- Email: [email protected]
Hospital management of suicidal behaviour and subsequent mortality: a prospective cohort study
Articles
Hospital management of suicidal behaviour and subsequent mortality: a prospective cohort study Nav Kapur, Sarah Steeg, Pauline Turnbull, Roger Webb, Helen Bergen, Keith Hawton, Galit Geulayov, Ellen Townsend, Jennifer Ness, Keith Waters, Jayne Cooper
Summary Background Self-poisoning and self-injury are associated with a high risk of suicide or death from any cause but the effect of routine aspects of hospital management on mortality risk is unknown. Methods We did a prospective cohort study using data for adults who had self-harmed presenting to five emergency departments in the UK between 2000 and 2010. We assessed the relation between four aspects of management (psychosocial assessment, medical admission, psychiatric admission, referral for mental health follow-up) and death by suicide or any cause within 12 months of presentation. Findings Of 38 415 individuals presenting with self-harm, 261 (0·7%) died by suicide and 832 (2·2%) died from any cause within 12 months. Most aspects of management were associated with a higher mortality risk in unadjusted analyses. Psychiatric admission was associated with the highest risks for both suicide (hazard ratio 2·35, 95% CI 1·59–3·45) and all-cause mortality (2·35, 2·04–2·72). After adjustment for baseline variables, the hazard ratios were generally smaller, particularly for psychiatric admission. There were significant interactions by sex, age, and history of self-harm. Interpretation This was an observational study and so we cannot infer causation. However, our finding that clinical services seem to reserve the most intensive levels of treatment for patients at highest risk is reassuring. Aspects of routine management might be associated with a lower mortality risk but these effects vary by clinical subgroup. Funding UK Department of Health.
Introduction Self-poisoning and self-injury are major health problems in many countries and are associated with an increased risk of subsequent early death.1 Many studies have focused on suicide as an outcome. Results of a systematic review2 showed that around one in 25 people who presented to hospital with self-harm died by suicide in the following 5 years. A clinical guideline that included a review of previous studies concluded that previous suicidal behaviour, suicidal intent, and being male were the strongest risk factors for suicide following self-harm.3 No protective factors had a consistent association across studies. The risk of death from other causes is also increased. In a cohort study, Kapur and colleagues reported that all-cause mortality in a population of people who had self-harmed was four times higher than expected.4 Individuals who died lost an average of more than 30 years of life,4 although the mortality risk was not uniform for different causes of death. Few studies have assessed the association between routine aspects of hospital care and subsequent mortality.5 Health services might reserve the most intensive forms of treatment for the highest risk patients. In this case, one might expect hospital management to be associated with an increased risk of death. In a Danish general population study,6 suicide risk varied in a stepwise pattern; patients with recent psychiatric admission had a higher incidence of suicide than did those who received outpatient mental
health treatment only, who in turn had a higher incidence than did those who received no treatment. These associations are not causal. This study is an example of confounding by indication, in which allocation to particular treatments was probably based on the underlying need. Routine interventions might also have a positive effect on outcome and reduce risk. One of the challenges in non-randomised studies is to estimate these treatment effects in the context of the sometimes much larger effects of selection or confounding by indication. One approach is to statistically adjust for baseline differences between people who do and do not receive particular forms of management.7 Kapur and colleagues7 (using repeat hospital presentation for self-harm rather than mortality as an outcome measure) showed that receiving a psychosocial assessment was associated with 40% reduction in the risk of repetition within 12 months in two of three centres. In a US study,8 recognition of psychiatric disorder was associated with a 34% reduction in repeat self-harm. There are limitations of using hospital presentations for repeat self-harm as an outcome measure; some people might repeat but choose not to present to health services.9 It is also likely that associations between management and outcome are not uniform3 and so assessing subgroup differences might be useful. We assessed the association between the hospital management of self-harm and subsequent mortality in a large cohort of patients in England.
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Lancet Psychiatry 2015 Published Online August 6, 2015 http://dx.doi.org/10.1016/ S2215-0366(15)00169-8 See Online/Comment http://dx.doi.org/10.1016/ S2215-0366(15)00212-6 Centre for Mental Health and Safety, Centre for Suicide Prevention, University of Manchester, Manchester, UK (Prof N Kapur FRCPsych, S Steeg BA, P Turnbull PhD, R Webb PhD, J Cooper PhD); Manchester Mental Health and Social Care Trust, Manchester, UK (Prof N Kapur); Centre for Suicide Research, University of Oxford, Department of Psychiatry, Warneford Hospital, Oxford, UK (H Bergen PhD, Prof K Hawton DSc, G Geulayov PhD); School of Psychology, University of Nottingham, Nottingham, UK (E Townsend PhD); and Derbyshire Healthcare NHS Foundation Trust, Royal Derby Hospital, Derby, UK (J Ness MSc, K Waters RMN) Correspondence to: Prof Nav Kapur, Centre for Mental Health and Safety, Centre for Suicide Prevention, University of Manchester, Manchester M13 9PL, UK [email protected]
1
Articles
Research in context Evidence before this study We searched PubMed and the Cochrane Library databases up to Dec 31, 2014, for self-harm, suicide, intervention, and related search terms. Previous studies suggest that specialist assessment of self-harm is associated with a reduced risk of future self-harm in some subgroups. However, few studies have considered suicide or other causes of mortality as outcomes. Randomised trials would need many thousands of participants to be adequately powered to detect differences in mortality. Observational studies are one alternative. In 2014, Carroll and colleagues reviewed observational studies of the management of self-harm presentations at hospitals and associations with patient outcomes. They found no association between receipt of psychosocial assessment, or referral for aftercare, and subsequent suicide. There was weak evidence that hospitals where higher proportions of patients were admitted to a medical bed after self-harm had a lower incidence of suicide. As an ecological study, however, there was little scope for addressing confounding. The included studies were done over a 40 year period and the nature of routine management probably varied between settings and over time. Added value of this study We examined the association between four aspects of routine management (psychosocial assessment, medical
Methods Study design and participants We did this prospective cohort study with data collected through the Multicentre Study of Self-Harm in England.10 Three centres (in Oxford, Manchester, and Derby) each have an established monitoring system to collect data about episodes of self-harm presenting to emergency departments. Information was collected from assessments done by psychiatric or emergency department staff and from clinical records. We included individuals presenting with self-harm to five emergency departments in the UK (three in Manchester, one in Oxford, and one in Derby) from Jan 1, 2000, to Dec 31, 2010 (including those who did not wait for assessment or treatment). Patients younger than 16 years at the time of self-harm were not included because the models of service provision for this group were distinct from those for adults. Data included sociodemographic information, clinical factors such as previous psychiatric treatment and selfharm, details of the self-harm episode itself, and subsequent management. We also included a standard measure of socioeconomic deprivation—the Index of Multiple Deprivation—based on census and administrative data sources and assigned according to an individual’s postcode of residence.11 We used consistent definitions of self-harm across centres to include all acts 2
admission, psychiatric admission, referral for mental health follow-up) and suicide and all-cause mortality in three English centres. Most aspects of management were associated with increased mortality risk in unadjusted analyses, particularly psychiatric admission. After adjustment for baseline variables, the hazard ratios were smaller, and for psychiatric admission suggested a possible protective effect on all-cause mortality. There were significant interactions by subgroup, suggesting that the association between management and outcome was not the same for all people. Implications of all the available evidence This was an observational study and we could not establish causation. The unadjusted results showed that psychiatric admission or mental health follow-up were associated with the highest risks of death. This finding is probably a result of confounding by indication, and suggests that hospitals might be assigning the most intensive management to the highest risk patients. Adjusted results suggest that for some high-risk individuals (notably men, adults older than 65 years, and those with a history of self-harm), admission to an inpatient psychiatric ward might reduce 12-month allcause mortality. These findings should be replicated in other settings using alternative research designs and novel analyses that minimise confounding.
of intentional self-poisoning or self-injury, irrespective of motivation.12 This definition is used in UK national guidance. Alternative classifications such as non-suicidal self-injury (to denote episodes without suicidal intent) and suicide attempt (to denote episodes with evidence of suicide intent) are used in some settings;13,14 however, intent can be difficult to quantify, both for the treating clinician and the patient, and can fluctuate during and between episodes.15 There is an increased risk of early mortality for individuals engaging in either act.15 Because terminology might vary between countries,15 we also considered a more tightly defined group of people who had poisoned themselves and would be consistently included in most definitions. All centres had a self-harm or mental health liaison team that worked 7 days a week to provide specialist psychosocial assessments. The Manchester service was provided across three hospital sites. Out-of-hours cover in all centres was provided by junior psychiatrists or crisis teams. We assessed four aspects of hospital management: psychosocial assessment by a mental health specialist, admission to a medical bed, referral for specialist community mental health follow-up, and admission to a psychiatric bed. Psychosocial assessment is an assessment of personal circumstances, social context, mental state, risk, and needs following self-harm.16 Specialist mental health follow-up included referrals to outpatient or
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Articles
community mental health teams, secondary care crisis services, and drug and alcohol teams. Individuals might have received more than one type of hospital management but for simplicity and to retain sufficient numbers of outcome events for the analysis, we did not break down these four categories further.
