Factors related to 30-day survival rate following accidental hypothermia – a retrospective single-center study from Northern Finland

Abstract

Background

Accidental hypothermia is a rare condition, and it is reported to cause 2–5 deaths per 100,000 in European studies and at least 1,500 deaths in the United States annually. There is limited evidence concerning concomitant factors, clinical findings, prehospital care, and treatment methods related to patients who experience accidental hypothermia.

Methods

This is a retrospective registry study with data consisting of patients over 18 years of age who were admitted to Oulu University Hospital in Finland with accidental hypothermia. Data were collected for 12 years (2008–2019). Patients were divided into survivors and non-survivors according to their 30-day survival.

Results

Of 315 initial patients, 241 were included in the final analysis: 206 survivors and 35 non-survivors. Alcohol abuse was the most common concomitant factor (49%), followed by age ≥ 65 years (42%). Indoor exposure was the most common etiology among non-survivors while outdoor exposure was most common among survivors (37% vs. 64%, p < 0.001). Daytime EMS dispatch was associated with higher mortality than nighttime dispatch (18% vs. 3%, p = 0.007). Lower recorded temperatures on the scene (27.0 °C vs. 30.1 °C) and in the emergency department (27.4 °C vs. 31.9 °C) were associated with decreased survival rate (p < 0.001 for both). Overall 30-day mortality was low (15%) and survival after cardiac arrest was high (47%). According to multivariable logistic regression analysis for concomitant factors and hypothermia etiology, submersion (odds ratio 0.27, p = 0.038) and trauma (odds ratio 0.32, p = 0.084) decreased the chance of 30-day survival.

Conclusions

Alcohol abuse and age ≥ 65 years were two of the most common concomitant factors for accidental hypothermia. Non-survivors were generally older and more likely to be found indoors. They were most often found during daytime hours, and their recorded temperature was generally lower. Submersion and trauma were independent risk factors for mortality.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13049-025-01491-3.

Background

Accidental hypothermia is a condition where the core temperature has involuntarily dropped to < 35 °C. Accidental hypothermia can be divided into primary hypothermia and secondary hypothermia. In general, primary hypothermia may develop when a healthy person is exposed to a cold environment. Wet and windy conditions and poor cold weather preparedness during outdoor activities increase the risk of primary hypothermia. Secondary hypothermia may be triggered by other causes than a cold environment and occurs typically in patients with chronic illnesses or other causes (e.g., stroke, infection, hypoglycemia, traumas, and alcohol or drug abuse) and typically indoors. The risk of secondary hypothermia is increased when thermoregulation is impaired, as in the case of elderly people, children, and patients with multimorbidity [1, 2].

Finland is a Northern European country with 5.6 million inhabitants. According to Finnish government official statistics, during the study period of 2008–2019 there were 847 (average 71 per year) reported deaths from accidental hypothermia and 172 (average 14 per year) deaths caused by drowning in cold water in Finland, classified in accordance with the International Classification of Diseases (ICD-10). The majority of deaths due to accidental hypothermia (69%) and drowning in cold water (91%) were associated with male sex. Death certificates indicated alcohol abuse was a concomitant factor in 37% of hypothermia deaths and 16% of deaths caused by drowning in cold water [3].

Many studies in various populations have reported that the incidence of accidental hypothermia is low, but the mortality rate is significant varying from 24.5% to 50.0% in the general population and from 39.0% to 62.5% in patients with hypothermic cardiac arrest (CA) [4–9]. The mortality rate of accidental hypothermia has increased in patients with multiple comorbidities and reduced ability to perform activities of daily living [8, 9]. In trauma patients, hypothermia can be a significant risk factor for mortality and complications, as it causes hemodynamic instability and impairs blood coagulation [1]. Hypothermic CA combined with asphyxia has been associated with higher mortality in submerged patients and avalanche victims [2, 4, 6]. However, if hypothermia develops before hypoxia and CA, and if the chain of survival functions optimally, hypothermic patients in CA have good chances of favorable recovery [2].Despite the significant mortality rate with accidental hypothermia, many studies have reported good neurological outcomes after hypothermic CA [6, 7, 10].

Hypothermic patient data is highly variable due to the rarity of cases, differences in emergency medical service (EMS) systems, and hospital treatment protocols. Further, there is a lack of evidence concerning concomitant factors, prehospital processes, and treatment methods before rewarming. The aim of this study was to explore the factors associated with hypothermia and overall 30-day survival rate in patients admitted to the emergency department (ED) for accidental hypothermia.

