The linguistic and interactional factors impacting recognition and dispatch in emergency calls for out-of-hospital cardiac arrest: a mixed-method linguistic analysis study protocol

Machine Learning-Based Text Analysis to Predict Severely Injured Patients in Emergency Medical Dispatch: Model Development and Validation

Background

Early recognition of severely injured patients in prehospital settings is of paramount importance for timely treatment and transportation of patients to further treatment facilities. The dispatching accuracy has seldom been addressed in previous studies.

Objective

In this study, we aimed to build a machine learning–based model through text mining of emergency calls for the automated identification of severely injured patients after a road accident.

Methods

Audio recordings of road accidents in Taipei City, Taiwan, in 2018 were obtained and randomly sampled. Data on call transfers or non-Mandarin speeches were excluded. To predict cases of severe trauma identified on-site by emergency medical technicians, all included cases were evaluated by both humans (6 dispatchers) and a machine learning model, that is, a prehospital-activated major trauma (PAMT) model. The PAMT model was developed using term frequency–inverse document frequency, rule-based classification, and a Bernoulli naïve Bayes classifier. Repeated random subsampling cross-validation was applied to evaluate the robustness of the model. The prediction performance of dispatchers and the PAMT model, in severe cases, was compared. Performance was indicated by sensitivity, specificity, positive predictive value, negative predictive value, and accuracy.

Results

Although the mean sensitivity and negative predictive value obtained by the PAMT model were higher than those of dispatchers, they obtained higher mean specificity, positive predictive value, and accuracy. The mean accuracy of the PAMT model, from certainty level 0 (lowest certainty) to level 6 (highest certainty), was higher except for levels 5 and 6. The overall performances of the dispatchers and the PAMT model were similar; however, the PAMT model had higher accuracy in cases where the dispatchers were less certain of their judgments.

Conclusions

A machine learning–based model, called the PAMT model, was developed to predict severe road accident trauma. The results of our study suggest that the accuracy of the PAMT model is not superior to that of the participating dispatchers; however, it may assist dispatchers when they lack confidence while making a judgment.

Validity and risk factor analysis for helicopter emergency medical services in Japan: a pilot study

Background

Some emergency departments use triage scales, such as the Canadian Triage and Acuity Scale and Japan Urgent Stroke Triage Score, to detect life-threatening situations. However, these protocols have not been used for aeromedical services. Therefore, we investigated the factors predicting these life-threatening situations in aeromedical services as a pilot study for establishing the protocol.

Method

We retrospectively evaluated helicopter emergency medical service cases from 1 April 2015 to 31 March 2020 at Gifu University Hospital using the mission records. We only evaluated cases dealing with suggested internal medicine issues. We excluded cases influenced by external factors such as trauma or cases that included hospital-to-hospital transportation, focusing only on prehospital care. We evaluated the validity of the medical emergencies based on the needs for emergency interventions and hospital admission and of the suggested diagnoses and associated risk factors.

Result

A total of 451 cases were suitable for inclusion in the study. In the analysis for all emergency calls, 235 (52.11%) cases needed emergency intervention and 300 (64.4%) required hospital admission. The suggested diagnosis was valid for 261 (57.87%) cases. After the first assessment by emergency medical technicians, 75 cases were removed. Analysis after this first assessment found that 52.31% cases required emergency intervention, 70.26% needed admission, and the suggested diagnosis was valid for 69.41% of cases. In the analysis of emergency calls, the multivariate analysis of some key variables identified age, playing sports, and gasping as risk factors for emergency intervention. Hospital admission risk factors included being age only. The suggested diagnosis was valid only for sports situations. In the analysis after the first assessment by an emergency medical technician, risk factors for emergency intervention included being age being male, playing sports, and gasping, and those for hospital admission was being age, being male, and experiencing stroke symptoms and/or disturbance of consciousness. The suggested diagnosis was valid only for sports situations.

Conclusion

Some ‘second’ keywords/phrases predict medical emergencies. Therefore, the dispatch commander should gather these keyword/phrases to assess.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12873-021-00471-x.

Dispatch of a helicopter emergency medicine service to patients with a sudden, unexplained loss of consciousness of medical origin

Background

Sudden loss of consciousness (LOC) in the prehospital setting in the absence of cardiac arrest and seizure activity may be a challenge from a dispatcher’s perspective: The aetiology is varied, with many causes being transient and mostly self-limiting, whereas other causes are potentially life threatening. In this study we aim to evaluate the dispatch of HEMS to patients with LOC of medical origin, by exploring to which patients with a LOC HEMS is dispatched, which interventions HEMS teams perform in these patients, and whether HEMS interventions can be predicted by patient characteristics.

