Machine learning model predicts short-term mortality among prehospital patients: A prospective development study from Finland.

AIM: To show whether adding blood glucose to the National Early Warning Score (NEWS) parameters in a machine learning model predicts 30-day mortality more precisely than the standard NEWS in a prehospital setting. METHODS: In this study, vital sign data prospectively collected from 3632 unselected prehospital patients in June 2015 were used to compare the standard NEWS to random forest models for predicting 30-day mortality. The NEWS parameters and blood glucose levels were used to develop the random forest models. Predictive performance on an unknown patient population was estimated with a ten-fold stratified cross-validation method. RESULTS: All NEWS parameters and blood glucose levels were reported in 2853 (79%) eligible patients. Within 30 days after contact with ambulance staff, 97 (3.4%) of the analysed patients had died. The area under the receiver operating characteristic curve for the 30-day mortality of the evaluated models was 0.682 (95% confidence interval [CI], 0.619-0.744) for the standard NEWS, 0.735 (95% CI, 0.679-0.787) for the random forest-trained NEWS parameters only and 0.758 (95% CI, 0.705-0.807) for the random forest-trained NEWS parameters and blood glucose. The models predicted secondary outcomes similarly, but adding blood glucose into the random forest model slightly improved its performance in predicting short-term mortality. CONCLUSIONS: Among unselected prehospital patients, a machine learning model including blood glucose and NEWS parameters had a fair performance in predicting 30-day mortality.

Spontaneous trigger words associated with confirmed out-of-hospital cardiac arrest: a descriptive pilot study of emergency calls.

BACKGROUND: According to the International Liaison Committee on Resuscitation (ILCOR), the trigger words used by callers that are associated with cardiac arrest constitute a scientific knowledge gap. This study was designed to find hypothetical trigger words in emergency calls in order to improve the specificity of out-of-hospital cardiac arrest recognition. METHODS: In this descriptive pilot study conducted in a Finnish hospital district, linguistic contents of 80 emergency calls of dispatcher-suspected or EMS-encountered out-of-hospital cardiac arrests between January 1, 2017 and May 31, 2017 were analysed. Spontaneous trigger words used by callers were transcribed and grouped into 36 categories. The association between the spontaneous trigger words and confirmed true cardiac arrests was tested with logistic regression. RESULTS: Of the suspected cardiac arrests, 51 (64%) were confirmed as true cardiac arrests when ambulance personnel met the patient. A total of 291 spontaneous trigger words were analysed. 'Is not breathing' (n = 9 [18%] in the true cardiac arrest group vs n = 1 [3%] in the non-cardiac arrest group, odds ratio [OR] 6.00, 95% confidence interval [CI] 0.72-50.0), 'the patient is blue' (n = 9 [18%] vs n = 1 [3%], OR 6.00, 95% CI 0.72-50.0), 'collapsed or fallen down' (n = 12 [24%] vs n = 2 [7%], OR 4.15, 95% CI 0.86-20.1) and 'is wheezing' (n = 17 [33%] vs n = 5 [17%], OR 2.40, 95% CI 0.78-7.40) were frequently used to describe true cardiac arrest. 'Is snoring' was associated with a false suspicion of cardiac arrest (n = 1 [2%] vs n = 6 [21%], OR 0.08, 95% CI 0.009-0.67). CONCLUSIONS: In our pilot study, no trigger word was associated with confirmed cardiac arrest. 'Is wheezing' was a frequently used spontaneous trigger word among later confirmed cardiac arrest victims.

Random forest machine learning method outperforms prehospital National Early Warning Score for predicting one-day mortality: A retrospective study.

AIM OF THE STUDY: The National Early Warning Score (NEWS) is a validated method for predicting clinical deterioration in hospital wards, but its performance in prehospital settings remains controversial. Modern machine learning models may outperform traditional statistical analyses for predicting short-term mortality. Thus, we aimed to compare the mortality prediction accuracy of NEWS and random forest machine learning using prehospital vital signs. METHODS: In this retrospective study, all electronic ambulance mission reports between 2008 and 2015 in a single EMS system were collected. Adult patients (≥ 18 years) were included in the analysis. Random forest models with and without blood glucose were compared to the traditional NEWS for predicting one-day mortality. A ten-fold cross-validation method was applied to train and validate the random forest models. RESULTS: A total of 26,458 patients were included in the study of whom 278 (1.0%) died within one day of ambulance mission. The area under the receiver operating characteristic curve for one-day mortality was 0.836 (95% CI, 0.810-0.860) for NEWS, 0.858 (95% CI, 0.832-0.883) for a random forest trained with NEWS variables only and 0.868 (0.843-0.892) for a random forest trained with NEWS variables and blood glucose. CONCLUSION: A random forest algorithm trained with NEWS variables was superior to traditional NEWS for predicting one-day mortality in adult prehospital patients, although the risk of selection bias must be acknowledged. The inclusion of blood glucose in the model further improved its predictive performance.

Professional firefighter and trained volunteer first-responding units in emergency medical service.

BACKGROUND: Although widely dispatched to out-of-hospital cardiac arrests, the performance of prehospital first-responding units in other medical emergencies is unknown. METHODS: In this retrospective, descriptive study, the general performance of 44 first-responding units in Pirkanmaa County, Finland, were examined. A subgroup analysis compared the first-responding units made up of professional firefighters and trained volunteers. RESULTS: First-responding units were dispatched to patients during 1622 missions between 1 January 2013 and 31 December 2013. The median time to reach the scene was 9 minutes in any mission. Overall, first responders evaluated 1015 patients and provided treatment or assisted ambulance personnel in 793 (78%) cases. The most common treatment modalities were assistance, such as carrying (22%) and the administration of supplemental oxygen (19%). There were 83 resuscitation attempts during the time period. In 42 of these, first-responding units initiated basic life support a median of 4 minutes prior to the arrival of ambulance personnel. Return of spontaneous circulation was achieved in 20% of cases. The subgroup analysis showed that trained volunteers administered oxygen more liberally than professional firefighters in stroke and chest pain mission (stroke: professional 9/236 cases [4%] vs layperson 26/181 cases [14%], P < 0.001; chest pain: professional 16/78 cases [21%] vs layperson 77/159 cases [48%], P < 0.001). CONCLUSION: First-responding units provided initial treatment or assistance to ambulance personnel in approximately half of the missions. Implementation of professional- and layperson-staffed first-responding units in emergency medical service system seems to be feasible.