© 2014 John Wiley & Sons A/S Published by John Wiley & Sons Ltd.
Bipolar Disorders 2014: 16: 867–874
BIPOLAR DISORDERS
Original Article
Post-discharge suicides of inpatients with bipolar disorder in Finland Isomets€a E, Sund R, Pirkola S. Post-discharge suicides of inpatients with bipolar disorder in Finland. Bipolar Disord 2014: 16: 867–874. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. Objectives: Suicide risk in psychiatric inpatients is known to be remarkably high after discharge. However, temporal patterns and risk factors among patients with bipolar disorder remain obscure. We investigated post-discharge temporal patterns of hazard and risk factors by type of illness phase among patients with bipolar disorder. Methods: Based on national registers, all discharges of patients with bipolar disorder from a psychiatric ward in Finland in 1987–2003 (n = 52,747) were identified, and each patient was followed up to postindex discharge or to suicide (n = 466). For discharges occurring in 1995–2003 (n = 35,946), factors modifying hazard of suicide during the first 120 days (n = 129) were investigated. Results: The temporal pattern of suicide risk depended on the type of illness phase, being highest but steeply declining after discharge with depression; less high and declining in mixed states; lower and relatively stable after mania. In Cox models, for post-discharge suicides (n = 65) after hospitalizations for bipolar depression (n = 9,635), the hazard ratio was 8.05 (p = 0.001) after hospitalization with a suicide attempt and 3.63 (p < 0.001) for male patients, but 0.186 (p = 0.001) for patients taking lithium. Suicides after mania (n = 28) or mixed episodes (n = 20) were predicted by male sex and preceding suicide attempts, respectively. Conclusions: Among inpatients with bipolar disorder, suicide risk is high and related strongly to the time elapsed from discharge after hospitalizations for depressive episodes, and less strongly after hospitalizations for mixed episodes. Intra-episodic suicide attempts and male sex powerfully predict suicide risk. Lower suicide rate after hospitalizations for depression among patients prescribed lithium is consistent with a preventive effect.
Suicide risk in patients with bipolar disorder is among the highest in all psychiatric patients. Based on data from Danish national psychiatric registers, it is estimated that about 8% of male patients and 4% of female patients diagnosed with bipolar disorder die by suicide (1). Risk of suicide is about 20-fold that of the general population (2–4). However, it likely varies markedly according to patients’ clinical state and the presence of other risk factors (5–7). Suicide risk in all psychiatric inpatients is known to be high immediately postdischarge, making this period an important target for preventive efforts (8, 9). This risk is also very
Erkki Isometsa€a,b,c, Reijo Sundd,e and Sami Pirkolaa,c a
Department of Psychiatry, University of Helsinki, Mood, Depression, and Suicidal Behavior Unit, National Institute for Health and Welfare, c Department of Psychiatry, Helsinki University Central Hospital (HUCH), dService Systems Research Unit, National Institute for Health and Welfare, eFaculty of Social Sciences, University of Helsinki, Helsinki, Finland b
doi: 10.1111/bdi.12237 Key words: attempted suicide – bipolar disorder – incidence – lithium – patient discharge – pharmacotherapy – risk factors – suicide Received 15 January 2014, revised and accepted for publication 9 May 2014 Corresponding author: Erkki T. Isometsa€, M.D., Ph.D. Department of Psychiatry Institute of Clinical Medicine University of Helsinki P.O. Box 22 Helsinki 00014 Finland Fax: +358-9-47163735 E-mail: [email protected]
high in patients with affective disorders overall (8–10), but the temporal patterns of post-discharge suicide risk and the factors modifying risk specifically among bipolar patients are poorly known. Suicidal acts by bipolar patients are most likely to take place during depressive or mixed episodes, and rarely during mania or hypomania or remission (11–13), and may be prevented by long-term lithium treatment more effectively than other pharmacotherapies (14–17). Thus, the level and temporal patterns of post-discharge suicide risk and the factors modifying them can be expected to differ markedly depending on the type of illness episode
867
Isomets€ a et al. leading to hospitalization, and possibly to be modified by type of pharmacotherapy. However, pertinent studies are scarce. The aim of this observational national study was to investigate temporal patterns of hazard for suicide according to type of illness phase, and factors modifying suicide risk after discharge from a psychiatric ward among patients with a diagnosis of bipolar disorder in Finland. Patients and methods
Data for this study were obtained from the research database collected for the MERTTU study project (18). This research database contains data for all patients identified from the Finnish Hospital Discharge Register whose principal discharge diagnosis was a mental disorder (corresponding to the current F-diagnoses in ICD-10) during 1980–2003. The accuracy of mental disorder diagnoses in the register is known to be good (19–22). Using the Finnish personal identity codes, data on sociodemographic variables from the census register of Statistics Finland were linked to the database. Data on reimbursements for purchases of psychopharmacological agents were obtained from the registers of the National Social Insurance Institution (available only for the years 1995–2003). Causes of deaths were obtained from the register maintained by Statistics Finland. Study design
We analyzed hospitalizations in a psychiatric hospital or a psychiatric ward within a general hospital during 1987–2003 for patients whose discharge diagnoses included bipolar disorder. The 1980–1986 ICD-8 era was not included due to incompatibility of the bipolar disorder diagnosis with the ICD-9 or ICD-10. In Finland the DSM-III-R criteria were used as part of the ICD-9, and the ICD-10 Finnish version includes the Diagnostic Criteria for Research (DCR); thus, operational criteria for psychiatric diagnosis have been available since 1987. The outcome of interest was suicide. More specifically, a patient was considered to have been diagnosed with bipolar disorder at the first discharge with a diagnosis of bipolar disorder (ICD-10: F31) or manic (F30) or mixed affective (F38.00) episode; this first, and all subsequent psychiatric hospitalizations during the study period were included, even if the diagnosis later was something other than bipolar disorder. Subsequent depressive (F33) episodes were considered bipolar, but patients’ hospitalizations for depression before the first bipolar diagnosis were
868
excluded from the analyses since event of death was logically impossible before the first bipolar diagnosis, and inclusion would therefore have introduced a bias. Suicides (X60-X84, Y870) of the population were identified from the causes of death data. The unit of analysis in this study was a discharge from a hospital, not a patient. Most patients had multiple discharges after hospitalizations for different types of illness. The hospitalizations were classified into four categories according the type of illness phase implicated by the clinical diagnosis, i.e., those after (i) mania, (ii) mixed episode, (iii) depression, or (iv) other. For each discharge, the related length of stay at the hospital and status of hospitalization [involuntary admission or admission because of attempted suicide (intra-episodic suicide attempt)] were determined. Hospitalizations were either voluntary or involuntary, the latter being based on the Finnish Mental Health Act. In Finland, a person can be ordered to undergo treatment in a psychiatric hospital against his or her will only: (i) if the person is diagnosed as mentally ill; (ii) if the person needs treatment for a mental illness which, if not treated, would become considerably worse or severely endanger the person’s health or safety or the health or safety of others; and (iii) if all other mental health services are inapplicable or inadequate. Sequentially three independent medical doctors have to agree about these conditions being met. The data contained additional information (pharmacotherapy reimbursements and extra fields in the hospital discharge records) from 1995 onwards, allowing us to determine the Global Assessment Scale (GAS) score at discharge for each hospitalization as well as purchases of lithium (ATC: N05AN), valproate (ATC: N03AG01), or antidepressive agents (ATC: N06A) before and after the hospitalization. If there were purchases during the three months preceding the admission to hospital, the patient was considered to be on medication. Medication during the hospitalization was known only for antidepressants. The medication status after discharge was somewhat challenging to determine using the purchase data because a patient could die or have a new hospitalization without purchases of prescribed medication. Particularly, if a patient committed suicide before purchasing prescribed medication, the preventive effect of pharmacotherapy would be overestimated because the medication status based on purchases would be negative even though the patient may have been on medication during the hospitalization and had a prescription for the post-hospital period. To avoid a bias
BD post-discharge suicides in Finland from not-on-medication to on-medication status on the day of purchase was used. As there were typically several discharges for each patient, an assumption of independent observations was relaxed using the clustered sandwich estimators for standard errors. The proportional hazards assumption was tested on the basis of Schoenfeld residuals. Model fit was assessed using the Cox–Snell residuals and the May–Hosmer test statistic. All models had adequate fit to the data (depression: p = 0.73; mania: p = 0.43, mixed: p = 0.67). A p-value below 0.05 was considered statistically significant. Software packages Survo MM (www. survo.fi), R v2.8 with the gss package (www. r-project.org), and Stata v10 (www.stata.com) were used for data processing and statistical analyses.
