Treat and release: an observational study of non-conveyed high-acuity dispatches in a Danish emergency medical system

This study was a population-based retrospective cohort study with a 30-day follow-up. Data were derived from the emergency medical dispatch centre (EMDC) on severe and immediate high-acuity calls of category A (lights and sirens) between January 1, 2019, and September 30, 2020, in Region Zealand, Denmark.

Region Zealand covers an area of 7274 km² and has approximately 840,000 inhabitants. The population density is 116 inhabitants/km². The Danish healthcare model is funded by direct and indirect taxes. The Danish EMS is organized in a three-tiered system: an ambulance staffed with two paramedics, a rapid response vehicle with a paramedic and a physician-staffed ground-based or helicopter-based mobile emergency care unit [14, 15]. The EMDC is staffed by medical professionals, predominantly nurses and paramedics. Medical dispatchers employ a criterion-based decision tool to dispatch prehospital resources and determine the urgency of the response [16]. The decision tool is divided into chief complaint categories; each categorized into more specific complaint groups and assigned a unique code. There are four levels of emergency that will require ambulance dispatch: A: Life-threatening or potentially life-threatening condition, immediate response required (high-acuity dispatch with lights and sirens), B: Urgent, but not life-threatening condition, C: Non-urgent condition that needs ambulance care, D: Non-urgent supine patient transport. When paramedics encounter a patient suitable for non-conveyance, the medical control physician is always contacted to confer and consolidate the decision to let the patient stay at home. If the physician at the control centre disagrees, the patient must be transported to a hospital for evaluation.

All patients assessed following a high-acuity dispatch were eligible for inclusion encompassing all age groups. Patients without valid personal identification numbers were excluded, as this was necessary for coupling with other registries [17]. All Danish citizens are assigned a unique personal identification number [17, 18]. Software from EMDC provided data on high-acuity calls within the study period.

Data on the personal identification number, day of week and time of day of the incident, ambulance response time, on-scene time and Danish Index criteria code were collected. Calls where the prehospital care provider chose not to convey a patient were manually identified. Additionally, non-conveyed calls were cross-referenced with the Danish National Patient Registry using the patients’ unique personal identification number [17, 18] to ascertain any subsequent hospital contact. Discharge diagnoses registered at the hospital contact and admissions to the hospital were coupled to the cases. Additionally, patients admitted within 3 h, regardless of conveyance, were identified, along with patients discharged within six hours from admission were identified. Finally, mortality was recorded from the Civil Registration System Registry [18]. Patients were considered admitted once they had a contact in the hospital, whether this was an ED contact where the patient was under evaluation or admitted to a specialised unit for a longer period of time.

The primary outcome was admission to the hospital within 48 h following dispatch. The secondary outcome was 30-day mortality. Further, admission within three hours and demographic characteristics among the included patients were assessed.

Descriptive statistics were reported as absolute numbers and percentages or medians and interquartile ranges. Comparative analyses were performed using non-parametric testing to examine subgroups as appropriate. Categorical data were analyzed using Fisher’s exact test. Multivariable regression models were used to estimate adjusted odds ratios (aOR) and corresponding 95% confidence intervals (CI). The analyses were adjusted for prognostic factors, including mortality, admission within 48 h, response time, age, sex, and chief complaints. A directed acyclic graph was provided to illustrate the basis of the included variables [19]. Patients who were terminally ill or sustained cardiac arrest were included in all analyses. Age was stratified according to 30-year intervals. The chief complaint was stratified into categories with > 1000 patients. Response time was defined as the time from the initial call to the dispatch centre to the arrival of the first EMS personnel. There was no imputation of missing data. Statistical significance was considered at a two-sided p-value of <0.05; all statistical tests were performed using R version 4.1.3 (2022-03-10).

Non-conveyed patients were significantly younger than conveyed patients, with a median age of 55 years [interquartile range (IQR) (31–74)] compared to 65 years [IQR (44–77)] (p < 0.001). Furthermore, non-conveyed patients experienced a shorter response time until the ambulance arrived, with a median time of 7:04 min [IQR (4–11)] compared to 7:48 min [IQR (4–11)]. Males accounted for 55% of both groups, n = 7333 in the non-conveyed group compared to n = 45,157 in the conveyed group. There was no significant difference between males and females regarding conveyance decisions, p = 0.17. The on-scene time was significantly longer for non-conveyed patients, with a median time of 35:00 min [IQR (25–48)] compared to 22:09 [IQR (16–22)] for conveyed patients, p < 0.001. The most common chief complaint was “chest pain”, which accounted for 2972 cases (22%) of non-conveyed patients compared to 19,147 cases (23%) of conveyed patients p < 0.001. In 1549 cases (12%) of non-conveyed patients, “impaired consciousness” was the dispatch criterion as compared to 13,129 (16%) of the conveyed patients, (p < 0.001). “Unclear problems” accounted for 2565 cases (19%) among non-conveyed patients compared to 11,048 cases (14%) among conveyed patients, p < 0.001 (Table 1).

