Telephone triage of cardiac emergency calls by dispatchers: a prospective study of 1386 emergency calls

Deep ensemble multitask classification of emergency medical call incidents combining multimodal data improves emergency medical dispatch

The objective of this work was to develop a predictive model to aid non-clinical dispatchers to classify emergency medical call incidents by their life-threatening level (yes/no), admissible response delay (undelayable, minutes, hours, days) and emergency system jurisdiction (emergency system/primary care) in real time. We used a total of 1 244 624 independent incidents from the Valencian emergency medical dispatch service in Spain, compiled in retrospective from 2009 to 2012, including clinical features, demographics, circumstantial factors and free text dispatcher observations. Based on them, we designed and developed DeepEMC², a deep ensemble multitask model integrating four subnetworks: three specialized to context, clinical and text data, respectively, and another to ensemble the former. The four subnetworks are composed in turn by multi-layer perceptron modules, bidirectional long short-term memory units and a bidirectional encoding representations from transformers module. DeepEMC² showed a macro F1-score of 0.759 in life-threatening classification, 0.576 in admissible response delay and 0.757 in emergency system jurisdiction. These results show a substantial performance increase of 12.5 %, 17.5 % and 5.1 %, respectively, with respect to the current in-house triage protocol of the Valencian emergency medical dispatch service. Besides, DeepEMC² significantly outperformed a set of baseline machine learning models, including naive bayes, logistic regression, random forest and gradient boosting (α = 0.05). Hence, DeepEMC² is able to: 1) capture information present in emergency medical calls not considered by the existing triage protocol, and 2) model complex data dependencies not feasible by the tested baseline models. Likewise, our results suggest that most of this unconsidered information is present in the free text dispatcher observations. To our knowledge, this study describes the first deep learning model undertaking emergency medical call incidents classification. Its adoption in medical dispatch centers would potentially improve emergency dispatch processes, resulting in a positive impact in patient wellbeing and health services sustainability.

Developing a Triage Telephone Program for Home Health Care Agencies

In the current managed care environment, innovative strategies for patient care will be the hallmarks for effective utilization of services. Protocol-based telephone triage is an effective way to clinically intervene. It is currently used successfully in a variety of patient care settings. Home health care agencies need to be proactive and to incorporate a telephone triage system into their after-hours programs to meet patient needs with appropriate services. Three key elements—structural, process, and outcome standards—are discussed as critical to the development and implementation of such a program.

Myocardial infarction: sex differences in symptoms reported to emergency dispatch

BACKGROUND

Emergency management of myocardial infarction (MI) is time-critical, because improved patient outcomes are associated with reduced time from symptom onset to definitive care. Previous studies have identified that women are less likely to present with chest pain.

OBJECTIVE

We sought to measure the effect of sex on symptoms reported to the ambulance dispatch and ambulance times for MI patients.

METHODS

The Western Australia Emergency Department Information System (EDIS) was used to identify patients with emergency department (ED) diagnoses of MI (ST-segment elevation MI and non-ST-segment elevation MI) who arrived by ambulance between January 1, 2008, and October 31, 2009. Their emergency telephone calls to the ambulance service were transcribed to identify presenting symptoms. Ambulance data were used to examine ambulance times. Sex differences were analyzed using descriptive and age-adjusted regression analysis.

RESULTS

Of 3,329 MI patients who presented to Perth EDs, 2,100 (63.1%) arrived by ambulance. After predefined exclusions, 1,681 emergency calls were analyzed. The women (n = 621; 36.9%) were older than the men (p < 0.001) and, even after age adjustment, were less likely to report chest pain (odds ratio [OR] = 0.70; 95% confidence interval [CI] 0.57, 0.88). After age adjustment, ambulance times did not differ between the male and female patients with chest pain. The women with chest pain were less likely than the men with chest pain to be allocated a "priority 1" (lights and sirens) ambulance response (men 98.3% vs. women 95.5%; OR = 0.39; 95% CI 0.18, 0.87).

CONCLUSION

Ambulance dispatch officers (and paramedics) need to be aware of potential sex differences in MI presentation in order to ensure appropriate ambulance response.