Methods

Study design and settings

This study was performed in Northern Finland at Oulu University Hospital. This is a tertiary-level teaching hospital that currently provides treatment for people in the Wellbeing Services County of North Ostrobothnia and university hospital-level treatment for Northern Finland in general. The population in the local wellbeing services county is approximately 416,000, and a total of 730,000 in Northern Finland [11]. The annual number of ED visits is approximately 29,800. The 609-bed hospital has a 34-bed mixed adult intensive care unit (ICU) [12]. Oulu University Hospital had no treatment protocol for accidental hypothermia during the study period and each patient was treated according to the attending physician’s judgment and national intensive care guidelines.

The EMS in the area is a four-level system including first responder units, basic- and advanced-level ambulances, and a physician-staffed helicopter emergency medical service (HEMS). All EMS units in Finland are dispatched via the emergency response center (112 Finland). During the study period, the EMS had their own specific national guidelines for treatment of accidental hypothermia, and also adhered to the national guidelines for resuscitation, including the resuscitation protocol for hypothermic patients [13, 14].

The study was approved by the Department of Operative Care of Oulu University Hospital (No. 152/2020). Due to the retrospective study design, an exemption from consent was obtained from the hospital ethics committee; the data had already been collected for clinical purposes and none of the patients were contacted during this study.

Study population

Patients included in the study were those classified according to the ICD-10 codes for hypothermia (T68) and drowning (T75.1) in the 12-year period between 1 January 2008, and 31 December 2019. All patients whose first recorded temperature on the scene or at hospital admission was below 35 °C were included in the study. Criteria for exclusion were age under 18 years, treatment withdrawal or living will, and hypothermia without external factors. Hypothermia without external factors is defined as a secondary hypothermia caused by internal factors that impair heat production or increase heat loss. Patients were classified as survivors and non-survivors according to 30-day survival.

Study parameters and data collection

This was a retrospective registry study, and the clinical data were collected from the hospital electronic patient record system (ESKO) and EMS documents (electronic and paper). We used a structured template for data collection and later transferred the information into digital form.

Collected patient data included the following demographic variables: age, sex and chronic illnesses (coronary heart disease, arterial hypertension, ischemic or hemorrhagic cerebral disease, chronic obstructive pulmonary disease, rheumatism, chronic kidney disease, chronic liver disease, epilepsy, degenerative brain disease, psychiatric disease, and alcoholism). A list of predetermined concomitant factors (drunkenness, intoxication with an alcohol substitute, medications or drugs, age ≥ 65 years, trauma, hypoglycemia, and psychiatric conditions) was created according to the findings of an earlier study [15]. Circumstance of hypothermia (outdoor exposure, indoor exposure, immersion, or submersion) was also collected. For this study immersion was defined as a patient floating in water without asphyxiation and submersion as floating in water with asphyxiation.

The first recorded temperature, Glasgow Coma Scale (GCS), arrhythmias, and CA before and after admission were collected. Core temperature measured from the esophagus, nasopharynx, or rectum was primarily used if available, otherwise the peripheral temperature from the eardrum was registered. Temperature was recorded as a numerical value if available and categorized as low if the temperature was below the thermometer’s measurement range. The stage of hypothermia was classified according to the first measured temperature as mild (35–32 °C), moderate (32–28 °C), or severe (< 28 °C). The first laboratory values were recorded at the emergency department and collected values included potassium, platelets, lactate, blood ethanol level, and arterial blood gases (pH, PaCO₂, PaO_2, and base excess [BE]) using pH–stat method.

Collected EMS data included dispatch time, type of transportation vehicle (ambulance, helicopter, or other), transport time and first transport destination, if available. The treatment methods recorded were intubation (lifeless and unconscious), resuscitation, CPR time (before invasive warming), the invasive warming method used, and the extracorporeal life support (ECLS) running time. The number of ICU admissions was also recorded.

