The study was a comparative cohort study, comparing a cohort exposed to the intervention with cohorts not exposed to the intervention (Fig. 1). The study was part of a larger prospective observational study of acute poisoning conducted at the Oslo Accident and Emergency Outpatient Clinic (OAEOC) in 2012 [19]. The cohorts were derived from this study and from similar studies done in 2003 [15] and 2008 [14]. All three studies registered all acute poisonings at the OAEOC during one year. The study periods were April 1st 2003 to March 31st 2004, April 15th 2008 to April 14th 2009, and October 1st 2011 to September 30th 2012. Patient inclusion criteria were identical in all the three study periods.
Setting
In Norway, patients cannot present directly to hospitals, but have to be assessed in primary care or by the ambulance service first. Primary care emergency services are provided by regular general practitioners during office hours, and by local casualty clinics during nights and weekends. The OAEOC is the main casualty clinic in Oslo. It serves the entire city (population 613,285 as per January 1st 2012 [20]) at all hours. There are facilities for short time observation, but diagnostic tools and treatment options are limited. The total number of consultations at the OAEOC was about 180,000 in 2003, and about 200,000 in 2008 and 2012. In 2012, about 3000 consultations were due to acute poisoning [19].
Participants
Patients 16–27 years of age presenting at the OAEOC during the study periods with an acute poisoning in which the main agent was a substance of abuse, were included. Substances of abuse were defined as any drug used for recreational purposes or with a known potential for addiction or abuse, including ethanol, prescription drugs and other substances. Patients with uncertain identity, i.e. without a Norwegian national identity number, were not included. The patients were divided into four groups based on being in the target age for the intervention (16–22 years) or not (23–27 years), and on being included prior to or after the implementation of the intervention. The patients eligible for the intervention were thus the younger post-intervention group (Fig. 1). The older groups (23–27 years) were included as comparisons. The rationale was that changes in referrals or repetition rates across the target age groups were likely also to show up in the older groups if due to other factors than the intervention. Hence, we decided to use groups of patients just older than the target age groups (16–22 years). The upper limit of 27 years was a compromise between being high enough to include an adequate number of patients and low enough not to include patients too different in sociodemographic characteristics due to being older.
In total, 1323 patients were included. The major toxic agents at the index episode were ethanol 823 (62 %) and opioids 215 (16 %). 719 (54 %) patients were male. Nearly all the poisonings were accidental overdoses (Table 1).
Intervention
The intervention program consisted of consultations with specially trained social workers. Patients in the target age treated for acute poisoning by substances of abuse were contacted by a social worker for a consultation before discharge. An appointment was also made for a telephone consultation within two weeks, unless the patient declined or the social worker deemed it unnecessary. Patients not contacted before discharge, were contacted by telephone within a week and offered similar consultations. Letters were sent to patients not reached by telephone, encouraging them to contact the intervention program. Parents were contacted if the patient was a minor (<18 years). In the consultations, both face to face and by telephone, the social workers used motivational interviewing [21, 22]. The main aim was to reduce the hazardous use of substances of abuse. In addition, AUDIT (Alcohol Use Disorders Identification Test) [23] and DUDIT (Drug Use Disorders Identification Test) [24] were used to identify patients in need of further follow-up, and to ensure that these patients were referred or already in relevant treatment. Over and above these elements, the consultation content was left to the discretion of the social worker and tailored to the individual patient.
Outcome measures
We compared proportions of patients referred to follow-up and repetition rates of poisoning. The comparisons were intention-to-treat analyses. Thus, the younger post-intervention group encompassed all patients eligible for the intervention, whether or not they were in contact with the program.
Referral to follow-up was defined as hospital admission, addiction clinic admission, referral to a psychiatric outpatient clinic, the patient’s general practitioner, child welfare services, social services, or ensuring that the patient already was in such treatment. Discharge with no follow-up or the patient self-discharging was also registered. When comparing proportions of patients referred to follow-up, patients from the 2003 cohort were not included, as referrals were not registered in this study period (Fig. 1).
The poisoning episode resulting in inclusion was considered the index episode. Subsequent episodes for the same patient in the same study period were considered repetition episodes.
Contact with the intervention program was defined as at least one consultation with program personnel, face to face or by telephone. A voicemail or just a short conversation was not considered a consultation.
Data collection
The poisoning episodes were recorded consecutively. The treating physician registered the following data on a pre-set registration form: date, age, sex, residence (2008 and 2012 studies only), main toxic agent, intention behind the poisoning (2008 and 2012 studies only), admission to hospital and referral to treatment or follow-up (2008 and 2012 studies only). We examined electronic patient records for patients in the younger post-intervention group, and registered all contacts with the program. Data on death and emigration were retrieved from the National Registry. Norwegian national identity numbers, unique for every inhabitant, were used to identify patients.
Toxic agents (ethanol, opioids, benzodiazepines, central stimulants, gamma-hydroxybutyrate (GHB), or other) were diagnosed by the physician treating the patient, based on all information available at the time: patient history; information from the police, the ambulance service, relatives or other companions; and clinical examination. No toxicological tests were available, apart from breath analysis for ethanol. Intention was clinically assessed by the treating physician, and registered as suicidal intention, accidental overdose by substances of abuse, mere accident, or unknown.
Ethics
The study was performed in accordance with the Helsinki declaration. Inclusion was based on verbal consent. The study was approved by the Regional Committee South East for Medical and Health Research Ethics (REK nr 2010/1129-1) and by the Oslo University Hospital Information Security and Privacy Office.
Statistics
Analyses were done in IBM SPSS version 21 (IBM Corp.). Pearson’s Chi square test or Fisher’s exact test (for expected cell counts of five or less) were used to compare frequencies.
Repetition rates were estimated by survival analysis. An event was defined as the first repetition episode. Time under observation was from the index episode to the event, or until censored by death, emigration, or the end of the relevant study period, whichever happened first. Time under observation was counted in number of days, as integers. If the first repetition occurred on the same day as the index episode, the time under observation was set to one day.
Cox regression analysis was used to estimate hazard ratios for factors associated with repetition of poisoning, with all cohorts pooled together. In the Cox regression analysis relevant variables were first analysed one by one. The variables analysed were age, sex, main toxic agent, hospitalisation, suicidal intention and self-discharge from the OAEOC. Ethanol was chosen as the reference group when estimating hazard ratios for main toxic agents, as it was the largest group. Factors associated with repetition in the univariate analyses with a significance level of p < 0.10 were included in the multivariate model. We did a separate Cox regression analysis to look for association between hazard of repetition and referral to follow-up. We used the same model, but only on the 2008 and 2012 cohorts, as referral to follow-up was not registered in 2003. We did not include hospitalisation in this analysis, as hospitalisation was categorised as being referred to follow-up.