Suicide is a preventable cause o\ue004 death. A\ue004ter about two centuries o\ue004 research in suicide prevention, the e\ue002ectiveness o\ue004 a number o\ue004 interventions has been demonstrated and various risk \ue004actors have been placed in perspective. Tus, \u201cit is no longer acceptable to state blandly that there is no convincing evidence \ue004or the e\ue002ectiveness o\ue004 suicide prevention measures\u2026\u201d and, even more importantly, \u201c\u2026 the unacceptable rate o\ue004 suicide worldwide can be reduced.\u201d1

WHO estimated that 877 000 deaths were due to suicide in the year 2002,2 the majority o\ue004 which (85%) occurred in low- and middle-income countries.3 Attempted suicide can be up

culturally di\ue001\ue001erent sites (Campinas, Brazil; Chennai, India; Colombo, Sri Lanka; Karaj, Islamic Republic o\ue001 Iran; and Yuncheng, China) participated, \ue001rom January 2002 to October 2005, in a randomized controlled trial to receive either treatment as usual, or treatment as usual plus brie\ue001 intervention and contact (BIC), which included patient education and \ue001ollow-up. Overall, 91% completed the study. The primary study outcome measurement was death \ue001rom suicide at 18-month \ue001ollow-up.

aDepartment o\ue001 Mental Health and Substance Abuse, World Health Organization, CH-1211 Geneva 27, Switzerland.
bDepartment o\ue001 Public Health Sciences, Karolinska Institute, Stockholm, Sweden.
cAustralian Institute \ue001or Suicide Research and Prevention, Gri\ue001\ue000th University, Brisbane, Queensland, Australia.
dTehran Psychiatric Institute, Mental Health Research Centre (IUMS), Tehran, Islamic Republic o\ue001 Iran.
eDepartment o\ue001 Psychiatry, FCM\u2013UNICAMP, Campinas, SP, Brazil.
\ue001Department o\ue001 Psychological Medicine, Faculty o\ue001 Medicine, University o\ue001 Colombo, Colombo, Sri Lanka.
gBeijing Suicide Research and Prevention Center, Beijing Hui Long Guan Hospital, Beijing, China.
hDepartment o\ue001 Psychiatry, Kotturpuram, Chennai, India.
iEstonian-Swedish Mental Health and Suicidology Institute, Estonian Centre o\ue001 Behavioural and Health Sciences, Tallinn, Estonia.
jDepartment o\ue001 Behavioural Medicine, School o\ue001 Family and Public Health Medicine, Nelson R Mandela School o\ue001 Medicine, University o\ue001 KwaZulu-Natal, Durban,

to 40 times more \ue004requent than com- pleted suicide.4,5 Many o\ue004 those who attempt suicide require medical atten- tion and they are at high risk \ue004or com- pleted suicide.6\u20138 Sel\ue004-in\ue001icted injuries represented 1.4% o\ue004 the global burden o\ue004 disease in 20022 and are expected to increase to 2.4% by 2020. As suicide is among the top three causes o\ue004 death in the population aged 15\u201334 years,9 there is a massive loss to societies o\ue004 young people in their productive years o\ue004 li\ue004e. Suicide mortality statistics are available at: http://www.who.int/mental_health/ prevention/suicide/country_reports/en/ index.html.

Tere have been several recent reviews o\ue004 interventions that may be considered e\ue002ective in reducing

suicides.1,10\u201312 Under the \ue004ramework o\ue004 universal, selective, and indicated interventions,13 the general population is targeted by universal interventions (e.g. restricting access to means o\ue004 sui- cide) and selective interventions \ue004ocus on high-risk subgroups (e.g. people with mental disorders), whereas those who have attempted suicide are con- sidered high-risk individuals and are there\ue004ore addressed with indicated interventions, which include a range o\ue004 behavioural therapies and approaches such as cognitive therapy.14

Among indicated interventions, various approaches have been tested to prevent subsequent suicidal behaviour by suicide attempters; extensive review articles are available.15,16 Usually, the

primary outcome measure used \ue004or these interventions was repeated suicide attempts. It is suggested that extrapola- tion \ue004rom attempted to completed sui- cide is valid.17 As completed suicide is a rare outcome in statistical terms, large numbers o\ue004 suicide attempters would be needed to demonstrate the e\ue002ective- ness o\ue004 an intervention in terms o\ue004 a reduction o\ue004 completed suicides. Te multisite study presented here tried to tackle this challenge by combining data \ue004rom di\ue002erent sites that had applied the same research protocol.

Previously, completed suicides were used as an outcome measure in a study that investigated the maintenance o\ue004 long-term contact (i.e. a total o\ue004 5 years and 24 contacts) with high suicide-risk psychiatric patients re\ue004using \ue004urther treatment.18\u201320 Te contact comprised regular short letters expressing concern \ue004or the person\u2019s well-being and inviting them to respond. Tis was associated with a signi\ue000cant reduction in suicide rates \ue004or at least 2 years a\ue004ter discharge \ue004rom the in-patient setting.