Methods

We performed retrospective cohort study of all patients with a reported unexplained LOC (e.g. not attributable to a circulatory arrest or seizures) attended by the Air Ambulance Kent, Surrey & Sussex (AAKSS), over a 4-year period (July 2013–December 2017). Primary outcome was defined as the number of HEMS-specific interventions performed in patients with unexplained LOC. Secondary outcome was the relation of clinical- and dispatch criteria with HEMS interventions being performed.

Results

During the study period, 127 patients with unexplained LOC were attended by HEMS. HEMS was dispatched directly to 25.2% of the patients, but mostly (74.8%) on request of the ground ambulance crews. HEMS interventions were performed in 65% of the patients (Prehospital Emergency Anaesthesia 56%, hyperosmolar therapy 21%, antibiotic/antiviral therapy 8%, vasopressor therapy 6%) and HEMS conveyed most patients (77%) to hospital. Acute neurological pathology was a prevalent underlying cause of unexplained LOC: 38% had gross pathology on their CT-scan upon arrival in hospital. Both GCS (r = − 0.60, p < .001) and SBP (r = 0.31, p < .001) were related to HEMS interventions being performed on scene. A GCS < 13 predicted the need for HEMS interventions in our population with a sensitivity of 94.9% and a specificity 75% (AUC 0.85).

Conclusion

HEMS dispatchers and ambulance personnel are able to identify a cohort of patients with unexplained LOC of medical origin who suffer from potentially life threatening (mainly neurological) pathology, in whom HEMS specific intervention are frequently required. Presenting GCS can be used to inform the triage process of patients with LOC at an early stage.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12873-020-00388-x.

Improving emergency call detection of Out-of-Hospital Cardiac Arrests in the Greater Paris area: Efficiency of a global system with a new method of detection

AIM

The detection of cardiac arrests by dispatchers allows telephone-assisted cardiopulmonary resuscitation (t-CPR) and improves Out-of-Hospital Cardiac Arrest (OHCA) survival. To enhance the OHCA detection rate, in 2012, the Paris Fire Brigade dispatch center created an original technique called "Hand On Belly" (HoB). The new algorithm that resulted has become a central point in a broader program for dispatch-assisted cardiac arrests.

METHODS

This is a repeated cross-sectional study with retrospective data of four 15-day call samples recorded from 2012 to 2018. We included all calls from OHCAs cared for by Basic Life Support (BLS) teams and excluded calls where the dispatcher was not in contact directly with a witness. The primary endpoint was the successful detection of an OHCA by the dispatcher; the secondary endpoints were successful t-CPR and measurements of the different time intervals related to the call. Logistic regressions were performed to assess parameters associated with detecting OHCAs and initiating t-CPR.

RESULTS

From 2012 to 2018, among the detectable OCHAs, the proportion correctly identified increased from 54% to 93%; the rate of t-CPRs from 51% to 84%. OHCA detection and t-CPR initiation were both associated with HoB breathing assessments (adjustedOR: 89, 95%CI: 31-299, and adjustedOR: 11.2, 95%CI: 1.4-149, respectively). Over the study period, the times to answering calls and the time to sending BLS teams were shorter than those recommended by international guidelines; however, the times to OHCA recognition and starting t-CPR delivery were longer.

CONCLUSIONS

The HoB effectively facilitated OHCA detection in our system, which has achieved very high performance levels.

The accuracy of medical dispatch - a systematic review

BACKGROUND

It is a challenge to dispatch Emergency medical Services (EMS) appropriately with limited resources and maintaining patient safety; this requires accurate dispatching systems. The objective of the current systematic review was to examine the evidence, according to GRADE, for medical dispatching systems to accurately dispatch EMS according to level of acuity and in recognition of specific conditions. A systematic search was performed trough PubMed, Web of Science, Embase (free text in all fields), Centre for Reviews and Dissemination (CRD), and Cochrane Central Register of Controlled Trials up to 16th of May, 2017. A combination of keywords and Medical Subject Heading (MeSH) terms relevant to "emergency medical dispatch criteria" were used, to search for articles published between 2012 and 2017. Publications were included according to the inclusion/exclusion criteria using the Systematic Reviews and Meta-Analyses (PRISMA) protocol. Level of evidence was evaluated in accordance with Grading of Recommendations Assessment, Development and Evaluation (GRADE). Articles included were those that provided evidence for at least one of the measures of dispatch system accuracy; i.e. sensitivity, specificity, positive and negative predictive and/or over- and under-triage. The search identified 1445 articles. After the removal of duplicates, 382 titles were reviewed for relevance and an additional 359 articles were excluded based on manuscript title and abstract. An additional five articles were excluded after review of the full text versions of the remaining articles. The current review included 18 publications which all were based on primary research.