Statistical analysis

The design of this study was descriptive and observational. The 12-year data collection period was chosen due to the low incidence of hypothermia. Two researchers manually checked the data twice to reduce human error. Relevant variables were analyzed using SPSS (IBM SPSS statistics version 28). Results are presented as counts with percentages, means and standard deviations (SD), or medians with interquartile range (IQR, 25th – 75th percentiles), depending on the variables. Missing values were treated as missing. The categorical variables were analyzed using the chi-square test or Fisher’s exact test. The independent samples t-test or Mann–Whitney U test were used for continuous variables. Two-tailed p-values are presented.

Multivariable adjusted logistic regression models were created to separately assess the impact of concomitant factors with more than 10 cases, i.e. hypothermia etiology, drunkenness, intoxication by substitute alcohol, medications or drugs, trauma, and psychiatric conditions on 30-day survival. In all models, age (< 65 vs. ≥ 65 years), sex (male vs. female), and number of diseases (0 vs. 1 vs. ≥ 2) were used as adjusting variables. The results of logistic regression analyses are presented as odds ratios (ORs) with 95% confidence intervals.

Results

Age, illnesses, concomitant factors, and hypothermia etiology

During the 12-year study period there were 358,141 ED admissions. Of these admissions, there were 315 patients identified by ICD-10 codes for hypothermia or drowning. The final analysis included 241 patients. The survivor group comprised 206 patients and the non-survivor group 35 patients. A recorded temperature over 35 °C was the most common reason for patient exclusion from the study (Fig. 1).

Fig. 1.

Flowchart of patients (*1 patient admitted twice due to accidental hypothermia)

Non-survivors were, on average, 6.3 years older than survivors and older age was a statistically significant factor for decreased 30-day survival (p = 0.008). At least one illness was present in 78% of the patients.

At least one concomitant factor was discovered in 87% of survivors and 68% of non-survivors, and this had a statistically significant effect on 30-day survival (p = 0.006). The most common concomitant factor was alcohol abuse (49%), followed by age ≥ 65 years (42%).

Our analysis revealed a significant difference in 30-day survival when comparing different exposure circumstances (p < 0.001). Indoor exposure was the most common circumstance among non-survivors (37%) while outdoor exposure among survivors was most common (64%). The mortalities among different exposure circumstances were as follows: submersion 52%, indoor 28%, outdoor 6% and immersion 6% (Table 1).

Table 1.

Demographic data, concomitant factors and hypothermia etiology categorized according to 30-day survival rate

Characteristic	All patients (n = 241)	Survivors (n = 206)	Non-survivors (n = 35)	p-value1)
Age, mean (SD)	60.1 (17.8)	59.2 (18.5)	65.5 (11.6)	0.008
Sex (male), n (%)	164 (68)	140 (68)	24 (69)	> 0.9
Chronic illnesses, n (%)				0.86
None	53 (22)	45 (22)	8 (23)
1	85 (35)	73 (35)	12 (34)
2	71 (29)	62 (30)	9 (26)
≥ 3	32 (13)	26 (13)	6 (17)
No significant previous medical history	53 (22)	45 (22)	8 (23)	> 0.9
MCC	38 (16)	34 (17)	4 (11)	0.48
HTA	84 (35)	72 (35)	12 (34)	> 0.9
Ischemic or hemorrhagic cerebral disease	22 (9)	18 (9)	4 (11)	0.54
COPD	11 (5)	8 (4)	3 (9)	0.20
Rheumatism	5 (2)	5 (2)	0	> 0.9
Chronic kidney disease	7 (3)	6 (3)	1 (3)	> 0.9
Chronic liver disease	10 (4)	9 (4)	1 (3)	> 0.9
Epilepsy	5 (2)	4 (2)	1 (3)	0.55
Degenerative brain disease	26 (11)	21 (10)	5 (14)	0.55
Psychiatric disease	47 (20)	41 (20)	6 (17)	0.82
Alcoholism	72 (30)	61 (30)	11 (31)	0.84
Concomitant factors, n (%)				0.006
None	37 (15)	26 (13)	11 (31)
1	137 (57)	124 (60)	13 (37)
≥ 2	67 (28)	56 (27)	11 (31)
Drunkenness	118 (49)	110 (53)	8 (23)	< 0.001
Intoxication of substitute alcohol	3 (1)	2 (1)	1 (3)	0.38
Intoxication of medications	16 (7)	15 (7)	1 (3)	0.48
Intoxication of drugs	6 (2)	6 (3)	0	0.60
Age 65 yrs and over, n (%)	101 (42)	80 (39)	21 (60)	0.026
Age 65–79 yrs	65 (27)	48 (23)	17 (49)	0.003
Age 80 yrs and over	36 (15)	32 (16)	4 (11)	0.62
Trauma	12 (5)	8 (4)	4 (11)	0.079
Hypoglycaemia	8 (3)	6 (3)	2 (6)	0.33
Psychiatric reasons	19 (8)	17 (8)	2 (6)	> 0.9
Other factors	23 (10)	19 (9)	4 (11)	0.75
Hypothermia etiology, n (%)				< 0.001
Indoor exposure	47 (20)	34 (17)	13 (37)
Outdoor exposure	141 (59)	132 (64)	9 (26)
Immersion	32 (13)	30 (15)	2 (6)
Submersion	21 (9)	10 (5)	11 (31)

