Discussion

In this study, we examined the proportion, predictors and timing of readmissions for STB in patients referred to the PER. Four main findings stand out: (a) 18.1% of patients with previous referrals were readmitted for suicide ideation/plan and 9.7% for suicide attempt; (b) there was high persistence in STB across PER referrals (38–44%); (c) being previously referred for depressed mood was associated with readmission for suicide ideation/plan; and (d) for 19–35% of the patients with previous referrals for STB or depressed mood, the time between referrals was less than 1 month.

In line with previous reports, substance misuse and STB were common clinical reasons for referrals among patients with multiple referrals^{Reference Zeppegno, Gramaglia, Castello, Bert, Gualano and Ressico25} and coexisted in PER patients.^{Reference Tadros, Sharon, Crum, Johnson, Quedado and Fang26} Against the scarcity of knowledge of the suicidal process,^{Reference Franklin, Ribeiro, Fox, Bentley, Kleiman and Huang27} there was a strong temporal association between suicide ideation/plan and suicide attempt, with respect to which four findings warrant attention. First, STB among PER patients was persistent and led to repeated referrals, confirming earlier findings,^{Reference Griffin, Kavalidou, Bonner, O'Hagan and Corcoran10,Reference Owens, Horrocks and House14} often within a short period. For example, patients with suicide attempts were five times more likely to have been previously referred to the PER because of suicide attempt, and for 18% of the patients, this happened within 1 month. Second, those previously referred for ideation/plan were at 66% higher risk of making the transition to suicide attempt, with 25% of patients making this transition within a month after previous referral. This was similar to the transition from depressed mood to suicide ideation/plan. These results suggest a clear transition trajectory through the STB spectrum in the PER, starting with depressive symptoms which may lead to suicide attempt in the emergency room. Third, a previous referral for suicide attempt was highly predictive of a consecutive referral to the PER for suicide attempt. Indeed, over one-third of patients previously referred for suicide attempt were referred again for the same reason and a quarter for suicide ideation/plan. Fourth, patients with substance use or psychotic symptoms were 52–55% less likely to return to the PER with any form of STB. Further analysis revealed substance use as being very persistent; hence, a previous referral for substance misuse was, more than anything, predictive of a referral to the PER again for substance misuse.

From the clinical perspective of the emergency room, our findings are a first step in enabling clinicians to identify those patients who carry the highest STB risk. Previous studies have shown that recent age-at-onset and controllability of ideation/plan are predictive of making a transition to suicide attempt.^{Reference Nock, Millner, Joiner, Gutierrez, Han and Hwang1} Although these are crucial elements in patients’ STB assessments, the rush of the emergency room often precludes clinicians from assessing such information reliably. Specifically, within a PER perspective, our data show that depressed mood precedes ideation/plan, and ideation/plan precedes suicide attempt, and that these transitions can occur within weeks. These are essential findings, stressing the need to take extreme care with such patients by, for instance, initiating a close treatment chain leading to altering the suicidal process.^{Reference Betz and Boudreaux7} Clearly, investing more resources at these crucial moments in the suicidal process, in the PER, may drastically alter the outcome for these patients.

Patients with repeated STB in the PER often elicit negative feelings from non-psychiatric and psychiatric staff.^{Reference Richard-Lepouriel, Weber, Baertschi, DiGiorgio, Sarasin and Canuto28} They are referred to as patients who do not belong in emergency departments, as ‘hard to treat’ and difficult patients,^{Reference Schmidt5} or as patients that systematically overuse the PER. However, we found that this specific group of patients was the most at risk for STB. How these two elements fit together may depend on local policies in emergency departments, but from the viewpoint of suicide prevention, patients with depressive problems and/or suicide ideation/plan do not constitute specific counterindications for an emergency room. On the contrary, our data show that referrals to the PER can constitute crucial opportunities to intervene in the suicidal process.

Although studies in the general population and those based on clinical reports have highlighted temporal associations between substance misuse and STB (either ideation/plan or attempt),^{Reference Hoertel, Faiz, Airagnes, Blanco, Pascal De Raykeer and Franco29} we could not find any evidence for such associations in our study. We found that those previously referred for substance misuse were about half as likely to be referred later with STB. We believe this was mainly due to the persistence of substance use problems and the particular context of the PER, where referral for substance misuse is prevalent (almost one-fourth of the population). We found a similar association between being referred earlier because of psychotic symptoms. This counterintuitive finding potentially challenges the traditional idea that psychosis and STB are closely related.^{Reference DeVylder, Lukens, Link and Lieberman30} However, it is important to keep in mind that our data concern snapshots at the PER. If patients manage to find professional support after the referral, they may no longer present at emergency facilities in case of crisis, as other options are known and available to them and their social network.