favoring pharmacotherapy, we assumed that the patient was on medication directly after discharge if (i) the patient had purchases within two months of discharge, or (ii) the patient was on medication before the hospitalization (or during the hospitalization for antidepressants) and died within two weeks without purchases. Statistical analysis
The crude incidence of suicide was calculated by dividing the number of events by the corresponding follow-up time in person-years. The survival curves were derived using the Kaplan–Meier estimator. Follow-up was from the day of discharge to the event of suicide, death, next psychiatric hospitalization, or 31 December 2003, whichever occurred first. The follow-up was considered censored for events other than suicide. To determine the probability of committing suicide after discharge from hospital on a certain day, smoothed hazard functions for post-discharge suicides and corresponding Bayesian confidence intervals were estimated using the penalized likelihood method (23). Cox’s proportional hazards model was used to determine the adjusted effects of observed variables potentially modifying the hazard of suicide after hospitalization for manic, depressive, or mixed episodes. Since the clinical course of bipolar disorder is episodic and pleomorphic, the validity of information on type of illness phase and clinical state at discharge declines progressively over time. These analyses focused only on the first 120 days after discharge, i.e., follow-up was considered censored after four months. About two-thirds of suicides during the first post-discharge year occur during this high-risk period. To control for possible bias in estimating the effect of post-discharge medication, a binary time-dependent covariate changing
1
Results Incidence and temporal patterns of post-discharge suicides in Finland in 1987–2003
The temporal patterns of hazard of suicide in patients with bipolar disorder in Finland in 1987–2003 are depicted in Figure 1. Overall, there were 466 suicides after 52,747 psychiatric hospital discharges in 13,581 patients (Table 1). The median number of hospitalizations for patients was two [mean 3.9, standard deviation (SD) 5.6]; 2,298 patients (16.9%) had hospitalizations only for depression. The hazard depended strongly on type of illness phase, being highest but steeply declining after discharge with depression, also high but less declining in mixed states, but lower and relatively stable after mania (Fig. 2). The incidence rate of completed suicide per 105 patient-years was 492 [95% confidence interval (CI): 449–539] during the post-discharge follow-up
Survival
0.995
Other Manic Mixed
0.99
0.985 Depression
0.98 0
100
200
300
Days since discharge
Fig. 1. Kaplan–Meier curves for post-discharge suicide mortality in bipolar disorder in Finland in 1987–2003 by type of illness episode.
869
Isomets€ a et al. Table 1. Psychiatric admissions of patients with bipolar disorder during 1987–2003 (used in estimation of hazard functions)
No of discharges No. of suicides No. of discharges in women Intra-episodic suicide attempts Involuntary admissions Admission length, days 0–13 14–27 28–41 42–120 120+
Manic n (%)
Mixed n (%)
Depressive n (%)
Other n (%)
16,403 115 (0.7) 8,140 (49.6) 55 (0.3) 7,848 (47.8)
9,891 80 (0.8) 5,624 (56.9) 51 (0.5) 1,964 (19.9)
14,233 213 (1.5) 8,375 (58.8) 110 (0.8) 1,632 (11.5)
12,220 58 (0.5) 6,773 (55.4) 24 (0.2) 3,390 (27.7)
3,957 (24.1) 3,919 (23.9) 3,263 (19.9) 4,701 (28.7) 563 (3.4)
4,376 (44.2) 2,077 (21.0) 1,257 (12.7) 1,721 (17.4) 460 (4.7)
5,109 (35.9) 3,428 (24.1) 2,033 (14.3) 3,018 (21.2) 645 (4.5)
5,368 (43.9) 2,203 (18.0) 1,336 (10.9) 2,482 (20.3) 831 (6.8)
period, and twofold for male patients [694 (618– 778)] compared with female patients [332 (286– 385)]. Overall, incidence rates declined over time after discharge following hospitalization, with most marked differences in rates after hospitalization for depression (Fig. 2). Hazard predictors of suicide during the first 120 days after discharge in Finland in 1995–2003
The information on suicides during the first 120 days after discharge in 1995–2003 is preA
sented in Table 2. In Cox models for suicides (n = 65) after hospitalizations (n = 9,635) for bipolar depression, the hazard ratio (HR) was 8.05 (95% CI: 2.49–26.04; p < 0.001) after hospitalization with an intra-episodic suicide attempt and 3.63 (95% CI: 2.12–6.23; p < 0.001) for male patients, but 0.186 (95% CI: 0.07–0.52; p = 0.001) for patients taking lithium. Valproate or antidepressants without concurrent mood stabilizer did not significantly influence the hazard. Suicides after mania (28 of 10,730 discharges) were predicted by male sex (HR = 4.00, 95% CI: B
0.00015
Hazard 0.00015
0.0001
Hazard
0.0001
0.00005
0.00005
0 100
200
0
300
100
Days since discharge
200
300
Days since discharge
C
D
0.00015
Hazard
Hazard 0.00015
0.0001
0.0001
0.00005
0.00005
0
0 100
200 Days since discharge
300
100
200
300
Days since discharge
Fig. 2. Survival function after discharge according to type of episode at hospitalization. (A) Manic episode. (B) Depressive episode. (C) Mixed episode. (D) Other episode.
