Validity and risk factor analysis for helicopter emergency medical services (HEMS): The emergency call dispatch of Japanese air ambulances (Doctor-HeliTM️)

Noriaki YAMADA (  hokken@go2.enjoy.ne.jp ) Gifu University School of Medicine Graduate School of Medicine: Gifu Daigaku Igakubu Daigakuin Igakukei Kenkyuka https://orcid.org/0000-0003-4714-5787 Yuichiro KITAGAWA Gifu University School of Medicine Graduate School of Medicine: Gifu Daigaku Igakubu Daigakuin Igakukei Kenkyuka Takahiro YOSHIDA Gifu University School of Medicine Graduate School of Medicine: Gifu Daigaku Igakubu Daigakuin Igakukei Kenkyuka Sho NACHI Gifu University School of Medicine Graduate School of Medicine: Gifu Daigaku Igakubu Daigakuin Igakukei Kenkyuka Hideshi OKADA Gifu University School of Medicine Graduate School of Medicine: Gifu Daigaku Igakubu Daigakuin Igakukei Kenkyuka Shinji OGURA Gifu University School of Medicine Graduate School of Medicine: Gifu Daigaku Igakubu Daigakuin Igakukei Kenkyuka

In medical diagnoses, a 'killer word' refers to words/phrases that predict life threatening and serious deceases, such as the sudden onset of chest pain, hemiplegia, unconsciousness, and collapsing.
In the emergency department, if the triage nurse/resident uses these words/phrases, these patients' evaluations and treatments are prioritized compared to other patients in the emergency room. However, we do not decide the triage level depending only on the keyword. Rather, we determine the triage level using patient statements and their medical history. For example, there are tools for clinical decision making, such as the Canadian Triage and Acuity Scale (CTAS) 1) . This scale is used for the rst triage, which is decided based on situation and symptoms. Prehospital CTAS 2) scores are used in North America. As another example, the JUST score 3) has been used for transport decisions in stroke care.
These scoring systems can help paramedics classify patients with suspected stroke.
Keyword responses are used for helicopter emergency medical service (HEMS) dispatch in Japan. This means that emergency medical communication centre operators dispatch HEMS according to these killer words. Therefore, we set up the key phrase tase killer word in the theory book for the operator. This system allows for rapid responses. However, the operator's response has an increasing probability of failure. Speed of response is important, but the validity of dispatch should be improved because HEMS are scarce. However, there are few studies investigating the accuracy of HEMS calls. We explored some studies regarding trauma dispatch and identi ed their criteria for selection/triage. 4) 5) In addition, we identi ed their various dispatch criteria/strategy for speci c statuses and diseases. 6)7) However, there are no studies that have systematically investigated the validity of dispatches. Therefore, we investigated this. In addition, by reviewing the records, we examined the trends of these predictive terms under speci c situations and evaluated previous records to improve our research quality and prediction of keyword phrases. In the future, we expect to establish a commander-assist scale and systems such as the JUST score. We performed this study as a pilot study to use its results to inform future studies.

Materials And Methods
This study was a single-centre, retrospective observational study. We evaluated the operated HEMS (Doctor-Heli™) cases from 1 April 2015 to 31 March 2020 in Gifu University Hospital using our mission records from the national database registry project, J-HEMS. All mission data records are stored as part of the National Registry. Using these data records, we focused on the keywords and chest, chest and back pain, predicting cardiovascular diseases, sudden onset of hemiplegia predicting stroke, collapsing/unconsciousness predicting cardiopulmonary arrest, and any other internal medicine status to order HEMS cases, excluding cases with trauma and other external causes such as heat stroke. This means we focused on internal medicine emergency cases. In addition, we only focused on prehospital care and excluded transportation from hospital-to-hospital cases. We also excluded cases that were not suitable for analysis. For example, cases of patients with congenital diseases.
We evaluated the validity of medical emergencies such as emergency intervention and the necessity of hospital admission. In addition, we evaluated the validity of the suggested diagnoses. To examine the characteristics of each step, we evaluated the emergency validity in the rst dispatch, and the second was examined after being assessed by an emergency medical technician (EMT), because if the patient status was not suitable for HEMS response, the HEMS order would be cancelled.
We evaluated validity from three viewpoints: needs emergency intervention, needs admission to hospital, and the validity of the suggested diagnoses. Then, we evaluated risk factors for each viewpoint by performing a multivariate logistic regression analysis including the following predictor variables: phrases from the order summary (in particular, years of age, gender, situation, symptoms, and other characteristics) and dependent variables included: needs emergency intervention, hospital admission, and validity of the suggested diagnoses.

