Discussion, conclusion and recommendations

Key findings

Discriminant function

The small proportion of opioid-related deaths limited our ability to identify a small high-risk subsection of the overall population. While there were differences between decedents from opioid overdose, decedents from all other causes, and those still alive, at individual level, people that died from opioid overdose were insufficiently different from members of the other two (more numerous) cohorts. In addition, a relatively high proportion of those that died from opioid poisoning had no healthcare event recorded in the year prior to death. This meant that we would need to monitor roughly one-third of the population in order to include approximately 75% of decedents (from opioid overdose) within that year. With mortality over 1 year of approximately 6% in that third of the population, and approximately 1–2% of these categorised as deaths from opioid overdose, it follows that reducing substantially or even eliminating entirely such deaths would have little effect on mortality rate in the identified ‘high-risk’ subsection.

Given the weak predictive link between deaths from and healthcare events related to opioid overdose, using routine data to identify with sufficient precision those at high risk of death from opioid overdose currently seems infeasible.

Feasibility randomised trial

Four sites participated in the trial and 299 clinical staff were trained. Sixty THN kits were supplied to patients during the 1-year recruitment. Eligible patients were not offered THN kits 164 times: ‘forgot’ (n = 136); ‘too busy’ (n = 15); suspected intentional overdose (n = 3).

In this trial, recruitment of clinical staff and distribution of THN kits were low, with considerable variation between participating sites. Distribution of kits was particularly low in the prehospital setting.

Qualitative

Service users supported the provision of THN kits and training in the emergency setting, as did the service providers. However, both raised concerns with regards to reluctance of the target group attending hospital and their ability to give consent. Both also suggested that a greater focus on family and friends of opioid users may be more appropriate and could improve uptake of THN kits.

Health economics

This study has shown that the methods for collecting costs relating to the naloxone kits and the training of NHS staff are feasible and can take into account differences in the form of training. However, we have been unable to assess the feasibility of collecting patient training times via staff recall. It is unlikely that this finding represents a serious barrier to the conduct of any future evaluation of THN. Cost-effectiveness is unlikely to be sensitive to uncertainties related to training costs,71 in a UK-based evaluation of THN provision in an emergency setting, so this should be recognised when designing future data collection; expensive or disruptive data collection is unlikely to be worthwhile.

Study limitations

1. Quality of coding: currently, many healthcare events have missing inadequate codes. There is a linked lack of granularity – currently events recorded in EDDS and PEDW have been categorised as opioid overdose-related or not. It is unclear that if there would be improvement in an extended classification of events.
2. We have noted above the major limitation that a considerable number of decedents from opioid overdose have no recorded events in routine data in the 12 months (or even in the 36 months) prior to the date of death. Our approach extends in principle to include further data sources, but we note the regulatory and logistical hurdles to doing so in practice. Further data sources include: the Welsh Longitudinal GP data set, potentially including data on smoking/alcohol usage; drug prescriptions; AS records (including records on reasons for emergency call, ambulance attendance and disposition; police and probation/justice data; and data from third-sector organisations working with opioid and other drug and substance users, and their peers.
3. The COVID-19 pandemic resulted in pauses in recruitment and an increase in strain on the ES which may have impacted recruitment of both staff and patients.
4. Unforeseen difficulties with Confidentiality Advisory Group (CAG) and NHS Digital approval to retrieve data for application of the discriminant function to a second population in England and to then retrieve outcomes for the population identified contributed to delays and the decision to discontinue this stage of the research. We were caught between CAG and NHS Digital for many months without clarity as to how to move forward. NHS Digital required detail of the discriminant function prior to approval; however, we were unable to produce the function without a means to test it on the English data.
5. We found some problems with variations in the quality of data. Regarding the staff training data, one centre produced high-quality data; two centres generated data with some discrepancies; and one centre failed to use the data collection tool at all. The number of patients identified as eligible for the trial was also vastly different across sites. While this may have been the case, it is more likely that some sites were more invested in the trial and proactively looking for eligible patients compared to others. In Site 2 ED, where the research nurse was redeployed following COVID-19 for several months, this seemed to be a particular issue. All of these problems are considered resolvable in any future trial.
6. Not all sites were able to provide staff for focus groups, which may have resulted in missing some perspectives of the service providers. It may have been that those who felt more positively towards the intervention were willing to take part in the qualitative aspect of the study. Within the study timetable and resources, and considering COVID-19 delays, we could not pursue this further.

Public and patient involvement

We involved public contributors throughout this study in order to ensure research relevance, quality and accountability and strengthen research rigour. We recruited two public contributors to join the TSC to provide experience-based expertise alongside input from members with academic, clinical and methodological skills. They were involved in all the TSC meetings. However, personal commitments and circumstances made it difficult for any individual to sustain their involvement for the full study period. To minimise the impact of the lack of continuity, we met public contributors in their settings and groups, allowing us to hear their stories, which reinforced for us the challenging circumstances of experiencing, or associating with, opioid addiction.

