Risk of suicide and fatal drug poisoning after discharge from the emergency department: a nested case-control study
- 1Office of Education, Research and Development, Landspitali, Reykjavik, Iceland
- 2Department of Preventive Medicine, University of Iceland, Reykjavik, Iceland
- Correspondence to Professor Vilhjalmur Rafnsson, Department of Preventive Medicine, University of Iceland, Stapi, Hringbraut, IS-101 Reykjavik, Iceland;
- Accepted 23 December 2008
Objectives The objectives were to study the risk of suicide and fatal drug poisoning among emergency department users who had been discharged home, based on the main diagnoses selected by the emergency physician upon discharge.
Methods This is a case-control study nested in a cohort of users of the emergency department who had been discharged. The cases of suicide (n=41) and fatal drug poisoning (n=21) were identified from the National Cause-of-death Registry, and five times as many controls were selected from users of the emergency department. Multivariate logistic regression analysis was used to calculate the OR and 95% CI adjusted for age and gender.
Results Frequent visits to the emergency department were significantly associated with suicide and fatal poisoning. The study period spanned 7 years. The OR for suicide among cases and controls was 7.84 for those diagnosed as having mental disorders, 96.89 for those with use of alcohol, 24.51 for those with drug intoxication and 2.69 for those with a non-causative diagnosis. The OR for fatal poisoning for cases and controls was 12.26 for those with use of alcohol, 37.22 for those with drug intoxication and 5.76 for those with the classification category factors influencing health status.
Conclusions The clinical implication is that patients with any combination of previous main diagnoses of mental disorder, alcohol use, drug intoxication, a non-causative diagnosis or with the classification category factors influencing health status should be evaluated and assessed for potential risk of suicide or fatal drug poisoning.
- Mental disorders
- alcohol use
- drug use
- non-causative diagnosis
- contacts with the health service
- frequent visits
A previous study of emergency department users showed that people who had committed suicide had frequently visited the emergency department.1 Results indicate that most suicide attempts are treated in the emergency department2 3 and, in the USA, an increasing proportion of mental healthcare is provided by emergency departments.3 Individuals who injure themselves and are treated in an emergency department are at high risk of repeated self-harm and suicide.4–7 Some of these studies of emergency departments conclude that suicidal ideation and planning may not be detected among their patients.6 7 Attention has understandably been directed at persons coming to the emergency department with self-harm because of their increased suicide risk. However, up to 60% of those who later commit suicide have attended the emergency department the year before the suicide but do not present themselves as cases of self-harm.5 Other risk factors for suicide identified among emergency department users who later commit suicide are alcohol and drug use, intoxication or overdose, mental disorders and physical injuries.4–7 There is therefore a need to search for risk indications presented by previous attendances of the users of the emergency department who later take their own life and, in order to be clinically important, these indications should be sought among the routinely recorded events at the emergency department.
The aim of this study was to evaluate the risk of suicide and fatal drug poisoning among emergency department users who had been discharged from hospital, with reference to the main diagnoses selected by the emergency physician upon discharge.
This case-control study is nested in a cohort of emergency department users at Landspitali, Hringbraut, Reykjavik, Iceland who were discharged from hospital. The primary source of data was computer records of attendances over the period 1995–2001. The study cohort comprised 19 259 individuals who had made 30 221 visits to the emergency department and been discharged. All residents of Iceland are included in the National Registry under a unique personal identification number allocated at birth (a 10-digit number that includes the day, month and year of birth). Each visit to the emergency department is recorded by the patient 's personal identification number, enabling automatic and accurate record linkages. This database on emergency department visits also includes information on gender, age, admission date and time, and one main discharge diagnosis according to the International Classification of Diseases (ICD-10).8 The main diagnosis at discharge is considered to reflect the medical conclusion regarding the patient's complaints on the occasion of the specific visit. In the present study no attempt was made to evaluate the quality of the diagnosis, but the category “non-causative diagnosis” (defined according to ICD-10) is likely to include an unknown number of somatoform disorders or medically unexplained symptoms. In Iceland, healthcare (including emergency department services) is financed by tax revenues and all residents are covered by national health insurance schemes. The fee which users must pay on visiting the emergency department only partly covers the cost of each visit and is similar to the amount charged to patients for visits to primary healthcare, general practitioners or specialists in out-of-hospital practices which are also covered by national health insurance.