The appropriateness of, and compliance with, telephone triage decisions: a systematic review and narrative synthesis

AIM

This paper is a report of the synthesis of evidence on the appropriateness of, and compliance with, telephone triage decisions.

BACKGROUND

Telephone triage plays an important role in managing demand for health care. Important questions are whether triage decisions are appropriate and patients comply with them.

DATA SOURCES

CINAHL, Cochrane Clinical Trials Database, Medline, Embase, Web of Science, and Psyc Info were searched between 1980-June 2010.

DESIGN

LITERATURE REVIEW

Rapid Evidence Synthesis.

REVIEW METHODS

The principles of rapid evidence assessment were followed.

RESULTS

We identified 54 relevant papers: 26 papers reported appropriateness of triage decision, 26 papers reported compliance with triage decision, and 2 papers reported both. Nurses triaged calls in most of the studies (n=49). Triage decisions rated as appropriate varied between 44-98% and compliance ranged from 56-98%. Variation could not be explained by type of service or method of assessing appropriateness. However, inconsistent definitions of appropriateness may explain some variation. Triage decisions to contact primary care may have lower compliance than decisions to contact emergency services or self care.

CONCLUSION

Telephone triage services can offer appropriate decisions and decisions that callers comply with. However, the association between the appropriateness of a decision and subsequent compliance requires further investigation and further consideration needs to be given to the minority of calls which are inappropriately managed. We suggest that a definition of appropriateness incorporating both accuracy and adequacy of triage decision should be encouraged.

Safety of telephone triage in out-of-hours care: a systematic review

OBJECTIVE

Telephone triage in patients requesting help may compromise patient safety, particularly if urgency is underestimated and the patient is not seen by a physician. The aim was to assess the research evidence on safety of telephone triage in out-of-hours primary care.

METHODS

A systematic review was performed of published research on telephone triage in out-of-hours care, searching in PubMed and EMBASE up to March 2010. Studies were included if they concerned out-of-hours medical care and focused on telephone triage in patients with a first request for help. Study inclusion and data extraction were performed by two researchers independently. Post-hoc two types of studies were distinguished: observational studies in contacts with real patients (unselected and highly urgent contacts), and prospective observational studies using high-risk simulated patients (with a highly urgent health problem).

RESULTS

Thirteen observational studies showed that on average triage was safe in 97% (95% CI 96.5-97.4%) of all patients contacting out-of-hours care and in 89% (95% CI 86.7-90.2%) of patients with high urgency. Ten studies that used high-risk simulated patients showed that on average 46% (95% CI 42.7-49.8%) were safe. Adverse events described in the studies included mortality (n = 6 studies), hospitalisations (n = 5), attendance at emergency department (n=1), and medical errors (n = 6).

CONCLUSIONS

There is room for improvement in safety of telephone triage in patients who present symptoms that are high risk. As these have a low incidence, recognition of these calls poses a challenge to health care providers in daily practice.

Early identification and delay to treatment in myocardial infarction and stroke: differences and similarities

Background

The two major complications of atherosclerosis are acute myocardial infarction (AMI) and acute ischemic stroke. Both are life-threatening conditions characterised by the abrupt cessation of blood flow to respective organs, resulting in an infarction. Depending on the extent of the infarction, loss of organ function varies considerably.

In both conditions, it is possible to limit the extent of infarction with early intervention. In both conditions, minutes count.

This article aims to describe differences and similarities with regard to the way patients, bystanders and health care providers act in the acute phase of the two diseases with the emphasis on the pre-hospital phase.

Method

A literature search was performed on the PubMed, Embase (Ovid SP) and Cochrane Library databases.

Results

In both conditions, symptoms vary considerably. Patients appear to suspect AMI more frequently than stroke and, in the former, there is a gender gap (men suspect AMI more frequently than women).

With regard to detection of AMI and stroke at dispatch centre and in Emergency Medical Service (EMS) there is room for improvement in both conditions. The use of EMS appears to be higher in stroke but the overall delay to hospital admission is shorter in AMI. In both conditions, the fast track concept has been shown to influence the delay to treatment considerably.

