Mortality and repeated poisoning after self-discharge during treatment for acute poisoning by substances of abuse: a prospective observational cohort study

We charted characteristics of patients self-discharging during treatment for acute poisoning by substances of abuse. Furthermore, we looked for associations between self-discharge and short-term mortality, and between self-discharge and repeated poisoning.

Prospective observational cohort study.

The study was done at the Oslo Accident and Emergency Outpatient Clinic (OAEOC) in Oslo, Norway. The OAEOC is a primary care emergency outpatient clinic, serving the entire city at all hours. There are about 200,000 consultations a year. The majority of patients treated for acute poisoning by substances of abuse in Oslo, are treated at the OAEOC [20]. These patients are assessed according to a systematic observation procedure, and the median observation time is four hours [19]. Oslo is the capital city of Norway (population 613,285 as per 1 January 2012 [21]).

We included all patients 12 years and older treated at the OAEOC for an acute poisoning by substances of abuse. Patients were included by the doctor treating them. Substances of abuse were defined as any potential substance of abuse including alcohol, prescription drugs, illegal drugs, and others. Patients were included consecutively during one year, from 1 October 2011 to 30 September 2012. Patients treated for multiple conditions were included if the poisoning in itself caused need for treatment or observation. Among 2733 eligible cases, 174 did not have a Norwegian national identity number and were excluded. In 216 cases the patient declined to participate. In the end, 2343 cases in 1731 patients were included.

For all included cases, a registration form was completed by the doctor treating the patient. Any information missing in the form was collected, if available, from the electronic medical records. Information on deaths from 1 October 2011 to 31 December 2012 was retrieved from the National Cause of Death Register, using the patients’ unique Norwegian national identity number.

For each case, we registered age, gender, main toxic agent, homelessness, suicidal intention, previous history of severe mental illness, time of presentation, time of discharge, and whether the patient self-discharged during treatment. Diagnoses of main toxic agents were made by the doctor treating the patient. Main toxic agent was defined as the agent assessed to be most toxic among the agents taken, assumed doses considered. The diagnoses were based on all information available then and there. Subsequently, main toxic agents were grouped by us as ethanol, opioids, stimulants, gamma-hydroxybutyrate (GHB), benzodiazepines, and others. Suicidal intention was registered according to the assessment of the doctor treating the patient. Previous history of severe mental illness was assessed and registered by the doctor treating the patient based on information from local medical records, from the patient and/or the patient’s companions. Severe mental illness encompassed psychosis, bipolar disorder and severe personality disorders. Homelessness was defined as being registered without a permanent address in the National Registry, which was accessed via the electronic medical records. Self-discharge was defined as leaving without being seen by a doctor, disappearing during treatment or leaving against medical advice. For patients disappearing during treatment, time of discharge was defined as the time when they were registered as missing from the clinic in the electronic medical records.

By using the patient’s unique Norwegian national identity number, we could identify patients presenting more than once during the inclusion period. The collected information was collated for each patient. Main toxic agent for a patient with more than one presentation was defined as the main toxic agent most frequently diagnosed in that patient’s poisoning episodes. In case of even frequencies, we used the toxic agent we considered most serious, in the following order: opioids, stimulants, GHB, benzodiazepines, ethanol, others.

The main outcome measures were short-time mortality, defined as death during the inclusion period or the following three months; repeated poisoning once during the inclusion period; and repeated poisoning several times during the inclusion period. We looked for associations between self-discharge and these outcomes. Furthermore, we described characteristics of the self-discharging patients.

Statistical analyses were done in SPSS version 25 and in an online calculator from EpiTools epidemiological calculators [22]. Chi-square test was used when comparing proportions. Mann-Whitney U-test was used when comparing continuous variables. We did a Cox regression analysis to identify factors associated with short-term mortality, and multiple logistic regression analyses to identify factors associated with repeated poisoning.

The study was performed in accordance with the Helsinki declaration. It was approved by the Regional Committee South East for Medical and Health Research Ethics (REK nr 2010/1129–1) and the Oslo University Hospital Information Security and Privacy Office.

Self-discharge was associated with a three-fold increased risk of repeated poisoning, and a five-fold increased risk of repeating more than once. Short-time mortality was twice as high among self-discharging patients, though this increase was not statistically significant. Self-discharging patients were older, more frequently homeless, and the main toxic agent more frequently was an opioid.

We found self-discharge to be an independent risk factor for repeated poisoning by substances of abuse. The increased risk of repetition was high compared with previous findings of increased readmission rates following discharge against medical advice in general hospital populations [1‐3].