Doctor-Heli™ in JAPAN
In Japan, HEMS, called Doctor-Heli™, is organized by the government. Generally, each prefectural government body organizes and nancially manages the HEMS while being supported by the national government. This means that the public government takes responsibility for this air ambulance system. However, the actual operation of the HEMS is assigned to each hospital. At the end of 2018, 43 public bodies organized HEMS, and 53 aircrafts/helicopters were in use in Japan. Gifu University Hospital is one of these assigned hospitals. Their HEMS covers Gifu prefecture and some parts of neighboring prefectures. The HEMS operation started in February 2011, and 4252 operations were performed until 31 March 2020. Annually, approximately 500-600 operations are performed at this hospital; approximately 50% of the operations are for prehospital care, approximately 40% are transported to advanced care, and approximately 10% are cancelled. In Japan, the patient and their family cannot directly call Doctor-Heli™.
When emergency medical communication centres receive emergency calls, if the operator deems it necessary to call Doctor-hell™, the dispatch commander orders a Doctor-Heli TM mission. This operator is a staff member of the re department but is not entirely trained systematically.
Each operating hospital has a set call strategy for orders. The keyword list in Gifu University Hospital and the prefecture is shown in Table 1. Each organizing body/facility has a keyword list. Keywords have similarities but there are differences depending on their situation.

Statistical analysis
Fundamental statistics were obtained from observation data, which were calculated using Microsoft Excel for MAC ver.16.45. Multivariate analysis was performed using SPSS (IBM).

Ethical considerations
This study was performed as a part of the national database registry project, J-HEMS, using the project's data and Gifu University Hospital's medical records. This study was approved by the institutional ethical review board of Gifu University/Gifu University Hospital. Informed consent of the recorded patient was obtained by opt-out on the website. Those who rejected were excluded. In addition, we were given permission to use institutional data from the Japan Society for Aeromedical services.

Results
From 1 April 2015 to 31 March 2020 Gifu University Hospital had 2387 cases. We excluded 873 cases due to transport between hospitals for advanced medical care; 1043 cases due to trauma, other external factor diseases, and mass casualty incidents; and 19 cases because we judged them as being unsuitable for analysis, such as cases with congenital diseases affecting decision making and data insu ciency.
Details are shown in Fig. 1. As a result, 451 cases were included for emergency call analysis, and 376 cases were included for emergency call analysis. Demographic data are shown in Fig. 1.

1: Analysis of validity for HEMS orders
We evaluated validity from three viewpoints: needs emergency intervention, needs admission to hospital, and the validity of the suggested diagnoses. Details are shown in Table 2

3: List of initial diagnoses in hospital
The list of suggested cardiovascular disease cases and the results are presented in Table 3.

3a) Cardiovascular diseases
In the analysis of emergency calls, 55.9% of all cases were cardiovascular diseases. ACS was diagnosed in 51 cases, 23.0% of all suggested cases, and 40% of diagnosed cardiovascular cases. In addition, 27 cases (12.7% of all suggested, 21.7% of all diagnosed cases) were aortic diseases.
In the analysis after the rst assessment by an EMT, 63.6% were cardiovascular diseases. ACS accounted for 26.2% of all suggested cases, and 13.8% of all suggested cases were aortic diseases.