Data (excluding age) are presented as n, % of all patients and % of survivors/non-survivors subgroups

¹⁾ p-values for comparisons between survivors and non-survivors

MCC: Coronary Artery Disease, HTA: Arterial Hypertension, COPD: Chronic Obstructive Pulmonary Disease

In the multivariable adjusted logistic regression model, submersion decreased, and outdoor exposure and immersion increased the likelihood of 30-day survival (Table 4).

Table 4.

Logistic regression analysis for concomitant factors with more than 10 cases and hypothermia etiology for 30-day survival rate. Age (< 65 vs 65- years), sex and number of diseases (0, 1 or 2-) were used as adjusting factors

Characteristic	n	Crude OR	Adjusted OR	95% C.I	p-value
Drunkenness	118	3.87	3.50	1.45 to 8.43	0.005
Intoxication of substitute alcohol, medications or drugs	22	1.77	1.07	0.22 to 5.20	0.94
Trauma	12	0.31	0.32	0.09 to 1.17	0.084
Psychiatric reasons	19	1.48	1.10	0.22 to 5.49	0.91
Hypothermia etiology
Indoor exposure	47	1.0 (ref)a	1.0 (ref)a
Outdoor exposure	141	5.61	5.72	2.18 to 14.97	< 0.001
Immersion	32	5.74	5.18	0.96 to 27.93	0.056
Submersion	21	0.35	0.27	0.08 to 0.93	0.038

^a) Ref ~ reference class

Prehospital processes

The patients were divided into two groups according to the EMS dispatch time of day (08:00–20:00 h and 20:00–08:00 h). In most of the cases (69%), EMS dispatch to patients with accidental hypothermia occurred during daytime hours and was associated with a significantly lower 30-day survival rate compared with that of the nighttime dispatch (p = 0.007). Other data are presented in Supplementary Table 1.

Temperature

The first temperature measurement was taken on the scene in 207 cases (86%). The mean initial temperature was 30.1 °C for survivors and 27.0 °C for non-survivors. In the ED, temperature was measured from 232 patients (96%), and it was 31.8 °C for survivors and 27.4 °C for non-survivors. The non-survivors' lower mean temperatures on the scene (− 3.1 °C) and in the ED (− 4.4 °C) were associated with decreased 30-day survival rate (p < 0.001 for both). Patients were divided into three groups according to hypothermia severity: 31% of the cases were mild, 37% moderate, and 32% severe. The severe category of hypothermia was associated with significantly decreased 30-day survival rate (p < 0.001) (Table 2).

Table 2.

Clinical and laboratory characteristics categorized according to 30-day survival rate. (^a) Include arrhythmias with and without cardiac arrest. ^b) Include patients with measured EtOH value > 0.0 ‰.)