In terms of diagnostic evaluation by the EMS, more supported instruments are available in AMI than in stroke. Knowledge of the importance of early treatment has been reported to influence delays in both AMI and stroke.

Conclusion

Both in AMI and stroke minutes count and therefore the fast track concept has been introduced. Time to treatment still appears to be longer in stroke than in AMI. In the future improvement in the early detection as well as further shortening to start of treatment will be in focus in both conditions. A collaboration between cardiologists and neurologists and also between pre-hospital and in-hospital care might be fruitful.

Inappropriate dispatcher decision for emergency medical service users with acute myocardial infarction

OBJECTIVES

Current guidelines recommend utilization of prehospital emergency medical services (EMSs) by patients with ST-elevation myocardial infarction (STEMI). The aims of this study were to estimate the percentage of inappropriate initial dispatcher decisions and determine their impact on delays in reperfusion therapy for EMS users with STEMI.

METHODS

As part of a prospective regional registry of patients with STEMI, we analyzed the original data for 245 patients who called a university hospital-affiliated EMS call center in France. The primary study outcome was time to reperfusion therapy calculated from the documented date and time of the first patient call.

RESULTS

The initial EMS dispatcher's decision was appropriate (ie, dispatching a mobile intensive care unit staffed by an emergency or critical care physician) for 171 (70%) patients and inappropriate for 74 (30%) patients. Inappropriate decisions included referring the patient to a family physician (n = 59), providing medical advice (n = 9), and dispatching an ambulance (n = 6). Inappropriate initial decisions resulted in increased median time to reperfusion for 140 patients receiving fibrinolysis (95 vs 53 minutes; P < .001) and 91 patients undergoing primary percutaneous coronary intervention (170 vs 107 minutes; P < .001). In-hospital mortality was not different between the 2 study groups (6.8% vs 9.9%; P = .42).

CONCLUSION

The initial dispatcher's decision is inappropriate for 30% of EMS users with STEMI and results in substantial delays in time to reperfusion therapy. Accuracy of telephone triage should be improved for patients who activate EMSs in response to symptoms suggestive of acute coronary syndrome.

The Ability of Emergency Medical Dispatch Codes of Medical Complaints to Predict ALS Prehospital Interventions

OBJECTIVE

The Medical Priority Dispatch System (MPDS) is an emergency medical dispatch (EMD) system that is commonly used to triage 9-1-1 calls and optimize paramedic and EMT dispatch. The objective of this study was to determine the sensitivity, specificity, and positive and negative predictive values of selected MPDS dispatch codes to predict the need for ALS medication or procedures.

METHODS

Patients with selected MPDS codes between November 1, 2003, and October 31, 2005, from a suburban California county were matched with their electronic patient care record. The records of all transported patients were queried for prehospital interventions and matched to their MPDS classification [Basic Life Support (BLS) versus Advanced Life Support (ALS)]. Patients who received prehospital interventions or medications were considered ALS Intervention. With true positive = ALS by MPDS + ALS Intervention, true negative = BLS by MPDS + BLS Interventions, false positive = ALS by MPDS + BLS Interventions, and false negative = BLS by MPDS + ALS Interventions, the screening performance of the San Mateo County EMD system was determined for selected complaint categories (abdominal pain, breathing problems chest pain, sick person, seizures, and unconscious/fainting).

RESULTS

There were a total of 64,647 medical calls, and 42,651 went through the EMD process; 31,187 went through the EMD process and were transported; 22,243 of these were matched to a patient care record. The sensitivity and specificity with 95% confidence intervals in () were as follows: all EMD calls 84 (83-85), 36 (35-36); abdominal pain, 53 (41-65), 47 (43-51); chest pain 99 (99-100), 2 (1-3); seizure 83 (77-88), 20 (17-23), sick 59 (53-64), 51 (49-54), and unconscious/fainting 99 (98-100), 2 (2-3).

CONCLUSION

In our EMS system, MPDS coding for all medical calls had high sensitivity and low specificity for the prediction of calls that required ALS intervention. Chest pain and unconscious/fainting calls were screened with very high sensitivity but very low specificity.

Can emergency medical dispatch codes predict prehospital interventions for common 9-1-1 call types?

