Approximately one third of European Drug Emergencies Network (Euro-DEN) Plus presentations to the EDs in the three cities (London, UK; Roskilde, Denmark and Basel, Switzerland) were coded using the ICD-10 coding systems, although there was variability in the proportion that were coded by centre. This study has demonstrated that the majority of primary and secondary codes applied related to acute recreational drug toxicity, but often they were not specific to the drug(s) used. The concordance of ICD-10 coding with drugs used showed high specificity but low sensitivity: those presentations coded with a specific drug ICD-10 code were likely to have taken that drug but the majority who had used the drug were not likely to be have the relevant ICD-10 code applied.

Previous studies have shown that around 50% of presentations with ARDT are discharged directly from the ED, and these presentations are unlikely to be coded using the ICD-10 coding system as this relates to patients admitted to hospital [2, 3]. In this study, in the two centres (Basel and London) with more uncoded Euro-DEN Plus presentations, nearly 90% of these had only been managed within the ED, consistent with those previously reported studies [4]. However, in Roskilde, Denmark, all of the presentations were coded as there is a requirement for all hospital presentations to be coded and reported to the National Register (Landspatientregisteret). Use of ICD-10 data in Denmark would potentially capture the burden of number of presentations; it would not allow true understanding of the burden for an individual drug, since the codes applied were not necessarily accurate for the drug(s) used. Further work is needed to understand whether other European countries utilise different methods for coding emergency presentations to see whether this variability in the capture of recreational drug presentations is replicated elsewhere.

Of the primary ICD-10 codes applied to the Euro-DEN Plus presentations, around 70% could be classified as a code that related to either the mental and behavioural or poisoning/toxicological effects of a recreational drug. Of the 213 presentations where a primary diagnostic code was recorded, only 150 (70.4%) received a primary ICD-10 diagnostic code that was related to ARDT. Some of the “acute drug toxicity”-related ICD-10 codes applied may not at initial review be consistent with the use of a recreational drug. One example is those presentations that had a primary ICD-10 code related to poisoning by anaesthetics agents; it is possible that some of these presentations related to the acute toxicity from either GBL/GHB or ketamine use since both GHB and ketamine have been used as anaesthetics and there is no specific ICD-10 code for these drugs. In a previous study, where clinical coders coded simulated ARDT presentations, the ICD-10 diagnostic code “T41.2 Poisoning by other and unspecified general anaesthetics” was used by 14% for a GBL-related collapse in a sauna [3]. The non-ARDT primary ICD-10 codes applied related to a range of other medical conditions, which could be considered as caused by the use of a recreational drugs/NPS such as acute mental health or neuropsychiatric disorders, cardiac toxicity or trauma. It is also interesting to note that the majority intent-related ICD-10 codes applied related to accidental rather than intentional poisoning. It is unknown whether coders used accidental codes as individuals are unlikely to use them with an intent to overdose, even if there is intentional use of a substance that could be associated with the risk of acute toxicity when used.

