Detecting inappropriate total duration of antimicrobial therapy using semi-automated surveillance

Abstract

Objectives Evaluation of the appropriateness of the duration of antimicrobial treatment is a cornerstone of antibiotic stewardship programs, but it is time-consuming. Furthermore, it is often restricted to antibiotics prescribed during hospital admission. This study aimed to determine whether mandatory prescription-indication registration at the moment of prescribing antibiotics enables reliable automated assessment of the duration of antibiotic therapy, including post-discharge duration, limiting the need for manual chart review to data validation. Methods Antibiotic prescription and admission data, from 1-6-2020 to 31-12-2021, were electronically extracted from the Electronic Medical Record of two hospitals using mandatory indication registration. All consecutively prescribed antibiotics of adult patients who received empiric therapy in the first 24 h of admission were merged to calculate the total length of therapy (LOT) per patient, broken down per registered indication. Endpoints were the accuracy of the data, evaluated by comparing the extracted LOT and registered indication with the clinical notes in 400 randomly selected records, and guideline adherence of treatment duration. Data were analysed using a reproducible syntax, allowing semi-automated surveillance. Results A total of 3,466 antibiotic courses were analysed. LOT was accurately retrieved in 96% of the 400 evaluated antibiotic courses. The registered indication did not match chart review in 17% of antibiotic courses, of which only half affected the assessment of guideline adherence. On average, in 44% of patients treatment was continued post-discharge, accounting for 60% (± 19%) of their total LOT. Guideline adherence ranged from 26 to 75% across indications. Conclusions Mandatory prescription-indication registration data can be used to reliably assess total treatment course duration, including post-discharge antibiotic duration, allowing semi-automated surveillance.

Identifying excessive length of antibiotic treatment duration for hospital-acquired infections: a semi-automated approach to support antimicrobial stewardship

“…As some of the validated treatment courses within the validated sample of UTI, RTI, IAI or other infections were also of extended duration or in readmitted patients these were also excluded. The error rate of extracted indications was comparable with the error rates we found in an earlier study [ 9 ]. In 21.0% of courses the extracted indication did not match the indication of the treatment course found in the EHR.…”

“…When considering only courses with post-discharge treatment, we observed that a similar proportion of treatment courses was given post-discharge. Our previous study showed that many infections are still treated too long and a considerable amount of this excessive treatment is given post-discharge [ 9 ]. Possible stewardship interventions could focus on antibiotic overuse at discharge for both CAI and HAI.…”

“…This institution uses Epic (Epic Systems Corporation) as its EHR and prescription software. For all in-hospital antimicrobial prescriptions, a mandatory indication registration tool is used to document the specific indication for which the antimicrobial agent is prescribed [ 8 , 9 ]. Approval from the institutional review board for this study was not required because retrospective, pseudonymized data was used for quality optimization purposes.…”

“…The possible indications were empirical therapy, targeted therapy or prophylaxis. In case of therapy, the prescriber had to select the main focus of infection, first on tract level, followed by a further specification for cases involving urinary tract infections (UTI) and respiratory tract infections (RTI) [ 8 , 9 ] (additional file 1 ).…”

“…Previously published literature demonstrates that healthcare provider-selected indication tools yield reliable information [ 6 – 8 ]. We previously showed that it is feasible to take data from a mandatory indication registration and use dedicated scripting to combine it with other raw data in the EHR, to extract the total length of treatment (LOT) for community-acquired infections (CAI) [ 9 ]. Scripts that are primarily based on the concatenation of sequential data blocks however cannot be used for processing data on hospital acquired infections (HAI).…”