- Design
- uncontrolled 24-month follow-up of a site-randomised trial, with interrupted time-series analysis
- Population
- paediatricians in a large statewide network; 5,275 pre-intervention and 16,662 post-intervention skin and soft tissue infection cases
- Primary outcome
- proportion of skin and soft tissue infection cases treated for 7 days or fewer
- Effect
- 62.5% to 88.5%, immediate level change (P < .001), with no decay over 21 months after feedback ceased (slope +0.2% per month)
A statewide paediatric network combined three months of performance feedback and education with a clinical decision support alert firing at the point of prescribing, aimed at shortening antibiotic courses for skin and soft tissue infection. The proportion of cases treated for seven days or fewer rose from 62.5% to 88.5%, with an immediate step change when the intervention began.
The important part is what happened next. The feedback and education stopped after three months; the alert kept running. Over the following 21 months there was no decay — the slope was fractionally positive at 0.2% a month.
That is unusual and worth naming, because decay is the normal fate of quality-improvement projects. Audit-and-feedback cycles produce a gain that erodes once attention moves on, and the literature is full of interventions that worked while someone was watching. This one suggests the durable component is the alert, not the teaching — the teaching moved practice, and the alert held it there.
The transferable design is a short active phase to establish the change and a low-friction reminder to sustain it, rather than repeated education campaigns. For any unit with an electronic prescribing system, that is buildable.
- Pair any prescribing-change campaign with a point-of-care prompt that outlasts it
- Keep the active teaching phase short; the step change here happened immediately
- Measure the outcome for two years, not three months — a gain that decays looks identical at week 12
- Choose one specific, countable behaviour (course length for one condition) rather than a general goal
- Where electronic prescribing is unavailable, consider a printed default on the prescription pad as the low-technology equivalent
The statistics, in plain English
This is an uncontrolled follow-up to a site-randomised trial, so the before-and-after comparison — 62.5% of 5,275 cases rising to 88.5% of 16,662 — has no concurrent control group and cannot exclude a secular trend in prescribing. The strength of the evidence lies in the shape rather than the size: an immediate level change at the intervention's start (P < .001), followed by a flat slope of +0.2% per month over 21 months. A secular trend would generally produce gradual change, not a step followed by a plateau.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for paediatrics, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free