Abstract
Background:
The National Emergency Medical Services (EMS) Information System (NEMSIS) Technical Assistance Center (TAC) collects and curates EMS activation level records for the United States. Originated as an outcomes assessment and service comparison tool, NEMSIS may have other high value clinical and public health uses.
Methods:
This study acquired a 100% activation level public dataset for 2019 from NEMSIS TAC and assessed item response quantities. Subsumption of NEMSIS terms within other controlled clinical vocabularies was also considered.
Results:
None of the assessed terminologies (LOINC, ICD10-CM, SNOMED-CT) could describe meaningful volumes of NEMSIS item response codes. The 2019 activation year dataset included 36,525 non-date/time or calculated distinct item responses for 43 activation descriptive items. Said item responses yielded 2,101,844,053 activation distinct non-blank responses. Several NEMSIS item responses had high clinical and public health value.
Conclusions:
NEMSIS can support multiple public health use cases in addition to EMS outcomes assessment. A comprehensive custom value set is appropriate to integrate NEMSIS item response codes into controlled terminologies, FHIR or hospital Electronic Health Record applications.
2. Introduction
Hospital care is relatively rare in the United States (US) when compared to non-hospital care[1,2]. In comparison, non-hospital care such as selfcare, outpatient, dental, poison control and urgent care are ubiquitous across the life course in the US. This difference in service volume is perhaps due to discrepancies in cost-effectiveness, acuity, referral requirements, insurances and a lack of specialized access restrictions in seeking non-hospital care[3–6]. Several authors have detailed the ‘clinical’ value of non-hospital data in predicting hospital and life course outcomes[7–9]. Given the relative volume of non-hospital services and the prevention value of ‘upstream’ intervention, the largest public health value may be found in non-hospital data standardization, data interoperability and terminology harmonization rather than hospital exclusive efforts[10–13].
Non-hospital clinical informatics remains largely under-described in peer review clinical (hospital) informatics journals, under-considered in controlled terminology development and left out of Common Data Model (CDM) development[14–17]. Non-hospital informatics may be under supported by national efforts for clinical terminology harmonization and hospital data integration. This inconsideration of non-hospital informatics has led to parallel development of terminology harmonization; one for hospital and another for non-hospital systems. These disparate terminologies should be integrated into hospital side data services if hospital data is to be informed, controlled and evaluated for non-hospital patient outcome contributions[18,19].
NEMSIS is the national data standard for EMS activation and outcome assessment in the US[20–22]. NEMSIS has successfully standardized all EMS activations for all states to a variable-item response model which is consistent over time. These variable-item responses form activation specific records and the TAC curates them to researchers as a national, activation level dataset. The success of the TAC in harmonizing EMS data practices necessitates an evaluation of existing controlled clinical terminologies which were not designed to support non-hospital data. To understand the potential for, and current informatics development of non-hospital data we investigate the National EMS Information System (NEMSIS) Technical Assistance Centers’ (TAC) Version 3 data dictionaries and activation level datasets.
This paper evaluates the contents of NEMSIS and how well Unified Medical Language System (UMLS) supported terminologies can approximate NEMSIS data dictionary variables and item responses. This paper describes an item response level map between NEMSIS and hospital Electronic Health Record (EHR) friendly vocabularies such as SNOMED-CT. Lastly, this paper further describes the item response frequency distributions contained in the NEMSIS Public 2019 dataset. If current clinical (hospital) terminologies cannot support EMS clinical data standards, an extension to common terminologies should be considered high priority and an ‘easy win’ for informatics[23–26].
3. Methods
3.1. NEMSIS Data Dictionary
This study uses four units of analysis, each with a different data collection effort. NEMSIS uses its own terminology code set to demarcate input terminal prompts (in an ambulance) which provides variables for EMS crews to populate with item responses. The NEMSIS V3 data dictionary used here was built from several PDF dictionaries, value sets and XML schemas which describe different aspects of NEMSIS data entry. Variable and item response descriptions were extracted from the dictionary into two study datasets. Variables with human readable descriptions (unit of analysis 1) and item response with human readable descriptions (unit of analysis 2) were considered as two separate units of analysis.
Both sets (unit of analysis 1 and 2) were submitted to MetaMap batch mode via National Library of Medicine’s indexing initiative services[27–31]. MetaMap is a complex tool that semantically parses human readable text ‘terms’ against prebuild language processing libraries and returns the UMLS Concept Unique ID (CUI) candidate matches for each submitted term. A submitted string, either variable name or item response can contain multiple ‘terms’ used for CUI matching. CUI’s are vocabulary agnostic and multiple UMLS vocabularies can contain the same CUI. All MetaMap supported vocabularies were considered for variable description unit of analysis. At item response level, only SNOMED-CT, LOINC, ICD10-CM and ICD10-PCS vocabularies were considered. This study considered the recall between CUIs returned from human readable variable names and item responses submitted to MetaMap that could pass manual (author) review as being related to EMS services.
Theoretically item responses and variable descriptions which are harvested from well documented vocabularies will be easily mapped by MetaMap to UMLS CUIs. MetaMap may discover worthy candidates to couch ‘organic’ terms in preexisting-controlled terminologies. NEMSIS codes are organic in most cases but may resemble controlled terminology codes. For example, there is an explicit LOINC value set for NEMSIS (182 terms) as well as ICD10-CM and SNOMED-CT subsets of item responses which are explicitly declared as such within NEMSIS. For example, to retrieve NEMSIS-LOINC codes search https://athena.ohdsi.org for the term ‘NEMSIS’, as NEMSIS-LOINC terms are postfixed with the string ‘<space>NEMSIS’.