Information on the vital status of the cohort was obtained by record linkage based on individual identification numbers from the National Registry and the National Cause-of-death Registry for the follow-up period (1995–2002). The cause of death in the National Cause-of-death Registry is obtained from death certificates. Both registries are maintained by Statistics Iceland. The causes of death were coded according to ICD-10.8 Two sets of cases were defined in this study. The first was designated as “suicide” and consisted of persons whose cause of death was one of two ICD-10 categories: Suicide and intentional injuries (ICD-10 codes X60–X84, 29 deaths) and Injuries of undetermined intent (ICD-10 codes Y10–Y34, 12 deaths). The second set of cases, “fatal drug poisoning”, comprised persons who had died of Accidental poisoning (ICD-10 codes X40–X45, 21 deaths). This procedure ensured that all people who were registered as having died from suicide and poisoning in the cohort during the follow-up period were included as cases in the study. The selection of controls was done according to the description by Rothman.9 We used five times as many controls as cases to strengthen the power of the study, and the controls were selected at random from the cohort. The controls were chosen from a unique risk set of people in the cohort who had attended the emergency department before or on the recorded day of death of the particular case, and the exposure variables were counted up to that date. The controls therefore represent the exposure condition of the cohort.
Some of the emergency department users had made a number of visits to the department each year or in different years during the inclusion period of the cohort. The total number of visits was counted per individual, starting with the first visit and ending at the day of death or the corresponding day for the controls. At the time of discharge from the emergency department the responsible doctor chose one diagnosis as the main one to be recorded in the computer file. The exposure categories were designed according to the main diagnosis which, in turn, was classified according to ICD-10. The following five exposure categories were defined a priori; three were according to previous studies and the two last categories were added as possible new risk indicators:
Mental disorders, the ICD-10 categories Mental and behavioural disorders F00–F99, except F10–F19.
Use of alcohol, the ICD-10 categories Mental and behavioural disorders due to use of alcohol F10 and Toxic effect of alcohol T51.
Drug intoxication, the ICD-10 categories Mental and behavioural disorders due to psychoactive substance use F11–F19, Poisoning by drug, medicaments and biological substances T36–T50, Accidental poisoning by and exposure to drugs X40–X44, and Intentional self-poisoning by and exposure to drugs X61–X63.
Non-causative diagnosis, the ICD-10 chapter Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified R00–R99;
Factors influencing health status, the ICD-10 chapter Factors influencing health status and contact with health services Z00–Z99.
A multivariate case-control analysis was performed using a logistic regression analysis.10 The adjusted OR and exact computation of 95% CI were calculated using the SPIDA software package.11 Case-control status was the dependent variable. Age was treated as a continuous variable expressed in years, and gender as a dichotomous variable. The number of visits to the emergency department was treated as a continuous variable. Whether an individual ever or never received the particular main diagnosis was treated as a dichotomous variable. Initially an analysis was performed to evaluate the number of visits to the emergency department adjusted for age and gender. In different separate analyses the other exposure categories were each treated as dichotomous variables and number of visits to the emergency department omitted.
The National Bioethics Committee and the Data Protection Commission approved the study.
Altogether there were 41 cases of suicide; the most common diagnosis was exposure to drugs (n=15, 36.6%), followed by hanging, strangulation and suffocation (n=9, 22.0%), poisoning due to other gases and vapours (n=6, 14.6%), injury from firearms (n=3, 7.3%), unspecified events (n=3, 7.3%), drowning (n=2, 4.9%) and one case each of poisoning by an unspecified chemical, jumping from a high place and injury by a sharp object. Of the 21 cases of fatal drug poisoning, all except two were due to drugs; these two were due to alcohol poisoning, with no distinction made between prescribed medicine and illegal drug use. There was no industrial poisoning in this cohort. The mean time between the last attendance and death for cases of suicide was 15.6 months (range 0–88) compared with 14.0 months (range 0–41) for cases of fatal poisoning.
