What is worth preventing after surgery in a low-resource system is not what is most lethal

Design

Secondary analysis of a prospective multicountry observational cohort using a logistic mixed-effects model

Population

10,748 adults undergoing surgery across African hospitals, data collected in 2016

Primary outcome

30-day in-hospital mortality and the attributable fraction for each severe complication

Effect

222 deaths (2.07%); pulmonary embolism OR 484 (95% CI 85.2-2744), bloodstream infection OR 34.3 (16.3-72.3); largest attributable fraction severe bloodstream infection at 13.7% (10-18.1)

Perioperative mortality in Africa runs at about twice the global average, and this analysis of 10,748 adults from the African Surgical Outcome Study asks which complications actually account for it. A logistic mixed-effects model, adjusted for patient and surgical factors, produced both odds ratios and attributable fractions - the proportion of all deaths a complication accounts for.

Two hundred and twenty-two patients (2.07%) died in hospital within 30 days. The strongest associations were pulmonary embolism (OR 484, 95% CI 85.2-2744), pulmonary oedema (OR 53.5, 95% CI 6.59-434) and bloodstream infection (OR 34.3, 95% CI 16.3-72.3). But the largest attributable fraction belonged to severe bloodstream infection, at 13.7% (95% CI 10-18.1) of all 30-day mortality.

That gap between the two measures is the whole point of the paper. Pulmonary embolism kills almost everyone it happens to and happens rarely, so its odds ratio is enormous and its contribution to the death toll is small. Bloodstream infection is less individually lethal and far more common, so it accounts for more deaths than anything else. A department choosing where to spend limited effort should follow the attributable fraction, not the odds ratio - which points at sepsis recognition, line care and antimicrobial stewardship rather than at thromboprophylaxis protocols. The same arithmetic applies to Indian district and tertiary practice, where the case mix and resource constraints look closer to this cohort than to a European one.

• Prioritise bloodstream infection prevention and early recognition over rarer, more dramatic complications
• Audit line insertion and care practices, which is where much of that infection burden originates
• Keep thromboprophylaxis protocols - a small attributable fraction is not a reason to drop them
• Failure to rescue, not complication rates alone, is what these numbers describe
• Data were collected in 2016; treat the ranking as a guide rather than a current measurement

Why it matters

It separates the complications that are most dangerous from the ones that account for most of the dying, and they are not the same list.

Don't overread it

Attributable fractions describe association, not the number of deaths prevention would avert.

The statistics, in plain English

An odds ratio of 484 with a confidence interval from 85 to 2,744 is a signal that the event is rare and almost always fatal, not that it is the main problem. The interval is that wide because very few patients had a pulmonary embolism. The attributable fraction is the more useful number for planning because it multiplies lethality by frequency: bloodstream infection has a far smaller odds ratio and accounts for 13.7% of all deaths, more than anything else measured. This is an association study, so it cannot show that preventing bloodstream infection would remove 13.7% of deaths - only that those deaths occurred in patients who had one.

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.