Clinical spectrum and risk factors for mortality among seawater and freshwater critically ill drowning patients: a French multicenter study

We conducted a 7 years retrospective multicenter study in 14 French ICUs (3 tertiary hospitals and 11 general hospitals) in the west of France (Additional file 1: Fig. S1). All consecutive adult patients (≥ 18 years old) admitted for drowning from January 2013 to January 2020 were identified through International Classification of Diseases (ICD) coding [13]. Drowning patients were defined as patients that experienced respiratory impairment from submersion or immersion in liquid in accordance with the WHO definition [2]. The ethics committee of the French Society of Intensive Care Medicine (CE SRLF 20–03) approved the study protocol. Informed consent was not required in compliance with French legislation on observational retrospective studies of anonymized data.

For each included patient, a standardized form was used to collect data on demographics, medical history (including the following psychiatric comorbid conditions: depressive disorders, anxiety disorders, bipolar disorders and psychotic disorders [14]). We also collected data on the drowning episode (season of the year, type of water, suspected mechanisms and circumstances, clinical findings at the scene (Coma Glasgow Scale (CGS) score, loss of consciousness, body temperature, cardiac arrest) and on initial management (duration of cardiopulmonary resuscitation (CPR) when performed, CPR before Emergency Medical Service (EMS)). Data collected on ICU admission included clinical parameters (mean blood pressure, pulse oximetry and heart rate) and biological parameters (PaO2, PaCO2, serum sodium level, and leukocyte counts). PaO2 to FiO2 ratio in non-mechanically ventilated patients was calculated by converting O2 flow to estimated FiO2 [15]. Finally, the severity of illness and organ failures were assessed using the Simplified Acute Physiology Score II (SAPS II) [16] and the Sequential Organ Failure Assessment (SOFA) score [17]. Acute respiratory distress syndrome (ARDS) was defined in accordance with international guidelines [18]. ICU clinical course and management data were also collected including the duration of invasive mechanical ventilation (MV), neuromuscular blockers agents and catecholamine use, prone positioning, acute kidney injury defined according to KDIGO criteria [19], renal replacement therapy requirement and pneumonia occurrence defined by the presence of a radiological pulmonary infiltrate persisting for more than 24 h compatible with the diagnosis of pneumonia associated with at least 3 of the following signs: Positive microbiological respiratory samples, purulent secretions, body temperature > 38 °C without other cause and leukocytes < 4000/mm³ or ≥ 10,000/mm³. Neurological status at hospital discharge or at day 28 when patients were still hospitalized was assessed by using the Cerebral Performance Category (CPC) score [20]. Finally, hospital survival status until day 28 was recorded.

Continuous data are reported as median [interquartile ranges (IQRs)] and categorical variables as number (%). Survival rates were established by the Kaplan–Meier method and compared by the log-rank test. For univariate analysis, patients’ characteristics were compared using Mann–Whitney test for continuous variables and the Fisher’s or the Chi-square test, when appropriate, for categorical variables. Regarding survival analysis, covariables achieving a p value < 0.1 in the non-adjusted analysis, with no more than 10% missing data, were entered in the adjusted analysis (Age, alcoholism, respiratory disease, drug use, presumed cardiac etiology for drowning, winter or summer seasons, cardiac arrest occurrence, CPR duration, GCS, temperature, loss of consciousness, event witnessed, resuscitation before EMS, PaCO2, invasive mechanical ventilation, SAPS2 and SOFA score). A multiple backward stepwise selection procedure eliminated those variables with an exit threshold set at p = 0.05. Interactions between variables were checked. To handle missing values as potential confounders, missing data were imputed using a multiple imputation with chained equations. Results are expressed by hazard ratios (HR) with their 95% confidence interval (CI). All statistical analyses were two-sided, and P values less than 0.05 were considered statistically significant. Analyses were performed using R software version 4.0.4 (https://​www.​rproject.​org).

Over the study period, 270 patients were admitted to ICU for drowning in participating ICUs of the west of France. Baseline characteristics of patients are listed in Table 1. Median inclusion number per center was 18 (IQR: 14–22). Patients were mainly male 161 (59.6%) with a median age of 68 (54–75) years and 72 patients (26.7%) had at least one psychiatric comorbidity. The presumed etiology of drowning was accidental for 151 patients (55.9%), a suicide attempt for 26 patients (10.1%). Drug or alcohol intoxication and a presumed cardiac origin were, respectively, observed in 30 and 48 patients. The overall day-28 mortality was 20.4% (55/270).

Drowning occurred in seawater for 199 patients (73.7%). When comparing baseline characteristics of the patients according to the salinity of the water, freshwater drowning patients were younger and suffered more often from psychiatric comorbidities (47.9 vs. 19.1%; p < 0.0001). The etiology of drowning appeared more frequently to be a suicide attempt in freshwater drowning patients (4.2 vs. 25.7%; p < 0.0001) (Table 1). As shown in Fig. 1, seawater drowning occurred more frequently during summer (79.4 vs. 38%; p < 0.0001), while a higher proportion of freshwater drowning occurred during winter and spring (respectively 22.5 vs. 5%; p < 0.0001 and 33.8 vs. 13.6%; p = 0.0004).

Freshwater drowning patients were more severe at the scene and in ICU than saltwater drowning patients (Table 1). They had more often an initial cardiac arrest (50.7 vs. 33.7%; p = 0.017), longer CPR and deeper conscious impairment at the drowning scene and, as a consequence, required more often mechanical ventilation (MV) (60.6 vs. 32.2%; p < 0.0001). SOFA score at ICU admission was higher in freshwater drowning patients (7 vs 2; p < 0.0001). A loss of consciousness was also more often observed in freshwater patients (83.1 vs. 69.8%; p = 0.04), and the events were less frequently witnessed in this population. When excluding patients that undergo an initial cardiac arrest, patients drowning in freshwater also appeared to be more severe (Additional 1: Table S1).

Freshwater drowning patients had worse CPC scores at hospital discharge and a higher 28-day mortality than saltwater drowning patients. Survival curves comparing seawater and freshwater drowning patients are represented in Fig. 2. Risk factors for 28-day mortality in the univariate analysis are presented in Table 2. By multivariate Cox regression, freshwater was found to be independently associated with 28-day mortality (adjusted Hazard Ratio (aHR) 1.85 [95% Confidence Interval (CI) 1.02–3.39], p = 0.04). The following variables were also independently associated with 28-day mortality: Occurrence of a drowning-related cardiac arrest (aHR 11.5 [95% CI 2.51–52.43], p = 0.0017), duration of cardiopulmonary resuscitation (aHR 1.05 [95% CI 1.03–1.07], p < 0.0001) and SOFA score at day 1 (aHR 1.2 [95% CI 1.11–1.3], p < 0.0001) (Table 3). Noteworthy, in freshwater drowning patients, mortality at day 28 appeared lower among patients that drowned in pools, while we observed higher mortality rates in patients that drowned in ponds (respectively, HR 0.19 [95% CI 0.06–0.64], p = 0.007 and HR 2.31 [95% CI 1.06–5.05], p = 0.03) (Additional 1: Table S2).