OBJECTIVE

The Medical Priority Dispatch System is an emergency medical dispatch (EMD) system that is widely used to categorize 9-1-1 calls and optimize resource allocation. This study evaluates the ability of EMD and non-EMD codes (calls not processed by EMD) to predict prehospital use of medications and procedures.

METHODS

All transported prehospital patients placed in an EMD or non-EMD category that exceeded 500 total calls from January 1, 2004, to December 31, 2006, in a suburban California county were matched with their prehospital electronic patient care record. These records (N = 69,541) were queried for the following prehospital interventions: basic life support (BLS) care only, intravenous line placement only, medication given, and procedures. Advanced life support (ALS) interventions were defined as the administration of a medications or a procedure. The numbers of medications and procedures that were performed on patients in each EMD code were measured.

RESULTS

Thirty-one of 141 EMD and non-EMD codes met inclusion criteria and comprised 73% of all calls during the study period. Non-EMD codes accounted for 48% of all calls in this study. Patients with shortness of breath, chest pain, diabetic problems, and altered mental status received the most medications. High rates of medication administration were also seen in the following codes: 17A (fall, 27%), 17B (fall, 14%), EMDX (unable to complete EMD process, 22%), MED (medical aid requested--details to follow, 26%), and MED3 (medical aid requested by police--code 3, 18%). Procedures were performed on only 0.9% of all calls, of which 75% were related to advanced airways. Higher rates of ALS interventions in higher-acuity categories (Alpha, Bravo, etc.) were seen in a number of EMD categories, including seizure, laceration/hemorrhage, sick, and traffic accident, but not seen in many categories, including abdominal pain, falls, and chest pain.

CONCLUSIONS

This study demonstrated only a modest ability of the EMD system to predict which patients would require ALS intervention. There were limited differences noted in the ALS rates between the different codes (Alpha, Bravo, etc.) in the same complaint category, bringing into question the utility of the multiple subgroups. Non-EMD codes made up a large portion of calls (48%) and should be included in future studies.

A study of police operated dispatch to acute coronary syndrome cases arising from 112 emergency calls in Aarhus county, Denmark

BACKGROUND

The accuracy of the Danish police operated "112" emergency call system was studied. Dispatch of the anaesthesiologist staffed mobile emergency care unit (MECU) to acute coronary syndrome (ACS) cases was used as an indicator of accuracy of dispatch to life threatening emergencies.

METHODS

This was an observational cohort study of patients given a 112 system report of heart attack and patients with a provisional diagnosis of ACS made on scene by the MECU. Sensitivity, specificity, and positive predictive value with 95% confidence intervals (CI) were calculated.

RESULTS

There were 341 reports of "heart attack" and 205 patients with ACS. Sensitivity was 75% (95% CI 68% to 80%) specificity 90% (89% to 92%) and positive predictive value 45% (40% to 50%).

CONCLUSION

The accuracy of 112 dispatch of the MECU was found to be moderate. We suggest more training of dispatch staff and medical supervision.

Accuracy in identifying patients with loss of consciousness in a police-operated emergency call centre - first step in the chain of survival

BACKGROUND

The first link in the 'chain of survival' is the activation of Emergency Medical Services (EMS). In the major part of Denmark, police officers operate the alarm 1-1-2 centre, including calls for EMS. Our aim was to study the police 1-1-2 operators' accuracy in identifying calls concerning patients with loss of consciousness as a key symptom of life-threatening conditions. 'Unconsciousness' was defined as patients with a Glasgow Coma Scale (GCS) score of < 9, scored by the on-scene anaesthesiologist from the Mobile Emergency Care Unit (MECU).

METHODS

This study was an observational cohort study based on data from the Police 1-1-2 Database and the Aarhus County Pre-hospital Database containing data from MECU cases during 6 months in 2004-05.

RESULTS

Two thousand, three hundred and forty-three emergency calls with MECU dispatch were identified. In 1655 cases, both 1-1-2 data and the GCS score were recorded. Two hundred and ninety-five patients were found with a GCS score of < 9 at MECU arrival, 243 of whom were reported 'unconscious' by 1-1-2, giving a sensitivity of 82%. One thousand, three hundred and sixty patients were found with a GCS score of > or = 9, 972 of whom were reported 'awake', giving a specificity of 72%. The positive predictive value (percentage of patients found with a GCS score of < 9 at MECU arrival amongst patients reported as 'unconscious') was 39%.