3.2. Hand Mapped SOMED-CT Item Responses
This study also sought to create a map between supported, controlled, verbose vocabularies and item NEMSIS response values using SNOMED-CT (unit of analysis 3). The custom value set was produced to evaluate gaps in candidate to CUI assignment for unit of analysis 1 and 2. Both total and partial coverage terms were included where partial terms are not entirely described by existing UMLS CUIs. Here the unit of analysis is the ability of SNOMED-CT to fully describe NEMSIS terms. If SNOMED-CT cannot fully describe NEMSIS terms then a novel, controlled terminology should be considered or a terminology extension.
3.3. Item Response Distributions
The 2019 public use NEMSIS file was requested and retrieved from NEMSIS TAC. The file was ‘built’ into a distributed table series and mounted in google cloud storage. The data was queried through google big query API service. The files were transformed from ASCII (‘~|~ ‘) to CSV in R prior to mounting on google cloud. Event context values were aggregated to describe the spread of national activations within item responses to characterize the 2019 EMS activation volume. Distinct activations per event are described either per activation or per observation, if a repeat measure (unit of analysis 4). Examples of repeat measures include heart rate monitors and ECG outputs. Distinct observation and volume are presented below for choice measures. Rational count values (true counts, times, dates) were not considered as they are already standardized.
4. Results
4.1. Variable Level MetaMap Recall
At variable level our NEMSIS dictionary contained 643 human readable text strings (variables). These terms were processed in MetaMap using all available vocabularies to evaluate how well MetaMap supported UMLS vocabularies can describe EMS terms. MetaMap returned 4,045 candidate CUIs. 1,784 CUIs passed review. Assignments ranged from 18 CUIs per term to zero, or ‘term not supported’ with 22 variable level terms completely unrecognizable by MetaMap. The quality of recall is important when considering recall and accuracy. Because ‘concepts’ compose terms a concept within term match does not mean that a term was adequately described with CUIs, however accurate the assignment.
4.2. Item Response Level Recall
At item response level NEMSIS contained 3,250 human readable text strings (item-responses). These terms were processed in MetaMap using only LOINC, SNOMED-CT, ICD10-CM and ICD10-PCS. These select vocabularies are commonly used in Hospital EHRs in the US. MetaMap returned 7,409 candidate CUIs, 5,114 of which passed review. Assignments ranged from 0 CUIs per term to 54 with 130 item responses completely unrecognizable by MetaMap.
Table one describes individual examples of hand mapped NEMSIS item response terms with their SNOMED-CT candidates. Table one details the complexity of reusing a terminology standard like SNOMED-CT to approximate an organic terminology like NEMSIS. Several levels of correspondence are considered including complete correspondence within one SNOMED-CT term (Set 3), incomplete correspondence (Sets, 4, 5 and 6) across multiple SNOMED-CT terms as well as no correspondence within SNOMED-CT (Sets 1 and 2).
Table 1.
NEMSIS choice variables, item responses with SNOMED-CT candidates 4.3 Item Response Volumes
| Set | NEMSIS Variable | Item Response (Code) | SNOMED-CT |Candidate | Status |
|---|---|---|---|
|
| |||
| 1 | Triage Classification for MCI Patient | Black – Deceased (2708009) | Color Coded Trauma Scales | Absent |
| 1 | Triage Classification for MCI Patient | Black – Deceased (2708009) | Deceased | Absent |
| 2 | Final Patient Acuity | Emergent Yellow (4219003) | Color Coded Trauma Scales | Absent |
| 2 | Final Patient Acuity | Emergent Yellow (4219003) | Emergent | Absent |
| 3 | Cardiac Rhythm Electrocardiography (ECG) | Sinus Arrhythmia (9901043) | 71792006|Nodal rhythm disorder | Present |
| 4 | Hospital Capability | Trauma Center Level 1 (9908021) | 36125001|Trauma center (environment) | Present |
| 4 | Hospital Capability | Trauma Center Level 1 (9908021) | 277733009|Level 1 (qualifier value) | Present |
| 5 | Hospital Capability | Trauma Center Level 2 (9908023) | 36125001|Trauma center (environment) | Present |
| 5 | Hospital Capability | Trauma Center Level 2 (9908023) | 277734003|Level 2 (qualifier value) | Present |
| 6 | Type of Facility | Other EMS Responder ground (1701029) | 74964007|Other (qualifier value) | Present |
| 6 | Type of Facility | Other EMS Responder ground (1701029) | 409971007|Emergency medical services (qualifier value) | Present |
| 6 | Type of Facility | Other EMS Responder ground (1701029) | Responder (ground) | Absent |
The 2019 activation year dataset included 36,525 non-date/time item responses for 43 activation descriptive items. Item responses yielded 2,101,844,053 activation distinct non-blank responses where each activation-ID could have any item response once. For example, an activation where a patient reported two distinct repeat measures three times, perhaps an arrest case that had a return of circulation but then arrested again would gain two responses, arrest and return of circulation counts. A sample of records describing the 2019 item responses are contained in Table 2. Note NEMSIS has several ‘blank’ item response options which are not counted as physically blank responses in this study (consider NEMSIS term 7701001|Not Applicable, 7701003|No Recorded or 7701005|Not Reporting).