Table 1 shows the mean age and number of admissions and the ranges for cases and controls. Logistic regression analysis showed that the OR among cases of suicide and controls according to age, gender and number of visits was 1.37 (95% CI 1.07 to 1.77). The number of visits to the emergency department was therefore significantly associated with suicide. The OR among cases of fatal drug poisoning and controls according to age, gender and number of admissions was 1.28 (95% CI 1.03 to 1.60). Thus, as for suicide, the number of visits to the emergency department was significantly associated with fatal drug poisoning.
Table 2 shows the adjusted OR for suicide according to age, gender, mental disorders, alcohol use, drug intoxication, non-causative diagnosis and the category factors influencing health status. The ORs were high for all the diagnoses and diagnostic categories and were significant for mental disorders, alcohol use, drug intoxication and non-causative diagnosis.
Table 3 shows the adjusted OR for fatal drug poisoning according to age, gender, mental disorders, alcohol use, drug intoxication, non-causative diagnosis and the category factors influencing health status. The ORs were high and significant for alcohol use, drug intoxication and factors influencing health status.
Our calculations show a significant association between the numbers of visits to the emergency department, mental disorders, use of alcohol, drug intoxication and non-causative diagnosis and the risk of suicide, adjusted for age and gender, among users of the emergency department who were discharged from hospital. We also found a significant association between the numbers of visits, use of alcohol, drug intoxication and the category factors influencing health status and the risk of fatal drug poisoning, adjusted for age and gender, among users of the emergency department who were discharged from hospital. The results concerning suicide are inconsistent with previous studies;4–7 however, the association between non-causative diagnosis and the risk of suicide is a new finding. In accordance with the size of the study, the confidence intervals are wide but the ORs in many cases are high.
Apart from the number of visits to the emergency department, the exposure variables in the present study were defined according to the main diagnosis selected by the physicians when discharging the patient home from the emergency department. So the risk factors identified in the study reflect the daily practices and the main diagnoses in the emergency department and not only that the visits to the emergency department were, for example, alcohol-related. The diagnoses at discharge were registered prior to and independent of the subsequent suicide or fatal poisoning; they were not derived from a retrospective evaluation of the hospital records. The users of the emergency department were discharged home without any specialised psychosocial assessment before leaving the department. It is unknown to us, and thus not the subject of this study, whether or what kind of treatment or follow-up patients received during the time interval from discharge until they took their own lives or during the corresponding time for the controls. The management of self-harm and self-poisoning patients has been considered most important. However, those who were not admitted to hospital wards were unlikely to receive psychosocial assessment according to studies in the UK.12–14 Acute poisonings are considered common15–17 and most frequently involve drugs and alcohol; most cases are not hospitalised but are discharged home after treatment in the emergency department.15–17 Patients with alcoholism, alcohol-related diseases and injuries frequently attend emergency departments,18 19 and alcohol use and abuse are major factors affecting the emergency department population.20
Our previous studies on users of the emergency department who were discharged from hospital also showed that the number of visits was associated with increased mortality, especially due to suicide and drug poisoning.1 In the present study we were able simultaneously to take into account mental diseases, alcohol use, drug intoxication, non-causative diagnosis and a new potential risk factor—that is, the category factors influencing health status, the ICD-10 chapter “Factors influencing health status and contact with health services” (codes Z00–Z99). This interesting category was used as a discharge diagnosis at the emergency department in approximately 4% of visits,21 and there are indications in a study from San Francisco that this category is also used in other emergency departments.22 According to ICD-10, this chapter is “provided for occasions when circumstances other than disease, injury or external cause classifiable to categories A00–Y89 are recorded as “diagnoses” or “problems”8 which, by definition, excludes the non-causative diagnosis of the chapter R00–R99.