870
BD post-discharge suicides in Finland Table 2. Psychiatric admissions of patients with bipolar disorder during 1995–2003 (used in estimation of Cox models)
No. of discharges Suicides (120 days) No. of discharges in women Intra-episodic suicide attempts Involuntary admissions Admission length, days 0–13 14–27 28–41 42–120 120+ Lithium (after discharge) Valproate (after discharge) Only antidepressant (after discharge)
Manic n (%)
Mixed n (%)
Depressive n (%)
Other n (%)
10,730 28 (0.3) 5,306 (49.4) 7 (0.1) 4,957 (46.2)
6,911 20 (0.3) 3,894 (56.3) 25 (0.4) 1,076 (15.6)
9,635 65 (0.7) 5,544 (57.5) 58 (0.6) 860 (8.9)
8,670 16 (0.2) 4,776 (55.1) 6 (0.1) 2,112 (24.4)
2,743 (25.6) 2,503 (23.3) 2,078 (19.4) 3,020 (28.1) 387 (3.6) 3,919 (36.5) 3,402 (31.7) 820 (7.6)
3,561 (51.5) 1,387 (20.1) 753 (10.9) 991 (14.3) 219 (3.2) 1,813 (26.2) 2,502 (36.2) 1,082 (15.7)
3,914 (40.6) 2,241 (23.3) 1,281 (13.3) 1,857 (19.3) 342 (3.5) 2,089 (21.7) 2,396 (24.9) 3,609 (37.5)
4,193 (48.4) 1,472 (17.0) 825 (9.5) 1,626 (18.8) 554 (6.4) 1,531 (17.7) 1,572 (18.1) 1,679 (19.4)
1.61–9.87; p = 0.003), and those after mixed episodes (20 of 6,911 discharges) by intra-episodic suicide attempts (HR = 11.1, 95% CI: 1.41–87.8; p = 0.02). Discussion
This study included all psychiatric hospitalizations with a clinical diagnosis of bipolar disorder in Finland in 1987–2003, and all suicides within one year of discharge were included. Specific risk factors could be reliably investigated among post-discharge suicides occurring in 1995–2003. As hypothesized, we found marked differences in temporal patterns of suicide risk depending on type of illness episode. Furthermore, an intra-episodic suicide attempt raised the hazard by up to 12-fold, and male sex in most analyses by about threefold. After hospitalization for depressive episodes, hazard for suicide was only one-seventh among those taking lithium, but not significantly lower with valproate. While this finding is consistent with many previous studies suggesting lithium to have a specific preventive effect, alternative explanations must also be considered. Our study had some major strengths. It was a national population study comprising all suicides within the 17-year period under investigation. The unit of analysis was discharge from a psychiatric hospital of a patient with a clinical diagnosis of bipolar disorder, and the hospitalizations were drawn from the Finnish National Hospital Discharge register. This information was linked to other registers, including Causes of Death obtained from Statistics Finland and reimbursement registers from the National Insurance Institution. The quality of these registers is generally considered excellent (19–22), and major inaccuracies are
unlikely. Given this setting, we were able to investigate suicides by type of illness episode, which due to the relatively low incidence of completed suicides necessitates large populations. We were also able to link information on pharmacotherapy and some other background variables. The study also had obvious limitations, and some methodological issues should be addressed. First, the study included only hospitalizations of patients with a clinical diagnosis of bipolar disorder. Bipolar disorder is known to have been significantly underdiagnosed in Finland during the study period (24, 25), with those with the most obvious classical clinical presentation of bipolar I disorder most likely correctly diagnosed, but no more than, at best, about one-half of patients with bipolar II disorder. Secondly, no information was available concerning the mental state of the subject at the time of the fatal act. While the study provides unbiased estimates of hazard after a hospitalization for a particular type of illness episode, a relapse of the same or another polarity may well have taken place, predisposing to suicide (11–13). Thirdly, only a limited range of all risk factors could be investigated. Specifically, information on preceding suicide attempts, the most robust indicator of elevated risk (1, 5–7), was limited to only those during the same hospitalization episode. Fourthly, it is important to note that the role of pharmacotherapy is evaluated based on the best available information on purchases of prescriptions before and after hospitalization. In order to purchase, one must be alive, a fact that may bias hazard estimates with regard to pharmacotherapy overall, but is unlikely to explain differences in patterns between classes of agents. There is no guarantee that the subject purchasing a prescription actually used it. We probably overestimated
871
Isomets€ a et al. the use of pharmacotherapy, likely resulting in conservative estimates of preventive effects. Fifthly, despite a massive national database covering 17 years, the number of completed suicides in some Cox model subgroup analyses remained low, rendering the study vulnerable to type II errors. Sixthly, one-sixth of the patients had hospitalizations exclusively for depressions, which is a rough proxy estimate for the proportion of diagnosed patients with bipolar II disorder. Thus, the vast majority of patients clinically diagnosed with bipolar disorder likely had bipolar I disorder. Seventhly, the study covered the years 1987–2003. Whether changes over time increasing the recognition and diagnosis of bipolar disorder, shortening the duration of hospitalizations (9), or increasing the availability of new pharmacotherapy options for the illness influence the patterns of suicide risk remains unknown. Finally, the generalizability of findings from Finland to other countries is uncertain, despite considerable similarity of epidemiological and register-based findings with regard to mental disorders. Finland has high suicide rates, being most comparable to those of other Nordic or northern European countries (26). Our analyses revealed the strong association of type of illness episode at hospitalization with subsequent risk for post-discharge suicide. While the incidence of suicide attempts is known to be highest in mixed episodes (13, 27), and a significant proportion of victims in a psychological autopsy study of bipolar patients were in a mixed episode at the time of suicide (11), post-discharge suicide risk was not highest among those hospitalized for a mixed episode. Suicide risk immediately after discharge was by far the highest among hospitalizations for bipolar depression. This subgroup comprised half of the post-discharge suicides and also most strongly exhibited the typical pattern of steep decline in risk during the first three months after discharge. The incidence of suicide was lower and temporal patterns unimpressive among those hospitalized for mania or other episodes. We can only speculate about the reasons for these partly unexpected patterns. A plausible explanation could be depressive relapse after discharge, which is the most common adverse outcome. In contrast, mixed episodes, when correctly diagnosed while hospitalized, may be more effectively treated with antimanic pharmacotherapy. Patients hospitalized for manic or mixed episodes generally receive adequate pharmacotherapy, whereas quality of treatment for bipolar depression may often be poor (28) and the available treatments leave room for improvement. The duration of mixed episodes is shorter than that of depressive episodes;
872
therefore, accumulated total risk for suicide attempts is higher for depressions, despite the highest incidence occurring in mixed episodes (13, 27). For purposes of suicide prevention, the implications are clear. In the post-discharge period for bipolar patients, those hospitalized for depression comprise the group with the highest immediate risk (in the first three months) for suicide. We also found robust evidence for the association of two risk factors with completed suicide during the post-discharge period: preceding suicide attempts and male gender. Both findings are unsurprising given that they are established risk factors for suicide overall and in bipolar disorder specifically (1, 5–7). However, the potency of these risk factors is marked within the high-risk subgroup of hospitalizations for depression. Male patients had an almost fourfold hazard of suicide during the first three months, so particular attention should be paid to preventive efforts in men. Further research on post-discharge suicides should investigate risk factors that may be particularly relevant for male patients, but remained uncovered in this study, such as substance abuse (5–7, 11) and availability of lethal methods (29). The hazard estimate for a preceding suicide attempt in those hospitalized for depression was even higher, at eightfold. Unlike most previous studies that have investigated preceding attempts during the patient’s lifetime, we investigated suicide attempts immediately preceding or during the same episode of hospitalization. It appears that these temporally close attempts may be even stronger indicators of shortterm risk than lifetime attempts. This pattern is consistent with the known clustering of preceding attempts close to eventual suicide among suicides overall (30). However, despite indicating an exceptionally high relative risk when present, the sensitivity of an intra-episodic suicide attempt as a risk indicator is low (4.2%). Due to the relative rarity of completed suicide even during this high-risk period, the positive predictive value of an intra-episodic suicide attempt was 3.7%. We investigated whether the hazard of suicide is related to purchasing lithium, valproate, or antidepressants preceding or after hospitalization. Of the three types of pharmacotherapy, only lithium seemed to modify risk. Purchasing lithium was associated with a markedly lower risk of suicide. This finding is fully concordant with other registerbased studies comparing suicide risk among individuals when using different pharmacotherapies (14), albeit not with a recent major randomized trial (31). Moreover, our estimate for effect size (HR 0.19) of reducing risk among patients hospitalized for depression is highly consistent
BD post-discharge suicides in Finland with that of two meta-analyses; Baldessarini et al. (15) reported an odds ratio of 0.20 and the recent meta-analysis of randomized evidence by Cipriani et al. (17) reported 0.13. Thus, our findings add to the literature on a preventive effect of lithium, suggesting that this effect also concerns the postdischarge period when hospitalized for depression. However, it is important to keep in mind the major methodological limitations of our study. We had no information on adherence to treatment or quality of treatment provided, both of which are known to be compromised in suicides occurring during lithium treatment (32). We only had data on purchases of lithium, from which we made assumptions of actual use. The characteristics of patients prescribed each of the medications may differ markedly, and it is possible that patients evaluated to be at risk of suicide are not prescribed lithium because of known toxicity in overdose. Therefore, many kinds of confounding factors may influence the findings. Obviously, despite adjusting for known differences in the models, a possibility of bias exists. Nevertheless, such factors did not result in our finding any preventive effect for valproate or antidepressants. It is noteworthy that we found no evidence for the latter two being associated with an elevated hazard either, despite both possibilities having been debated in the literature (33). Overall, the finding of a possible preventive effect for lithium is consistent with results of most randomized studies (17). Conclusions
Among inpatients with bipolar disorder, the risk of completed suicide is high and related strongly to time elapsed from discharge after hospitalizations for depressive episodes, and less strongly after hospitalizations for mixed episodes. Preceding intra-episodic suicide attempts and male sex are powerful predictors of suicide risk. Lower suicide rate among patients taking lithium is consistent with a preventive effect. Acknowledgements Funding was provided by the National Institute for Health and Welfare, Helsinki, Finland. At the time of data collection, the MERTTU project was supported by a grant from the Academy of Finland.