Table 2.
Choice health care system measures by NEMSIS variable, item response, item response code and distinct activation volumes, NEMSIS 2019
| Indication of whether or not there were any patient specific barriers to serving the patient at the scene | |
|
| |
| Item Response (Code) | Distinct Activations |
| Physically Impaired (3101015) | 4,43,095 |
| Language (3101007) | 3,68,038 |
| Unconscious (3101027) | 3,55,336 |
| Psychologically Impaired (3101019) | 3,52,173 |
| Uncooperative (3101029) | 3,41,280 |
| Obesity (3101011) | 2,60,180 |
| Speech Impaired (3101023) | 1,39,483 |
| Hearing Impaired (3101005) | 1,13,477 |
| Developmentally Impaired (3101003) | 1,01,067 |
| State of Emotional Distress (3101031) | 74,187 |
|
| |
| The reason the unit chose to deliver or transfer the patient to the destination | |
|
| |
| Item Response (Code) | Distinct Activations |
| Closest Facility (4220001) | 1,08,05,530 |
| Patients Choice (4220015) | 99,94,080 |
| Patients Physicians Choice (4220017) | 33,49,943 |
| Protocol (4220019) | 18,03,880 |
| Insurance Status/Requirement (4220007) | 3,33,525 |
| Law Enforcement Choice (4220009) | 1,20,035 |
| Diversion (4220003) | 1,08,984 |
|
| |
| Choice Delay Types by Delay Class | |
|
| |
| Item Response (Code) | Distinct Activations |
| (Response) Distance (2209005) | 6,35,941 |
| (Activation) Distance (2211005) | 2,23,081 |
| (Response) Directions/Unable to Locate (2209003) | 70,342 |
| (Scene) Directions/Unable to Locate (2210007) | 36,673 |
| (Turn-Around) ED Overcrowding / Transfer of Care (2212009) | 5,02,224 |
|
| |
| Alcohol/Drug Use Indicators | |
|
| |
| Item Response (Code) | Distinct Activations |
| Patient Admits to Alcohol Use (3117005) | 13,36,697 |
| Patient Admits to Drug Use (3117007) | 8,01,636 |
| “Smell of Alcohol on Breath” (3117011) | 4,45,032 |
| Alcohol Containers/Paraphernalia at Scene (3117001) | 2,59,748 |
| Drug Paraphernalia at Scene (3117003) | 1,56,675 |
4.4. Medications and Adverse Events
EMS data can support pharmacovigilance efforts where a medication of exposure, a clinical condition of interest and an Adverse Event (AE) is declared. Not unlike FAERS (Federal Adverse Event Reporting System) NEMSIS documents these features on an activation-by-activation basis. NEMSIS uses RxNorm CUI, which do not necessarily represent standardized RxNorm clinical drugs. Volumes in Table 3 should not be understood as ‘exclusive’ and additional volumes for the same pharmaco-active substance may well exist in several activation aggregated volumes. In this study NEMSIS contained 14,733,840 activations distinct ‘drug description’ events where drugs are described by the Rx-Norm CUI human readable string name provided by NEMSIS. Volumes are further confounded by multiple medication administrations within an activation.
Table 3.
Choice medication and adverse event measures by NEMSIS variable, item response, item response code and distinct activation volumes, NEMSIS 2019
| Response to Medication | |||
|
| |||
| Distinct Activations | Improved (9916001) | Unchanged (9916003) | Worse (9916005) |
| 74,18,819 | 57,48,622 | 26,445 | |
|
| |||
| Select Medications | Distinct Activations | Select AE from EMS Medications | Distinct Activations |
|
| |||
| Oxygen | 44,61,938 | Altered Mental Status (3708001) | 67,118 |
| Fentanyl | 10,03,435 | Respiratory Distress (3708035) | 40,752 |
| Ondansetron | 9,47,556 | Hypoxia (3708023) | 25,889 |
| Nitroglycerin | 7,69,992 | Nausea (3708029) | 24,269 |
| Aspirin | 7,67,982 | Hypotension (3708019) | 14,011 |
| EPINEPHrine 0.1 MG/ML | 6,64,168 | Apnea (3708003) | 11,566 |
| Sodium Chloride | 6,59,229 | Vomiting (3708041) | 9,868 |
| Naloxone | 4,69,585 | Injury (3708025) | 7,078 |
| Albuterol | 4,67,334 | Tachycardia (3708037) | 4,835 |
| Midazolam | 2,80,444 | Tachypnea (3708039) | 4,300 |
| Morphine | 1,62,704 | Hypertension (3708015) | 3,949 |
| Ipratropium | 1,45,796 | Bradypnea (3708009) | 3,128 |
| Ketamine | 1,34,536 | Bradycardia (3708007) | 2,709 |
| Heparin | 64,644 | Bleeding (3708005) | 2,276 |
4.5. Arrest, Resuscitation and Witnesses
Table 4 describes cardiac arrest values. Cardiac arrest is an important end point for out-of-hospital mortality. Who witnessed the arrest, the type of resuscitation attempted, the outcome of the resuscitation and rhythm detail is considered. Typically, rhythm is assessed using an ECG/EKG machine as a declared procedure under procedure item responses. This table assumes repeat measures and multiple states. For example, ‘Artifact’ finding on ECG is not exclusive with other ECG findings and may cooccur with other findings within an activation.