In addition, a \u201ctele-help/tele-check\u201d service (i.e. an alarm system that can be activated to call \ue004or help and a service that contacts a person twice a week \ue004or assessment o\ue004 their needs and to provide emotional support) could sig- ni\ue000cantly reduce the number o\ue004 suicide deaths in the elderly, who typically have an elevated risk o\ue004 suicide com- pared with an age-adjusted number \ue004or the general population.21,22

Tese two examples demonstrate that it is possible to reduce the suicide rate in populations at risk by keeping in regular contact with patients. Brie\ue004 interventions \ue004or alcohol problems are another promising type o\ue004 intervention that have not been previously applied to suicidal behaviours.23\u201325 Tese are designed to address the speci\ue000c be- haviour o\ue004 drinking with in\ue004ormation, \ue004eedback, health education and practi- cal advice and \ue004ocus in order to raise awareness o\ue004 the problem and advise change. Tey were \ue004ound to be e\ue002ective in reducing alcohol-related problems, to be more e\ue002ective than no counsel- ling, o\ue004ten as e\ue002ective as more extensive treatment, and \ue004easible within relatively brie\ue004 contacts. Repeated \ue004ollow-up visits were recognized as a \ue004actor \ue004avouring behaviour change and maintenance.

Te multisite intervention study on suicidal behaviours (SUPRE-MISS), launched by WHO in 2000, evalu-

ated an innovative intervention in a large randomized controlled trial, that brought together the elements o\ue004 in\ue004or- mation, education, and practical advice \ue004rom brie\ue004 interventions with the main- tenance o\ue004 long-term \ue004ollow-up contact on a regular basis. It used completed suicides as the primary outcome mea- sure because the reduction in suicide mortality is the most convincing evi- dence \ue004or the e\ue002ectiveness o\ue004 suicide prevention.26 Te multisite randomized controlled trial o\ue004 di\ue002erent treatment strategies \ue004or suicide attempters rep- resented one component o\ue004 SUPRE- MISS, which, overall, aimed at increas- ing knowledge about suicidal behav- iours and e\ue002ective interventions \ue004or suicide attempters.27,28 Tis paper pres- ents the results \ue004rom the \ue000ve sites that completed the randomized controlled trial \ue004ully according to the protocol.

Between January 2002 and April 2004, \ue000ve participating sites (Campinas, Brazil; Chennai, India; Colombo, Sri Lanka; Karaj, Islamic Republic o\ue004 Iran; and Yuncheng, China) applied the same protocol and recruited a total o\ue004 1867 suicide attempters, with an overall drop-out rate o\ue004 9% [brie\ue004 intervention and contact (BIC): 5.4%; treatment as usual (\ue003AU): 12.5%] at the 18-month \ue004ollow-up time point (Fig. 1). Te \ue004ollow-up period lasted until 31 Octo- ber 2005.

Te suicide attempters were iden- ti\ue000ed by medical sta\ue002 in one or more emergency care settings within a de- \ue000ned catchment area with a population o\ue004 at least 250 000. Te study tried to include all suicide attempters con- secutively seen at the emergency care departments. Inadequate recording o\ue004 emergency department visits, inten- tional misreporting o\ue004 suicide attempts as accidental, \ue004ailure o\ue004 the emergency department sta\ue002 to noti\ue004y research sta\ue002, and rapid departure \ue004rom the emer- gency departments o\ue004 patients made it difcult to include all eligible patients, once medically stable. However, no more than an estimated 5% o\ue004 cases were lost that way. Te rate o\ue004 re\ue004usal o\ue004 enrolment was 7%. Other reasons \ue004or exclusion were death in the ward, clinical conditions not allowing an in- terview, leaving against medical order, residence in a di\ue002erent catchment area

or language problems. At any rate, the age and sex o\ue004 the enrolled patients did not di\ue002er \ue004rom those assessed \ue004or eligibility.

Te research protocol was ap- proved by the relevant ethics commit- tee in each site and all patients enrolled in the randomized controlled trial gave written in\ue004ormed consent. Te baseline interviews were conducted \ue004ace-to-\ue004ace by trained psychiatrists, medical doc- tors, psychologists or psychiatric nurses, a maximum o\ue004 3 days a\ue004ter the emer- gency department admission.

All enrolled participants (N = 1867) were randomly assigned to BIC (n = 922) or \ue003AU (n = 945). An allocation sequence based on a random-number table was used to randomly assign all enrolled subjects to BIC or \ue003AU; the allocation sequence was maintained in a separate location to prevent clinician bias. Te subjects were blinded as to their assignment to speci\ue000c treatment groups. In the consent \ue004orm, subjects were asked to agree to a \ue004ollow-up, without speci\ue000cation o\ue004 the number and time o\ue004 contacts. Tis in\ue004ormation was given only a\ue004ter subjects had been randomly assigned to their group. Completed suicide was the primary outcome measure applied. For a sig- ni\ue000cance level o\ue004 95% (two-sided) and power o\ue004 80%, assuming 3% suicides in the \ue003AU and 1% in the BIC group at 18-months \ue004ollow-up, a total o\ue004 1730 subjects was needed.