Characteristic	All patients (n = 241)	Survivors (n = 206)	Non-survivors (n = 35)	p-value1)
First temperature (°C)
Out-of-hospital, mean (SD)  n = 207	29.7 (3.7)	30.1 (3.5)	27.0 (4.3)	< 0.001
In the ED, mean (SD)  n = 232	31.2 (4.3)	31.8 (3.9)	27.4 (4.5)	< 0.001
Rate of hypothermia, n (%)				< 0.001
Mild (35–32 °C)	75 (31)	71 (34)	4 (11)
Moderate (32–28 °C)	89 (37)	79 (38)	10 (29)
Severe (< 28 °C)	77 (32)	56 (27)	21 (60)
Glasgow coma scale
Out-of-hospital, mean (SD)  n = 229	10.4 (4.9)	11.1 (4.6)	6.2 (4.3)	< 0.001
In the ED, mean (SD)  n = 237	11.2 (4.7)	12.0 (4.2)	6.0 (4.6)	< 0.001
Some monitored arrhythmia experienced, n (%) a)	108 (45)	78 (38)	30 (86)	< 0.001
First monitored arrhythmia; without cardiac arrest, n (%) n = 65				0.12
FA	39 (60)	37 (64)	2 (29)
SIN brady	22 (34)	18 (31)	4 (57)
Nodal	2 (3)	2 (3)	0
Other	2 (3)	1 (2)	1 (14)
Other injury in addition of hypothermia, n (%)	30 (12)	24 (12)	6 (17)	0.40
First laboratory values, mean (SD)
Potassium (mmol/l), n = 228	4.2 (1.1)	4.0 (0.8)	5.0 (1.9)	0.007
Platelets (× 10 e9/l), n = 227	218 (98)	228 (96)	152 (89)	< 0.001
pH, n = 189	7.20 (0.22)	7.24 (0.16)	7.00 (0.34)	< 0.001
PaCO2 (kPa), n = 187	5.9 (2.1)	5.7 (1.8)	6.9 (3.2)	0.062
PaO2 (kPa), n = 187	21.7 (17.2)	18.7 (13.7)	36.1 (24.5)	< 0.001
BE (mmol/l), n = 185	−10.0 (8.7)	−8.6 (7.6)	−17.1 (10.3)	< 0.001
Lactate (mmol/l), n = 176	5.6 (4.4)	5.2 (4.2)	7.3 (5.1)	0.025
EtOH (‰), n = 98 b)	2.41 (1.14)	2.43 (1.12)	2.25 (1.40)	0.70

Data (excluding first temperature, GCS scores and first laboratory values) are presented as n, % of all patients and % of survivors/non-survivors subgroups

¹⁾ p-values for comparisons between survivors and non-survivors

ED: Emergency Department, FA: Atrial fibrillation, SIN brady: Sinus bradycardia, Nodal: Nodal rhythm, PaCO₂: Partial Pressure of Carbon Dioxide in Arterial Blood, PaO₂: Partial Pressure of Oxygen in Arterial Blood, BE: Base excess, EtOH: Ethyl alcohol, GCS: Glasgow coma scale

Consciousness

Level of consciousness was recorded in 229 cases (95%) by EMS personnel on the scene and in 237 cases (98%) in the ED. The mean GCS scores on the scene were 11.1 for survivors and 6.2 for non-survivors (p < 0.001), while in the ED the mean GCS scores were 12.0 for survivors and 6.0 for non-survivors (p < 0.001). The average GCS score increased during transport from the scene to the ED by 0.9 points for survivors and decreased 0.2 points for non-survivors (Table 2).

Laboratory characteristics

There were several differences in the first laboratory values between survivors and non-survivors. On admission the pH and base excess (BE) values were significantly lower, and lactate and PaO₂ values were higher in non-survivors compared with survivors. In non-survivors, potassium levels were significantly higher and platelet numbers were lower (Table 2).

Cardiac arrest

43 patients (18%) had a CA in the prehospital or hospital phase with resuscitation attempted, and 20 of them (47%) survived (p < 0.001). The primary rhythm with CA was monitored in 38 patients (88%) and was not recorded in 5. The most common primary rhythm was asystole in 20 patients (53%), and 20% of them survived. The most favorable arrhythmia with CA was pulseless electrical activity (PEA) in 8 patients (21%) with a 63% survival rate, which was slightly better than ventricular fibrillation (VF) with a 60% survival rate (p < 0.036) (Table 3).

Table 3.

Treatments in EMS and in hospital categorized according to 30-day survival rate. (^a) One lifeless patient’s airway secured by SAD. ^b) Five patients with cardiac arrest have not made a note of the first monitored rhythm. ^c) In addition two patients received ECMO after CPB warming.)