Procedures We linked self-harm data to records held by the Data Linkage Service17 on the basis of name, sex, date of birth, National Health Service number, and postcode of last known address. These records provided current information on the status of patients in England, Wales, and Scotland, including information on mortality and coroners’ verdicts up to 2012. The ICD codes used to classify a death as suicide included intentional self-harm (X60–X84) and undetermined intent (Y10–Y34), consistent with research practice and public policy in the UK.18 A small proportion (n=842) of individuals could not be traced and so were excluded from the analyses.
Outcomes We studied two main outcomes: death by suicide and death by any cause within 12 months of a hospital presentation for self-harm. Although suicide was the main outcome in many studies, all-cause mortality is also much higher than expected in people who have selfharmed.4 We used a 12 month time period because we think that associations between hospital management and outcome over a longer period were unlikely to be direct and could be a result of interventions or changes to life circumstances in the intervening period. We excluded deaths that occurred as part of the self-harm episode (for example, death of a patient who presented to hospital alive following an antidepressant overdose but died in hospital some hours later).
Statistical analysis We analysed survival with Cox’s proportional hazards regression. Survival time began from the date of the selfharm episode and ended at either death, other exit such as emigration, or end of the follow-up period. To investigate the association between aspects of management and outcome, we calculated hazard ratios for suicide and all-cause mortality within 12 months of hospital presentation for self-harm. We plotted KaplanMeier survival curves comparing time to suicide and allcause mortality between individuals admitted to psychiatry and all others. For individuals who presented to hospital more than once during the study period, we considered only the most recent episode and the management received during that episode.19 Because allocation to treatment was probably heavily influenced by patient characteristics (confounding by indication), we adjusted the models for baseline demographic and clinical characteristics: sex, age, method of self-harm, drugs used in self-poisoning, previous or
current receipt of psychiatric treatment, history of selfharm, and deprivation score tertile. We selected these potential confounders as ones that might affect risk of suicide or death from other causes.3 We grouped missing demographic or clinical data into a distinct category within each variable. We repeated the adjusted analyses on complete data only as sensitivity analyses. We corrected SEs for clustering by hospital. We checked the proportional hazards assumption regarding the effect of the variables of interest over time with Schoenfeld residuals; the results were non-significant, suggesting that the proportional hazards assumption held. We also calculated interaction terms to investigate whether the association between hospital management and outcome varied by subgroup. We considered the following subgroups: sex, age group (<25 years, 25–44 years, 45–64 years, ≥65 years), ethnic origin (not white, white), method of self-harm (self-poisoning, selfinjury), previous and current psychiatric treatment, and previous self-harm. These were selected a priori and were based on factors which might be expected to modify associations between treatment and mortality outcomes on the basis of previous research. When the p value for the interaction term was less than 0·1, we present hazard ratios (HRs) for subgroups separately (comparing mortality in people in that subgroup who had and who had not received particular forms of management). We used this relatively relaxed threshold for the p value for interaction terms because we anticipated that the number of deaths would be small in relation to the size of the database and we did not wish to miss potentially important interactions.20 We used Stata (version 12) for all the statistical analyses. Psychosocial assessment (n=22144) Men Women Age <25 years
Medical inpatient admission (n=17227)
Referral for mental health follow-up (n=9979)
Psychiatric inpatient admission (n=1992)
9280 (41·9%)
7064 (41·0%)
4213 (42·2%)
965 (48·5%)
12 854 (58·0%)
10 158 (59·0%)
5765 (57·8%)
1025 (51·5%)
7511 (33·9%)
5702 (33·0%)
3063 (30·7%)
353 (17·7%)
Age 25–44 years
10 033 (45·3%)
7653 (44·4%)
4641 (46·5%)
936 (47·0%)
Age 45–64 years
3836 (17·3%)
3117 (18·1%)
1889 (18·9%)
418 (21·0%)
764 (3·5%)
755 (4·4%)
386 (3·9%)
285 (14·3%)
8506 (85·2%)
1591 (79·9%)
Age ≥65 years Method of self-poisoning
19 316 (87·2%)
15 760 (91·5%)
Previous self-harm
11 941 (53·9%)
8080 (46·9%)
6195 (62·0%)
1119 (56·2%)
8255 (37·3%)
5284 (30·7%)
5464 (57·8%)
1315 (66·0%)
10 227 (46·2%)
6768 (39·3%)
5827 (58·4%)
1256 (63·1%)
In current psychiatric treatment Previous psychiatric treatment Area-level deprivation tertile* Low deprivation
8139 (36·8%)
7673 (44·5%)
4093 (41·0%)
891 (44·7%)
Medium deprivation
6824 (30·8%)
5065 (29·3%)
2976 (29·8%)
625 (31·4%)
High deprivation
5956 (26·9%)
3592 (20·9%)
2378 (23·8%)
357 (17·9%)
Data are n (%). Categories of management not mutually exclusive. *Lowest tertile of deprivation is the most affluent.
Table 1: Baseline characteristics
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
3
Articles
The funders had no role in study design, data collection, analysis, or interpretation, writing of the report, or the decision to submit for publication. The views and opinions expressed in this Article are those of the authors and do not necessarily reflect those of the funders.
Results The sample included 38 415 individuals who had presented with self-harm during the 11-year study period and been followed up with respect to their mortality status. Median age was 30 years (IQR 22–41, range 16–97) and 22 013 (57%) were women. Self-poisoning was the most common method of self-harm (n=30 784, 80%). Of those who self-poisoned, 14 078 (46%) used paracetamol, 7654 (25%) used an antidepressant, and 3828 (10%) used a benzodiazepine (individuals may have used more than one substance and therefore categories are not mutually exclusive). Mean data completeness for variables was high (86%). 3113 (8·1%) of 38415 individuals did not wait for assessment, 22 144 (57·6%) received a psychosocial assessment, 9979 (26·0%) were referred for specialist mental health follow-up, and 1992 (5·2%) were admitted to a psychiatric bed (table 1). Of those who received an assessment, 11 592 (52·3%) were not referred for mental health
follow-up care or admitted to a psychiatric bed. In one centre, medical admission data were only available for more recent years (which reduced the sample size for this variable) but overall 17 227 (61·3%) of 28 119 individuals were admitted to a medical bed. Previous self-harm and previous or ongoing contact with services were common in patients who received specialist follow-up or psychiatric in-patient care. Older patients were over-represented in the psychiatric admission group. People from the least deprived areas were most likely to receive all types of active management (table 1). 261 (0·7%, 95% CI 0·6–0·8) individuals died by suicide and 832 (2·2%, 2·0–2·3) died from any cause within 12 months of a hospital presentation for A
Not admitted Admitted
100 90 80 100
70 Survival (%)
Role of the funding source
60
99·5
50 99
40 30
98·5
A
20
0
4 2·35 (1·59–3·45) 1·98 (1·29–3·04) Hazard ratio
0
1·42 (1·15–1·75)
200
300
400
0
3
2
100
10
Number at risk Not admitted 36 423 Admitted 1992
1·39 (1·18–1·65)
100
200
300
36 074 1955
35 880 1933
35 711 1907
400
B
1
100 90
0 80
B
100
70
3 Hazard ratio
1·61 (1·01–2·56) 2
2·35 (2·04–2·72)
1·50 (1·29–1·75) 1·16 (0·93–1·46)
Survival (%)
4 60
99
50
98
40
97
30
96 95
20
1
0
100
200
300
400
10 0 0
0 Psychosocial assessment
Medical inpatient Referral for mental admission health follow-up Hospital management
Psychiatric inpatient admission
Figure 1: Unadjusted hazard ratios for suicide (A) and death from any cause (B) within 12 months of presentation for self-harm by level of hospital management Error bars are 95% CIs.
