PROD-ALERT: Psychiatric restraint open data-analysis using logarithmic estimates on reporting trends

Restraint reporting varies, which undermines regulation, obfuscates analyses, and incentivises minimisation. The English Mental Health Units Use of Force Act 2018, "Seni's Law" mandates reporting. This paper analysed open data from all psychiatric and learning disability institutions in England from September 2020 to August 2021. We correlated logarithms of "people restrained per month", against "bed days" per month and "people under legal mental health detention" per month, per institution. We designated institutions reporting some restraint for at least 11 of 12 months as reporting "completely" and used their trend to infer rates from non-"complete" institutions. Allowance was made for size. Our a priori manual can be shared on request.

Logarithms of people restrained per month and bed-days per month correlated among complete reporters: R² 0.90 (2.s.f). Persons detained per month also correlated with restraint: R² 0.78. "Partial" institutions reported intermittently. "Joiner" institutions reported firstly null, then substantive reporting. "Null" institutions (including the largest) reported no restraint. Precisely-reporting institutions with high inverse variance between months reported similar restraint-rates but less-precise reported lower rates. In institutions reporting no restraint, two independent "true rate" estimations, by bed-days or people detained, correlated across institutions: R² 0.95. Inference from size suggested non-complete reporters restrained 1,774 people in England per month 95% CI (1,449-2,174).

Restraint remains under-reported. Institutional size explains most restraint variation among complete reporting institutions, 90% of R². Institutional restraint reports can be compared per-bed per-month. Rates of people detained are a useful independent "checking" comparator in England.