Being retrospective in nature, the present study had some notable strengths and limitations. The main strength of our study was its magnitude; overall, we covered a time frame of nearly 20 years, including approximately 25 000 PER referrals. Although our data dis not include all PER referrals at UZ Leuven in this time frame, we weighted the data to be representative of the general PER population. Moreover, we carefully imputed missing values to allow the most accurate estimates of uncertainty. By contrast, many PER studies are hampered by telescoping bias, i.e. patients’ experience that previous referrals to the PER would be recalled as more recent compared to their current referrel.^{Reference Janssen, Chessa and Murre31} Finally, we found that information on prior referrals and reason for referral hold predictive value for STB risk assessment. This is especially relevant for emergency rooms without psychiatric facilities, or facilities with limited resources, or those emergency room settings which prohibit staff from collecting detailed information regarding patients’ STB.

Our study also had notable limitations. First, the setting for this study, an academic teaching hospital emergency service, may limit the generalisability of our findings. In addition, the pathways through the STB spectrum were described from the perspective of the PER and may not be straightforwardly generalisable to the general population. Variables such as family history, hopelessness or history of NSSI were not systematically collected in a structured way and thus not considered.

Furthermore, as NSSI was not a distinct class of reasons for referral (NSSI was added to the semi-structured interview as a separate category in February 2020), we were unable to differentiate STBs from NSSI. Although it is unclear how NSSI was classified by clinicians throughout the two decades of data collection, additional analyses on records since 2020 revealed that only 1.63% (95% CI: 1.32–1.97%) of referrals to the PER were due to NSSI. Its clinical relevance is unequivocal – indeed, that is the reason NSSI was added as a separate category in 2020 – and it may be critical in prognosis and outcome. We assume that the main findings of the present study would not be substantially different if NSSI had been included as a separate category. In any case, the potentially specific transition course of NSSI will form the object of further study on the more recent referrals, as more data become available.

In the same vein, the linkage of suicide ideation and plan means that distinctions between an impulsive suicide attempt versus those planned were more difficult to identify in the data. Against the rush of a large emergency room, it is clinically difficult to make clear distinctions between ideation and plan. Similarly, data for completed suicides or estimates of such based upon the data were unavailable. From a clinical and scientific perspective, these should form the focus of further studies.

In addition, although missing data strategies such as multiple imputation techniques were implemented, imputations were valid to the extent that the missingness is at random. The last limitation is that our study focused on patients aged >17 years, potentially excluding the most vulnerable group for STB onset and transition.^{Reference Cwik, O'Keefe and Haroz32}

The main finding of the present study was that in the context of the PER, STB is a persistent condition, showing a clear progressive pattern in subsequent referrals; moreover, the time between these referrals can be extremely short. In addition, we established that at least for a subgroup of patients referred to the PER with depressive symptoms, there was a risk of transitioning to STB. From these findings, three clear recommendations follow. First, immediate risk assessment is essential to identify patients in the emergency room at risk of self-harm or suicide; staff training and education may ensure that emergency room staff will be able to recognise and manage patients who are suicidal when they are referred to the PER. Although we did not examine specific factors associated with persistence and progressive transitioning in this study, we did establish that they represent an unambiguous risk. Second, given that progressive transitions are likely and can occur in a short time frame, follow-up and continuity of care for these patients is essential.

Third, suicide prevention efforts in the emergency room should not be limited to building capacity for what is commonly referred to as ‘a suicide risk assessment’, that is, an assessment of suicide ideation, plans, attempts or NSSI.^{Reference Betz and Boudreaux33} Given our findings, staff should also be aware that depressive symptoms (even in the absence of any STB) may lead to a STB referral within a month, indicating the need for continuity of care and more elaborate follow-up of these patients. Indeed, early intervention in STB is a key challenge, and safety planning should be initiated before the actual occurrence of STB. Unfortunately, so far, the potential of such early intervention has been limited because there are no specific factors that consistently precede STB.^{Reference Walsh, Sheehan and Liu34} A more refined assessment that goes beyond merely establishing depressive symptoms may lead to the early identification of patients at risk of entering a pathway to STB. Potentially, such an assessment could, in combination with more recent developments in machine learning and artificial intelligence – with prediction models that can access the entire medical and psychiatric history of patients and allow for complicated non-linear relations and interactions – lead to decision support that could identify cases at risk with more sensitivity and specificity.