4
Number at risk Not admitted 36 423 Admitted 1992
100
200
300
400
Time (days) 36 074 1955
35 880 1933
35 711 1907
Figure 2: Kaplan-Meier curves for suicide (A) and death from any cause (B) according to psychiatric admission status
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Articles
Discussion Aspects of routine hospital management after self-harm were associated with an increased risk of death in our unadjusted analyses. This association was particularly strong for psychiatric admission, which was associated with more than twice the risk of suicide or death from
A 4
Hazard ratio
3 1·41 (0·96–2·06) 2
1·48 (0·99–2·22) 1·12 (0·67–1·85)
1·35 (1·22–1·49)
1
0
B 4
3 Hazard ratio
self-harm. In unadjusted analysis, all aspects of management were associated with a high risk of suicide relative to not receiving the treatment (figure 1). The highest risks were associated with referral for specialist community mental health follow-up and admission to psychiatric in-patient care (figure 1A). Medical admission, referral for specialist outpatient care and psychiatric in-patient admission were associated with a significantly higher risk of death from any cause. The highest risk was associated with admission to psychiatric in-patient care (figure 1B). The risk for suicide seemed to be highest in the period early after admission (figure 2). Adjustment had little effect on the HRs for specialist assessment or medical admission, but reduced the HRs for referral for mental health follow-up and psychiatric admission, such that the associations were consistent with chance (figure 3). In general, adjustment reduced the hazard ratios of death from any cause, particularly for psychiatric admission. The HRs for both suicide and all-cause mortality were similar for the subset of individuals with complete data (data not shown). One exception was the association between medical inpatient admission and death: the adjusted HRs in the complete data analysis were smaller with wide CIs for suicide (HR 0·93, 95% CI 0·61–1·40) and for all-cause mortality (1·04, 0·83–1·28). For suicide deaths, we detected no significant interactions by subgroup apart from for psychiatric in-patient admission and age (table 2). Psychiatric admission might have been associated with a reduced suicide risk in participants aged 65 years and older (by contrast with the pattern in younger age groups) but with comparatively small numbers of deaths and the 95% CIs overlapped unity. For all-cause mortality, we showed evidence of interactions by age, sex, ethnic origin, and history of self-harm (table 2). Psychosocial assessment might have been associated with a reduced all-cause mortality risk in those with a history of self-harm. Medical admission was associated with a higher risk of death in people who had harmed themselves for the first time. For patients referred to specialist follow-up, risk of all-cause mortality was higher for those without a history of self-harm and those of white ethnic origin. Psychiatric inpatient admission was associated with reduced mortality risks in men, older adults, and those with a history of self-harm but might have been associated with increased risk in younger people. The findings of our sensitivity analysis including only self-poisoning episodes were much the same as those for the main analysis (appendix).
1·28 (0·84–1·97) 2
1·29 (1·01–1·64) 1·06 (0·95–1·18)
0·88 (0·72–1·07)
1
0 Psychosocial assessment
Medical inpatient admission
Referral for mental health follow-up
Psychiatric inpatient admission
Hospital management
Figure 3: Adjusted hazard ratios for suicide (A) and death from any cause (B) within 12 months of presentation for self-harm by level of hospital management Error bars are 95% CIs. Hazard ratios adjusted for sex, age, method of self-harm, drugs used in self-poisoning, previous or current receipt of psychiatric treatment, history of self-harm, and deprivation score tertile. Standard errors were corrected for clustering by hospital.
any cause within 12 months. Rather than suggesting a detrimental effect of hospital management, these findings probably suggest a selection effect (confounding by indication), with clinical services preferentially providing treatment to those most at risk. The unadjusted HRs reflect both selection effects and effects of treatment, and one approach to isolating the effects of management is to statistically adjust for differences in baseline factors between groups. The adjusted HR for psychiatric inpatient admission suggested that it might have been associated with a reduced all-cause mortality risk. However, this effect was not statistically significant and was not present for suicide mortality, and therefore it should be interpreted cautiously. Other forms of management (eg, psychosocial assessment, referral for specialist follow-up) continued to be associated with an increased risk of suicide or all-cause mortality. In our view, this finding is probably as a result of incomplete adjustment for confounding factors. The relation between hospital management and subsequent mortality was not uniform across groups. There were significant interactions by sex, age, and history of self-harm. Some forms of management were associated with a lower mortality risk in particular subgroups (eg,
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
See Online for appendix
5
Articles
Adjusted hazard ratio (95% CI)
pinteraction
Association with suicide Psychiatric inpatient admission Age (years)
<0·001
<25
2·54 (0·69–9·40)
25–44
1·13 (0·75–1·69)
45–64
1·19 (0·32–4·34)
≥65
0·70 (0·41–1·21)
Association with death from any cause Psychosocial assessment Any previous self-harm?
0·052
Yes
0·90 (0·81–1·01)
No
1·55 (0·99–2·42)
Medical inpatient admission Any previous self-harm?
0·04
Yes
1·22 (0·95–1·57)
No
1·63 (1·18–2·25)
Referral for mental health follow-up Ethnic origin
0·005
Non-white
0·74 (0·23–2·52)
White
1·46 (0·95–2·24)
Any previous self-harm?
0·001
Yes
1·23 (0·72–2·10)
No
1·65 (0·99–2·75)
Psychiatric in-patient admission Sex
0·098
Female
1·02 (0·65–1·60)
Male
0·79 (0·64–0·98)
Age (years)
0·0001
<25
2·09 (1·40–3·10)
25–44
1·00 (0·74–1·35)
45–64
1·14 (0·69–1·86)
≥65
0·68 (0·54–0·85)
Any previous self-harm?
0·006
Yes
0·75 (0·58–0·97)
No
0·90 (0·54–1·48)
Showing only those subgroups for which pinteraction<0·1. Hazard ratios compare risk of death in individuals who received a particular form of management to individuals in the same subgroup who did not receive that form of management (adjusted for baseline characteristics).
Table 2: The association between clinical management and suicide and death from any cause by subgroups
psychosocial assessment in those with a history of selfharm; psychiatric admission in men, older people, and those with a history of self-harm). By contrast, higher HRs were associated with medical admission for people without a past history of self-harm and with psychiatric admission for young people. Again, these associations could be a result of residual confounding, but they also might suggest that admission is not beneficial for all. This possibility could be explored in future work. We had a large sample with little loss to follow-up, minimising bias. This study is the first to our knowledge 6
to assess the association between the management that patients with self-harm received in hospital and their subsequent mortality risk. Previous studies have used repeat presentation to hospital with self-harm as an outcome, which could partly be a result of help-seeking behaviour. However, our findings should be interpreted in the context of some methodological limitations. This was an observational study without random allocation to different forms of treatment. Although we tried to statistically adjust for differences in baseline characteristics, residual confounding was probably still present. We cannot conclude that the difference in outcome between the groups was a result of the treatment they received in hospital. Equally, because we collected data for only broad categories of management, we cannot be certain of the nature, intensity, or quality of the interventions received in individual cases. Furthermore, although we knew when individuals were referred for specialist mental health follow-up, we could not know if this care was actually received. The categories of management we assessed were not mutually exclusive and so the associations might not be a result of the management in isolation. However, a previous study7 has shown that people who receive combinations of management approaches have a similar risk of repeated self-harm to those who receive a single approach. The data were obtained from three English centres, which might not be representative of other settings nationally or internationally, and we included only people presenting to hospital after self-harm. Data were collected over several years and it is possible that management changed over time. When we assessed this possibility retrospectively, the treatment that patients received stayed much the same, with the exception that the proportion of patients admitted to a medical bed increased from 38% in the earlier part of the study to 54% in the latter part of the study (possibly because of national waiting time restrictions in emergency departments). This is similar to the pattern across England.21 All-cause mortality also included people who died by suicide. When we assessed all nonsuicide deaths in a post-hoc analysis, the HRs were similar to those for all-cause mortality (data not shown). Our finding of an association between aspects of management and risk of death is consistent with the results of a Danish population-based case-control study,6 which showed a stepwise relation between suicide risk and level of psychiatric treatment. In the Danish study, people who had been admitted to inpatient care had the highest risks, followed by those who had emergency department contact, and then those with outpatient contact only. The survival curve suggested that the risks of suicide in our study were at their highest soon after admission. This finding is consistent with previous research of mental health patients.22
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Articles