Non-conveyed patients had a crude 30-day mortality rate of 2% compared to 6% among conveyed patients, p < 0.001. Admission to the hospital within 3 h of the primary contact was observed in 11% (1507 cases) of the non-conveyed patients. In the cohort of the initially conveyed patients, 95% of the patients (77,310 patients) were still admitted after three hours following their initial contact, p < 0.001. Admission within 48 h was seen in 3010 cases (22%) of non-conveyed patients compared to 77,853 cases (95%) in conveyed patients, p < 0.001 (Table 2).

Within the multivariable analysis, non-conveyed patients showed a decreased likelihood of presenting with potentially fatal symptoms (aOR 0.22 [95% CI 0.19–0.26)] or presenting symptoms related to psychiatry and suicide, aOR 0.29 (95% CI 0.23–0.37). Non-conveyed patients were less likely to be unconscious, aOR 0.51 (95% CI 0.43–0.61). Non-conveyed patients showed an increased likelihood of presenting with unclear symptoms aOR 1.29 (95% CI 1.14–1.46). Additionally, the risk of mortality was increased within the non-conveyed patients, aOR 1.21 (95% CI 1.05–1.40) and finally, non-conveyed patients showed a decreased likelihood of admission within 48 h from the initial call, aOR 0.01 (0.01–0.01) (Fig. 2).

Being categorized by the EMDC as having “Respiratory symptoms” and/or “unclear symptoms” was associated with an increased likelihood of mortality, aOR 2.75 (95% CI 1.88–4.03) and aOR 1.74 (95% CI 1.20–2.52). Similarly, male sex was associated with an increased likelihood of mortality aOR 1.37 (95% CI 1.06–1.77). The strongest predictor for mortality was admission within 48 h, aOR 7.73 (95% CI 5.94–10.14) (Fig. 3).

Patients aged 61 to 80 were most likely to be admitted within 48 h, aOR 1.19 (95% CI 1.04–1.36). Additionally, male sex was associated with an increased likelihood of admission within 48 h aOR 1.10 (95% CI 1.01–1.20). On the contrary, patients presenting with “altered mental status” or “paralysis”, “chest pain”, or “unclear symptoms” were less likely to be admitted at any time within 48 h. Respiratory symptoms did not contribute significantly to this model (Fig. 3).

In total, 3147 cases (65.4%) of the initially non-conveyed patients were discharged within 6 h from admission compared to 26,979 cases (33.4%) of the conveyed patients, p < 0.001. Furthermore, 50% of initially non-conveyed patients were admitted for 4 h or less. In comparison, 50% of conveyed patients were admitted for 11 h or less (Fig. 4). The initially non-conveyed patients, who were later admitted, had a median length of stay (LOS) of 4.4 h (IQR 1.1–20.2) in contrast to the median length of 12.1 h for conveyed patients (IQR 4.2–20.2), p = 0.001.

The most frequent chief complaint among admitted patients was “chest pain” which accounted for 19,184 cases (20.1%). “Impaired consciousness” accounted for 13,267 cases (13.9%) and unclear problems for 11,019 (11.6%). With respect to discharge diagnoses, the most frequent were symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified, followed by injury, poisoning and certain other consequences of external causes accounting for 24,992 cases (30.9%) and 11,970 cases (14.8%). The third most frequent discharge diagnosis was cardiovascular disease, with 11,185 cases (13.8%). One-fourth of patients presenting with chest pain as the chief complaint received a discharge diagnosis of cardiovascular disease, while approximately half of the patients with respiratory problems as their chief complaint received a discharge diagnosis of respiratory disease (Supplementary Fig. 1).

A strength of this study was the large sample size and population-based study design. Further, the comprehensiveness of the employed registries ensured thorough follow-up, mitigating the potential for selection bias. Additionally, equitable availability of prehospital health care services ensures that the process of inclusion remains unaffected by variations in socioeconomic status. The latter may pose implications for the decision of non-conveyance, and future research could benefit from adjusting for this potential confounder. While the data for this study were comprehensive, the fact that the reason for non-conveyance was not available is a limitation. Further, Region Zealand EMS operates with trained paramedics and physician support available for conferences prior to non-conveyance decisions, which may limit generalizability to regions with different EMS competencies available. Caution is advised when extrapolating findings beyond Denmark, highlighting the need for further research. Further, the non-randomized design represents an inherent susceptibility to selection bias, potentially influencing the extent to which the findings can be extrapolated. In this present study, it was not possible to clarify the causes behind non-conveyed decisions; in addition to this, there was no information regarding patients’ perspectives on non-conveyance. Within the group of initially non-conveyed patients, 11% were admitted within 3 h of the non-conveyance decision. Unfortunately, it was beyond the scope of the present study to evaluate this subgroup. However, it could have been relevant to investigate the means of transport in this group. Additionally, this study only investigated high-acuity dispatches, which challenge the generalisability to the general population.