Table 4.
Choice cardiac arrest measures by NEMSIS variable, item response, item response code and distinct activation volumes, NEMSIS 2019.
| Indication of who the cardiac arrest was witnessed by | |
|
| |
| Item Response (Code) | Distinct Activations |
| Not Witnessed (3004001) | 2,42,806 |
| Witnessed by Family Member (3004003) | 86,291 |
| Witnessed by Healthcare Provider (3004005) | 73,274 |
| Witnessed by Bystander (3004007) | 36,740 |
|
| |
| Indication of an attempt to resuscitate the patient who is in cardiac arrest | |
|
| |
| Item Response (Code) | Distinct Activations |
| Initiated Chest Compressions (3003005) | 3,01,857 |
| Attempted Ventilation (3003003) | 2,10,974 |
| Attempted Defibrillation (3003001) | 83,338 |
| Not Attempted-Considered Futile (3003007) | 81,579 |
| Not Attempted-DNR Orders (3003009) | 16,178 |
| Not Attempted-Signs of Circulation (3003011) | 7,559 |
|
| |
| Documentation of the type/technique of CPR used by EMS | |
|
| |
| Item Response (Code) | Distinct Activations |
| Compressions-Manual (3009001) | 2,03,424 |
| Ventilation-Bag Valve Mask (3009013) | 1,66,325 |
| Compressions-Intermittent with Ventilation (3009009) | 90,291 |
| Compressions-External Plunger Type Device (3009005) | 62,232 |
| Compressions-External Thumper Type Device (3009007) | 15,654 |
| Compressions-External Band Type Device (3009003) | 11,111 |
| Ventilation-Impedance Threshold Device (3009015) | 9,367 |
| Compressions-Other Device (3009011) | 5,693 |
| Ventilation-Mouth to Mouth (3009017) | 1,522 |
| Ventilation-Pocket Mask (3009019) | 982 |
|
| |
| Indication whether or not there was any return of spontaneous circulation | |
|
| |
| Item Response (Code) | Distinct Activations |
| No (3012001) | 2,93,114 |
| Yes, Prior to Arrival at the ED (3012005) | 67,199 |
| Yes, Sustained for 20 consecutive minutes (3012007) | 22,453 |
| Yes, At Arrival at the ED (3012003) | 14,338 |
|
| |
| The patients cardiac rhythm upon delivery or transfer to the destination | |
|
| |
| Item Response (Code) | Distinct Activations |
| Asystole (9901003) | 84,042 |
| PEA (9901035) | 44,892 |
| Artifact (9901005) | 30,473 |
| Sinus Tachycardia (9901049) | 20,141 |
| Sinus Rhythm (9901047) | 18,833 |
| Ventricular Fibrillation (9901067) | 9,292 |
| Other (9901031) | 7,040 |
| Atrial Fibrillation (9901007) | 5,586 |
| Sinus Bradycardia (9901045) | 3,844 |
| Agonal/Idioventricular (9901001) | 3,330 |
| Paced Rhythm (9901033) | 3,182 |
| Unknown AED Non-Shockable Rhythm (9901063) | 2,481 |
| Sinus Arrhythmia (9901043) | 1,914 |
| Junctional (9901019) | 1,351 |
| STEMI Anterior Ischemia (9901051) | 1,279 |
| STEMI Inferior Ischemia (9901053) | 1,256 |
| Ventricular Tachycardia (With Pulse) (9901069) | 1,139 |
| Right Bundle Branch Block (9901041) | 740 |
| Ventricular Tachycardia (Pulseless) (9901071) | 726 |
| STEMI Lateral Ischemia (9901055) | 683 |
| Unknown AED Shockable Rhythm (9901065) | 422 |
| Premature Ventricular Contractions (9901039) | 399 |
| Left Bundle Branch Block (9901021) | 347 |
| Supraventricular Tachycardia (9901059) | 335 |
| AV Block-1st Degree (9901011) | 321 |
| AV Block-3rd Degree (9901017) | 265 |
| STEMI Posterior Ischemia (9901057) | 168 |
| Atrial Flutter (9901009) | 138 |
| AV Block-2nd Degree-Type 2 (9901015) | 117 |
| Torsades De Points (9901061) | 114 |
| Premature Atrial Contractions (9901037) | 88 |
| Non-STEMI Inferior Ischemia (9901025) | 78 |
| Non-STEMI Anterior Ischemia (9901023) | 74 |
| Non-STEMI Lateral Ischemia (9901027) | 62 |
| AV Block-2nd Degree-Type 1(9901013) | 41 |
| Non-STEMI Posterior Ischemia (9901029) | 19 |
4.6. Injuries
Injury codes are described below. Typically synonymous with ICD10-CM codes, NEMSIS provides their own descriptions which may alter or confound interpretation and use. For example, the NEMSIS term ‘fall on moving sidewalk’ is defined in ICD10-CM as ‘Fall on same level from slipping, tripping and stumbling’ for code ‘W01’ and Collision NOS (V89.9) is defined as ‘Person injured in unspecified vehicle accident’ in ICD10-CM. The ‘reuse’ of controlled terminology codes is observed in NEMSIS. There were 2,303 distinct injury codes across 43,500,191 activation declarations. Choice injuries are detailed in Table 5 along with the descending rank of distinct activation volumes.