With respect to the possible effects of routine aspects of management following self-harm, the evidence from randomised trials is weak. In one small UK study,23 77 patients were randomised to a brief medical admission or discharge home. Repetition, symptoms, or functioning at 1 week and 16 weeks did not differ between groups. However, the inclusion criteria (patients who had “no immediate physical or mental health care needs”) meant that the participants were a highly selected group representing only around one in six people who had self-harmed during the study period. A larger trial24 from the Netherlands found that psychiatric admission to a specialist ward (with an option to contact the unit after discharge) and problemsolving after-care did not reduce repetition or improve psychological functioning in patients who had selfharmed compared with usual care. In this study admissions were brief (1–4 days) and a quarter of patients in the usual treatment group were also admitted to a psychiatric bed.24 Observational studies have also examined the possible effects of hospital management. A review5 used aggregatelevel data from studies between 1970, and 2012, to examine the association between aspects of care and selfharm repetition or suicide. There was little evidence for the effect of psychosocial assessment or referral for aftercare, but based on five studies, the researchers reported that settings that admitted many patients to medical beds following self-harm seemed to have a lower incidence of suicide. Studies using individual-level data have particularly shown the possible protective effects of psychosocial assessment7,16 or the recognition of psychiatric disorder.8 However, both aggregate-level and individual-level studies are prone to confounding and these findings should be interpreted cautiously. Investigating the potential effects of routine treatments and management strategies for patients who self-harm is challenging and a variety of methods should be used. Randomised trials are a gold standard but previous studies have often been underpowered and have tended to recruit selected samples of patients.3 Observational studies can be much larger but are unable to provide causal explanations. Novel analytical methods such as use of the propensity score and instrumental variable analysis25,26 might increase the utility of observational data. Few qualitative studies have been done but such research has the potential to generate powerful explanatory insights.9 A greater understanding of the extent to which the content and quality of routine care following self-harm vary, and of the levels of care actually received following a referral, are also likely to improve our understanding about the effects in different settings. Future studies done in more centres could explore differences in associations by setting and model of service provision. With respect to the possible reduction in all-cause mortality after psychiatric admission, it could be that
admission offered an opportunity to assess and treat comorbid medical conditions, perhaps particularly in older patients. Recent policy initiatives have shown the importance of improving the physical health of psychiatric patients in England.21 Alcohol use contributes to the excess mortality in self-harm populations.27 The identification and subsequent treatment of alcohol disorders is another potential intervention to reduce mortality. Patients admitted to a psychiatric ward are also likely to receive specialist outpatient follow-up, which could affect outcomes. Our findings suggested that the mortality outcomes in relation to management after self-harm are similar when self-poisoning is considered in isolation. This finding could increase the generalisability of our findings to settings that use different classifications of suicidal behaviour. Our finding that clinical services seem to reserve the most intensive levels of treatment for the patients at highest risk is reassuring. However, a more important question is whether such treatment makes a difference to eventual outcome. Our findings on the role of psychiatric admission might warrant particular attention. In the UK, psychiatric admission after self-harm is rare, at around 7%, which might suggest that clinicians are reluctant to consider this management option.28 The availability of psychiatric beds is likely to be an important determinant of admission.29 There is even a view that psychiatric admission might actively contribute to suicide.30 By contrast, our results suggest that, at least for some high-risk patients, psychiatric admission might be a life-saving intervention. Contributors NK and SS designed the study. All authors collected and interpreted data. SS analysed the data. NK wrote the first draft of the report. All authors revised the report and approved the final version. Declaration of interests NK was Chair of the Guideline Development Group for the National Institute for Health and Care Excellence (NICE) self-harm guidelines (longer-term management). He is Chair of the NICE Depression in Adults guidelines. KH is a National Institute for Health Research Senior Investigator. Acknowledgments Funded by the UK Department of Health under the NHS R&D Programme (DH/DSH2008). NK and SS had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The views and opinions expressed within this paper are those of the authors and do not necessarily reflect those of the Department of Health. References 1 World Health Organization. Preventing suicide: a global imperative. 2014. http://www.who.int/mental_health/suicide-prevention/ world_report_2014/en/ (accessed Sept 26, 2014). 2 Carroll R, Metcalfe C, Gunnell D. Hospital presenting self-harm and risk of fatal and non-fatal repetition: systematic review and meta-analysis. PLoS One 2014; 9: e89944. 3 NICE. The long term care and treatment of self-harm. Clinical Guideline 133. 2011. 4 Bergen H, Hawton K, Waters K, et al. Premature death after self-harm: a multicentre cohort study. Lancet 2012; 380: 1568–74. 5 Carroll R, Metcalfe C, Gunnell D. Hospital management of self-harm patients and risk of repetition: systematic review and meta-analysis. J Affect Disord 2014; 168: 476–83.
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
7
Articles
6
7
8
9
10
11
12
13
14
15
16
17 18
8
Hjorthoj CR, Madsen T, Agerbo E, Nordentoft M. Risk of suicide according to level of psychiatric treatment: a nationwide nested case-control study. Soc Psychiatry Psychiatr Epidemiol 2014; 49: 1357–65. Kapur N, Steeg S, Webb R, et al. Does clinical management improve outcomes following self-harm? Results from the Multicentre Study of Self-Harm in England. PLoS One 2013; 8: e70434. Olfson M, Marcus SC, Bridge JA. Emergency department recognition of mental disorders and short-term outcome of deliberate self-harm. Am J Psychiatry 2013; 170: 1442–50. Hunter C, Chantler K, Kapur N, Cooper J. Service user perspectives on psychosocial assessment following self-harm and its impact on further help-seeking: a qualitative study. J Affect Disord 2013; 145: 315–23. Bergen H, Hawton K, Waters K, Cooper J, Kapur N. Epidemiology and trends in non-fatal self-harm in three centres in England: 2000–2007. Br J Psychiatry 2010; 197: 493–98. Department for Communities and Local Government. The English Indices of Deprivation 2007. https://www.gov.uk/government/ collections/english-indices-of-deprivation (accessed Aug 3, 2015). Hawton K, Bergen H, Casey D, et al. Self-harm in England: a tale of three cities — multicentre study of self-harm. Soc Psychiatry Psychiatr Epidemiol 2007; 42: 513–21. Silverman MM, Berman AL, Sanddal ND, O’Carroll PW, Joiner TE Jr. Rebuilding the Tower of Babel: a revised nomenclature for the study of suicide and suicidal behaviors—part 1: background, rationale, and methodology. Suicide Life Threat Behav 2007; 37: 248–63. Silverman MM, Berman AL, Sanddal ND, O’Carroll PW, Joiner TE Jr. Rebuilding the Tower of Babel: a revised nomenclature for the study of suicide and suicidal behaviors—part 2: suiciderelated ideations, communications, and behaviors. Suicide Life Threat Behav 2007; 37: 264–77. Kapur N, Cooper J, O’Connor RC, Hawton K. Non-suicidal self-injury v. attempted suicide: new diagnosis or false dichotomy? Br J Psychiatry 2013; 202: 326–28. Kapur N, Murphy E, Cooper J, et al. Psychosocial assessment following self-harm: results from the multi-centre monitoring of self-harm project. J Affect Disord 2008; 106: 285–93. Health and Social Care Information Centre. Data Linkage Service. http://www.hscic.gov.uk/datalinkage (accessed Aug 3, 2015). Health Improvement Analytical Team. Statistical update on suicide: January 2014. http://www.gov.uk/government/uploads/system/ uploads/attachment_data/file/278120/Suicide_update_Jan_2014_ FINAL_revised.pdf (accessed June 18, 2015).
19
20
21
22 23
24
25
26 27
28
29
30
Bergen H, Hawton K, Waters K, Cooper J, Kapur N. Psychosocial assessment and repetition of self-harm: the significance of single and multiple repeat episode analyses. J Affect Disord 2010; 127: 257–65. Magruder KM, Ouyang B, Miller S, Tilley BC. Retention of under-represented minorities in drug abuse treatment studies. Clin Trials 2009; 6: 252–60. NHS England. Mental health staff to introduce physical ‘MOT’ to reduce deaths. 2014. http://www.england.nhs.uk/2014/05/15/ physical-mot/ (accessed March 23, 2014). Hunt IM, Kapur N, Webb R, et al. Suicide in recently discharged psychiatric patients: a case-control study. Psychol Med 2009; 39: 443–49. Waterhouse J, Platt S. General-hospital admission in the management of parasuicide - a randomized controlled trial. Br J Psychiatry 1990; 156: 236–42. VanderSande R, VanRooijen L, Buskens E, et al. Intensive in-patient and community intervention versus routine care after attempted suicide—a randomised controlled intervention study. Br J Psychiatry 1997; 171: 35–41. Stukel TA, Fisher ES, Wennberg DE, Alter DA, Gottlieb DJ, Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias - Effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA 2007; 297: 278–85. Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med 1997; 127: 757–63. Bergen H, Hawton K, Webb R, et al. Alcohol-related mortality following self-harm: a multicentre cohort study. J R Soc Med Open 2014; 5: 2054270414533326. Cooper J, Steeg S, Bennewith O, et al. Are hospital services for self-harm getting better? An observational study examining management, service provision and temporal trends in England. BMJ Open 2013; 3: e003444 . Keown P, Mercer G, Scott J. Retrospective analysis of hospital episode statistics, involuntary admissions under the Mental Health Act 1983, and number of psychiatric beds in England 1996–2006. BMJ 2008; 337: a1837. Large MM, Ryan CJ. Disturbing findings about the risk of suicide and psychiatric hospitals. Soc Psychiatry Psychiatr Epidemiol 2014; 49: 1353–55.
