Coercive measures to manage disruptive or violent behaviour are accepted as standard practice in mental healthcare, but systematic knowledge of potentially harmful outcomes is insufficient.

To examine the association of mechanical restraint with several predefined somatic harmful outcomes.

We conducted a population-based, observational cohort study linking data from the Danish national registers from 2007 to 2019. The primary analyses investigated the association of mechanical restraint with somatic adverse events, using panel regression analyses (within-individual analysis) to account for repeated exposures and outcomes. Secondary between-group analyses were performed with a control group exposed to types of coercion other than mechanical restraint.

The study population comprised 13 022 individuals. We report a statistically significant association of mechanical restraint with thromboembolic events (relative risk 4.377, number needed to harm (NNH) 8231), pneumonia (relative risk 5.470, NNH 3945), injuries (relative risk 2.286, NNH 3240) and all-cause death (relative risk 5.540, NNH 4043) within 30 days after mechanical restraint. Estimates from the between-group analyses (comparing the exposed group with a control group of 22 643 individuals) were non-significant or indicated increased baseline risk in the control group. A positive dose–response analysis for cardiac arrest, injury and death supported a causative role of mechanical restraint in the reported associations.

Although the observed absolute risk increases were small, the derived relative risks were non-negligible considering that less restrictive interventions are available. Clinicians and decision makers should be aware of the excess risk in future decisions on the use of mechanical restraint versus alternative interventions.