Characteristic	All patients (n = 241)	Survivors (n = 206)	Non-survivors (n = 35)	p-value1)
Intubation, n (%) n = 55				0.011
Lifeless a)	35 (64)	15 (48)	20 (83)
Unconscious	20 (36)	16 (52)	4 (17)
Resuscitation, n (%)	43 (18)	20 (10)	23 (66)	< 0.001
First monitored arrhythmia; with cardiac arrest, n (%) n = 38 b)				0.036
ASY	20 (53)	4 (27)	16 (70)
VF	10 (26)	6 (40)	4 (17)
PEA	8 (21)	5 (33)	3 (13)
CPR time (min), median (IQR), n = 28	42.5 (10–153.5)	30 (10.5–126)	120 (12–170)	0.52
Invasive warming, n = 52c)				0.002
CoolGard device	25 (48)	23 (62)	2 (13)
CPB	23 (44)	12 (32)	11 (73)
ECMO c)	3 (6)	1 (3)	2 (13)
Bladder lavage	1 (2)	1 (3)	0
CPB warming time (min), median (IQR), n = 23	158 (120.5–222)	179.5 (129–225)	144 (112–204)	0.58
Admitted to ICU, n (%)	130 (54)	110 (53)	20 (57)	0.72

CoolGard is an intravascular cooling and warming device

Data (excluding CPR time and CPB warming time) are presented as n, % of all patients and % of survivors/non-survivors subgroups

¹⁾ p-values for comparisons between survivors and non-survivors

IQR ~ Interquartile range, 25th – 75th percentiles

ASY Asystole, VF: Ventricular fibrillation, PEA Pulseless electrical activity, CPR Cardiopulmonary resuscitation, CPB Cardiopulmonary bypass, ECMO Extracorporeal membrane oxygenation, ICU Intensive care unit, EMS Emergency Medical Service, SAD Supraglottic airway device

Warming and ICU admission

Fifty-two patients (22%) received invasive warming treatment and of these, 25 patients were warmed with a CoolGard device, 23 with cardiopulmonary bypass (CPB), 3 with extracorporeal membrane oxygenation (ECMO), and 1 patient received bladder lavage. CPB was used to warm the most severely affected patients and 12 (52%) of these survived (Table 3).

Fifty-four percent of patients (130) were admitted to the ICU. Mean treatment time in the ICU was 3 days for survivors and 2 days for non-survivors.

Logistic regression analysis of survival

The results of the multivariable adjusted logistic regression analyses are presented in Table 4. According to the models, drunkenness, outdoor exposure, and immersion increased 30-day survival, whereas submersion and trauma increased the risk of 30-day mortality.

Discussion

Alcohol abuse and age ≥ 65 years were the most common concomitant factors for accidental hypothermia. Indoor exposure and daytime EMS dispatch were associated with a decreased 30-day survival rate. A greater severity of hypothermia was clearly correlated with decreased survival. Submersion and trauma were independent risk factors for mortality.

When designing this study, our hypothesis was that there are identifiable concomitant factors for hypothermia that are common among patients. Two of the most common factors were alcohol abuse and age ≥ 65 years. Similar findings of alcohol’s relation to hypothermia were also documented in a Danish study using data collected during 1996–2016, in which accidental hypothermia was most commonly associated with alcohol-related diagnoses. In the same study, old age was associated with higher incidence and mortality in hypothermic patients [8]. In a Japanese study published in 2021, alcohol intoxication was a predisposing factor in only 4.8% of patients, but 81% were over 65 years of age, and old age (> 75 years) was a significant factor for poor prognosis [9]. Similar findings of alcohol abuse as a concomitant factor have been published in Swedish and Polish studies, with incidences of 34% and 68%, respectively [16, 17]. In our data, trauma was a concomitant factor in only 5% of patients, which is lower than we predicted prior to data collection. This finding could be explained by human error in recording patient information at admission, because applying the diagnostic code for hypothermia may be overlooked when treating trauma patients.

Drunkenness benefited survival in our series. Parallel findings were reported in a Japanese study published in 2021, where 55 of 57 patients survived with alcohol intoxication as the reason for their exposure and resulting hypothermia [9]. Another Japanese study published in 2020 reported that intoxication was the probable cause of accidental hypothermia in 14% of cases, and 96% of these patients survived. However, in that study the causes of intoxication were not distinguished [18]. In our series, survivors with alcohol intoxication were younger (75% aged under 65 years) and most of them (71%) had mild or moderate hypothermia. The majority of those survivors were exposed in outdoor conditions (70%).