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Hospital management of suicidal behaviour and subsequent mortality: a prospective cohort study Nav Kapur, Sarah Steeg, Pauline Turnbull, Roger Webb, Helen Bergen, Keith Hawton, Galit Geulayov, Ellen Townsend, Jennifer Ness, Keith Waters, Jayne Cooper
Summary Background Self-poisoning and self-injury are associated with a high risk of suicide or death from any cause but the effect of routine aspects of hospital management on mortality risk is unknown. Methods We did a prospective cohort study using data for adults who had self-harmed presenting to five emergency departments in the UK between 2000 and 2010. We assessed the relation between four aspects of management (psychosocial assessment, medical admission, psychiatric admission, referral for mental health follow-up) and death by suicide or any cause within 12 months of presentation. Findings Of 38 415 individuals presenting with self-harm, 261 (0·7%) died by suicide and 832 (2·2%) died from any cause within 12 months. Most aspects of management were associated with a higher mortality risk in unadjusted analyses. Psychiatric admission was associated with the highest risks for both suicide (hazard ratio 2·35, 95% CI 1·59–3·45) and all-cause mortality (2·35, 2·04–2·72). After adjustment for baseline variables, the hazard ratios were generally smaller, particularly for psychiatric admission. There were significant interactions by sex, age, and history of self-harm. Interpretation This was an observational study and so we cannot infer causation. However, our finding that clinical services seem to reserve the most intensive levels of treatment for patients at highest risk is reassuring. Aspects of routine management might be associated with a lower mortality risk but these effects vary by clinical subgroup. Funding UK Department of Health.
Introduction Self-poisoning and self-injury are major health problems in many countries and are associated with an increased risk of subsequent early death.1 Many studies have focused on suicide as an outcome. Results of a systematic review2 showed that around one in 25 people who presented to hospital with self-harm died by suicide in the following 5 years. A clinical guideline that included a review of previous studies concluded that previous suicidal behaviour, suicidal intent, and being male were the strongest risk factors for suicide following self-harm.3 No protective factors had a consistent association across studies. The risk of death from other causes is also increased. In a cohort study, Kapur and colleagues reported that all-cause mortality in a population of people who had self-harmed was four times higher than expected.4 Individuals who died lost an average of more than 30 years of life,4 although the mortality risk was not uniform for different causes of death. Few studies have assessed the association between routine aspects of hospital care and subsequent mortality.5 Health services might reserve the most intensive forms of treatment for the highest risk patients. In this case, one might expect hospital management to be associated with an increased risk of death. In a Danish general population study,6 suicide risk varied in a stepwise pattern; patients with recent psychiatric admission had a higher incidence of suicide than did those who received outpatient mental
health treatment only, who in turn had a higher incidence than did those who received no treatment. These associations are not causal. This study is an example of confounding by indication, in which allocation to particular treatments was probably based on the underlying need. Routine interventions might also have a positive effect on outcome and reduce risk. One of the challenges in non-randomised studies is to estimate these treatment effects in the context of the sometimes much larger effects of selection or confounding by indication. One approach is to statistically adjust for baseline differences between people who do and do not receive particular forms of management.7 Kapur and colleagues7 (using repeat hospital presentation for self-harm rather than mortality as an outcome measure) showed that receiving a psychosocial assessment was associated with 40% reduction in the risk of repetition within 12 months in two of three centres. In a US study,8 recognition of psychiatric disorder was associated with a 34% reduction in repeat self-harm. There are limitations of using hospital presentations for repeat self-harm as an outcome measure; some people might repeat but choose not to present to health services.9 It is also likely that associations between management and outcome are not uniform3 and so assessing subgroup differences might be useful. We assessed the association between the hospital management of self-harm and subsequent mortality in a large cohort of patients in England.
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Lancet Psychiatry 2015 Published Online August 6, 2015 http://dx.doi.org/10.1016/ S2215-0366(15)00169-8 See Online/Comment http://dx.doi.org/10.1016/ S2215-0366(15)00212-6 Centre for Mental Health and Safety, Centre for Suicide Prevention, University of Manchester, Manchester, UK (Prof N Kapur FRCPsych, S Steeg BA, P Turnbull PhD, R Webb PhD, J Cooper PhD); Manchester Mental Health and Social Care Trust, Manchester, UK (Prof N Kapur); Centre for Suicide Research, University of Oxford, Department of Psychiatry, Warneford Hospital, Oxford, UK (H Bergen PhD, Prof K Hawton DSc, G Geulayov PhD); School of Psychology, University of Nottingham, Nottingham, UK (E Townsend PhD); and Derbyshire Healthcare NHS Foundation Trust, Royal Derby Hospital, Derby, UK (J Ness MSc, K Waters RMN) Correspondence to: Prof Nav Kapur, Centre for Mental Health and Safety, Centre for Suicide Prevention, University of Manchester, Manchester M13 9PL, UK [email protected]
1
Articles
Research in context Evidence before this study We searched PubMed and the Cochrane Library databases up to Dec 31, 2014, for self-harm, suicide, intervention, and related search terms. Previous studies suggest that specialist assessment of self-harm is associated with a reduced risk of future self-harm in some subgroups. However, few studies have considered suicide or other causes of mortality as outcomes. Randomised trials would need many thousands of participants to be adequately powered to detect differences in mortality. Observational studies are one alternative. In 2014, Carroll and colleagues reviewed observational studies of the management of self-harm presentations at hospitals and associations with patient outcomes. They found no association between receipt of psychosocial assessment, or referral for aftercare, and subsequent suicide. There was weak evidence that hospitals where higher proportions of patients were admitted to a medical bed after self-harm had a lower incidence of suicide. As an ecological study, however, there was little scope for addressing confounding. The included studies were done over a 40 year period and the nature of routine management probably varied between settings and over time. Added value of this study We examined the association between four aspects of routine management (psychosocial assessment, medical
Methods Study design and participants We did this prospective cohort study with data collected through the Multicentre Study of Self-Harm in England.10 Three centres (in Oxford, Manchester, and Derby) each have an established monitoring system to collect data about episodes of self-harm presenting to emergency departments. Information was collected from assessments done by psychiatric or emergency department staff and from clinical records. We included individuals presenting with self-harm to five emergency departments in the UK (three in Manchester, one in Oxford, and one in Derby) from Jan 1, 2000, to Dec 31, 2010 (including those who did not wait for assessment or treatment). Patients younger than 16 years at the time of self-harm were not included because the models of service provision for this group were distinct from those for adults. Data included sociodemographic information, clinical factors such as previous psychiatric treatment and selfharm, details of the self-harm episode itself, and subsequent management. We also included a standard measure of socioeconomic deprivation—the Index of Multiple Deprivation—based on census and administrative data sources and assigned according to an individual’s postcode of residence.11 We used consistent definitions of self-harm across centres to include all acts 2
admission, psychiatric admission, referral for mental health follow-up) and suicide and all-cause mortality in three English centres. Most aspects of management were associated with increased mortality risk in unadjusted analyses, particularly psychiatric admission. After adjustment for baseline variables, the hazard ratios were smaller, and for psychiatric admission suggested a possible protective effect on all-cause mortality. There were significant interactions by subgroup, suggesting that the association between management and outcome was not the same for all people. Implications of all the available evidence This was an observational study and we could not establish causation. The unadjusted results showed that psychiatric admission or mental health follow-up were associated with the highest risks of death. This finding is probably a result of confounding by indication, and suggests that hospitals might be assigning the most intensive management to the highest risk patients. Adjusted results suggest that for some high-risk individuals (notably men, adults older than 65 years, and those with a history of self-harm), admission to an inpatient psychiatric ward might reduce 12-month allcause mortality. These findings should be replicated in other settings using alternative research designs and novel analyses that minimise confounding.
of intentional self-poisoning or self-injury, irrespective of motivation.12 This definition is used in UK national guidance. Alternative classifications such as non-suicidal self-injury (to denote episodes without suicidal intent) and suicide attempt (to denote episodes with evidence of suicide intent) are used in some settings;13,14 however, intent can be difficult to quantify, both for the treating clinician and the patient, and can fluctuate during and between episodes.15 There is an increased risk of early mortality for individuals engaging in either act.15 Because terminology might vary between countries,15 we also considered a more tightly defined group of people who had poisoned themselves and would be consistently included in most definitions. All centres had a self-harm or mental health liaison team that worked 7 days a week to provide specialist psychosocial assessments. The Manchester service was provided across three hospital sites. Out-of-hours cover in all centres was provided by junior psychiatrists or crisis teams. We assessed four aspects of hospital management: psychosocial assessment by a mental health specialist, admission to a medical bed, referral for specialist community mental health follow-up, and admission to a psychiatric bed. Psychosocial assessment is an assessment of personal circumstances, social context, mental state, risk, and needs following self-harm.16 Specialist mental health follow-up included referrals to outpatient or
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Articles
community mental health teams, secondary care crisis services, and drug and alcohol teams. Individuals might have received more than one type of hospital management but for simplicity and to retain sufficient numbers of outcome events for the analysis, we did not break down these four categories further.
