Design: secondary analysis of a multi-country prospective observational cohort using a logistic mixed-effects model with attributable fraction calculation
Population: 10,748 adult surgical patients across African hospitals; 222 in-hospital deaths within 30 days
Primary outcome: severe postoperative complications independently associated with 30-day in-hospital mortality, and their attributable fractions
Effect: pulmonary embolism odds ratio 484 (95% CI 85.2 to 2,744); bloodstream infection odds ratio 34.3 with the highest attributable fraction, 13.7% (95% CI 10 to 18.1)

The African Surgical Outcome Study collected 10,748 adult surgical patients prospectively in 2016; 222 (2.07%) died in hospital within 30 days. This analysis asked which severe complications were independently associated with death, and then went a step further and calculated the attributable fraction — the share of all deaths a complication accounts for.

The strongest associations were pulmonary embolism (adjusted odds ratio 484, 95% CI 85.2 to 2,744), pulmonary oedema (53.5, 95% CI 6.59 to 434) and bloodstream infection (34.3, 95% CI 16.3 to 72.3). But the largest attributable fraction belonged to severe bloodstream infection at 13.7% (95% CI 10 to 18.1) — because it is common enough for its odds ratio to matter at population scale, while pulmonary embolism is not.

That distinction is the point of the paper and it is worth transferring. A ranking by odds ratio tells you what is most lethal when it happens; a ranking by attributable fraction tells you where to spend effort. If one in seven postoperative deaths runs through bloodstream infection, then line care, timely source control, antibiotic stewardship and the recognition of the deteriorating patient are where a perioperative service buys the most survival — not another thromboprophylaxis protocol.

The relevance to Indian practice is direct. This is a multi-country low- and middle-income cohort with a perioperative mortality around twice the global average, operating under constraints closer to an Indian district hospital than a European tertiary centre. Failure to rescue, not the anaesthetic, is where these deaths sit.

Audit central and peripheral line insertion and care as a mortality intervention, not an infection-control formality
Build escalation triggers for the deteriorating postoperative patient into the ward observation chart
Rank your own improvement priorities by how common a complication is, not only by how lethal
Keep thromboprophylaxis, but recognise it addresses a rare rather than a high-yield cause of death here
Record postoperative complications prospectively; without the denominator none of this is measurable locally

Why it matters

It separates what is most lethal from what kills most people, and the two point at different improvement programmes.

The statistics, in plain English

An odds ratio of 484 with limits from 85.2 to 2,744 is a statistical artefact of a very rare complication with a very high case fatality: it tells you pulmonary embolism is nearly always fatal here, not that it explains many deaths. The attributable fraction corrects for that, and 13.7% (10 to 18.1) for bloodstream infection is both precise and actionable. This is observational data from 2016, so it describes association and priority, not the effect of any intervention.