Most patients in our study (59%) were exposed to hypothermia outdoors, and the second most common circumstance was indoor exposure (20%). Parallel to our findings, a French study that investigated prognostic factors for 81 ICU-treated hypothermia patients found that hypothermia from indoor exposure had notably higher mortality compared with that from outdoor exposure (44% vs. 6%) [19]. A Japanese study from 2020 compared in-hospital mortality for hypothermia patients according to exposure circumstances (537 patients). Indoor exposure was by far the most common (78%) and had higher in-hospital mortality (28.2% vs. 10.9%) compared with outdoor exposure [20]. Also parallel to our findings, worse outcomes for indoor exposure patients were assumed to be due to older age and higher numbers of chronic and underlying illnesses (secondary hypothermia). The number of indoor exposure patients is expected to increase due to the aging population [19, 21].

In our study, submersion had the worst outcomes at 52% mortality, and it was an independent risk factor for mortality as an etiology of hypothermia. Comparing our findings with those of other publications including submerged patients with hypothermia, a 48% survival rate is quite high. A Norwegian study from 2014 investigated survival of 34 patients with accidental hypothermia, 19 of which due to submersion with CA treated with extracorporeal life support. In their data, survival after submersion was 31.6%. On the other hand, in their study 6 of all 9 survivors had submersion as the circumstance of hypothermia [6]. A Dutch study from 2010 reported a survival rate as high as 55.2% among patients who had experienced hypothermia due to submersion [4]. Based on the findings from our study and other similar studies, it seems that submersion as an etiology of hypothermia is associated with poor prognosis. On the other hand, the prognosis is still better than CA with normothermia overall.

In our series, patients with daytime EMS dispatch had significantly higher 30-day mortality compared with the nighttime dispatch group (18% vs. 3%). We speculate that the difference could be explained by the greater number of activities that increase the risk of hypothermia in daytime; for example, ice fishing, out-of-area skiing and accidents during other outdoor activities. Our data supports this interpretation. According to the etiology of accidental hypothermia, all submersion cases and 75% of outdoor exposure cases among non-survivors were dispatched during the daytime. In addition, non-survivors with daytime EMS dispatch were slightly younger and had fewer concomitant illnesses, but they more frequently had trauma as a concomitant factor. On the other hand, it is easier to find patients that are exposed to primary hypothermia in the daytime. Also, patients with various intoxications who are exposed to secondary hypothermia during nighttime might not be discovered until the morning. In our study, 83% of non-survivors EMS dispatched during the daytime had drunkenness as a concomitant factor, and this finding supports the above-mentioned assumption.

Measuring a patient’s temperature is necessary when determining hypothermia severity, which is an important prognostic factor as deeper hypothermia is a predisposing factor for CA [1]. In our data the initial recorded temperature on the scene was available for only 86% of patients, suggesting that there is a lack in the quantity or usage of low-reading thermometers within Finnish EMS units. A Swedish study from 2017 surveyed the availability of equipment for treating hypothermia in prehospital services (road ambulance service, HEMS, and search and rescue). A total of 255 units answered and less than half were equipped with low-reading thermometers [22]. The European Resuscitation Council (ERC) has noticed this same problem, and in the CPR guideline 2021 they recommend using low-reading thermometers when evaluating hypothermic patients [2]. In our data, the severity of hypothermia by measured temperature was clearly inversely correlated with 30-day survival when three hypothermia groups were compared: as the severity increased from mild to moderate to severe, mortality increased from 5 to 11% to 27%, respectively. The number of patients was divided evenly between those three groups and the finding can therefore be considered credible. In our study, the recorded temperature in the emergency room was slightly higher than when the patient was found. This difference may possibly be attributed to an ability to maintain patient body temperature during transportation. Our prehospital guidelines recommend active external rewarming using chemical heating packs or heating blankets. However, it is also possible that this difference is caused by measurement inaccuracies, as a valid core temperature measurements are difficult to obtain.

The prognosis of hypothermia-related CA is known to be better than that of CA in normothermic patients. In systematic reviews of hypothermic patients, witnessed CA has a considerably higher survival rate (73%) than unwitnessed CA (27%) [23, 24]. Our data includes patients with both witnessed and unwitnessed CAs, but they could not be classified reliably for analysis. In our study, the 30-day survival rate for all CA patients was as high as 47%. Further, asystole was the most common initial rhythm in CA (53%), with a 20% survival rate, and PEA was rarest but associated with the highest survival rate (63%). These findings are in line with a systematic review of patients with unwitnessed hypothermic CA [23]. A review of witnessed CA reported ventricular fibrillation as the most common initial rhythm, with 84% survival. Our unclassified data supports the findings of a favorable outcome for VF, with a survival rate of 60% [24].