Procedures We linked self-harm data to records held by the Data Linkage Service17 on the basis of name, sex, date of birth, National Health Service number, and postcode of last known address. These records provided current information on the status of patients in England, Wales, and Scotland, including information on mortality and coroners’ verdicts up to 2012. The ICD codes used to classify a death as suicide included intentional self-harm (X60–X84) and undetermined intent (Y10–Y34), consistent with research practice and public policy in the UK.18 A small proportion (n=842) of individuals could not be traced and so were excluded from the analyses.
Outcomes We studied two main outcomes: death by suicide and death by any cause within 12 months of a hospital presentation for self-harm. Although suicide was the main outcome in many studies, all-cause mortality is also much higher than expected in people who have selfharmed.4 We used a 12 month time period because we think that associations between hospital management and outcome over a longer period were unlikely to be direct and could be a result of interventions or changes to life circumstances in the intervening period. We excluded deaths that occurred as part of the self-harm episode (for example, death of a patient who presented to hospital alive following an antidepressant overdose but died in hospital some hours later).
Statistical analysis We analysed survival with Cox’s proportional hazards regression. Survival time began from the date of the selfharm episode and ended at either death, other exit such as emigration, or end of the follow-up period. To investigate the association between aspects of management and outcome, we calculated hazard ratios for suicide and all-cause mortality within 12 months of hospital presentation for self-harm. We plotted KaplanMeier survival curves comparing time to suicide and allcause mortality between individuals admitted to psychiatry and all others. For individuals who presented to hospital more than once during the study period, we considered only the most recent episode and the management received during that episode.19 Because allocation to treatment was probably heavily influenced by patient characteristics (confounding by indication), we adjusted the models for baseline demographic and clinical characteristics: sex, age, method of self-harm, drugs used in self-poisoning, previous or
current receipt of psychiatric treatment, history of selfharm, and deprivation score tertile. We selected these potential confounders as ones that might affect risk of suicide or death from other causes.3 We grouped missing demographic or clinical data into a distinct category within each variable. We repeated the adjusted analyses on complete data only as sensitivity analyses. We corrected SEs for clustering by hospital. We checked the proportional hazards assumption regarding the effect of the variables of interest over time with Schoenfeld residuals; the results were non-significant, suggesting that the proportional hazards assumption held. We also calculated interaction terms to investigate whether the association between hospital management and outcome varied by subgroup. We considered the following subgroups: sex, age group (<25 years, 25–44 years, 45–64 years, ≥65 years), ethnic origin (not white, white), method of self-harm (self-poisoning, selfinjury), previous and current psychiatric treatment, and previous self-harm. These were selected a priori and were based on factors which might be expected to modify associations between treatment and mortality outcomes on the basis of previous research. When the p value for the interaction term was less than 0·1, we present hazard ratios (HRs) for subgroups separately (comparing mortality in people in that subgroup who had and who had not received particular forms of management). We used this relatively relaxed threshold for the p value for interaction terms because we anticipated that the number of deaths would be small in relation to the size of the database and we did not wish to miss potentially important interactions.20 We used Stata (version 12) for all the statistical analyses. Psychosocial assessment (n=22144) Men Women Age <25 years
Medical inpatient admission (n=17227)
Referral for mental health follow-up (n=9979)
Psychiatric inpatient admission (n=1992)
9280 (41·9%)
7064 (41·0%)
4213 (42·2%)
965 (48·5%)
12 854 (58·0%)
10 158 (59·0%)
5765 (57·8%)
1025 (51·5%)
7511 (33·9%)
5702 (33·0%)
3063 (30·7%)
353 (17·7%)
Age 25–44 years
10 033 (45·3%)
7653 (44·4%)
4641 (46·5%)
936 (47·0%)
Age 45–64 years
3836 (17·3%)
3117 (18·1%)
1889 (18·9%)
418 (21·0%)
764 (3·5%)
755 (4·4%)
386 (3·9%)
285 (14·3%)
8506 (85·2%)
1591 (79·9%)
Age ≥65 years Method of self-poisoning
19 316 (87·2%)
15 760 (91·5%)
Previous self-harm
11 941 (53·9%)
8080 (46·9%)
6195 (62·0%)
1119 (56·2%)
8255 (37·3%)
5284 (30·7%)
5464 (57·8%)
1315 (66·0%)
10 227 (46·2%)
6768 (39·3%)
5827 (58·4%)
1256 (63·1%)
In current psychiatric treatment Previous psychiatric treatment Area-level deprivation tertile* Low deprivation
8139 (36·8%)
7673 (44·5%)
4093 (41·0%)
891 (44·7%)
Medium deprivation
6824 (30·8%)
5065 (29·3%)
2976 (29·8%)
625 (31·4%)
High deprivation
5956 (26·9%)
3592 (20·9%)
2378 (23·8%)
357 (17·9%)
Data are n (%). Categories of management not mutually exclusive. *Lowest tertile of deprivation is the most affluent.
Table 1: Baseline characteristics
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
3
Articles
The funders had no role in study design, data collection, analysis, or interpretation, writing of the report, or the decision to submit for publication. The views and opinions expressed in this Article are those of the authors and do not necessarily reflect those of the funders.
Results The sample included 38 415 individuals who had presented with self-harm during the 11-year study period and been followed up with respect to their mortality status. Median age was 30 years (IQR 22–41, range 16–97) and 22 013 (57%) were women. Self-poisoning was the most common method of self-harm (n=30 784, 80%). Of those who self-poisoned, 14 078 (46%) used paracetamol, 7654 (25%) used an antidepressant, and 3828 (10%) used a benzodiazepine (individuals may have used more than one substance and therefore categories are not mutually exclusive). Mean data completeness for variables was high (86%). 3113 (8·1%) of 38415 individuals did not wait for assessment, 22 144 (57·6%) received a psychosocial assessment, 9979 (26·0%) were referred for specialist mental health follow-up, and 1992 (5·2%) were admitted to a psychiatric bed (table 1). Of those who received an assessment, 11 592 (52·3%) were not referred for mental health
follow-up care or admitted to a psychiatric bed. In one centre, medical admission data were only available for more recent years (which reduced the sample size for this variable) but overall 17 227 (61·3%) of 28 119 individuals were admitted to a medical bed. Previous self-harm and previous or ongoing contact with services were common in patients who received specialist follow-up or psychiatric in-patient care. Older patients were over-represented in the psychiatric admission group. People from the least deprived areas were most likely to receive all types of active management (table 1). 261 (0·7%, 95% CI 0·6–0·8) individuals died by suicide and 832 (2·2%, 2·0–2·3) died from any cause within 12 months of a hospital presentation for A
Not admitted Admitted
100 90 80 100
70 Survival (%)
Role of the funding source
60
99·5
50 99
40 30
98·5
A
20
0
4 2·35 (1·59–3·45) 1·98 (1·29–3·04) Hazard ratio
0
1·42 (1·15–1·75)
200
300
400
0
3
2
100
10
Number at risk Not admitted 36 423 Admitted 1992
1·39 (1·18–1·65)
100
200
300
36 074 1955
35 880 1933
35 711 1907
400
B
1
100 90
0 80
B
100
70
3 Hazard ratio
1·61 (1·01–2·56) 2
2·35 (2·04–2·72)
1·50 (1·29–1·75) 1·16 (0·93–1·46)
Survival (%)
4 60
99
50
98
40
97
30
96 95
20
1
0
100
200
300
400
10 0 0
0 Psychosocial assessment
Medical inpatient Referral for mental admission health follow-up Hospital management
Psychiatric inpatient admission
Figure 1: Unadjusted hazard ratios for suicide (A) and death from any cause (B) within 12 months of presentation for self-harm by level of hospital management Error bars are 95% CIs.