CONCLUSIONS

The 18 articles addressed the identification of cardiac arrest, stroke, medical priority and major trauma using different dispatching systems. The results of the current review show that there is a very low to low overall level of evidence for the accuracy of medical dispatching systems. We suggest that it is necessary to create a consensus on common standards for reporting before consensus can be reached for the level of accuracy in medical dispatching systems.

'We're going to do CPR': A linguistic study of the words used to initiate dispatcher-assisted CPR and their association with caller agreement

BACKGROUND

In emergency ambulance calls for out-of-hospital cardiac arrest (OHCA), dispatcher-assisted cardiopulmonary resuscitation (CPR) plays a crucial role in patient survival. We examined whether the language used by dispatchers to initiate CPR had an impact on callers' agreement to perform CPR.

METHODS

We analysed 424 emergency calls relating to cases of paramedic-confirmed OHCA where OHCA was recognised by the dispatcher, the caller was with the patient, and resuscitation was attempted by paramedics. We investigated the linguistic choices used by dispatchers to initiate CPR, and the impact of those choices on caller agreement to perform CPR.

RESULTS

Overall, CPR occurred in 85% of calls. Caller agreement was low (43%) when dispatchers used terms of willingness ("do you want to do CPR?"). Caller agreement was high (97% and 84% respectively) when dispatchers talked about CPR in terms of futurity ("we are going to do CPR") or obligation ("we need to do CPR"). In 38% (25/66) of calls where the caller initially declined CPR, the dispatcher eventually secured their agreement by making several attempts at initiating CPR.

CONCLUSION

There is potential for increased agreement to perform CPR if dispatchers are trained to initiate CPR with words of futurity and/or obligation.

'She's sort of breathing': What linguistic factors determine call-taker recognition of agonal breathing in emergency calls for cardiac arrest?

BACKGROUND

In emergency ambulance calls, agonal breathing remains a barrier to the recognition of out-of-hospital cardiac arrest (OHCA), initiation of cardiopulmonary resuscitation, and rapid dispatch. We aimed to explore whether the language used by callers to describe breathing had an impact on call-taker recognition of agonal breathing and hence cardiac arrest.

METHODS

We analysed 176 calls of paramedic-confirmed OHCA, stratified by recognition of OHCA (89 cases recognised, 87 cases not recognised). We investigated the linguistic features of callers' response to the question "is s/he breathing?" and examined the impact on subsequent coding by call-takers.

RESULTS

Among all cases (recognised and non-recognised), 64% (113/176) of callers said that the patients were breathing (yes-answers). We identified two categories of yes-answers: 56% (63/113) were plain answers, confirming that the patient was breathing ("he's breathing"); and 44% (50/113) were qualified answers, containing additional information ("yes but gasping"). Qualified yes-answers were suggestive of agonal breathing. Yet these answers were often not pursued and most (32/50) of these calls were not recognised as OHCA at dispatch.

CONCLUSION

There is potential for improved recognition of agonal breathing if call-takers are trained to be alert to any qualification following a confirmation that the patient is breathing.

'Tell me exactly what's happened': When linguistic choices affect the efficiency of emergency calls for cardiac arrest

BACKGROUND

Clear and efficient communication between emergency caller and call-taker is crucial to timely ambulance dispatch. We aimed to explore the impact of linguistic variation in the delivery of the prompt "okay, tell me exactly what happened" on the way callers describe the emergency in the Medical Priority Dispatch System^®.

METHODS

We analysed 188 emergency calls for cases of paramedic-confirmed out-of-hospital cardiac arrest. We investigated the linguistic features of the prompt "okay, tell me exactly what happened" in relation to the format (report vs. narrative) of the caller's response. In addition, we compared calls with report vs. narrative responses in the length of response and time to dispatch.

RESULTS

Callers were more likely to respond with a report format when call-takers used the present perfect ("what's happened") rather than the simple past ("what happened") (Adjusted Odds Ratio [AOR] 4.07; 95% Confidence Interval [95%CI] 2.05-8.28, p<0.001). Reports were significantly shorter than narrative responses (9s vs. 18s, p<0.001), and were associated with less time to dispatch (50s vs. 58s, p=0.002).

CONCLUSION

These results suggest that linguistic variations in the way the scripted sentences of a protocol are delivered can have an impact on the efficiency with which call-takers process emergency calls. A better understanding of interactional dynamics between caller and call-taker may translate into improvements of dispatch performance.