Table 5.
Ranked injury measures by NEMSIS variable, item response, item response code and distinct activation volumes, NEMSIS 2019
| Fall Injuries | ||
|
| ||
| Item Response (Code) | Distinct Activations | Rank |
| Fall on moving sidewalk (W01) | 11,55,585 | 1 |
| Accidental fall NOS (W19) | 3,39,171 | 3 |
| Other slipping, tripping and stumbling and falls (W18) | 2,10,951 | 4 |
| Fall from non-moving wheelchair, nonmotorized scooter and motorized mobility scooter (W05) | 23,106 | 36 |
|
| ||
| Vehicle Injuries | ||
|
| ||
| Item Response (Code) | Distinct Activations | Rank |
| Car accident NOS (V49.9) | 6,87,629 | 2 |
| Collision NOS (V89.9) | 1,52,226 | 6 |
| Person injured in collision between other specified motor vehicles (traffic), initial encounter (V87.7XXA) | 56,181 | 16 |
| Crashing of motor vehicle, undetermined intent (Y32) | 46,126 | 18 |
| Car occupant injured in collision with car, pick-up truck or van (V43) | 38,089 | 22 |
| Pedal cycle accident NOS, nontraffic (V19.3) | 24,768 | 32 |
|
| ||
| Assault and Firearms Injuries | ||
|
| ||
| Item Response (Code) | Distinct Activations | Rank |
| Assault by bodily force (Y04) | 1,56,123 | 5 |
| Homicide (attempted) NOS (Y09) | 66,449 | 13 |
| Assault by stabbing NOS (X99.9) | 39,349 | 20 |
| Unspecified firearm discharge, undetermined intent (Y24.9) | 11,970 | 58 |
| Intentional self-harm by unspecified firearm discharge (X74.9) | 7,544 | 70 |
| Discharge of gun for single hand use, undetermined intent (Y22) | 6,718 | 76 |
| Assault by other and unspecified firearm and gun discharge (X95) | 4,876 | 92 |
|
| ||
| Bite Injuries | ||
|
| ||
| Item Response (Code) | Distinct Activations | Rank |
5. Discussion
Within mapping efforts, item responses fared better by recall than variables. This is most likely due to NEMSIS using controlled terminologies supported by UMLS for some item responses, but not for variable names. However, neither mapping effort was able to comprehensively account for EMS terminology. Integration of EMS terms into UMLS and controlled, supported terminologies may improve EMS data integration and bulk FHIR efforts. Hospital Electronic Health Record friendly terminologies such as SNOMED-CT and ICD could produce EMS specific value sets as LOINC has, to better support emergency department handoff and hospital to EMS data workflows[32]. This study finds that a separate value set for EMS is appropriate as UMLS under supports EMS at present.
Table 2 offers a complex view of the care system in the United States. Patients have barriers to care which may be unknown at transfer to a nursing home or emergency department 33,34. Transfer sites are determined by several factors including insurance coverage, law enforcement decisions and distance. Evidence of substance use (not necessarily attributable to the patient) is regularly observed, complicating mental state attribution and management. Untold volumes of time are spent not caring for patients but managing transfers of patients to emergency medicine departments. Nemesis 2019 suggests that over 500,000 activations were delayed due to transfer complications; Emergency Department overcrowding and diversion are explicitly described as well. Disability, obesity and speaking a language other than English are both disparity exacerbating and ‘essential information’ that should be transmitted to the receiving facility; potentially via FHIR or bulk FHIR.
EMS data can support understandings of multi-facility health care systems as well as intrasystem causes of systemic disfunction[33–36]. The lack of a national standard for the identification of locations within buildings and the weakness of street address recall during an emergency further congest the care system. Some congestion is systematic, for example if a facility in a network has poor turnaround time due to the handoff facility being overcrowded there is a knock-on effect for EMS availability network wide (turn around delay). Diversion may seem like a logical solution, but diversion further complicates patient delivery within a network. Assessment and management of network capacity could benefit from a post-facility, activation level dataset like NEMSIS.
Table 3 describes choice medications and adverse events. Adverse events in table 3 are not necessarily attributable to the medications listed. Note Oxygen and Fentanyl are the two most common NDC/ RxNorm CUIs in NEMSIS 2019 at distinct activation level. The presence of complex therapeutics like Heparin being administered in an EMS setting is undoubtedly a complex issue well discussed elsewhere[37–40]. Hypoxia, respiratory distress, and cardiology AE are perhaps life threatening and potentially more dangerous than the patient chief complaint which originated the activation. Within EMS the identification, management and prevention of AE is high value. Towards the discovery of novel or candidate AE, EMS data is most likely and under used resource. Note there are over 450 thousand activations that reportedly administered naloxone, the opioid overdose antidote. EMS is a potential source of detail for the US opioid epidemic, which claimed over 80,000 lives in 2021[41].