4
Number at risk Not admitted 36 423 Admitted 1992
100
200
300
400
Time (days) 36 074 1955
35 880 1933
35 711 1907
Figure 2: Kaplan-Meier curves for suicide (A) and death from any cause (B) according to psychiatric admission status
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Articles
Discussion Aspects of routine hospital management after self-harm were associated with an increased risk of death in our unadjusted analyses. This association was particularly strong for psychiatric admission, which was associated with more than twice the risk of suicide or death from
A 4
Hazard ratio
3 1·41 (0·96–2·06) 2
1·48 (0·99–2·22) 1·12 (0·67–1·85)
1·35 (1·22–1·49)
1
0
B 4
3 Hazard ratio
self-harm. In unadjusted analysis, all aspects of management were associated with a high risk of suicide relative to not receiving the treatment (figure 1). The highest risks were associated with referral for specialist community mental health follow-up and admission to psychiatric in-patient care (figure 1A). Medical admission, referral for specialist outpatient care and psychiatric in-patient admission were associated with a significantly higher risk of death from any cause. The highest risk was associated with admission to psychiatric in-patient care (figure 1B). The risk for suicide seemed to be highest in the period early after admission (figure 2). Adjustment had little effect on the HRs for specialist assessment or medical admission, but reduced the HRs for referral for mental health follow-up and psychiatric admission, such that the associations were consistent with chance (figure 3). In general, adjustment reduced the hazard ratios of death from any cause, particularly for psychiatric admission. The HRs for both suicide and all-cause mortality were similar for the subset of individuals with complete data (data not shown). One exception was the association between medical inpatient admission and death: the adjusted HRs in the complete data analysis were smaller with wide CIs for suicide (HR 0·93, 95% CI 0·61–1·40) and for all-cause mortality (1·04, 0·83–1·28). For suicide deaths, we detected no significant interactions by subgroup apart from for psychiatric in-patient admission and age (table 2). Psychiatric admission might have been associated with a reduced suicide risk in participants aged 65 years and older (by contrast with the pattern in younger age groups) but with comparatively small numbers of deaths and the 95% CIs overlapped unity. For all-cause mortality, we showed evidence of interactions by age, sex, ethnic origin, and history of self-harm (table 2). Psychosocial assessment might have been associated with a reduced all-cause mortality risk in those with a history of self-harm. Medical admission was associated with a higher risk of death in people who had harmed themselves for the first time. For patients referred to specialist follow-up, risk of all-cause mortality was higher for those without a history of self-harm and those of white ethnic origin. Psychiatric inpatient admission was associated with reduced mortality risks in men, older adults, and those with a history of self-harm but might have been associated with increased risk in younger people. The findings of our sensitivity analysis including only self-poisoning episodes were much the same as those for the main analysis (appendix).
1·28 (0·84–1·97) 2
1·29 (1·01–1·64) 1·06 (0·95–1·18)
0·88 (0·72–1·07)
1
0 Psychosocial assessment
Medical inpatient admission
Referral for mental health follow-up
Psychiatric inpatient admission
Hospital management
Figure 3: Adjusted hazard ratios for suicide (A) and death from any cause (B) within 12 months of presentation for self-harm by level of hospital management Error bars are 95% CIs. Hazard ratios adjusted for sex, age, method of self-harm, drugs used in self-poisoning, previous or current receipt of psychiatric treatment, history of self-harm, and deprivation score tertile. Standard errors were corrected for clustering by hospital.
any cause within 12 months. Rather than suggesting a detrimental effect of hospital management, these findings probably suggest a selection effect (confounding by indication), with clinical services preferentially providing treatment to those most at risk. The unadjusted HRs reflect both selection effects and effects of treatment, and one approach to isolating the effects of management is to statistically adjust for differences in baseline factors between groups. The adjusted HR for psychiatric inpatient admission suggested that it might have been associated with a reduced all-cause mortality risk. However, this effect was not statistically significant and was not present for suicide mortality, and therefore it should be interpreted cautiously. Other forms of management (eg, psychosocial assessment, referral for specialist follow-up) continued to be associated with an increased risk of suicide or all-cause mortality. In our view, this finding is probably as a result of incomplete adjustment for confounding factors. The relation between hospital management and subsequent mortality was not uniform across groups. There were significant interactions by sex, age, and history of self-harm. Some forms of management were associated with a lower mortality risk in particular subgroups (eg,
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
See Online for appendix
5
Articles
Adjusted hazard ratio (95% CI)
pinteraction
Association with suicide Psychiatric inpatient admission Age (years)
<0·001
<25
2·54 (0·69–9·40)
25–44
1·13 (0·75–1·69)
45–64
1·19 (0·32–4·34)
≥65
0·70 (0·41–1·21)
Association with death from any cause Psychosocial assessment Any previous self-harm?
0·052
Yes
0·90 (0·81–1·01)
No
1·55 (0·99–2·42)
Medical inpatient admission Any previous self-harm?
0·04
Yes
1·22 (0·95–1·57)
No
1·63 (1·18–2·25)
Referral for mental health follow-up Ethnic origin
0·005
Non-white
0·74 (0·23–2·52)
White
1·46 (0·95–2·24)
Any previous self-harm?
0·001
Yes
1·23 (0·72–2·10)
No
1·65 (0·99–2·75)
Psychiatric in-patient admission Sex
0·098
Female
1·02 (0·65–1·60)
Male
0·79 (0·64–0·98)
Age (years)
0·0001
<25
2·09 (1·40–3·10)
25–44
1·00 (0·74–1·35)
45–64
1·14 (0·69–1·86)
≥65
0·68 (0·54–0·85)
Any previous self-harm?
0·006
Yes
0·75 (0·58–0·97)
No
0·90 (0·54–1·48)
Showing only those subgroups for which pinteraction<0·1. Hazard ratios compare risk of death in individuals who received a particular form of management to individuals in the same subgroup who did not receive that form of management (adjusted for baseline characteristics).
Table 2: The association between clinical management and suicide and death from any cause by subgroups
psychosocial assessment in those with a history of selfharm; psychiatric admission in men, older people, and those with a history of self-harm). By contrast, higher HRs were associated with medical admission for people without a past history of self-harm and with psychiatric admission for young people. Again, these associations could be a result of residual confounding, but they also might suggest that admission is not beneficial for all. This possibility could be explored in future work. We had a large sample with little loss to follow-up, minimising bias. This study is the first to our knowledge 6
to assess the association between the management that patients with self-harm received in hospital and their subsequent mortality risk. Previous studies have used repeat presentation to hospital with self-harm as an outcome, which could partly be a result of help-seeking behaviour. However, our findings should be interpreted in the context of some methodological limitations. This was an observational study without random allocation to different forms of treatment. Although we tried to statistically adjust for differences in baseline characteristics, residual confounding was probably still present. We cannot conclude that the difference in outcome between the groups was a result of the treatment they received in hospital. Equally, because we collected data for only broad categories of management, we cannot be certain of the nature, intensity, or quality of the interventions received in individual cases. Furthermore, although we knew when individuals were referred for specialist mental health follow-up, we could not know if this care was actually received. The categories of management we assessed were not mutually exclusive and so the associations might not be a result of the management in isolation. However, a previous study7 has shown that people who receive combinations of management approaches have a similar risk of repeated self-harm to those who receive a single approach. The data were obtained from three English centres, which might not be representative of other settings nationally or internationally, and we included only people presenting to hospital after self-harm. Data were collected over several years and it is possible that management changed over time. When we assessed this possibility retrospectively, the treatment that patients received stayed much the same, with the exception that the proportion of patients admitted to a medical bed increased from 38% in the earlier part of the study to 54% in the latter part of the study (possibly because of national waiting time restrictions in emergency departments). This is similar to the pattern across England.21 All-cause mortality also included people who died by suicide. When we assessed all nonsuicide deaths in a post-hoc analysis, the HRs were similar to those for all-cause mortality (data not shown). Our finding of an association between aspects of management and risk of death is consistent with the results of a Danish population-based case-control study,6 which showed a stepwise relation between suicide risk and level of psychiatric treatment. In the Danish study, people who had been admitted to inpatient care had the highest risks, followed by those who had emergency department contact, and then those with outpatient contact only. The survival curve suggested that the risks of suicide in our study were at their highest soon after admission. This finding is consistent with previous research of mental health patients.22
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
Articles