The main issue identified with this study is likely under detection of presentations through the utilisation of searching by ICD-10 codes. In a previous retrospective review of 484 known acute recreational toxicity presentations to an ED in London, UK, only 13% had been coded with a primary ICD-10 code that related to ARDT and overall looking at levels of coding, only approximately 20% of presentations had one or more ICD-10 codes applied that could relate to ARDT [2]. There was high specificity to detect specific drugs, since when an ICD-10 code was applied to a presentation, the majority of those presentations involved the use of that drug. However, there was low sensitivity since the majority of those who had actually used a specific drug did not have the relevant drug-specific ICD-10 code applied to the presentation. NPS will not have specific ICD-10 codes, since they have not been used for as long as the current ICD codes have been. Mephedrone was the most widely used and available NPS at the time of these Euro-DEN Plus presentations; since there is no ICD-10 code for mephedrone, it is not possible to look at coded presentations to detect mephedrone use. In part, this is due to ICD-10 having been endorsed for use by the World Health Organisation, prior to widespread use of NPS. However, when we look at the known mephedrone presentations, only around a third had an ICD-10 code applied that related to the use of a stimulant recreational drug (which would be the most appropriate for mephedrone in the absence of a specific ICD-10 code). This is different to the previous clinical coding study of simulated ARDT presentations, where around 68% applied a primary ICD-10 code of “poisoning by psychostimulants with abuse potential” to a patient with anxiety and palpitations after the use of mephedrone [3]. There have been no changes to the coding practice in the study centres since the cases used in this study, and therefore we would not expect there to have been any improvement in quality of local coding. However, following this study, the proposed ICD-11 coding system was released on the 18th June 2018 [9]. The proposed ICD-11 coding system appears to have incorporated more coding categories related to the use of recreational drugs and novel psychoactive substances, including proposed specific codes for the synthetic cathinones, MDMA and related drugs and the synthetic cannabinoids. These changes may in part help with some of the coding issues that we have identified in this and previous studies, enabling better understanding of the burden of health care utilisation related to the use of a wider range of substances. However, despite the release of the proposed ICD-11 coding system, the expectation is that this will be endorsed by the 72nd World Health Assembly in May 2019 and therefore it is unlikely to be used in routine coding reporting until 2022 onwards.

The European Monitoring Centre for Drugs and Drug Addiction (EMCDDA) through its REITOX national focal point network early warning system collates information on the prevalence of use and harms associated with the use of both classical recreational drugs and novel psychoactive substances [10]. A survey of these national focal points has demonstrated only ~ 50% of countries had any systematic data collection system in place to detect the acute harms of classical recreational drugs and novel psychoactive substances [11]. Those that did largely relied on information from poison information centres and/or data from ICD-10/other clinical coding systems, with limitations on the detail of information provided or the accuracy of detection of cases. The Euro-DEN Plus project enables a detailed data collection currently, in May 2018, within 31 hospitals in 21 European and neighbouring countries, which can provide detailed information to the EMCDDA on recreational drugs/NPS associated with acute toxicity and describe the pattern of acute toxicity seen with different drugs/NPS [4, 8, 12–14]. As such data from the Euro-DEN Plus project now forms part of the annual EMCDDA European Drug Report, adding the existing key indicators around drug detection, drug-related deaths and use of treatment services for drug dependency [15–17]. In the USA, The Toxicology Investigators Consortium (ToxIC) project collects data on cases directly reviewed by medical/clinical toxicologists from a number of hospitals across the USA [18–21]. This has some similarities to the Euro-DEN Plus project, in that a proportion of the cases in the Euro-DEN Plus project have been seen by a medical/clinical toxicologist during their hospital attendance. However, the Euro-DEN Plus project has the advantage of collecting all cases presenting to each individual participating centre irrespective of whether they have been seen at the bedside by a medica or clinical toxicologist. This means cases presenting outside of the time that a medical/clinical toxicologist is on site will also be captured. Additionally, it captures potentially less severe cases who may be discharged directly by the emergency department physicians, rather than needing to involve a medical/clinical toxicologist.

The main limitation of the Euro-DEN Plus dataset is that drug(s) used is based on self-report or clinicians’ interpretation of the presentation for the majority of the cases. The use of self-reported and/or clinical interpretation of drugs involved reflects real-life clinical practice in many hospitals, where the majority of patients do not have analytical confirmation undertaken. A previous sub-group analysis of Euro-DEN Plus presentations with analytical confirmation of drugs used showed there was good correlation between self-reported use of drugs such as heroin, methadone, cocaine, and amphetamines [22]. There was less correlation between cases of NPS-related toxicity, as NPS cannot be detected with the usual drug screening immunoassays, and some substances were not routinely screened for due to limited detection windows in biological matrices (e.g. volatile nitrites and GHB/GBL).