Disclosures The authors of this paper do not have any commercial associations that might pose a conflict of interest in connection with this manuscript.
References 1. Nordentoft M, Mortensen PB, Pedersen CB. Absolute risk of suicide after first hospital contact in mental disorder. Arch Gen Psychiatry 2011; 68: 1058–1064. 2. Tondo L, Isacsson G, Baldessarini R. Suicidal behaviour in bipolar disorder: risk and prevention. CNS Drugs 2003; 17: 491–511. € 3. Osby U, Brandt L, Correia N, Ekbom A, Sparen P. Excess mortality in bipolar and unipolar disorder in Sweden. Arch Gen Psychiatry 2001; 58: 844–850. 4. Angst J, Angst F, Gerber-Werder R, Gamma A. Suicide in 406 mood-disorder patients with and without long-term medication: a 40 to 44 years’ follow-up. Arch Suicide Res 2005; 9: 279–300. 5. Hawton K, Hawton K, Sutton L, Haw C, Sinclair J, Harriss L. Suicide and attempted suicide in bipolar disorder: a systematic review of risk factors. J Clin Psychiatry 2005; 66: 693–704. 6. Pompili M, Gonda X, Serafini G et al. Epidemiology of suicide in bipolar disorders: a systematic review of the literature. Bipolar Disord 2013; 15: 457–490. 7. Isometsa E. Suicidal behavior in mood disorders – who, when and why? Can J Psychiatry 2014; 59: 120–130. 8. Qin P, Nordentoft M. Suicide risk in relation to psychiatric hospitalization: evidence based on longitudinal registers. Arch Gen Psychiatry 2005; 62: 427–432. 9. Pirkola S, Sohlman B, Heil€ a H, Wahlbeck K. Reductions in post-discharge suicide after psychiatric deinstitutionalization – a nationwide register study. Psychiatr Serv 2007; 52: 221–226. 10. Høyer EH, Licht RW, Mortensen PB. Risk factors of suicide in inpatients and recently discharged patients with affective disorders. A case–control study. Eur Psychiatry 2009; 24: 317–321. 11. Isomets€ a ET, Henriksson MM, Aro HM, L€ onnqvist JK. Suicide in bipolar disorder in Finland. Am J Psychiatry 1994; 151: 1020–1024. 12. Valtonen H, Suominen K, Mantere O, Lepp€ am€ aki S, Arvilommi P, Isomets€ a E. Suicidal behaviour during different phases of bipolar disorder. J Affect Disord 2007; 97: 101–107. 13. Valtonen HM, Suominen K, Haukka J. Differences in incidence of suicide attempts during phases of bipolar I and II disorders. Bipolar Disord 2008; 10: 588–596. 14. Goodwin FK, Fireman B, Simon GE, Hunkeler EM, Lee J, Revicki D. Suicide risk in bipolar disorder during treatment with lithium and divalproex. JAMA 2003; 290: 1467–1473. 15. Baldessarini RJ, Tondo L, Davis P, Pompili M, Goodwin FK, Hennen J. Decreased risk of suicides and attempts during long-term lithium treatment: a meta-analytic review. Bipolar Disord 2006; 8: 625–639. 16. Kessing LV, Sønderg ard L, Kvist K, Andersen PK. Suicide risk in patients treated with lithium. Arch Gen Psychiatry 2005; 62: 860–866. 17. Cipriani A, Hawton K, Stockton S, Geddes JR. Lithium in the prevention of suicide in mood disorders: updated systematic review and meta-analysis. BMJ 2013; 346: f3646. 18. Pirkola S, Sund R, Sailas E, Wahlbeck K. Community mental-health services and suicide rate in Finland: a nationwide small-area analysis. Lancet 2009; 373: 147–153. 19. Kiesepp€ a T, Partonen T, Kaprio J, L€ onnqvist J. Accuracy of register- and record-based bipolar I disorder diagnoses in Finland: a study of twins. Acta Neuropsychiatr 2000; 12: 106–109.
873
© 2014 John Wiley & Sons A/S Published by John Wiley & Sons Ltd.
Bipolar Disorders 2014: 16: 867–874
BIPOLAR DISORDERS
Original Article
Post-discharge suicides of inpatients with bipolar disorder in Finland Isomets€a E, Sund R, Pirkola S. Post-discharge suicides of inpatients with bipolar disorder in Finland. Bipolar Disord 2014: 16: 867–874. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. Objectives: Suicide risk in psychiatric inpatients is known to be remarkably high after discharge. However, temporal patterns and risk factors among patients with bipolar disorder remain obscure. We investigated post-discharge temporal patterns of hazard and risk factors by type of illness phase among patients with bipolar disorder. Methods: Based on national registers, all discharges of patients with bipolar disorder from a psychiatric ward in Finland in 1987–2003 (n = 52,747) were identified, and each patient was followed up to postindex discharge or to suicide (n = 466). For discharges occurring in 1995–2003 (n = 35,946), factors modifying hazard of suicide during the first 120 days (n = 129) were investigated. Results: The temporal pattern of suicide risk depended on the type of illness phase, being highest but steeply declining after discharge with depression; less high and declining in mixed states; lower and relatively stable after mania. In Cox models, for post-discharge suicides (n = 65) after hospitalizations for bipolar depression (n = 9,635), the hazard ratio was 8.05 (p = 0.001) after hospitalization with a suicide attempt and 3.63 (p < 0.001) for male patients, but 0.186 (p = 0.001) for patients taking lithium. Suicides after mania (n = 28) or mixed episodes (n = 20) were predicted by male sex and preceding suicide attempts, respectively. Conclusions: Among inpatients with bipolar disorder, suicide risk is high and related strongly to the time elapsed from discharge after hospitalizations for depressive episodes, and less strongly after hospitalizations for mixed episodes. Intra-episodic suicide attempts and male sex powerfully predict suicide risk. Lower suicide rate after hospitalizations for depression among patients prescribed lithium is consistent with a preventive effect.
