%0 Journal Article %T Psychosocial therapy and causes of death after deliberate self-harm: a register-based, nationwide multicentre study using propensity score matching %J Psychological medicine %D 2016 %A Birkbak, J. %A Stuart, E. A. %A Lind, B. D. %A Qin, P. %A Stenager, E. %A Larsen, K. J. %A Wang, August G. %A Nielsen, Ann Colleen %A Pedersen, C. M. %A Winsløv, J-h %A Langhoff, C. %A Mühlmann, C. %A Nordentoft, Merete %A Erlangsen, A. %V 46 %N 16 %P 3419-3427 %X BACKGROUND: Psychosocial therapy after deliberate self-harm might be associated with reduced risk of specific causes of death.

METHOD: In this matched cohort study, we included patients, who after an episode of deliberate self-harm received psychosocial therapy at a Suicide Prevention Clinic in Denmark between 1992 and 2010. We used propensity score matching in a 1:3 ratio to select a comparison group from 59 046 individuals who received standard care. National Danish registers supplied data on specific causes of death over a 20-year follow-up period.

RESULTS: At the end of follow-up, 391 (6.9%) of 5678 patients in the psychosocial therapy group had died, compared with 1736 (10.2%) of 17 034 patients in the matched comparison group. Lower odds ratios of dying by mental or behavioural disorders [0.54, 95% confidence interval (CI) 0.37-0.79], alcohol-related causes (0.63, 95% CI 0.50-0.80) and other diseases and medical conditions (0.61, 95% CI 0.49-0.77) were noted in the psychosocial therapy group. Also, we found a reduced risk of dying by suicide as well as other external causes, however, not by neoplasms and circulatory system diseases. Numbers needed to treat were 212.9 (95% CI 139.5-448.4) for mental or behavioural disorders as a cause of death, 111.1 (95% CI 79.2-210.5) for alcohol-related causes and 96.8 (95% CI 69.1-161.8) for other diseases and medical conditions.

CONCLUSIONS: Our findings indicate that psychosocial therapy after deliberate self-harm might reduce long-term risk of death from select medical conditions and external causes. These promising results should be tested in a randomized design.

Language: en

%G en %I Cambridge University Press %@ 0033-2917 %U http://dx.doi.org/10.1017/S0033291716001872