Paediatric sepsis mortality varied more across regions than illness severity explains

Design

retrospective cohort study of national Medicaid claims with hierarchical modelling

Population

36,552 sepsis episodes in US publicly insured children under 19, 2021-2023

Primary outcome

death within 30 days of sepsis onset

Effect

3.9% 30-day mortality after community-acquired sepsis; regional standardised mortality ratios 0.00-3.82 with 26 of 104 regions outliers; children's hospital in region aOR 0.76 (95% CI 0.61-0.94)

A cohort of all US publicly insured children under 19 hospitalised with sepsis between 2021 and 2023 identified 36,552 episodes, an incidence of 45.1 per 100,000 person-years. Thirty-day mortality after community-acquired sepsis was 3.9%, and 16.7% of episodes were followed by rehospitalisation within 30 days.

The finding of interest is geographical. After adjusting each region's expected deaths for patient demographics and illness severity, standardised mortality ratios ranged from 0.00 to 3.82, and there were more high- and low-mortality outlier regions than simulation predicted by chance — 12 and 14 respectively out of 104. Two region-level features were associated with lower mortality: the presence of a children's hospital, and a higher regional sepsis incidence.

The second of those is the more interesting. Higher incidence plausibly reflects volume and familiarity rather than more disease, which is the classic volume-outcome relationship reappearing in a condition where it is rarely measured.

The finding is an argument about systems, not about individual practice. In Indian terms it maps onto the question of where a septic child should be taken rather than what should be done when they arrive — and the answer this supports is a centre that sees these children often.

• Where retrieval is possible, a centre that regularly manages paediatric sepsis is preferable to the nearest one
• Do not delay first-hour antibiotics and fluid resuscitation for transfer; stabilise, then move
• Track your own unit's paediatric sepsis volume and outcomes; the variation here was invisible without adjustment
• Plan the 30-day follow-up — one in six survivors was rehospitalised within a month
• Treat organ dysfunction and complex chronic conditions as the mortality risk factors they were shown to be

The statistics, in plain English

Standardised mortality ratios are observed deaths divided by the number a model predicts from that region's case mix, so a ratio of 3.82 means nearly four times the expected deaths — but the model can only adjust for what was recorded, and this is administrative claims data with a diagnosis-code definition of sepsis. Coding differences between regions would produce exactly this pattern without any difference in care. The adjusted odds ratios for a children's hospital (0.76; 95% CI 0.61-0.94) and for higher regional incidence (0.91 per episode per 10,000 person-years; 0.86-0.97) are region-level associations, which cannot be read as effects on an individual child.

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.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.