Cardiac arrest values demonstrate that most arrests are not witnessed by anyone; perhaps driven by social, precariat or elder isolation[42–44]. As self-administered resuscitation is a non-option, care planning for high-risk individuals could consider living situation and the inherent hazard of an unwitnessed pre-hospital arrest[45,46]. Chest compressions are more common than ventilation; perhaps suggesting that ‘rescue breathing’ and its controversies have not fully subsided[47–51]. Device mediated resuscitation is much more rare than manual resuscitation, though bag breathing is more common than mouth-to-mouth. A return of circulation is seldom observed with 293 thousand activations reporting no return of circulation; Asystole is the most common state of rhythm at handoff when resuscitation is attempted followed by PEA (pulseless electrical activity).
Falls, Assaults (including firearms) and car strikes are observed injury causes in NEMSIS data. Bites are remarkably more common than expected, especially human-to-human. Falls were the most populated volume group considered followed by car attacks/accidents and human vs. human assaults. Though mechanism is not always declared within ‘assaults’, ‘firearm injury’ is observed, as is ‘self-inflicted firearm injury’ in NEMSIS data. The national tally of firearm mortality is known; but firearm morbidity is difficult to describe. NEMSIS could discover additional cases of gunshot wounds where the injury is known to EMS but unknown to downstream clinical surveillance and registered multiple cause of death certificates.
6. Conclusion
NEMSIS is under supported by ‘clinical’ terminologies and a controlled EMS extension is warranted for EHR supporting terminologies.
Figure 1. Accuracy counts and density of CUI candidate assignments to and from NEMSIS Terms, Variable (A) and Item Response (B) level.
The accuracy of the mapping is plotted as count and density plots for both variable level (A) and item response level(B). Candidate accuracy is the percent of the candidates that were accurately fit to (multiple) NEMSIS terms, while (NEMSIS) Term accuracy is the percent of the (NEMSIS) Terms accurately fit to (multiple) candidates. Term Accuracy A was densest at .25, while Term Accuracy B was densest at 1, or complete agreement. Candidate Accuracy A and B mirror each other’s density distributions, where both are U shaped suggesting that candidates were either high value or no value with very few being ‘mixed’ value. Count accuracies are varied but A and B sets appear consistent within counts.
7. Acknowledgement
This research was carried out by staff of the National Library of Medicine (NLM), National Institutes of Health with support from NLM.
Footnotes
Declaration of interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
9. Bibliography
- [1].McDermott KW, Sun R. Trends in Hospital Inpatient Stays in the United States, 2005–2014 n.d.:18.
- [2].Persons with hospital stays in the past year, by selected characteristics: United States, selected years 1997–2018; Table 040. 2019.
- [3].National Ambulatory Medical Care Survey: 2018 National Summary 2018:40.
- [4].Black L, Adjaye-Gbewonyo D. Urgent Care Center and Retail Health Clinic Utilization Among Adults: United States, 2019. National Center for Health Statistics (U.S.); 2021. 10.15620/cdc:106463. [DOI] [PubMed] [Google Scholar]
- [5].Gummin DD, Mowry JB, Beuhler MC, Spyker DA, Bronstein AC, Rivers LJ, et al. 2020 Annual Report of the American Association of Poison Control Centers’ National Poison Data System (NPDS): 38th Annual Report. Clin Toxicol (Phila) 2021;59:1282–501. 10.1080/15563650.2021.1989785. [DOI] [PubMed] [Google Scholar]
- [6].Nasseh K, Vujicic M. Dental care utilization steady among working-age adults and children, up slightly among the elderly. 2016.
- [7].Sočan M, Erčulj V, Lajovic J. Early Detection of Influenza-Like Illness Through Medication Sales. Cent Eur J Public Health 2012;20:156–62. 10.21101/cejph.a3735. [DOI] [PubMed] [Google Scholar]
- [8].Liu TY, Sanders JL, Tsui F-C, Espino JU, Dato VM, Suyama J. Association of Over-The-Counter Pharmaceutical Sales with Influenza-Like-Illnesses to Patient Volume in an Urgent Care Setting. PLOS ONE 2013;8:e59273. 10.1371/journal.pone.0059273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Magruder SF, Lewis SH, Najmi A, Florio E. Progress in Understanding and Using Over-the-Counter Pharmaceuticals for Syndromic Surveillance. Morbidity and Mortality Weekly Report 2004;53:117–22. [PubMed] [Google Scholar]
- [10].Dowker SR, Nallamothu BK. Uberization of Prehospital Communications—Are We Ready? JAMA Cardiology 2020;5:1321–2. 10.1001/jamacardio.2020.2990. [DOI] [PubMed] [Google Scholar]
