This was a nationwide population-based cohort study including all calls to the national emergency number (1-1-2) regarding patients aged ≤ 15 years followed by dispatch of either ground or helicopter EMS. The study period was 1 January 2016 to 31 December 2021 plus 30 days of follow-up. Interhospital transfers were excluded. Reporting of the study follows the ‘Strengthening the Reporting of Observational Studies in Epidemiology’ guidelines [15].

The Danish EMS serve 5,806,081 inhabitants in mixed rural, semi-rural, and urban areas (ground area 42,962 km2). Children aged ≤ 15 years comprised 18% of the population amid the study period (per 1 January 2019) [16]. Emergency medical dispatch centres are staffed by health professionals who manage the dispatch of both ground and helicopter EMS units according to the level of urgency (A-E, A = lights and sirens, E = telephone counselling only). All dispatch centres use the criteria based decision support tool Danish Index for Emergency Care [17]. The health professionals who answer emergency calls use the index to question the caller to decide on an appropriate main reason for the call and urgency level (i.e. dispatch criterion), which then trigger a certain response. For each call, the patient can only be labelled with one criterion. Ambulances and rapid response vehicles, staffed by emergency medical technicians and paramedics, can be assisted by 24-hour operative physician-staffed mobile emergency care units and/or physician-staffed helicopter EMS. Generally, a physician-staffed unit is engaged in about 20–25% of ambulance dispatches. Further details of the organisation of the prehospital system in Denmark have been described in Supplementary Material 1 and in previous works [18].

All calls to the national emergency number (1-1-2) regarding patients aged ≤ 15 years were collected from. The Danish Quality Database for Prehospital Emergency Medical Services under The Danish Clinical Quality Program – National Clinical Registries. Reporting to the registry is mandatory for all Danish EMS [19]. The Danish Civil Registration System contains a personal identification number (PIN) for all residents in Denmark and holds information on migration and vital status [20]. The Danish National Patient Register [21] contains data from all inpatient and outpatient hospital visits and visits to private practitioners in all medical specialities, including ICD-10 diagnoses [22]. Registries were linked using PINs and timestamps of the first ambulance that had arrived at a hospital. For details about the record linkage process, we refer to Supplementary Material 1.

Patients without a valid PIN could not be identified as children of a certain age and were therefore not eligible for the study. Patients with missing PINs might introduce a selection bias, as registering a PIN on the medical record might not be prioritised in critical emergencies. Therefore, we crosschecked mortality with selected variables from prehospital electronic patient medical records for all patients regardless of whether a PIN was present (specific variables are designated in Table 1). Consequently, we were able to investigate the direction and magnitude of this possible selection bias in our mortality analysis.

Exposures were reasons for emergency call and the outcome was 7-day mortality, and outcomes were presented stratified by comorbidity subgroups. Dispatch criteria were used to define reasons for emergency call. Criteria were grouped into:

1)‘Trauma’ if criteria were:

Large scale accident

Fire or electrical injury

Drowning

Diving accident

Hypothermia – Hyperthermia

Traffic accident

Accidents (not traffic-related)

Minor wounds – fractures – injuries

Violence – abuse

2) ‘Paediatric out-of-hospital cardiac arrest (POHCA)’ if criteria were:

Unconscious adult (after puberty) or Unconscious child (before puberty)

Plus

The patient had received cardiopulmonary resuscitation from either a bystander or from healthcare professionals outside of a hospital according to current reporting standards [23, 24].

3) ‘Suspected death’ (a specific dispatch criterion that describes the ‘finding of a lifeless person’).

4) ‘Medical symptoms’ if criteria were any of the remaining criteria in the Danish Index for Emergency:

Care [17]: ‘Allergic reaction’, ‘Fever’, ‘Poisoning in children’, ‘Headache’, ‘Breathing difficulties’, ‘Psychiatry – suicidal’, ‘Abdominal pain – back pain’, ‘Seizure’, ‘Sick child’, ‘Foreign body in airway’, ‘Ordered mission’, ‘Unclear problem’, ‘Bleeding – non-traumatic’, ‘Chest pain – heart disease’, ‘Diabetes’, ‘Animal and insect bites’, ‘Childbirth’, ‘Gynaecology – pregnancy’, ‘Skin – rash’, ‘Chemicals – gases’, ‘Impaired consciousness – paralysis – vertigo’, ‘Alcohol – poisoning – overdose’, ‘Urinary system’, ‘Ear – nose – throat’ or ‘Eye’.

5) ‘Missing criteria’.

The association between the reasons for emergency call and the outcome 7-day mortality was stratified by comorbidity subgroups. The subgroups were formed based on clinical experience, previous literature [25, 26] and publicly available information on the expected prognosis for each of the diseases and conditions in the ICD-10 classification system. An example of publicly available information was an international web portal for rare diseases, www.orpha.net.First we downloaded a list of all ICD-10 diagnoses and then manually reviewed each one and put them into one of three categories: ‘Comorbidity’ (previous or chronic disease, condition, or perinatal complication), ‘Severe chronic comorbidity’ (diseases or conditions with potentially considerably reduced lifespan) or ‘None’. If patients had not had any visits in those five years, i.e. no previous diagnoses, they were classified as comorbidity subgroup ‘None’.

We then searched all inpatient and outpatient hospital visits and visits to private practitioners from the five years preceding each emergency call for all of the unique patients.

The average number of visits was calculated to underline disease severity in these predefined subgroups. Patients with ‘Severe chronic comorbidity’ had had median (IQR) 88 (45–164) visits in the five years preceding their last emergency call while patients in the ‘Comorbidity’ subgroup had had 16 (9–31) visits.

Summary statistics were used for the grouped exposure, reasons for emergency call. For mortality analysis, patients were considered at risk from the date of emergency call until either death or the end of follow-up. Patients were lost to follow-up if they had no current address in Denmark, PIN was annulled, patient disappeared or travelled abroad as indicated in the Civil Registration System [20]. No data were imputed. All cause mortality estimates were calculated using last-time events only by modified Poisson regression with robust variance estimation [27].

A unique patient could have had more than one emergency call during the study period. These ‘repeat calls’ could influence the mortality analyses, which is why we performed sensitivity analyses using first-time events only and all events, respectively. This means that mortality is reported both at a mission level, i.e. the denominator was number of emergency calls (= missions), and at an individual patient level, i.e. the denominator was unique patients.

In a Cox proportional hazards model with time since inclusion as time scale, we estimated 7-day mortality between the dispatch criteria categories mentioned above in each of the three comorbidity subgroups. Assumptions of proportional hazards were assessed visually by a log-minus-log plot and using Schoenfeld residuals. Hazard rate ratios were reported as crude and adjusted for sex (dichotomous, no missing data) and age (categorical, no missing data) with ‘Trauma’ as reference. Age was grouped according to the Danish Regions’ Paediatric Triage Model used by all of the Danish EMS. Two-sided p values < 0.05 were considered statistically significant. Statistical analyses were performed with Stata/MP 17.0 (StataCorp LLC, TX 77845, USA).