Use of the comprehensive population registries in Iceland adds to the strength of our study. The universal use of personal identification numbers has made record linkage possible and, together with the registered time and hour of visits to the emergency department, enabled us to count every individual's attendances. Record linkage with the National Registry enabled us to ascertain whether individuals were residents of Iceland and to access vital and emigration status for all cohort members. The National Registry and National Cause-of-death Registry are nationwide registries, and use of the latter enables identification of the causes of death according to the death certificate and thus accurately identifies the cases. All death certificates in Iceland are issued by a physician and, if the deceased person's physician refuses to sign the death certificate because he or she is unable to state the cause of death or due to the circumstances of the death (unexplained, unusual, suspicious, due to intoxication, or following an accident), this is reported to the police and the medical examiner. The latter will, depending upon the circumstances, take care of any post mortem and forensic investigation, after which the death certificate is issued.23 Information on the quality of the recording of the cause of death on death certificates in Iceland is not available. However, when evaluating death registration at a global level, the registration data from Iceland were categorised as high-quality data overall and ranked in the same category as data from 23 developed countries including the USA and the UK.24
One of the limitations of this study is the sole use of the main diagnosis at discharge from the emergency department; many of these users of the department surely also had other diagnoses. However, the registered main diagnosis at discharge is considered to reflect the conclusion of the physician concerning the patient's complaints on the occasion of the specific visit. There have been repeated claims of underreporting of suicide in official records,25 and this may be an inherent weakness of studies based on such data. If such misclassification is present in this study, it would bias the observed effect towards the null value. This potential limitation of the study may be unimportant in the light of the high OR found for the various variables. In this case-control study it was not possible to take into consideration the time which elapsed between death and attendance, but diagnoses with a short time span to death may be stronger predictors than those with a long time to death after attendance. The difference in absolute number of attendances between cases and control was not large even though the frequency of attendances had statistically significant ORs, which indicates a need for further evaluation of the clinical usefulness.
To our knowledge, this is the first published case-control study on suicide and fatal drug poisoning based on a cohort of users of emergency departments using clinical diagnoses as exposure variables. The case series from Leeds which studied people committing suicide during 1994–7 and their contact with mental health services and accident and emergency departments found that 39% of those who took their own life had attended accident and emergency departments and that 39% of these had attended because of self-harm.5 Only a few of the patients who self-harmed were discharged home without being assessed by a psychiatrist,5 and alcohol or other intoxicating drugs were mentioned as important. The cohort study from Albuquerque, New Mexico7 analysed the association of several diagnoses and potential suicide risk with suicide, but no distinction was made between admitted and discharged patients. A diagnosis of self-harm, suicidal ideation, overdose and mental disorders followed by diagnoses of alcohol use, substance use and drug use carried the highest risk for suicide.7 The result of an interesting prospective survey from an emergency department in Dallas, Texas6 showed that a suicidal tendency is not readily detected in a busy emergency department among non-psychiatric patients.
The clinical implication is that a patient with any combination of previous main diagnoses of mental disorder, alcohol use, drug intoxication and the new exposure categories “non-causative diagnosis” and “factors influencing health status” should be evaluated and assessed for risk of suicide or fatal drug poisoning. Each main diagnosis which is significantly associated with suicide or fatal poisoning is an independent risk indicator for either suicide or fatal poisoning. The cases of suicide and fatal drug poisoning had a similar pattern in terms of the risk indicators under study. Increased frequency of attendance is also important in this assessment.
Future studies should aim at further evaluation of the importance of the newly confirmed risk indicators and analysis of subcategories (ie, non-causative diagnosis and factors influencing health status). Moreover, these diagnoses may be clinically known as medically unexplained symptoms which warrant further studies on the possible overlapping of these diagnoses with somatoform or dissociative disorders classified under F40–F48 in the ICD-10.
We thank Helgi Sigvaldason for his assistance in the statistical analysis. This work was supported by the Landspitali University Hospital Research Fund, the Landspitali and the Science Fund of the Icelandic Nurses' Association.
Funding This study was supported by a grant from the Landspitali University Hospital Research Fund and the Icelandic Nurses' Association Science Fund.
Competing interests None declared.
Ethics approval This study was conducted with the approval of the The National Bioethic Committee,Vegmúli 3, 108 Reykjavik, Iceland, #03-071-V1.
Provenance and peer review Not commissioned; externally peer reviewed.