Study design and setting

We conducted a cross-sectional study in Lyon, France, using data obtained from the departmental EMCC during the period from January 1, 2019, to December 31, 2019. The corresponding geographical area has a population of 1.9 million inhabitants, and an average of 700,000 calls are managed annually.

Prehospital EMS in France operate on a 24-hour physician-led system. Access to the EMCC is available nationwide through a single free national telephone number “15”. Initial call receptions are handled by nonphysician professionals known as assistants, whose purpose is to promptly identify immediate life-threatening cases and collect essential information, including patient identity and location. Subsequently, calls are transferred to one attending physician, who may be a general practitioner or an emergency physician, depending on the initial severity assessment. In comparison to other nations where dispatcher decisions are guided by standardized protocols like the “Medical Priority Dispatch System”, in French EMCC, the decision-making process relies primarily on the judgment of physicians.

In cases where no critical situations are identified, a paramedic ambulance may be dispatched to the scene to implement basic life support (BLS). This includes administering oxygen through a cannula or mask but does not extend to advanced procedures or the use of medication, even if a critical condition such as cardiac arrest unexpectedly arises. In such instances, the paramedics will relay a situation report to the EMCC to seek further support. In contrast, both a paramedic ambulance and a MMT are systematically dispatched to the scene for suspected life-threatening cases. The MMT consists of an emergency physician or an anesthesiologist-intensivist physician, a nurse, an ambulance driver, and a medical resident in academic centers. MMTs can be transported by ground ambulance or helicopter and are strategically distributed throughout the country at hospital-based locations (Fig. 1).

Following a basic clinical evaluation, both FB and paramedics are required to contact the dispatching physician for further decision-making. They determine whether the patient is stable and suitable for transportation or if additional assistance is needed. In cases requiring additional assistance, a MMT may be dispatched later (delayed MMT group). Similarly, in cases where immediate MMT dispatch is warranted, the on-scene physician and the dispatching physician collaborate to determine the most appropriate healthcare facility for patient(s) referral.

Population

We included consecutive calls related to patients aged 18 years or older who were labeled as experiencing “acute respiratory distress” and recorded as such in the electronic medical chart record system. We excluded calls that were not handled by a physician dispatcher and those that did not result in the dispatch of an MMT. Additionally, patients who did not require or refuse transportation, those admitted to private centers (because of limited data accessibility), and those identified in cardiac arrest upon the first healthcare provider arrival were also excluded from the analysis. Patients with missing data related to a variable of the propensity score, or outcome were also excluded. Finally, MMT interventions conducted within healthcare facilities were excluded from the analysis due to the potential confounding effects of preliminary treatments initiated prior to MMT arrival, which could impact patient outcomes. The included patients were subsequently divided into two groups based on whether they received immediate MMT dispatch at the time of the initial emergency call or delayed MMT dispatch after a paramedic team evaluation and review.

Outcomes

The primary outcome was mortality on day 0, while the secondary outcomes were mortality on day 7 and 30.

Variables

Patient demographics, comorbidities, prehospital clinical findings, timings, prehospital management, and patient outcomes were extracted from patient electronic medical charts by a post-graduate year 3 emergency medicine resident (LC). Comorbidities were divided into four categories: cardiovascular (chronic high blood pressure, chronic heart failure, ischemic cardiomyopathy, arrhythmia), neurological (past history of stroke), pulmonary diseases (chronic obstructive pulmonary disease, asthma, pulmonary embolism, pulmonary fibrosis) and active cancer. Patients level of dependency was also collected based on the AGGIR (Autonomie Gérontologie Groupes Iso Ressources) scale from 1 to 6 [14, 15]. High-level of dependency patients were defined as GIR 1 and 2 patients. GIR 1 pertains to individuals confined to bed or chair, with severely impaired cognitive functions, requiring the continuous presence of a caregiver or end-of-life individuals. GIR 2 corresponds individuals confined to bed or chair, whose cognitive functions are not entirely impaired, and whose condition demands assistance for most daily activities or individuals with impaired cognitive functions, but who can move around and require constant monitoring. Vital status was obtained from the French death registry.

Statistical analysisDescription and comparison of the two populations

Quantitative variables were expressed by their medians and interquartile ranges (IQR). Qualitative variables were expressed by their frequencies and percentages. Patient characteristics were compared between the two populations (immediate MMT and delayed MMT) using the Wilcoxon rank sum test for quantitative variables, while Pearson’s chi-squared test or Fisher’s exact test was used for qualitative variables. Time to event data was presented using Kaplan-Meier survival curves.

Calculation of the propensity score and the matching method

Patients from the two groups were matched on a propensity score to mitigate bias caused by confounders. A logistic regression model with all second-order interactions was used to estimate the propensity score. A total of 7 covariates were selected a priori according to their clinical relevance: patient’s age, permanent residency in care home, comorbidities: cardiovascular disease, pulmonary embolism, chronic obstructive pulmonary disease, asthma, and distance between the patient location and the MMT base. For the propensity score, the dependent variable was the delay before MMT dispatch defined as delayed or immediate MMT dispatch. A 1:1 optimal propensity score matching without replacement was conducted.

Bias reduction through propensity score matching was assessed by calculating standardized absolute mean differences (SADs) in each baseline characteristic between the two populations. An SAD of less than 0.1 was considered acceptable to indicate a negligible difference between the two populations.

Evaluation of outcomes

For each outcome (mortality on day 0, 7 and 30), we estimated the effect size of immediate MMT (compared to delayed MMT) using parametric g-formula [16]on the matched sample. For each outcome, a logistic regression model was fitted on the matched dataset, adjusted for the covariates used to conduct the matching procedure. Then, we predicted the counterfactual outcomes for each subject (two predictions per individual: with delayed MMT and immediate MMT). The final estimate was the mean of individual level effect sizes. Standard errors were computed using the delta method. All analysis were conducted in R 4.1.0. Matching was conducted using the MatchIt package [17]with “optimal [18, 19]” setting, parametric g-formula computation was conducted using the marginal effects package.

Sensisitivy analyses

Mortality at day 0 was evaluated across various sub-populations of the study to assess the sensitivity of the results to deviations from positivity (patients for whom it is highly unlikely that a dispatch occurred at the call), which could lead to a biased estimation of the estimated effect.

The selection criteria (age > 85, high level of dependency [GIR 1 or 2 patients]) were applied before the matching procedure. Relative risks were estimated using the same procedure as described for the main analysis.