Suicide risk in patients with bipolar disorder is among the highest in all psychiatric patients. Based on data from Danish national psychiatric registers, it is estimated that about 8% of male patients and 4% of female patients diagnosed with bipolar disorder die by suicide (1). Risk of suicide is about 20-fold that of the general population (2–4). However, it likely varies markedly according to patients’ clinical state and the presence of other risk factors (5–7). Suicide risk in all psychiatric inpatients is known to be high immediately postdischarge, making this period an important target for preventive efforts (8, 9). This risk is also very
Erkki Isometsa€a,b,c, Reijo Sundd,e and Sami Pirkolaa,c a
Department of Psychiatry, University of Helsinki, Mood, Depression, and Suicidal Behavior Unit, National Institute for Health and Welfare, c Department of Psychiatry, Helsinki University Central Hospital (HUCH), dService Systems Research Unit, National Institute for Health and Welfare, eFaculty of Social Sciences, University of Helsinki, Helsinki, Finland b
doi: 10.1111/bdi.12237 Key words: attempted suicide – bipolar disorder – incidence – lithium – patient discharge – pharmacotherapy – risk factors – suicide Received 15 January 2014, revised and accepted for publication 9 May 2014 Corresponding author: Erkki T. Isometsa€, M.D., Ph.D. Department of Psychiatry Institute of Clinical Medicine University of Helsinki P.O. Box 22 Helsinki 00014 Finland Fax: +358-9-47163735 E-mail: [email protected]
high in patients with affective disorders overall (8–10), but the temporal patterns of post-discharge suicide risk and the factors modifying risk specifically among bipolar patients are poorly known. Suicidal acts by bipolar patients are most likely to take place during depressive or mixed episodes, and rarely during mania or hypomania or remission (11–13), and may be prevented by long-term lithium treatment more effectively than other pharmacotherapies (14–17). Thus, the level and temporal patterns of post-discharge suicide risk and the factors modifying them can be expected to differ markedly depending on the type of illness episode
867
Bipolar Disorders 2014: 16: 867–874
BIPOLAR DISORDERS
Original Article
Post-discharge suicides of inpatients with bipolar disorder in Finland Isomets€a E, Sund R, Pirkola S. Post-discharge suicides of inpatients with bipolar disorder in Finland. Bipolar Disord 2014: 16: 867–874. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. Objectives: Suicide risk in psychiatric inpatients is known to be remarkably high after discharge. However, temporal patterns and risk factors among patients with bipolar disorder remain obscure. We investigated post-discharge temporal patterns of hazard and risk factors by type of illness phase among patients with bipolar disorder. Methods: Based on national registers, all discharges of patients with bipolar disorder from a psychiatric ward in Finland in 1987–2003 (n = 52,747) were identified, and each patient was followed up to postindex discharge or to suicide (n = 466). For discharges occurring in 1995–2003 (n = 35,946), factors modifying hazard of suicide during the first 120 days (n = 129) were investigated. Results: The temporal pattern of suicide risk depended on the type of illness phase, being highest but steeply declining after discharge with depression; less high and declining in mixed states; lower and relatively stable after mania. In Cox models, for post-discharge suicides (n = 65) after hospitalizations for bipolar depression (n = 9,635), the hazard ratio was 8.05 (p = 0.001) after hospitalization with a suicide attempt and 3.63 (p < 0.001) for male patients, but 0.186 (p = 0.001) for patients taking lithium. Suicides after mania (n = 28) or mixed episodes (n = 20) were predicted by male sex and preceding suicide attempts, respectively. Conclusions: Among inpatients with bipolar disorder, suicide risk is high and related strongly to the time elapsed from discharge after hospitalizations for depressive episodes, and less strongly after hospitalizations for mixed episodes. Intra-episodic suicide attempts and male sex powerfully predict suicide risk. Lower suicide rate after hospitalizations for depression among patients prescribed lithium is consistent with a preventive effect.
Suicide risk in patients with bipolar disorder is among the highest in all psychiatric patients. Based on data from Danish national psychiatric registers, it is estimated that about 8% of male patients and 4% of female patients diagnosed with bipolar disorder die by suicide (1). Risk of suicide is about 20-fold that of the general population (2–4). However, it likely varies markedly according to patients’ clinical state and the presence of other risk factors (5–7). Suicide risk in all psychiatric inpatients is known to be high immediately postdischarge, making this period an important target for preventive efforts (8, 9). This risk is also very
Erkki Isometsa€a,b,c, Reijo Sundd,e and Sami Pirkolaa,c a
Department of Psychiatry, University of Helsinki, Mood, Depression, and Suicidal Behavior Unit, National Institute for Health and Welfare, c Department of Psychiatry, Helsinki University Central Hospital (HUCH), dService Systems Research Unit, National Institute for Health and Welfare, eFaculty of Social Sciences, University of Helsinki, Helsinki, Finland b
doi: 10.1111/bdi.12237 Key words: attempted suicide – bipolar disorder – incidence – lithium – patient discharge – pharmacotherapy – risk factors – suicide Received 15 January 2014, revised and accepted for publication 9 May 2014 Corresponding author: Erkki T. Isometsa€, M.D., Ph.D. Department of Psychiatry Institute of Clinical Medicine University of Helsinki P.O. Box 22 Helsinki 00014 Finland Fax: +358-9-47163735 E-mail: [email protected]
high in patients with affective disorders overall (8–10), but the temporal patterns of post-discharge suicide risk and the factors modifying risk specifically among bipolar patients are poorly known. Suicidal acts by bipolar patients are most likely to take place during depressive or mixed episodes, and rarely during mania or hypomania or remission (11–13), and may be prevented by long-term lithium treatment more effectively than other pharmacotherapies (14–17). Thus, the level and temporal patterns of post-discharge suicide risk and the factors modifying them can be expected to differ markedly depending on the type of illness episode
867
Isomets€ a et al. leading to hospitalization, and possibly to be modified by type of pharmacotherapy. However, pertinent studies are scarce. The aim of this observational national study was to investigate temporal patterns of hazard for suicide according to type of illness phase, and factors modifying suicide risk after discharge from a psychiatric ward among patients with a diagnosis of bipolar disorder in Finland. Patients and methods
Data for this study were obtained from the research database collected for the MERTTU study project (18). This research database contains data for all patients identified from the Finnish Hospital Discharge Register whose principal discharge diagnosis was a mental disorder (corresponding to the current F-diagnoses in ICD-10) during 1980–2003. The accuracy of mental disorder diagnoses in the register is known to be good (19–22). Using the Finnish personal identity codes, data on sociodemographic variables from the census register of Statistics Finland were linked to the database. Data on reimbursements for purchases of psychopharmacological agents were obtained from the registers of the National Social Insurance Institution (available only for the years 1995–2003). Causes of deaths were obtained from the register maintained by Statistics Finland. Study design
We analyzed hospitalizations in a psychiatric hospital or a psychiatric ward within a general hospital during 1987–2003 for patients whose discharge diagnoses included bipolar disorder. The 1980–1986 ICD-8 era was not included due to incompatibility of the bipolar disorder diagnosis with the ICD-9 or ICD-10. In Finland the DSM-III-R criteria were used as part of the ICD-9, and the ICD-10 Finnish version includes the Diagnostic Criteria for Research (DCR); thus, operational criteria for psychiatric diagnosis have been available since 1987. The outcome of interest was suicide. More specifically, a patient was considered to have been diagnosed with bipolar disorder at the first discharge with a diagnosis of bipolar disorder (ICD-10: F31) or manic (F30) or mixed affective (F38.00) episode; this first, and all subsequent psychiatric hospitalizations during the study period were included, even if the diagnosis later was something other than bipolar disorder. Subsequent depressive (F33) episodes were considered bipolar, but patients’ hospitalizations for depression before the first bipolar diagnosis were
868
excluded from the analyses since event of death was logically impossible before the first bipolar diagnosis, and inclusion would therefore have introduced a bias. Suicides (X60-X84, Y870) of the population were identified from the causes of death data. The unit of analysis in this study was a discharge from a hospital, not a patient. Most patients had multiple discharges after hospitalizations for different types of illness. The hospitalizations were classified into four categories according the type of illness phase implicated by the clinical diagnosis, i.e., those after (i) mania, (ii) mixed episode, (iii) depression, or (iv) other. For each discharge, the related length of stay at the hospital and status of hospitalization [involuntary admission or admission because of attempted suicide (intra-episodic suicide attempt)] were determined. Hospitalizations were either voluntary or involuntary, the latter being based on the Finnish Mental Health Act. In Finland, a person can be ordered to undergo treatment in a psychiatric hospital against his or her will only: (i) if the person is diagnosed as mentally ill; (ii) if the person needs treatment for a mental illness which, if not treated, would become considerably worse or severely endanger the person’s health or safety or the health or safety of others; and (iii) if all other mental health services are inapplicable or inadequate. Sequentially three independent medical doctors have to agree about these conditions being met. The data contained additional information (pharmacotherapy reimbursements and extra fields in the hospital discharge records) from 1995 onwards, allowing us to determine the Global Assessment Scale (GAS) score at discharge for each hospitalization as well as purchases of lithium (ATC: N05AN), valproate (ATC: N03AG01), or antidepressive agents (ATC: N06A) before and after the hospitalization. If there were purchases during the three months preceding the admission to hospital, the patient was considered to be on medication. Medication during the hospitalization was known only for antidepressants. The medication status after discharge was somewhat challenging to determine using the purchase data because a patient could die or have a new hospitalization without purchases of prescribed medication. Particularly, if a patient committed suicide before purchasing prescribed medication, the preventive effect of pharmacotherapy would be overestimated because the medication status based on purchases would be negative even though the patient may have been on medication during the hospitalization and had a prescription for the post-hospital period. To avoid a bias
BD post-discharge suicides in Finland from not-on-medication to on-medication status on the day of purchase was used. As there were typically several discharges for each patient, an assumption of independent observations was relaxed using the clustered sandwich estimators for standard errors. The proportional hazards assumption was tested on the basis of Schoenfeld residuals. Model fit was assessed using the Cox–Snell residuals and the May–Hosmer test statistic. All models had adequate fit to the data (depression: p = 0.73; mania: p = 0.43, mixed: p = 0.67). A p-value below 0.05 was considered statistically significant. Software packages Survo MM (www. survo.fi), R v2.8 with the gss package (www. r-project.org), and Stata v10 (www.stata.com) were used for data processing and statistical analyses.