Interpretation

We found that THN kit administration was overall low in this study, with considerable variation between sites. Reasons given for this were related to the emergency care setting – staff were busy and under pressure; patients were undergoing emergency treatment, mainly for overdoses, and were often not fit or willing to consent to receipt of the kit or the training. There may be ways to overcome these barriers – the protocol for administration of kits could, for instance, be more flexible. A greater focus on relatives and friends may increase the chance of success of TNH provision in the emergency setting. Research conducted in the USA reported that nearly half of the kits distributed by ESs were given to family members with the patient themselves being the second largest group to receive the kit.78 A study assessing the acceptance of nasal naloxone in the ED reported low uptake, with barriers such as difficulties identifying the ‘right’ patient, access to the kits and lack of clarity as to when to offer the kit due to the patient typically not waiting for formal discharge so they may not get a chance to be offered a kit.79 The barriers mentioned are similar to those mentioned by service users and providers in this study and indicate that although there is a general consensus that the ED is a suitable setting to provide THN kits due to the contact with the patient group which may not receive contact with the health service in other means, there are issues with regard to the feasibility of the intervention. These findings are in line with other research which found the ED to be a suitable point for THN kit distribution and training but reported ED staff didn’t have enough time for training and patient identification workflow which could hinder the implementation of this intervention in the ED.38,80 A more recent study assessing methods of increasing THN prescribing in the ED found that although barriers remain, with improved, targeted staff training, the use of works aid such as best-practice advisory tools can increase the prescription of THN kits in the ED.81 Interestingly, we found that both the service users and service providers believed in focusing on the relatives and friends of opioid users; however, a 2022 European study assessing the attitudes and likelihood of using THN kits reported that opioid users were significantly more likely to witness an overdose and use a THN kit compared to the family.44 It may also be possible to identify patients for administration of a THN kit at the time of follow-up, rather than during the emergency episode.

There is overall support from both ED staff and opioid users to provide THN kits and training in the emergency setting. Further findings in this study highlighted issues with regards to recruiting eligible staff in some sites; we were unable to reach the target of training 50% or more of eligible staff in two of the four intervention sites, and therefore did not meet he progression criteria. It is however worth noting that Site 2 ED, which only recruited 8% of staff, and Site 1 ED, which recruited 55% of staff, are similar-sized EDs with similar throughputs and have used an identical approach to the recruitment of staff and contact with the academic research team, but have had very different outcomes in training staff, recording training data and were also unable to provide staff to contribute to focus groups. This indicates that the issues with regards to staff recruitment may lie within the EDs themselves, potentially including pressures of the department, staff support for the project and other factors which could not be accounted for in this study but may be work investigating for future research. It could be beneficial for future research to tailor recruitment and training to each ED and provide different support to each ED depending on their requirements. Furthermore, barriers in recruiting ED staff identified from this research have been documented and include general pressures of the ED including large numbers of patients with challenging circumstances of admittance and unpredictable working environment. Other barriers were identified to be specific to research including a lack of face-to-face communication between the academic research team and research nurses recruiting, as well as a lack of support for research nurses during recruitment which can cause a sense of unease and concern about their confidence in identifying eligible staff and patients.70

Methodologically, this is a challenging area for research and evaluation. There are four main interrelated issues, some of which relate to the specific target population for the intervention, and some of which apply more generally to the conduct of trials in emergency care:

1. Identification of trial population: The THN kits are intended for peer administration. Therefore, the person who benefits from the naloxone may not be the recipient of the kit. Standard RCT recruitment strategies will not work, an alternative has to be found to capture outcomes. We tried to do this through attempting to identify a high-risk population to include in a trial; however, we found that sensitivity and specificity of a discriminant function were both low, and it would not be feasible to use this method in a trial.
2. Consenting patients to any trial in the emergency setting is challenging. In practice, with ethical approval, we often consent them to treatment at the point of recruitment, and to follow up at a later point. In this study we knew from our previous feasibility study49 that follow-up rates would be extremely low and we would not be able to count on consenting patients to follow up after the emergency episode itself.
3. Further to this point, it is very challenging to include self-reported outcomes in a trial in the population. For this reason, we proposed using anonymised linked routine data outcomes, but were unable to define the high-risk population to include in outcome comparisons.
4. With opioid-related death being a relatively rare event, and low THN administration levels, data linkage becomes problematic, as individuals may be at risk of becoming identifiable.

Recommendations for research

The evidence base for distribution of THN in terms of safety, costs and effects is very thin – in any setting, but particularly so in the emergency care environment. It remains the case that the safety and effectiveness of this intervention needs to be evaluated, and that those who suffer an opioid-related emergency may be at highest risk. However, in the light of our findings related to feasibility of the intervention as designed in this study, and our proposed trial methods, in order of priority we recommend the following research be undertaken:

1. What modifications could be made to the intervention to increase its uptake in the emergency setting?
   1. Could the intervention be distributed to family or friends of the opioid user?
   2. Could the intervention be distributed at follow up, rather than during the initial emergency episode?
   3. How can the rate of missed recruitments be reduced? – how could we make sure healthcare staff are more invested in trials? – particularly in this patient group.
2. Is there another setting in which THN could be distributed and evaluated? For example, could the intervention be feasible in the primary care setting?
3. What are the self-reported outcomes of patients who are given THN kits?
4. How can a randomised trial be designed for this population in this setting? How can those who may benefit from the naloxone in the THN kits be identified and followed up? Is there an alternative rigorous study design that could be used?

Conclusion

Rigorous research evidence about safety, costs and effectiveness of THN is still needed to inform policy and practice. We recognise the emergency setting may be a good starting point for identification of people at high risk of harm from opioids and, potentially, for THN kit provision. However, in this feasibility study we did not meet our preset progression criteria related to staff and patient recruitment and to the identification of high-risk patients for outcome comparisons. Therefore, we conclude that a full RCT based on the intervention and methods that we tested is not feasible.