3b) Stroke
In the analysis for emergency calls, 55.6% of all suggested cases were strokes. Intracranial hemorrhage (ICH) occurred in 30 cases, 22.2% of all suggested cases, and 40% were diagnosed stroke cases. In addition, 29 cases (21.4% of all suggested, 38.6% of all diagnosed cases) were ischemic stroke, and 13 cases (9.6% of all suggested, 17.3% of all diagnosed cases) were subarachnoid hemorrhages (SAHs). In the analysis after the rst assessment by an EMT, 69.4% were stroke cases. ICH accounted for 27.8% of all suggested cases. In addition, 26.8% of all suggested cases were ischemic strokes (12% of all diagnosed cases) and were SAHs.

Risk analysis
In this study, we analyzed the factors that affect clinical decisions and outcomes. To reveal which phrases correspond to which complaints/symptoms, affecting clinical results such as emergency interventions, we analyzed various phrases from medical and operation records.
We performed a multivariate logistic regression analysis which included predictor variables: some phrases from order summaries (in particular, age, gender, situation, symptoms, and other characteristics), and dependent variables: needing emergency intervention, hospital admission, and validity of the suggested diagnoses.
In the analysis of the emergency calls, the risk factors for emergency intervention were years old, the situation was under sports, and the symptom was gasping. For hospital admission the risk factor was only years old. Validity for suggested diagnosis was only situations: under sports.
In the analysis of the rst assessment by an EMT, the risk factors for emergency intervention were years old, being male, situation: under sports, and gasping for air symptoms. For hospital admission the risk factors were years old, being male, having stroke symptoms or experiencing disturbance of consciousness. For validity of suggested diagnoses the risk factors was only situations under sports. The details are shown in Table 4.
We also analyzed suggested disease groups for cardiovascular diseases and strokes. a) Analysis for cardiovascular diseases In the analysis of emergency calls, the risk factors for emergency intervention were years old and situation: under sports, for hospital admission they were only years old and being male, and for validity of suggested diagnoses the only risk factor was only situations: under sports only.
In the second analysis, the risk factors for emergency intervention were years old, situation: under sports, for hospital admission they were years old and being male, and for validity of suggested diagnoses the only risk factor was situations: under sports. Details of the analysis are shown in Supplementary Table 1.

b) Analysis for stroke
In the analysis of emergency calls, the risk factors for emergency intervention were gasping for air, and downgrade factors were disturbance of consciousness and an emergency call from a family member. The risk for hospital admission was only gasping for air. The validity of the suggested diagnosis was also only gasping for air. The downgrade factor for validity of the suggested diagnoses was only a disturbance of consciousness.
After the rst assessment by an EMT, there were no risk factors for emergency intervention and hospital admission. The downgrade factor for validity of the suggested diagnoses was only a disturbance of consciousness. Details of the analysis are shown in Supplementary Table 2.