favoring pharmacotherapy, we assumed that the patient was on medication directly after discharge if (i) the patient had purchases within two months of discharge, or (ii) the patient was on medication before the hospitalization (or during the hospitalization for antidepressants) and died within two weeks without purchases. Statistical analysis
The crude incidence of suicide was calculated by dividing the number of events by the corresponding follow-up time in person-years. The survival curves were derived using the Kaplan–Meier estimator. Follow-up was from the day of discharge to the event of suicide, death, next psychiatric hospitalization, or 31 December 2003, whichever occurred first. The follow-up was considered censored for events other than suicide. To determine the probability of committing suicide after discharge from hospital on a certain day, smoothed hazard functions for post-discharge suicides and corresponding Bayesian confidence intervals were estimated using the penalized likelihood method (23). Cox’s proportional hazards model was used to determine the adjusted effects of observed variables potentially modifying the hazard of suicide after hospitalization for manic, depressive, or mixed episodes. Since the clinical course of bipolar disorder is episodic and pleomorphic, the validity of information on type of illness phase and clinical state at discharge declines progressively over time. These analyses focused only on the first 120 days after discharge, i.e., follow-up was considered censored after four months. About two-thirds of suicides during the first post-discharge year occur during this high-risk period. To control for possible bias in estimating the effect of post-discharge medication, a binary time-dependent covariate changing
1
Results Incidence and temporal patterns of post-discharge suicides in Finland in 1987–2003
The temporal patterns of hazard of suicide in patients with bipolar disorder in Finland in 1987–2003 are depicted in Figure 1. Overall, there were 466 suicides after 52,747 psychiatric hospital discharges in 13,581 patients (Table 1). The median number of hospitalizations for patients was two [mean 3.9, standard deviation (SD) 5.6]; 2,298 patients (16.9%) had hospitalizations only for depression. The hazard depended strongly on type of illness phase, being highest but steeply declining after discharge with depression, also high but less declining in mixed states, but lower and relatively stable after mania (Fig. 2). The incidence rate of completed suicide per 105 patient-years was 492 [95% confidence interval (CI): 449–539] during the post-discharge follow-up
Survival
0.995
Other Manic Mixed
0.99
0.985 Depression
0.98 0
100
200
300
Days since discharge
Fig. 1. Kaplan–Meier curves for post-discharge suicide mortality in bipolar disorder in Finland in 1987–2003 by type of illness episode.
869
Isomets€ a et al. Table 1. Psychiatric admissions of patients with bipolar disorder during 1987–2003 (used in estimation of hazard functions)
No of discharges No. of suicides No. of discharges in women Intra-episodic suicide attempts Involuntary admissions Admission length, days 0–13 14–27 28–41 42–120 120+
Manic n (%)
Mixed n (%)
Depressive n (%)
Other n (%)
16,403 115 (0.7) 8,140 (49.6) 55 (0.3) 7,848 (47.8)
9,891 80 (0.8) 5,624 (56.9) 51 (0.5) 1,964 (19.9)
14,233 213 (1.5) 8,375 (58.8) 110 (0.8) 1,632 (11.5)
12,220 58 (0.5) 6,773 (55.4) 24 (0.2) 3,390 (27.7)
3,957 (24.1) 3,919 (23.9) 3,263 (19.9) 4,701 (28.7) 563 (3.4)
4,376 (44.2) 2,077 (21.0) 1,257 (12.7) 1,721 (17.4) 460 (4.7)
5,109 (35.9) 3,428 (24.1) 2,033 (14.3) 3,018 (21.2) 645 (4.5)
5,368 (43.9) 2,203 (18.0) 1,336 (10.9) 2,482 (20.3) 831 (6.8)
period, and twofold for male patients [694 (618– 778)] compared with female patients [332 (286– 385)]. Overall, incidence rates declined over time after discharge following hospitalization, with most marked differences in rates after hospitalization for depression (Fig. 2). Hazard predictors of suicide during the first 120 days after discharge in Finland in 1995–2003
The information on suicides during the first 120 days after discharge in 1995–2003 is preA
sented in Table 2. In Cox models for suicides (n = 65) after hospitalizations (n = 9,635) for bipolar depression, the hazard ratio (HR) was 8.05 (95% CI: 2.49–26.04; p < 0.001) after hospitalization with an intra-episodic suicide attempt and 3.63 (95% CI: 2.12–6.23; p < 0.001) for male patients, but 0.186 (95% CI: 0.07–0.52; p = 0.001) for patients taking lithium. Valproate or antidepressants without concurrent mood stabilizer did not significantly influence the hazard. Suicides after mania (28 of 10,730 discharges) were predicted by male sex (HR = 4.00, 95% CI: B
0.00015
Hazard 0.00015
0.0001
Hazard
0.0001
0.00005
0.00005
0 100
200
0
300
100
Days since discharge
200
300
Days since discharge
C
D
0.00015
Hazard
Hazard 0.00015
0.0001
0.0001
0.00005
0.00005
0
0 100
200 Days since discharge
300
100
200
300
Days since discharge
Fig. 2. Survival function after discharge according to type of episode at hospitalization. (A) Manic episode. (B) Depressive episode. (C) Mixed episode. (D) Other episode.