CONCLUSIONS

The accuracy was moderate with room for improvement. The positive predictive value was low, indicating over-triage of MECU.

Improving the first link in the chain of survival: the Antwerp experience

OBJECTIVE

To evaluate the effects of a training programme for emergency medical dispatchers held between 1996 and 1997 in the Belgian city of Antwerp on decision-making regarding tiered emergency medical system responses (mobile intensive care unit teams or basic life support ambulances).

MATERIALS AND METHODS

Two training sessions based on standardized dispatching protocols were given to the dispatchers. To evaluate the effect of this training, 3000 emergency medical system calls were reviewed. Before and after each training session, audio recordings registered in the dispatch centre, hospital records and in-field provider registrations were analysed. The need for advanced life support was retrospectively assessed.

RESULTS

Increased dispatching sensitivity: the proportion of calls in which a mobile intensive care unit team was not dispatched along with the basic life support ambulance, although necessary, dropped from 64% at baseline to 55 and 40%, respectively, after the first and second training sessions. The main effect of the training sessions was a reduction in the amount of errors caused by lack of information. The specificity of the dispatching process, e.g. avoiding sending out mobile intensive care unit teams unnecessarily, was decreased by the first training session but returned to baseline after the second session.

CONCLUSION

Results suggest that only repetitive training efforts can result in increased dispatching sensitivity without decreasing dispatching specificity. We recommend the implementation of medically supervised and standardized dispatching protocols in a continuous education programme.

A systematic review of the evidence supporting the use of priority dispatch of emergency ambulances

OBJECTIVES

Systematic reviews of the literature assist in the location, appraisal, and synthesis of available evidence. This systematic review aimed to 1) assess the existing literature evaluating the effect of the priority dispatch of emergency ambulances on clinical outcome and ambulance utilization and 2) assess the relative effectiveness of sources of literature relevant to prehospital care.

METHODS

Systematic review. The quality of each paper was assessed using a standardized seven-point scoring schedule. Sources used were: Medline, the Cumulative Index to Nursing & Allied Health Literature (CINAHL), Bath Information & Data Services (BIDS), bibliographic searching, contacting researchers active in the field, and hand-searching relevant journals. Key words used were: "ambulance," "prioritisation," "dispatch," and "triage."

RESULTS

Three hundred twenty-six papers were identified: 64 (19.6%) were related to the prioritization of emergency ambulances, and only 20 (6.1%) contained original data. The overall quality of publications was poor, seven (35%) papers having a quality score > or = 4. Only half were identified by electronic databases, 55% were identified by people working in the field, and two (10%) were identified by hand-searching (some papers were identified by more than one source). Two high-quality papers support the concept that criteria-based dispatch (CBD) improves clinical outcome; two other papers support CBD's role in improving ambulance utilization.

CONCLUSIONS

There is very little evidence to support the effect of the prioritization of emergency ambulances on patient outcome. Electronic databases identify only approximately half of all relevant prehospital literature. Future systematic reviews in this area should use electronic databases, supplemented by contact with appropriate experts.

Do health care professionals report sudden cardiac arrest better than laymen?

OBJECTIVE

To compare the emergency calls made by health care providers and by laymen reporting a non-traumatic cardiac arrest, and to evaluate the handling of these calls by dispatchers.

METHODS

The study was conducted prospectively over a 1-year period in 1996. The callers (N=328) where divided in to three groups based on profession: I, doctors and nurses (N=33); II, other health care providers (N=19); and III, laymen (N=276). Main outcome measures where the information given by the caller, use of the dispatching protocol, recognition of the cardiac arrest, and survival to hospital.

