Variation in compulsory psychiatric inpatient admission in England: a cross-sectional, multilevel analysis

S. Weich, O. Mcbride, Liz Twigg, P. Keown, E. Cyhlarova, D. Crepaz-keay, H. Parsons, J. Scott, K. Bhui

Research output: Contribution to journalArticlepeer-review

318 Downloads (Pure)


Background: Rates of compulsory admission have increased in England in recent decades, and this trend is accelerating. Studying variation in rates between people and places can help identify modifiable causes.

Objectives: To quantify and model variances in the rate of compulsory admission in England at different spatial levels and to assess the extent to which this was explained by characteristics of people and places.

Design: Cross-sectional analysis using multilevel statistical modelling.

Setting: England, including 98% of Census lower layer super output areas (LSOAs), 95% of primary care trusts (PCTs), 93% of general practices and all 69 NHS providers of specialist mental health services.

Participants: 1,287,730 patients.

Main outcome measure: The study outcome was compulsory admission, defined as time spent in an inpatient mental illness bed subject to the Mental Health Act (2007) in 2010/11. We excluded patients detained under sections applying to emergency assessment only (including those in places of safety), guardianship or supervision of community treatment. The control group comprised all other users of specialist mental health services during the same period.

Data sources: The Mental Health Minimum Data Set (MHMDS). Data on explanatory variables, characterising each of the spatial levels in the data set, were obtained from a wide range of sources, and were linked using MHMDS identifiers.

Results: A total of 3.5% of patients had at least one compulsory admission in 2010/11. Of (unexplained) variance in the null model, 84.5% occurred between individuals. Statistically significant variance occurred between LSOAs [6.7%, 95% confidence interval (CI) 6.2% to 7.2%] and provider trusts (6.9%, 95% CI 4.3% to 9.5%). Variances at these higher levels remained statistically significant even after adjusting for a large number of explanatory variables, which together explained only 10.2% of variance in the study outcome. The number of provider trusts whose observed rate of compulsory admission differed from the model average to a statistically significant extent fell from 45 in the null model to 20 in the fully adjusted model. We found statistically significant associations between compulsory admission and age, gender, ethnicity, local area deprivation and ethnic density. There was a small but statistically significant association between (higher) bed occupancy and compulsory admission, but this was subsequently confounded by other covariates. Adjusting for PCT investment in mental health services did not improve model fit in the fully adjusted models.

Conclusions: This was the largest study of compulsory admissions in England. While 85% of the variance in this outcome occurred between individuals, statistically significant variance (around 7% each) occurred between places (LSOAs) and provider trusts. This higher-level variance in compulsory admission remained largely unchanged even after adjusting for a large number of explanatory variables. We were constrained by data available to us, and therefore our results must be interpreted with caution. We were also unable to consider many hypotheses suggested by the service users, carers and professionals who we consulted. There is an imperative to develop and evaluate interventions to reduce compulsory admission rates. This requires further research to extend our understanding of the reasons why these rates remain so high.

Funding: The National Institute for Health Research Health Services and Delivery Research programme.
Original languageEnglish
Pages (from-to)90
JournalHealth Services and Delivery Research
Issue number49
Publication statusPublished - Dec 2014


Dive into the research topics of 'Variation in compulsory psychiatric inpatient admission in England: a cross-sectional, multilevel analysis'. Together they form a unique fingerprint.

Cite this