Te \ue003AU modality was carried out according to the norms prevailing in the respective emergency department. At 18 months a\ue004ter discharge, the subjects were \ue004ollowed-up using the same \ue004orm used by the BIC group. \ue003ypically, the treatment provided in the participat- ing sites would not cover routine or systematic psychiatric or psychological assessment or help besides the treat- ment o\ue004 somatic symptoms. I\ue004 there were no complications, the patients were normally discharged a\ue004ter somatic treatment. Tere was no routine or sys- tematic approach o\ue004 re\ue004erral to outpa- tient \ue004acilities or a psychiatric unit.

Te BIC treatment modality in- cluded, in addition to \ue003AU, a 1-hour individual in\ue004ormation session as close to the time o\ue004 discharge as possible and, a\ue004ter discharge, nine \ue004ollow-up contacts (phone calls or visits, as appropriate) according to a speci\ue000c time-line up to

Re\ue001usal (7%);
death in the ward/clinical conditions
not allowing interview (15%);
leaving against medical order (70%);
residence in a di\ue001\ue001erent catchment
area (3%);
language problems (5%).

Allocated to
BIC
(n = 922)

18 months (at 1, 2, 4, 7 and 11 week(s), and 4, 6,12 and 18 months), conducted by a person with clinical experience (e.g. doctor, nurse, psychologist). Te individual in\ue004ormation session was conducted according to a written protocol which all sites adhered to. It included in\ue004ormation about suicidal behaviour as a sign o\ue004 psychological and/or social distress, risk and protec- tive \ue004actors, basic epidemiology, rep- etition, alternatives to suicidal behav- iours, and re\ue004erral options. Whenever an interviewer realized that a patient needed more intensive treatment, the relevant re\ue004erral to help was made, when available and i\ue004 judged necessary (Fig. 1).

Te questionnaire29 \ue004or the compre- hensive assessment o\ue004 all suicide at- tempters enrolled was commonly ap- plied across all sites, translated into the local language o\ue004 each site, adapted to take into account cultural speci\ue000cities, and pilot-tested to assess \ue004ace and con- tent validity. It was largely based on the European Parasuicide Study Interview Schedule (EPSIS),30 which had been applied in the WHO/EURO Multi- centre Study on Suicidal Behaviour. It covered sociodemographic items, in\ue004ormation about the current suicide attempt, a series o\ue004 variables on clinical in\ue004ormation (e.g. mental and physical health status, traumatic experiences, alcohol and drug use) and included several sel\ue004-report scales.

For recording \ue004ollow-up contacts with the patients, a short one-page questionnaire was applied. Questions included whether the patient was still alive; i\ue004 not, what the cause o\ue004 death had been (as reported by in\ue004ormants); i\ue004 yes, whether he/she had committed any \ue004urther suicide attempts; how the patient \ue004elt; whether he/she \ue004elt the need \ue004or any support and whether he/ she had sought support. Te protocol is accessible on the web in English, French and Spanish (available at: http:// www.who.int/mental_health/resources/ suicide/en/index.html).29

In each site, data entry, cleaning, veri- \ue000cation and con\ue000dentiality were con- ducted under the direction o\ue004 the principal investigator. Te site-speci\ue000c data were sent to WHO where they were re-checked and compiled into one

database and an overall analysis across all sites was undertaken. Te sociodemo- graphic characteristics describe all ran- domized subjects in the BIC and \ue003AU treatment groups, both when enrolled and when analysed at the 18-month \ue004ollow-up. Di\ue002erences in mortality are presented \ue004or those analysed at \ue004ollow- up. Selected variables at baseline were compared to determine any di\ue002erences between the two treatment groups. Di\ue004- \ue004erences in mortality at 18-months were assessed with thec\u00b2 statistics at a signi\ue000- cance level o\ue004 0.05 (two-sided).

No subjects at all were lost during \ue004ollow-up in Yuncheng; 3%, 4%, 11%, and 15% were lost at the \ue000nal \ue004ollow- up in Campinas, Karaj, Colombo, and Chennai respectively.

Te suicide attempters enrolled were typically \ue004emale, single, with second- ary education and employed (\ue003able 1).

Among the enrolled cases, more than one-third (41% in the \ue003AU; 35% in the BIC group) o\ue004 the suicide attempt- ers put their li\ue004e into danger, almost one-quarter (22% in the \ue003AU; 24% in the BIC group) intentionally ingested alcohol or drugs to \ue004acilitate and imple- ment the suicide attempt, and about one-\ue000\ue004th (around 20% in both groups) had made a previous suicide attempt. No di\ue002erences in the sociodemo- graphic variables and items related to the current attempt between the \ue003AU and BIC groups were \ue004ound among the subjects analysed at the 18-month \ue004ollow-up, which is crucial to the comparison o\ue004 the two groups at this time.

More deaths o\ue004 any cause occurred in the \ue003AU than in the BIC group up to the \ue004ollow-up at 18 months (\ue003able 2); this di\ue002erence was signi\ue000cant (c\u00b2 = 4.360;

stroke, cancer, urinary in\ue004ection, acute respiratory \ue004ailure, AIDS, liver cirrhosis, old age and suicide.