With respect to the possible effects of routine aspects of management following self-harm, the evidence from randomised trials is weak. In one small UK study,23 77 patients were randomised to a brief medical admission or discharge home. Repetition, symptoms, or functioning at 1 week and 16 weeks did not differ between groups. However, the inclusion criteria (patients who had “no immediate physical or mental health care needs”) meant that the participants were a highly selected group representing only around one in six people who had self-harmed during the study period. A larger trial24 from the Netherlands found that psychiatric admission to a specialist ward (with an option to contact the unit after discharge) and problemsolving after-care did not reduce repetition or improve psychological functioning in patients who had selfharmed compared with usual care. In this study admissions were brief (1–4 days) and a quarter of patients in the usual treatment group were also admitted to a psychiatric bed.24 Observational studies have also examined the possible effects of hospital management. A review5 used aggregatelevel data from studies between 1970, and 2012, to examine the association between aspects of care and selfharm repetition or suicide. There was little evidence for the effect of psychosocial assessment or referral for aftercare, but based on five studies, the researchers reported that settings that admitted many patients to medical beds following self-harm seemed to have a lower incidence of suicide. Studies using individual-level data have particularly shown the possible protective effects of psychosocial assessment7,16 or the recognition of psychiatric disorder.8 However, both aggregate-level and individual-level studies are prone to confounding and these findings should be interpreted cautiously. Investigating the potential effects of routine treatments and management strategies for patients who self-harm is challenging and a variety of methods should be used. Randomised trials are a gold standard but previous studies have often been underpowered and have tended to recruit selected samples of patients.3 Observational studies can be much larger but are unable to provide causal explanations. Novel analytical methods such as use of the propensity score and instrumental variable analysis25,26 might increase the utility of observational data. Few qualitative studies have been done but such research has the potential to generate powerful explanatory insights.9 A greater understanding of the extent to which the content and quality of routine care following self-harm vary, and of the levels of care actually received following a referral, are also likely to improve our understanding about the effects in different settings. Future studies done in more centres could explore differences in associations by setting and model of service provision. With respect to the possible reduction in all-cause mortality after psychiatric admission, it could be that
admission offered an opportunity to assess and treat comorbid medical conditions, perhaps particularly in older patients. Recent policy initiatives have shown the importance of improving the physical health of psychiatric patients in England.21 Alcohol use contributes to the excess mortality in self-harm populations.27 The identification and subsequent treatment of alcohol disorders is another potential intervention to reduce mortality. Patients admitted to a psychiatric ward are also likely to receive specialist outpatient follow-up, which could affect outcomes. Our findings suggested that the mortality outcomes in relation to management after self-harm are similar when self-poisoning is considered in isolation. This finding could increase the generalisability of our findings to settings that use different classifications of suicidal behaviour. Our finding that clinical services seem to reserve the most intensive levels of treatment for the patients at highest risk is reassuring. However, a more important question is whether such treatment makes a difference to eventual outcome. Our findings on the role of psychiatric admission might warrant particular attention. In the UK, psychiatric admission after self-harm is rare, at around 7%, which might suggest that clinicians are reluctant to consider this management option.28 The availability of psychiatric beds is likely to be an important determinant of admission.29 There is even a view that psychiatric admission might actively contribute to suicide.30 By contrast, our results suggest that, at least for some high-risk patients, psychiatric admission might be a life-saving intervention. Contributors NK and SS designed the study. All authors collected and interpreted data. SS analysed the data. NK wrote the first draft of the report. All authors revised the report and approved the final version. Declaration of interests NK was Chair of the Guideline Development Group for the National Institute for Health and Care Excellence (NICE) self-harm guidelines (longer-term management). He is Chair of the NICE Depression in Adults guidelines. KH is a National Institute for Health Research Senior Investigator. Acknowledgments Funded by the UK Department of Health under the NHS R&D Programme (DH/DSH2008). NK and SS had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. The views and opinions expressed within this paper are those of the authors and do not necessarily reflect those of the Department of Health. References 1 World Health Organization. Preventing suicide: a global imperative. 2014. http://www.who.int/mental_health/suicide-prevention/ world_report_2014/en/ (accessed Sept 26, 2014). 2 Carroll R, Metcalfe C, Gunnell D. Hospital presenting self-harm and risk of fatal and non-fatal repetition: systematic review and meta-analysis. PLoS One 2014; 9: e89944. 3 NICE. The long term care and treatment of self-harm. Clinical Guideline 133. 2011. 4 Bergen H, Hawton K, Waters K, et al. Premature death after self-harm: a multicentre cohort study. Lancet 2012; 380: 1568–74. 5 Carroll R, Metcalfe C, Gunnell D. Hospital management of self-harm patients and risk of repetition: systematic review and meta-analysis. J Affect Disord 2014; 168: 476–83.
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8
7
Articles
6
7
8
9
10
11
12
13
14
15
16
17 18
8
Hjorthoj CR, Madsen T, Agerbo E, Nordentoft M. Risk of suicide according to level of psychiatric treatment: a nationwide nested case-control study. Soc Psychiatry Psychiatr Epidemiol 2014; 49: 1357–65. Kapur N, Steeg S, Webb R, et al. Does clinical management improve outcomes following self-harm? Results from the Multicentre Study of Self-Harm in England. PLoS One 2013; 8: e70434. Olfson M, Marcus SC, Bridge JA. Emergency department recognition of mental disorders and short-term outcome of deliberate self-harm. Am J Psychiatry 2013; 170: 1442–50. Hunter C, Chantler K, Kapur N, Cooper J. Service user perspectives on psychosocial assessment following self-harm and its impact on further help-seeking: a qualitative study. J Affect Disord 2013; 145: 315–23. Bergen H, Hawton K, Waters K, Cooper J, Kapur N. Epidemiology and trends in non-fatal self-harm in three centres in England: 2000–2007. Br J Psychiatry 2010; 197: 493–98. Department for Communities and Local Government. The English Indices of Deprivation 2007. https://www.gov.uk/government/ collections/english-indices-of-deprivation (accessed Aug 3, 2015). Hawton K, Bergen H, Casey D, et al. Self-harm in England: a tale of three cities — multicentre study of self-harm. Soc Psychiatry Psychiatr Epidemiol 2007; 42: 513–21. Silverman MM, Berman AL, Sanddal ND, O’Carroll PW, Joiner TE Jr. Rebuilding the Tower of Babel: a revised nomenclature for the study of suicide and suicidal behaviors—part 1: background, rationale, and methodology. Suicide Life Threat Behav 2007; 37: 248–63. Silverman MM, Berman AL, Sanddal ND, O’Carroll PW, Joiner TE Jr. Rebuilding the Tower of Babel: a revised nomenclature for the study of suicide and suicidal behaviors—part 2: suiciderelated ideations, communications, and behaviors. Suicide Life Threat Behav 2007; 37: 264–77. Kapur N, Cooper J, O’Connor RC, Hawton K. Non-suicidal self-injury v. attempted suicide: new diagnosis or false dichotomy? Br J Psychiatry 2013; 202: 326–28. Kapur N, Murphy E, Cooper J, et al. Psychosocial assessment following self-harm: results from the multi-centre monitoring of self-harm project. J Affect Disord 2008; 106: 285–93. Health and Social Care Information Centre. Data Linkage Service. http://www.hscic.gov.uk/datalinkage (accessed Aug 3, 2015). Health Improvement Analytical Team. Statistical update on suicide: January 2014. http://www.gov.uk/government/uploads/system/ uploads/attachment_data/file/278120/Suicide_update_Jan_2014_ FINAL_revised.pdf (accessed June 18, 2015).
19
20
21
22 23
24
25
26 27
28
29
30
Bergen H, Hawton K, Waters K, Cooper J, Kapur N. Psychosocial assessment and repetition of self-harm: the significance of single and multiple repeat episode analyses. J Affect Disord 2010; 127: 257–65. Magruder KM, Ouyang B, Miller S, Tilley BC. Retention of under-represented minorities in drug abuse treatment studies. Clin Trials 2009; 6: 252–60. NHS England. Mental health staff to introduce physical ‘MOT’ to reduce deaths. 2014. http://www.england.nhs.uk/2014/05/15/ physical-mot/ (accessed March 23, 2014). Hunt IM, Kapur N, Webb R, et al. Suicide in recently discharged psychiatric patients: a case-control study. Psychol Med 2009; 39: 443–49. Waterhouse J, Platt S. General-hospital admission in the management of parasuicide - a randomized controlled trial. Br J Psychiatry 1990; 156: 236–42. VanderSande R, VanRooijen L, Buskens E, et al. Intensive in-patient and community intervention versus routine care after attempted suicide—a randomised controlled intervention study. Br J Psychiatry 1997; 171: 35–41. Stukel TA, Fisher ES, Wennberg DE, Alter DA, Gottlieb DJ, Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias - Effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA 2007; 297: 278–85. Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med 1997; 127: 757–63. Bergen H, Hawton K, Webb R, et al. Alcohol-related mortality following self-harm: a multicentre cohort study. J R Soc Med Open 2014; 5: 2054270414533326. Cooper J, Steeg S, Bennewith O, et al. Are hospital services for self-harm getting better? An observational study examining management, service provision and temporal trends in England. BMJ Open 2013; 3: e003444 . Keown P, Mercer G, Scott J. Retrospective analysis of hospital episode statistics, involuntary admissions under the Mental Health Act 1983, and number of psychiatric beds in England 1996–2006. BMJ 2008; 337: a1837. Large MM, Ryan CJ. Disturbing findings about the risk of suicide and psychiatric hospitals. Soc Psychiatry Psychiatr Epidemiol 2014; 49: 1353–55.
www.thelancet.com/psychiatry Published online August 6, 2015 http://dx.doi.org/10.1016/S2215-0366(15)00169-8