870
BD post-discharge suicides in Finland Table 2. Psychiatric admissions of patients with bipolar disorder during 1995–2003 (used in estimation of Cox models)
No. of discharges Suicides (120 days) No. of discharges in women Intra-episodic suicide attempts Involuntary admissions Admission length, days 0–13 14–27 28–41 42–120 120+ Lithium (after discharge) Valproate (after discharge) Only antidepressant (after discharge)
Manic n (%)
Mixed n (%)
Depressive n (%)
Other n (%)
10,730 28 (0.3) 5,306 (49.4) 7 (0.1) 4,957 (46.2)
6,911 20 (0.3) 3,894 (56.3) 25 (0.4) 1,076 (15.6)
9,635 65 (0.7) 5,544 (57.5) 58 (0.6) 860 (8.9)
8,670 16 (0.2) 4,776 (55.1) 6 (0.1) 2,112 (24.4)
2,743 (25.6) 2,503 (23.3) 2,078 (19.4) 3,020 (28.1) 387 (3.6) 3,919 (36.5) 3,402 (31.7) 820 (7.6)
3,561 (51.5) 1,387 (20.1) 753 (10.9) 991 (14.3) 219 (3.2) 1,813 (26.2) 2,502 (36.2) 1,082 (15.7)
3,914 (40.6) 2,241 (23.3) 1,281 (13.3) 1,857 (19.3) 342 (3.5) 2,089 (21.7) 2,396 (24.9) 3,609 (37.5)
4,193 (48.4) 1,472 (17.0) 825 (9.5) 1,626 (18.8) 554 (6.4) 1,531 (17.7) 1,572 (18.1) 1,679 (19.4)
1.61–9.87; p = 0.003), and those after mixed episodes (20 of 6,911 discharges) by intra-episodic suicide attempts (HR = 11.1, 95% CI: 1.41–87.8; p = 0.02). Discussion
This study included all psychiatric hospitalizations with a clinical diagnosis of bipolar disorder in Finland in 1987–2003, and all suicides within one year of discharge were included. Specific risk factors could be reliably investigated among post-discharge suicides occurring in 1995–2003. As hypothesized, we found marked differences in temporal patterns of suicide risk depending on type of illness episode. Furthermore, an intra-episodic suicide attempt raised the hazard by up to 12-fold, and male sex in most analyses by about threefold. After hospitalization for depressive episodes, hazard for suicide was only one-seventh among those taking lithium, but not significantly lower with valproate. While this finding is consistent with many previous studies suggesting lithium to have a specific preventive effect, alternative explanations must also be considered. Our study had some major strengths. It was a national population study comprising all suicides within the 17-year period under investigation. The unit of analysis was discharge from a psychiatric hospital of a patient with a clinical diagnosis of bipolar disorder, and the hospitalizations were drawn from the Finnish National Hospital Discharge register. This information was linked to other registers, including Causes of Death obtained from Statistics Finland and reimbursement registers from the National Insurance Institution. The quality of these registers is generally considered excellent (19–22), and major inaccuracies are
unlikely. Given this setting, we were able to investigate suicides by type of illness episode, which due to the relatively low incidence of completed suicides necessitates large populations. We were also able to link information on pharmacotherapy and some other background variables. The study also had obvious limitations, and some methodological issues should be addressed. First, the study included only hospitalizations of patients with a clinical diagnosis of bipolar disorder. Bipolar disorder is known to have been significantly underdiagnosed in Finland during the study period (24, 25), with those with the most obvious classical clinical presentation of bipolar I disorder most likely correctly diagnosed, but no more than, at best, about one-half of patients with bipolar II disorder. Secondly, no information was available concerning the mental state of the subject at the time of the fatal act. While the study provides unbiased estimates of hazard after a hospitalization for a particular type of illness episode, a relapse of the same or another polarity may well have taken place, predisposing to suicide (11–13). Thirdly, only a limited range of all risk factors could be investigated. Specifically, information on preceding suicide attempts, the most robust indicator of elevated risk (1, 5–7), was limited to only those during the same hospitalization episode. Fourthly, it is important to note that the role of pharmacotherapy is evaluated based on the best available information on purchases of prescriptions before and after hospitalization. In order to purchase, one must be alive, a fact that may bias hazard estimates with regard to pharmacotherapy overall, but is unlikely to explain differences in patterns between classes of agents. There is no guarantee that the subject purchasing a prescription actually used it. We probably overestimated
871
Isomets€ a et al. the use of pharmacotherapy, likely resulting in conservative estimates of preventive effects. Fifthly, despite a massive national database covering 17 years, the number of completed suicides in some Cox model subgroup analyses remained low, rendering the study vulnerable to type II errors. Sixthly, one-sixth of the patients had hospitalizations exclusively for depressions, which is a rough proxy estimate for the proportion of diagnosed patients with bipolar II disorder. Thus, the vast majority of patients clinically diagnosed with bipolar disorder likely had bipolar I disorder. Seventhly, the study covered the years 1987–2003. Whether changes over time increasing the recognition and diagnosis of bipolar disorder, shortening the duration of hospitalizations (9), or increasing the availability of new pharmacotherapy options for the illness influence the patterns of suicide risk remains unknown. Finally, the generalizability of findings from Finland to other countries is uncertain, despite considerable similarity of epidemiological and register-based findings with regard to mental disorders. Finland has high suicide rates, being most comparable to those of other Nordic or northern European countries (26). Our analyses revealed the strong association of type of illness episode at hospitalization with subsequent risk for post-discharge suicide. While the incidence of suicide attempts is known to be highest in mixed episodes (13, 27), and a significant proportion of victims in a psychological autopsy study of bipolar patients were in a mixed episode at the time of suicide (11), post-discharge suicide risk was not highest among those hospitalized for a mixed episode. Suicide risk immediately after discharge was by far the highest among hospitalizations for bipolar depression. This subgroup comprised half of the post-discharge suicides and also most strongly exhibited the typical pattern of steep decline in risk during the first three months after discharge. The incidence of suicide was lower and temporal patterns unimpressive among those hospitalized for mania or other episodes. We can only speculate about the reasons for these partly unexpected patterns. A plausible explanation could be depressive relapse after discharge, which is the most common adverse outcome. In contrast, mixed episodes, when correctly diagnosed while hospitalized, may be more effectively treated with antimanic pharmacotherapy. Patients hospitalized for manic or mixed episodes generally receive adequate pharmacotherapy, whereas quality of treatment for bipolar depression may often be poor (28) and the available treatments leave room for improvement. The duration of mixed episodes is shorter than that of depressive episodes;
872
therefore, accumulated total risk for suicide attempts is higher for depressions, despite the highest incidence occurring in mixed episodes (13, 27). For purposes of suicide prevention, the implications are clear. In the post-discharge period for bipolar patients, those hospitalized for depression comprise the group with the highest immediate risk (in the first three months) for suicide. We also found robust evidence for the association of two risk factors with completed suicide during the post-discharge period: preceding suicide attempts and male gender. Both findings are unsurprising given that they are established risk factors for suicide overall and in bipolar disorder specifically (1, 5–7). However, the potency of these risk factors is marked within the high-risk subgroup of hospitalizations for depression. Male patients had an almost fourfold hazard of suicide during the first three months, so particular attention should be paid to preventive efforts in men. Further research on post-discharge suicides should investigate risk factors that may be particularly relevant for male patients, but remained uncovered in this study, such as substance abuse (5–7, 11) and availability of lethal methods (29). The hazard estimate for a preceding suicide attempt in those hospitalized for depression was even higher, at eightfold. Unlike most previous studies that have investigated preceding attempts during the patient’s lifetime, we investigated suicide attempts immediately preceding or during the same episode of hospitalization. It appears that these temporally close attempts may be even stronger indicators of shortterm risk than lifetime attempts. This pattern is consistent with the known clustering of preceding attempts close to eventual suicide among suicides overall (30). However, despite indicating an exceptionally high relative risk when present, the sensitivity of an intra-episodic suicide attempt as a risk indicator is low (4.2%). Due to the relative rarity of completed suicide even during this high-risk period, the positive predictive value of an intra-episodic suicide attempt was 3.7%. We investigated whether the hazard of suicide is related to purchasing lithium, valproate, or antidepressants preceding or after hospitalization. Of the three types of pharmacotherapy, only lithium seemed to modify risk. Purchasing lithium was associated with a markedly lower risk of suicide. This finding is fully concordant with other registerbased studies comparing suicide risk among individuals when using different pharmacotherapies (14), albeit not with a recent major randomized trial (31). Moreover, our estimate for effect size (HR 0.19) of reducing risk among patients hospitalized for depression is highly consistent
BD post-discharge suicides in Finland with that of two meta-analyses; Baldessarini et al. (15) reported an odds ratio of 0.20 and the recent meta-analysis of randomized evidence by Cipriani et al. (17) reported 0.13. Thus, our findings add to the literature on a preventive effect of lithium, suggesting that this effect also concerns the postdischarge period when hospitalized for depression. However, it is important to keep in mind the major methodological limitations of our study. We had no information on adherence to treatment or quality of treatment provided, both of which are known to be compromised in suicides occurring during lithium treatment (32). We only had data on purchases of lithium, from which we made assumptions of actual use. The characteristics of patients prescribed each of the medications may differ markedly, and it is possible that patients evaluated to be at risk of suicide are not prescribed lithium because of known toxicity in overdose. Therefore, many kinds of confounding factors may influence the findings. Obviously, despite adjusting for known differences in the models, a possibility of bias exists. Nevertheless, such factors did not result in our finding any preventive effect for valproate or antidepressants. It is noteworthy that we found no evidence for the latter two being associated with an elevated hazard either, despite both possibilities having been debated in the literature (33). Overall, the finding of a possible preventive effect for lithium is consistent with results of most randomized studies (17). Conclusions
Among inpatients with bipolar disorder, the risk of completed suicide is high and related strongly to time elapsed from discharge after hospitalizations for depressive episodes, and less strongly after hospitalizations for mixed episodes. Preceding intra-episodic suicide attempts and male sex are powerful predictors of suicide risk. Lower suicide rate among patients taking lithium is consistent with a preventive effect. Acknowledgements Funding was provided by the National Institute for Health and Welfare, Helsinki, Finland. At the time of data collection, the MERTTU project was supported by a grant from the Academy of Finland.