- [11].Eastridge BJ, Mabry R, Blackbourne LH, Butler FK. We don’t know what we don’t know: prehospital data in combat casualty care. US Army Medical Department Journal 2011:11–5. [PubMed] [Google Scholar]
- [12].Holcomb JB, Niles SE, Miller CC, Hinds D, Duke JH, Moore FA. Prehospital Physiologic Data and Lifesaving Interventions in Trauma Patients. Military Medicine 2005;170:7–13. 10.7205/MILMED.170.1.7. [DOI] [PubMed] [Google Scholar]
- [13].Monahan AC, Feldman SS. Models Predicting Hospital Admission of Adult Patients Utilizing Prehospital Data: Systematic Review Using PROBAST and CHARMS. JMIR Medical Informatics 2021;9:e30022. 10.2196/30022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Wibring K, Magnusson C, Axelsson C, Lundgren P, Herlitz J, Andersson Hagiwara M. Towards definitions of time-sensitive conditions in prehospital care. Scand J Trauma Resusc Emerg Med 2020;28:7. 10.1186/s13049-020-0706-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Andersen SNL, Brandsborg CM, Pape-Haugaard L. Use of Semantic Interoperability to Improve the Urgent Continuity of Care in Danish ERs. Stud Health Technol Inform 2021;281:203–7. 10.3233/SHTI210149. [DOI] [PubMed] [Google Scholar]
- [16].Edeler B, Majeed RW, Ahlbrandt J, Stöhr MR, Stommel F, Brenck F, et al. LOINC in prehospital emergency medicine in Germany - experience of the `DIRK´-project. Methods Inf Med 2014;53:87–91. 10.3414/ME12-02-0015. [DOI] [PubMed] [Google Scholar]
- [17].Bodenreider O, Cornet R, Vreeman DJ. Recent Developments in Clinical Terminologies - SNOMED CT, LOINC, and RxNorm. Yearb Med Inform 2018;27:129–39. 10.1055/s-0038-1667077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Lee MJ, Martinez AJ, Rutledge LM, Maull KI. Informatics in Prehospital Care. In: Maull KI, Augenstein JS, editors. Trauma Informatics, New York, NY: Springer; 1998, p. 31–44. 10.1007/978-1-4612-1636-0_2. [DOI] [Google Scholar]
- [19].Gausche-Hill M, Krug S, Wright J. Emergency Medical Services (EMS) 2050: A Vision for the Future of Pediatric Prehospital Care. Prehospital Emergency Care 2021;25:91–4. 10.1080/10903127.2020.1734123. [DOI] [PubMed] [Google Scholar]
- [20].Ehlers J, Fisher B, Peterson S, Dai M, Larkin A, Bradt L, et al. Description of the 2020 NEMSIS Public-Release Research Dataset. Prehospital Emergency Care 2022;0:1–15. 10.1080/10903127.2022.2079779. [DOI] [PubMed] [Google Scholar]
- [21].Mann NC, Kane L, Dai M, Jacobson K. Description of the 2012 NEMSIS Public-Release Research Dataset. Prehospital Emergency Care 2015;19:232–40. 10.3109/10903127.2014.959219. [DOI] [PubMed] [Google Scholar]
- [22].Dawson DE. National Emergency Medical Services Information System (NEMSIS). Prehospital Emergency Care 2006;10:314–6. 10.1080/10903120600724200. [DOI] [PubMed] [Google Scholar]
- [23].Anantharaman V, Swee Han L. Hospital and emergency ambulance link: using IT to enhance emergency pre-hospital care. International Journal of Medical Informatics 2001;61:147–61. 10.1016/S1386-5056(01)00137-X. [DOI] [PubMed] [Google Scholar]
- [24].Mears GD, Pratt D, Glickman SW, Brice JH, Glickman LT, Cabañas JG, et al. The North Carolina EMS Data System: A Comprehensive Integrated Emergency Medical Services Quality Improvement Program. Prehospital Emergency Care 2010;14:85–94. 10.3109/10903120903349846. [DOI] [PubMed] [Google Scholar]
- [25].Krafft T, García Castrillo-Riesgo L, Edwards S, Fischer M, Overton J, Robertson-Steel I, et al. European Emergency Data Project (EED Project): EMS data-based health surveillance system. Eur J Public Health 2003;13:85–90. 10.1093/eurpub/13.suppl_1.85. [DOI] [PubMed] [Google Scholar]
- [26].Timbi-Sisalima C, Rodas EB, Salamea JC, Sacoto H, Monje-Ortega D, Robles-Bykbaev V. An Intelligent Ecosystem for Providing Support in Prehospital Trauma Care in Cuenca, Ecuador. Stud Health Technol Inform 2015;216:329–32. [PubMed] [Google Scholar]
- [27].Aronson AR. Effective mapping of biomedical text to the UMLS Metathesaurus: the MetaMap program. Proc AMIA Symp 2001:17–21. [PMC free article] [PubMed] [Google Scholar]
- [28].Humphreys BL, Tuttle MS. Something New and Different: The Unified Medical Language System. Stud Health Technol Inform 2022;288:100–12. 10.3233/SHTI210985. [DOI] [PubMed] [Google Scholar]
- [29].Weinzierl MA, Maldonado R, Harabagiu SM. The impact of learning Unified Medical Language System knowledge embeddings in relation extraction from biomedical texts. J Am Med Inform Assoc 2020;27:1556–67. 10.1093/jamia/ocaa205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Aronson AR, Lang F-M. An overview of MetaMap: historical perspective and recent advances. J Am Med Inform Assoc 2010;17:229–36. 10.1136/jamia.2009.002733. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Chiaramello E, Pinciroli F, Bonalumi A, Caroli A, Tognola G. Use of “off-the-shelf” information extraction algorithms in clinical informatics: A feasibility study of MetaMap annotation of Italian medical notes. J Biomed Inform 2016;63:22–32. 10.1016/j.jbi.2016.07.017. [DOI] [PubMed] [Google Scholar]