Discussion
In Japan, the number of emergency calls and HEMS dispatch continue to increase.8) This means that HEMS order numbers are likely to increase. However, it is quite di cult to increase the number of HEMS because of its very high cost. In addition, keyword responses are consulted when HEMS are dispatched in Japan, as the operator in emergency medical communication centres' head o ce or crew leader order HEMS referring to these killer words. Therefore, we set up the key phrase tase killer word in the theory book for the dispatch commander. This system can ensure rapid responses. However, this system can overestimate and increase the number of unnecessary cases. Therefore, we must develop the theory to guide decisions regarding which cases are high priorities.
First, we discuss the validity of the HEMS dispatch. It is di cult to de ne validity. In this study, we set admission as the relation between suggested diagnosis and initial diagnosis in the hospital, and necessity of admission as 'correct dispatch'. Regarding objectives for HEMS, we provide medical and de nitive care as soon as possible. Therefore, we evaluated possible factors of emergency interventions. Referring to previous studies, there are no studies evaluating the validity of dispatch. However, we found symptom-based research on emergency phone protocol. Ellensen et al. 9) investigated emergency medical communication centres' dispatch resources and transport for stroke patients in Norway. According to their results, the validity of suspected stroke was only 45.6% from the emergency call protocol. Burman 10) investigated data on the epidemiology of acute chest pain outside the hospitals in Norway. They highlighted that NACA-scores indicated that 26% of the patients were in a life-threatening medical situation. Judging from these studies, our analysis of the validity of using keywords when dispatching HEMS is warranted.
However, we believe this validity is insu cient. HEMS are scarce resources. Therefore, we performed a multivariate analysis with predictor variables being the phrases from order summaries (in particular, age, gender, situation, symptoms, and other characteristics). Dependent variables were the need for emergency intervention, hospital admission, and the validity of the suggested diagnosis. As a result, some keywords were identi ed as predicting factors.
This result is not only for the overview analysis, but also for the disease group analysis. Referring to previous studies, Munro 11) investigated the improvement accuracy of HEMS intervention using an algorithm approach and concluded that when aided by a bespoke algorithm the accuracy of HEMS dispatch improved.
Additionally, a similar approach is suggested for each symptom and disease group analysis. For example, Pedersen et al. 12) investigated chest pain in acute ambulance transport in the Central Denmark Region, and presented its pro le and the factors in uencing a patient being discharged without a severe cardiac diagnosis and surviving 30 days after a chest pain event. Ellensen et al. 9) investigated emergency medical communication centres' dispatch resources and transport for stroke patients in Norway, and highlighted possible factors associated with stroke prediction. In our study, a similar trend was found. There are possible factors associated with stroke prediction, for example, in the analysis being an elderly male and participating in sports were the predictive factors for emergency intervention.
From our results, risk factors and downgrading factors from the multivariate analysis were general and were not speci c to a patient's medical history. Therefore, it is not di cult for the communication centre operators to gather these medical histories if these risk factors are listed. In fact, Grusd et al. 13) attempted to analyse whether dispatch triage tools could reliably identify patients who only required transport by analyzing electronic and paper records of an ambulance service from four random days in 2012. They concluded that the Norwegian index was able to predict which patients do not need immediate medical treatment. This study explains 'downgrade' factors but using predicting systems could be bene cial.
Based on the ndings of Grusd et al. 13) , we suggest the following steps.
1. HEMS should be ordered based on the keywords listed in the guidelines.
3. HEMS personnel stay on the line while the emergency medical communication centre operators gather a second keyword. 4. HEMS operators receive this information, and then score and grade the case to ensure there is no overlapping.
This system could enhance to speed and decide the priority of each case and should be investigated in future studies. The nal goal should be to establish a scoring tool such as the EDACS in ACS 14) to improve HEMS in Japan.

Limitation Of This Study
This single-centre study focused only on one prefecture in Japan. Therefore, the results of this study only re ect the trend of this prefecture, and do not re ect the national Japanese trend. In addition, we only analyzed the order records of one hospital whose information on emergency calls and activities we had access to. Therefore, this information is limited and cannot provide generalizable results. Furthermore, we did not focus on 'underestimation' cases. This means that this study did not include the cases that were not called in, hence we could not determine the validity of cancelled orders.

Conclusion
There are some keyword/phrases that predict medical emergencies. Thus, HEMS dispatch commanders should gather these keyword/phrases. However, we found some trends in HEMS orders.
It is necessary for us to perform further analyses using a national database and to establish a set of guidelines to enhance the validity of clinical decision making.

Declarations
Ethics approval and consent to participate This study was performed as a part of National database registry project called J-HEMS, using this data and using medical record in our facility. This study was approved by institutional ethical review board in Gifu University/Gifu University Hospital. In addition, we are allowed to use our institutional data by Japan Society for Aeromedical services.
Informed consent of the recorded patient was obtained by opt-out on the website. Those who rejected were excluded.

Consent for publication
Informed consent of the recorded patient was obtained by opt-out on the website. Those who rejected were excluded.
Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.