Disclosures The authors of this paper do not have any commercial associations that might pose a conflict of interest in connection with this manuscript.
References 1. Nordentoft M, Mortensen PB, Pedersen CB. Absolute risk of suicide after first hospital contact in mental disorder. Arch Gen Psychiatry 2011; 68: 1058–1064. 2. Tondo L, Isacsson G, Baldessarini R. Suicidal behaviour in bipolar disorder: risk and prevention. CNS Drugs 2003; 17: 491–511. € 3. Osby U, Brandt L, Correia N, Ekbom A, Sparen P. Excess mortality in bipolar and unipolar disorder in Sweden. Arch Gen Psychiatry 2001; 58: 844–850. 4. Angst J, Angst F, Gerber-Werder R, Gamma A. Suicide in 406 mood-disorder patients with and without long-term medication: a 40 to 44 years’ follow-up. Arch Suicide Res 2005; 9: 279–300. 5. Hawton K, Hawton K, Sutton L, Haw C, Sinclair J, Harriss L. Suicide and attempted suicide in bipolar disorder: a systematic review of risk factors. J Clin Psychiatry 2005; 66: 693–704. 6. Pompili M, Gonda X, Serafini G et al. Epidemiology of suicide in bipolar disorders: a systematic review of the literature. Bipolar Disord 2013; 15: 457–490. 7. Isometsa E. Suicidal behavior in mood disorders – who, when and why? Can J Psychiatry 2014; 59: 120–130. 8. Qin P, Nordentoft M. Suicide risk in relation to psychiatric hospitalization: evidence based on longitudinal registers. Arch Gen Psychiatry 2005; 62: 427–432. 9. Pirkola S, Sohlman B, Heil€ a H, Wahlbeck K. Reductions in post-discharge suicide after psychiatric deinstitutionalization – a nationwide register study. Psychiatr Serv 2007; 52: 221–226. 10. Høyer EH, Licht RW, Mortensen PB. Risk factors of suicide in inpatients and recently discharged patients with affective disorders. A case–control study. Eur Psychiatry 2009; 24: 317–321. 11. Isomets€ a ET, Henriksson MM, Aro HM, L€ onnqvist JK. Suicide in bipolar disorder in Finland. Am J Psychiatry 1994; 151: 1020–1024. 12. Valtonen H, Suominen K, Mantere O, Lepp€ am€ aki S, Arvilommi P, Isomets€ a E. Suicidal behaviour during different phases of bipolar disorder. J Affect Disord 2007; 97: 101–107. 13. Valtonen HM, Suominen K, Haukka J. Differences in incidence of suicide attempts during phases of bipolar I and II disorders. Bipolar Disord 2008; 10: 588–596. 14. Goodwin FK, Fireman B, Simon GE, Hunkeler EM, Lee J, Revicki D. Suicide risk in bipolar disorder during treatment with lithium and divalproex. JAMA 2003; 290: 1467–1473. 15. Baldessarini RJ, Tondo L, Davis P, Pompili M, Goodwin FK, Hennen J. Decreased risk of suicides and attempts during long-term lithium treatment: a meta-analytic review. Bipolar Disord 2006; 8: 625–639. 16. Kessing LV, Sønderg ard L, Kvist K, Andersen PK. Suicide risk in patients treated with lithium. Arch Gen Psychiatry 2005; 62: 860–866. 17. Cipriani A, Hawton K, Stockton S, Geddes JR. Lithium in the prevention of suicide in mood disorders: updated systematic review and meta-analysis. BMJ 2013; 346: f3646. 18. Pirkola S, Sund R, Sailas E, Wahlbeck K. Community mental-health services and suicide rate in Finland: a nationwide small-area analysis. Lancet 2009; 373: 147–153. 19. Kiesepp€ a T, Partonen T, Kaprio J, L€ onnqvist J. Accuracy of register- and record-based bipolar I disorder diagnoses in Finland: a study of twins. Acta Neuropsychiatr 2000; 12: 106–109.
873
© 2014 John Wiley & Sons A/S Published by John Wiley & Sons Ltd.
Bipolar Disorders 2014: 16: 867–874
BIPOLAR DISORDERS
Original Article
Post-discharge suicides of inpatients with bipolar disorder in Finland Isomets€a E, Sund R, Pirkola S. Post-discharge suicides of inpatients with bipolar disorder in Finland. Bipolar Disord 2014: 16: 867–874. © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. Objectives: Suicide risk in psychiatric inpatients is known to be remarkably high after discharge. However, temporal patterns and risk factors among patients with bipolar disorder remain obscure. We investigated post-discharge temporal patterns of hazard and risk factors by type of illness phase among patients with bipolar disorder. Methods: Based on national registers, all discharges of patients with bipolar disorder from a psychiatric ward in Finland in 1987–2003 (n = 52,747) were identified, and each patient was followed up to postindex discharge or to suicide (n = 466). For discharges occurring in 1995–2003 (n = 35,946), factors modifying hazard of suicide during the first 120 days (n = 129) were investigated. Results: The temporal pattern of suicide risk depended on the type of illness phase, being highest but steeply declining after discharge with depression; less high and declining in mixed states; lower and relatively stable after mania. In Cox models, for post-discharge suicides (n = 65) after hospitalizations for bipolar depression (n = 9,635), the hazard ratio was 8.05 (p = 0.001) after hospitalization with a suicide attempt and 3.63 (p < 0.001) for male patients, but 0.186 (p = 0.001) for patients taking lithium. Suicides after mania (n = 28) or mixed episodes (n = 20) were predicted by male sex and preceding suicide attempts, respectively. Conclusions: Among inpatients with bipolar disorder, suicide risk is high and related strongly to the time elapsed from discharge after hospitalizations for depressive episodes, and less strongly after hospitalizations for mixed episodes. Intra-episodic suicide attempts and male sex powerfully predict suicide risk. Lower suicide rate after hospitalizations for depression among patients prescribed lithium is consistent with a preventive effect.
Suicide risk in patients with bipolar disorder is among the highest in all psychiatric patients. Based on data from Danish national psychiatric registers, it is estimated that about 8% of male patients and 4% of female patients diagnosed with bipolar disorder die by suicide (1). Risk of suicide is about 20-fold that of the general population (2–4). However, it likely varies markedly according to patients’ clinical state and the presence of other risk factors (5–7). Suicide risk in all psychiatric inpatients is known to be high immediately postdischarge, making this period an important target for preventive efforts (8, 9). This risk is also very
Erkki Isometsa€a,b,c, Reijo Sundd,e and Sami Pirkolaa,c a
Department of Psychiatry, University of Helsinki, Mood, Depression, and Suicidal Behavior Unit, National Institute for Health and Welfare, c Department of Psychiatry, Helsinki University Central Hospital (HUCH), dService Systems Research Unit, National Institute for Health and Welfare, eFaculty of Social Sciences, University of Helsinki, Helsinki, Finland b
doi: 10.1111/bdi.12237 Key words: attempted suicide – bipolar disorder – incidence – lithium – patient discharge – pharmacotherapy – risk factors – suicide Received 15 January 2014, revised and accepted for publication 9 May 2014 Corresponding author: Erkki T. Isometsa€, M.D., Ph.D. Department of Psychiatry Institute of Clinical Medicine University of Helsinki P.O. Box 22 Helsinki 00014 Finland Fax: +358-9-47163735 E-mail: [email protected]
high in patients with affective disorders overall (8–10), but the temporal patterns of post-discharge suicide risk and the factors modifying risk specifically among bipolar patients are poorly known. Suicidal acts by bipolar patients are most likely to take place during depressive or mixed episodes, and rarely during mania or hypomania or remission (11–13), and may be prevented by long-term lithium treatment more effectively than other pharmacotherapies (14–17). Thus, the level and temporal patterns of post-discharge suicide risk and the factors modifying them can be expected to differ markedly depending on the type of illness episode
867