The increased repetition rate among patients self-discharging during treatment for acute poisoning of abuse calls for concern, as risk of fatal overdose increases with increasing numbers of previous non-fatal overdoses [23‐25]. In line with previous studies in general hospital populations [1, 2], we found an increased risk of death following self-discharge. However, the increase was not statistically significant in our study, probably a consequence of too small a sample for the short observation time.

The self-discharge rate of 15% among our patients was high compared to the rates of 1–3% reported in general hospital populations [1‐5]. The rate was more comparable to the rate of 17% previously reported among patients treated for recreational drug toxicity [14], but higher than the rates of 6–11% reported among patients treated for acute poisoning in general [10‐13]. Though higher rates of self-discharge have been reported among patients treated for self-harm, 23% among adolescents in a Korean study [26], and ranging from eight to 39% in a UK study encompassing 22 general hospitals [9], the rate of 15% marks our patient group as prone to self-discharge.

There are clear indications that self-discharging patients are a troubled and vulnerable group. Garland et al. found readmission rates and mortality to be increased immediately following discharge against medical advice, then declining but still increased at least as long as six months later, leading them to speculate that the increases were related both to the acute condition and to characteristics of the patient [1]. In our study, self-discharging patients more often were opioid users, homeless, and of older age. In a Canadian study of hospitalised patients using illicit drugs, unstable employment, incarceration, daily heroin injection and younger age were associated with leaving against medical advice [27]. An Australian study of patients with deliberate self-poisoning found absconding patients more likely to have a mental illness and to be unemployed [28]. Among trauma patients in California, readmissions were more likely to be for psychiatric reasons following discharge against medical advice compared to regular discharge [3].

In a qualitative study, communication breakdown was found to be a major reason for self-discharge, encompassing health service providers using language that was hard to understand for patients, inconsistent messages from doctors and nurses, and sheer rudeness from health service providers [29]. Accordingly, clear and consistent communication in an easy to understand language along with good bedside manners, might partly remedy the situation. However, self-discharge is a complex phenomenon, associated with serious consequences. Self-discharging patients might benefit from special attention, for instance by way of post-discharge telephone contact or home visits from outreach services.

Somewhat surprisingly, in our study self-discharge occurred most frequent on weekday late afternoons and evenings, while during the lesser staffed night shifts larger proportions were medically discharged. However, adolescents and young adults with ethanol poisoning dominate among patients with acute poisoning at the OAEOC at night, especially during weekends [20]. During the inclusion period, these patients were eligible for a targeted follow-up program at the OAEOC [30], possibly also affecting the discharge process.

The information on deaths is reliable, as all included patients were traced in the National Cause of Death Register using their unique Norwegian national identity number. This is a major strength. However, our study was probably underpowered concerning factors associated with death due to the short period of observation for fatalities.

To add power to the Cox regression analysis, we chose to use death within the inclusion period or the following three months as a measure of short-time mortality, hence an observation time varying from three to fifteen months between patients. Patients with several poisonings during the inclusion period are more likely to present for the first time earlier in the inclusion period than non-repeating patients, thus having longer observation times. As self-discharge was associated with repeated poisoning, self-discharging patients might also tend to have longer observation times. This might overestimate the short-time mortality among self-discharging patients when comparing proportions between the groups. However, this ratio was in the same range as the unadjusted hazard ratio. Accordingly, we do not think that this possible bias had any impact on our results, as variation in observation time is handled by the Cox regression analysis.

For the subgroup of self-discharging patients merely disappearing during treatment, time of discharge was defined as when they were registered as missing from the clinic. Patients with acute poisoning by substances of abuse are observed at short intervals by the nursing staff at the OAEOC. When a patient is not found, this is marked in the electronic medical records. Thus, observation time is systematically somewhat overestimated for this subgroup, but on average probably by less than 15 min.

Our study included the majority of acute poisonings by substances of abuse during one year in a European capital city. However, about 200 patients with acute poisoning by substances of abuse are triaged for direct hospital treatment by the ambulance services every year, thus bypassing the OAEOC [19]. These are more severe poisonings. Furthermore, in about 700 cases per year, the patient is left on scene after treatment by the ambulance service [31]. These are mainly opioid overdoses. Consequently, the repetition rate is probably underestimated in our study, especially for patients with opioid overdoses.

The registration of previous history of severe mental illness was based on the information available then and there to the doctor treating the patient. Consequently, the prevalence is probably underestimated. Furthermore, the category of severe mental illness was not rigorously defined and encompassed a number of different diagnoses. Similarly, diagnoses of toxic agents and suicidal intention were made by the doctor treating the patient. Toxicological laboratory analyses were not done. On the other hand, the diagnoses were made in real clinical situations, and the patients were managed accordingly.