was a 15-fold variation in hospital-level DDDs/1,000 bed-days and comparable or greater variation in broad-spectrum, parenteral, and Reserve antibiotic use. After adjusting for a wide range of admission factors to reflect varying case-mix across hospitals, the adjusted probability of 30-day mortality changed by -0.010% (95% CI: -0.064 to +0.044) for each increase in hospital-level antibiotic use of 500 DDDs/1,000 bed-days. Analyses focusing on other metrics of antibiotic use, sub-populations, and 14-day mortality also showed no consistent association with the adjusted probability of death.DiscussionWe find no evidence that the wide variation in antibiotic use across NHS hospitals is associated with case-mix adjusted mortality risk in acute/general medicine inpatients. Our results indicate that hospital antibiotic use in the acute/general medicine population could be safely cut by up to one-third.What is already known on this topicPrevious studies have reported wide variation in both recommended antibiotic prescribing duration and total antibiotic consumption among acute hospitals.In hospitals with more acute patients, systematic under-treatment might reasonably be expected to harm patients, and though a growing body of evidence shows reducing hospital antibiotic overuse may be done safely, there is a lack of good data to indicate how much it may be possible to safely reduce useExamination of the possibility that substantially driving down antibiotic use could compromise clinical outcomes is needed to reassure practitioners and the public that substantially reducing antibiotic use is safe.What this study addsAfter adjusting for a wide range of admission factors to reflect varying case-mix across acute hospitals, we observed no consistent association between 24 metrics of hospital-level antibiotic use and the adjusted probability of death in a large national cohort of over 36 million acute/general medicine inpatientsThese findings indicate that at many hospitals patients are receiving considerably more antibiotics than necessary to treat their acute infections, and we estimate system-wide reductions of up to one-third of antibiotic defined-daily-doses (DDDs) could be achieved safely among medical admissions.The magnitude of the antibiotic reductions that could be safely achieved dwarf the 1% year-on-year reductions required of NHS hospitals.