There are some limitations to keep in mind when interpreting our findings. First, our study design is retrospective and thus we cannot be sure all available hypothermic patients were included in our analysis. We were only able to include patients with ICD-10 codes of either T68 (hypothermia) or T75.1 (drowning). There are many other conditions besides drowning that would carry an increased risk of hypothermia. Unfortunately, it was not technically possible to find all patients with a core temperature below 35 degrees on the scene or on admission. Because patients’ chart data were originally gathered for clinical purposes, there is a risk of selection bias and confounding bias. However, we used a structured approach in collecting all data. The rate of missing data was 1.7–34.9%, including GCS values, temperature values, dispatch times, transport times, first monitored arrhythmia of CA, and CPR times, and missing values were treated as missing. Nevertheless, the impact of unmeasured confounders remains, and some information is bound to be missing from charts in acute care settings (e.g., it is difficult to collect proper anamnesis during triage, and trauma patient ICD-10 coding can be insufficient). Second, in our study, in most cases the patients’ recorded temperature was measured from the eardrum in prehospital phase, and those are not true measurements of core temperature. Obtaining an accurate core temperature is challenging and depends strongly on the mode of measurement. Third, we used a convenient sample size, and a long time period was needed due to the rarity of hypothermia among hospitalized patients. During the 12-year period there might have been a few changes in treatment practices that could have affected outcomes, such as transporting CA patients straight to an ECLS centre and starting to use ECMO for rewarming. Fourth, our data are almost solely from Caucasian adults, and the geographic area is sparsely populated. Finally, we only recorded 30-day survival. Neurological status is also important to determine the extent of harm and the depth of hypothermia. We were not able to record neurological outcome in this retrospective analysis.

The study’s main strength is our relatively large patient population, with a uniform electronic patient data system in a single center. Still, our study cannot determine causation, only association. Thus, future prospective multicenter studies with complete datasets are required to address these limitations. Future research should also focus on the optimization of hypothermia management in the hospital. The generalizability of this study is supported by the large catchment area of a tertiary university hospital and inclusion of both indoor and outdoor exposure.

Conclusion

Alcohol abuse and age ≥ 65 years were two of the most common concomitant factors for accidental hypothermia. Only one-third of patients had severe hypothermia, and the overall mortality rate was low. Non-survivors were generally older and more likely to be found indoors. They were most often found during daytime hours, and their recorded temperature was generally lower. Submersion and trauma were independent risk factors for mortality. However, alcohol abuse, outdoor exposure and immersion were the predictors of survival. Thus, our findings suggest older people and those with indoor exposure potential deserve special attention and careful clinical evaluation since they may be at higher risk for adverse outcomes.

Abbreviations

BE

Base excess

CA

Cardiac arrest

CPB

Cardiopulmonary bypass

CPR

Cardiopulmonary resuscitation

ECLS

Extracorporeal life support

ECMO

Extracorporeal membrane oxygenation

ED

Emergency department

EMS

Emergency medical service

ERC

European Resuscitation Council

ESKO

The name of the hospital electronic patient record system

GCS

Glasgow Coma Scale

ICD-10

International Classification of Diseases (10^th Revision)

ICU

Intensive care unit

HEMS

Helicopter emergency medical service

PEA

Pulseless electrical activity

VF

Ventricular fibrillation

Authors’ contributions

JP, JN, TK, SS, PO, MH, and TA participated in the study design. JP and JN collected and analyzed the patient data and wrote the initial draft. TK collected the patient data. SS collected the data on hospital processes. PO did the logistic regression analysis. JP, JN, PO, MH and TA drafted the final manuscript. MH and TA supervised the study and the writing of the manuscript. All authors interpreted the data, helped to form the scientific content of the manuscript, and read and approved the final manuscript.

Funding

Open Access funding provided by University of Oulu (including Oulu University Hospital). This study was financially supported by the State Funding for University-Level Health Research, Oulu University Hospital, and the Wellbeing Services County of North Ostrobothnia.

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Due to the retrospective study design, an exemption from consent was obtained from the hospital ethics committee; the data had already been collected for clinical purposes and none of the patients were contacted during this study (Registration number 152/2020).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.