If the observed associations were causal, the adjusted absolute difference in estimated 30-d mortality for admission to a SU with a ratio of 1.5 nurses/ten beds (15.2%, 95% CI 13.8%–16.4%) versus 3.0 nurses/ten beds (11.2%, 95% 10.2%–12.1%) on a weekend was 4.0% (95% CI 1.7%–6.3%). The number needed to harm through admission to a SU with 1.5 nurses/ten beds compared to 3.0 nurses/ten beds was 25 (95% CI 16–59).