RESULTS

Doctors and nurses told the dispatcher spontaneously what had happened in 67% of the calls when total strangers to the patient told it in 72%. Group I gave no information about the vital signs in 24% of the calls, group II in 0% and group III in 6%. Of the 52 phone calls made by groups I and II, in six cases the patient was not in cardiac arrest, in four the patient had already irreversible signs of death and in four only transportation to another hospital was requested for a patient in cardiac arrest. Of the professionals calling, 49 (94%) were on duty at the time of the call. The cardiac arrest was recognized by the dispatcher in group I in 70%, in group II in 74% and in group III in 73%. There where no statistical differences between the groups.

CONCLUSIONS

Our data do suggest that health care professionals, excluding those in emergency medicine, are not better than laymen in evaluating an emergency situation correctly, and when the caller is a doctor or a nurse the dispatcher seems to trust the evaluation of the situation to be correct and rarely asks any clarifying questions about vital signs of the patient.

The strength of specific EMS dispatcher questions for identifying patients with important clinical field findings

INTRODUCTION

There is growing interest in more efficiently matching emergency medical services (EMS) resources to patient need. Emergency medical services dispatchers may be asked to distinguish between callers with an immediate need for EMS and those who may safely use alternative services. New dispatcher protocols are required or existing protocols must be shown to be reliable for this new task.

OBJECTIVE

To examine whether answers to currently asked dispatcher questions in one urban center can identify callers with important clinical field findings (ICFFs).

METHODS

Audio recordings of EMS dispatcher-caller conversations within three nature codes (falls, sick, trauma) were retrospectively reviewed. Specifically scripted "cardinal" questions, asked of all callers, identify what happened, whether the patient is breathing okay, and whether the patient is conscious. "Key" questions are specific to each nature code and further specify patient circumstances. Compliance with protocol and caller answers were documented. Researchers developed a list of ICFFs that, if present on the corresponding EMS record, were judged to justify an immediate EMS response. Logistic regression was used to analyze the relationship between caller answers and the presence of ICFFs. A p-value of 0.10 was used.

RESULTS

Of 430 recordings, 383 (89%) were usable. Falls: 103 (26%); trauma: 136 (37%); sick: 144 (37%). The caller was the patient 41 (11%) times. There were 198 (52%) females in the sample. There was no matching EMS record for 96 (25%) cases. An ICFF was determined to be present in 191 (67%) of the 287 recordings with matching EMS data. Compliance across the cardinal and key questions ranged from 62% to 88%. Age alone was suggestive of a patient who may be identified at dispatch as having an ICFF [adjusted OR 1.01 (90% CI: 0.999-1.025), p < 0.10]. No other key or cardinal questions were related to ICFFs.

CONCLUSION

Cardinal questions are most often asked. Implied or volunteered information is often relied upon to answer key questions. Key questions for certain nature codes are not answered about one third of the time. Increasing age may suggest a stronger likelihood for an ICFF to be identified at dispatch. Otherwise, in this sample, caller answers to currently asked questions do not appear useful if the goal is to identify at dispatch those without an ICFF.

Out-of-hospital cardiac arrests in Amsterdam and its surrounding areas: results from the Amsterdam resuscitation study (ARREST) in 'Utstein' style

The purpose of this study was to describe the chain of survival in Amsterdam and its surroundings and to suggest areas for improvement. To ensure accurate data, collection was made by research personnel during the resuscitation, according to the Utstein recommendations. Between June 1, 1995 and August 1, 1997 all consecutive cardiac arrests were registered. Patient characteristics, resuscitation characteristics and time intervals were analyzed in relation to survival. From the 1046 arrests with a cardiac etiology and where resuscitation was attempted, 918 cases were not witnessed by EMS personnel. The analysis focussed on these 918 patients of whom 686 (75%) died during resuscitation, 148 (16%) died during hospital admission and 84 patients (9%) survived to hospital discharge. Patient and resuscitation characteristics associated with survival were: age, VF as initial rhythm, witnessed arrest and bystander CPR. EMS arrival time was significantly shorter for survivors (median 9 min) compared to non-survivors (median 11 min). In 151 cases the police was also alerted and arrived 5 min (median) earlier than EMS personnel. Using the OPC/CPC good functional health was observed in 50% of the survivors and moderate performance in 29%. All links in the chain of survival must be strengthened, but equipping the police with semi-automatic defibrillators may be the most useful intervention to improve survival.