- [32].Martin TJ, Ranney ML, Dorroh J, Asselin N, Sarkar IN. Health Information Exchange in Emergency Medical Services. Appl Clin Inform 2018;9:884–91. 10.1055/s-0038-1676041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Randall Williams G, Hertelendy AJ. The expanding scope of EMS: decreasing emergency department overcrowding in the United States. Journal of Paramedic Practice 2014;6:614–8. 10.12968/jpar.2014.6.12.614. [DOI] [Google Scholar]
- [34].Cooney DR, Millin MG, Carter A, Lawner BJ, Nable JV, Wallus HJ. Ambulance Diversion and Emergency Department Offload Delay: Resource Document for the National Association of EMS Physicians Position Statement. Prehospital Emergency Care 2011;15:555–61. 10.3109/10903127.2011.608871. [DOI] [PubMed] [Google Scholar]
- [35].Olshaker JS, Rathlev NK. Emergency Department overcrowding and ambulance diversion: The impact and potential solutions of extended boarding of admitted patients in the Emergency Department. The Journal of Emergency Medicine 2006;30:351–6. 10.1016/j.jemermed.2005.05.023. [DOI] [PubMed] [Google Scholar]
- [36].Schneider S, Zwemer F, Doniger A, Dick R, Czapranski T, Davis E. Rochester, New York A Decade of Emergency Department Overcrowding. Academic Emergency Medicine 2001;8:1044–50. 10.1111/j.1553-2712.2001.tb01113.x. [DOI] [PubMed] [Google Scholar]
- [37].Dasararaju R, Singh N, Mehta A. Heparin induced thrombocytopenia: review. Expert Rev Hematol 2013;6:419–28. 10.1586/17474086.2013.814446. [DOI] [PubMed] [Google Scholar]
- [38].McMahon AW, Pratt RG, Hammad TA, Kozlowski S, Zhou E, Lu S, et al. Description of hypersensitivity adverse events following administration of heparin that was potentially contaminated with oversulfated chondroitin sulfate in early 2008. Pharmacoepidemiol Drug Saf 2010;19:921–33. 10.1002/pds.1991. [DOI] [PubMed] [Google Scholar]
- [39].Signorelli SS, Scuto S, Marino E, Giusti M, Xourafa A, Gaudio A. Anticoagulants and Osteoporosis. Int J Mol Sci 2019;20:E5275. 10.3390/ijms20215275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].O’Connor R, Persse D, Zachariah B, Ornato JP, Swor RA, Falk J, et al. Acute coronary syndrome: pharmacotherapy. Prehosp Emerg Care 2001;5:58–64. 10.1080/10903120190940353. [DOI] [PubMed] [Google Scholar]
- [41].Ahmad FB, Rossen LM, Sutton P. Provisional drug overdose death counts. 2021.
- [42].Courtin E, Knapp M. Social isolation, loneliness and health in old age: a scoping review. Health Soc Care Community 2017;25:799–812. 10.1111/hsc.12311. [DOI] [PubMed] [Google Scholar]
- [43].Freedman A, Nicolle J. Social isolation and loneliness: the new geriatric giants: Approach for primary care. Can Fam Physician 2020;66:176–82. [PMC free article] [PubMed] [Google Scholar]
- [44].Xia N, Li H. Loneliness, Social Isolation, and Cardiovascular Health. Antioxid Redox Signal 2018;28:837–51. 10.1089/ars.2017.7312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [45].Sasson C, Rogers MAM, Dahl J, Kellermann AL. Predictors of survival from out-of-hospital cardiac arrest: a systematic review and meta-analysis. Circ Cardiovasc Qual Outcomes 2010;3:63–81. 10.1161/CIRCOUTCOMES.109.889576. [DOI] [PubMed] [Google Scholar]
- [46].Zhan L, Yang LJ, Huang Y, He Q, Liu GJ. Continuous chest compression versus interrupted chest compression for cardiopulmonary resuscitation of non-asphyxial out-of-hospital cardiac arrest. Cochrane Database Syst Rev 2017;3:CD010134. 10.1002/14651858.CD010134.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [47].Berg DD, Berg RA. When Should Rescue Breathing Be Removed From the ABCs of CPR? Critical Care Clinics 2012;28:155–65. 10.1016/j.ccc.2011.12.001. [DOI] [PubMed] [Google Scholar]
- [48].Shekhar A, Mercer C, Ball R. Pre-hospital use of cpr-assist devices and return of spontaneous circulation: nationwide experience from the nemsis database. Journal of the American College of Cardiology 2020;75:3462–3462. 10.1016/S0735-1097(20)34089-4. [DOI] [Google Scholar]
- [49].Berg RA. Role of mouth-to-mouth rescue breathing in bystander cardiopulmonary resuscitation for asphyxial cardiac arrest. Crit Care Med 2000;28:N193–195. 10.1097/00003246-200011001-00005. [DOI] [PubMed] [Google Scholar]
- [50].Berg RA, Hilwig RW, Kern KB, Babar I, Ewy GA. Simulated mouth-to-mouth ventilation and chest compressions (bystander cardiopulmonary resuscitation) improves outcome in a swine model of prehospital pediatric asphyxial cardiac arrest. Crit Care Med 1999;27:1893–9. 10.1097/00003246-199909000-00030. [DOI] [PubMed] [Google Scholar]
- [51].Krarup NH, Løfgren B. [Rescue breathing is still a part of basic life support]. Ugeskr Laeger 2009;171:3531–5. [PubMed] [Google Scholar]