Creating an Infrastructure for Comparative Effectiveness Research in Emergency Medical Services

Abstract

Objectives

Emergency medical services (EMS) providers deliver the initial care for millions of people in the United States each year. The Institute of Medicine noted a deficit in research necessary to improve prehospital care, created by the existence of data silos, absence of long-term outcomes, and limited stakeholder engagement in research. This article describes a regional effort to create a high-performing infrastructure in southwestern Pennsylvania addressing these fundamental barriers.

Methods

Regional EMS records from 33 agencies in January 2011 were linked to hospital-based electronic health records (EHRs) in a single nine-hospital system, with manual review of matches for accuracy. The use of community stakeholder engagement was included to guide scientific inquiry, as well as 2-year follow up for patient-centered outcomes.

Results

Local EMS medicine stakeholders emphasized the limits of single-agency EMS research, and suggested that studies focus on improving cross-cutting, long-term outcomes. Guided by this input, more than 95% of EMS records (2,675 out of 2,800) were linked to hospital-based EHRs. More than 80% of records were linked to 2-year mortality, with more deaths among EMS patients with prehospital hypotension (30.5%) or respiratory distress (19.5%) than chest pain (5.4%) or non-specific complaints (9.4%).

Conclusions

A prehospital comparative effectiveness research infrastructure composed of patient-level EMS data, EHRs at multiple hospitals, long-term outcomes, and community stakeholder perspectives is feasible and may be scalable to larger regions and networks. The lessons learned and barriers identified offer a roadmap to answering community and policy-relevant research questions in prehospital care.

INTRODUCTION

Emergency medical services (EMS) provide essential care for millions in the United States each year, including hundreds of thousands who need time-sensitive care or who are at high risk of death.¹ As emphasized in the Institute of Medicine report Emergency Medical Services at the Crossroads,² many policy-relevant questions about EMS care are unanswered, leading to a narrow knowledge base about the effectiveness of the design, deployment, and interventions performed by EMS. Although notable advances in the prehospital care of those with cardiac arrest, stroke, and trauma expand the evidence base, these patients account for the minority of transports, with limited ongoing research into other patient conditions.³

Answering broader EMS research questions requires an infrastructure to perform prehospital comparative effectiveness research, for which the existing infrastructure is deficient. Foremost are the data silos for each step in the emergency care system. Distinct prehospital databases may be present within the same county or township,⁴ and these are infrequently linked to the subsequent in-hospital, electronic medical records, or post-discharge assessments.⁵ Although the National EMS Information System (NEMSIS) was established in 2001 to tackle these problems, it is plagued by variable participation and maturity.^6,7 EMS research infrastructure is also challenged by data privacy concerns, specifically those surrounding adherence to the Health Insurance Portability and Accountability Act (HIPAA). Currently, HIPAA protects identifiable patient information not only in clinical practice but also in research, and therefore may impede the linking of prehospital and hospital-based records in the absence of appropriate waivers. Finally, EMS sits at the interface of multiple community and health care stakeholders, all of whom deserve a voice in authentic research partnerships that are difficult to build.⁸

Based on the recommendations of the Institute of Medicine and the EMS Agenda for the Future,^2,9 we sought to address the barriers above by developing and testing a conceptual model for the infrastructure necessary to conduct policy-relevant comparative effectiveness research in EMS (Figure 1). This infrastructure has the following key components: 1) stakeholder partnership to guide research priorities, 2) patient-level linking of emergency care records across data silos, including electronic medical records in a single, multi-hospital health system, and 3) long-term follow-up beyond hospital discharge with patient-centered outcomes. We report empirical data on the feasibility of this infrastructure, describe lessons learned, and outline key policy initiatives that may facilitate extension of this infrastructure to a national scale.

Figure 1.

Schematic of high-performing EMS comparative effectiveness research infrastructure.

METHODS

Study Design

This was a retrospective case study of 33 EMS agencies operating in western Pennsylvania. Each agency receives medical direction and oversight from the University of Pittsburgh Medical Center (UPMC). The University of Pittsburgh Medical Center Quality Improvement Review Committee approved this case study as a quality improvement project without the requirement for informed consent (#0001074, 0001246); it does not meet the federal definition of research according to 45 CFR 46.102(d) by review to the University of Pittsburgh Institutional Review Board.

Study Setting and Population

In Allegheny County alone, these 33 EMS systems serve an estimated catchment of 1.2 million residents, and are supported by nine UPMC hospitals with more than 525,000 admissions per year. The electronic medical records of participating ground EMS agencies are stored in an electronic patient care report (ePCR) program (emsCharts, Inc. Pittsburgh, PA). emsCharts is a NEMSIS-compliant ePCR program that records detailed information on patient care and incorporates custom reporting software that allows aggregate data to be obtained across multiple agencies that are overseen by a single medical oversight facility. Aggregate data for all participating EMS agencies were obtained through emsCharts simultaneously to create a single dataset. We did not need to merge or link EMS records first as they all could be queried in the same database. We included all records from January 2011 from EMS agencies for patients who were 18 years of age or older and were transported from the scene to a hospital.

Study Protocol

We performed a deterministic record linkage of patient encounters in emsCharts to the Medical Abstract Reporting System (MARS) of the UPMC health system.¹⁰ MARS is an electronic medical record that combines multiple clinical data sources (e.g. laboratory, pharmacy) into a relational database directly accessible by university programmers. It does not include any prehospital data, outpatient clinic visits, or data outside the health system. Separate from our linkage, UPMC hospitals electronically supplement EMS demographic information into the emsCharts program for patients transported to these facilities through an electronic HL7 feed. This activity supports required elements for EMS agency billing. We also linked records to hospital-level data in the 2009 American Hospital Association survey to understand generalizability of our hospital system (e.g. intensive care unit beds, hospital beds). To demonstrate feasibility of linking prehospital records to long-term outcomes, we matched records to the Social Security Death Master File in April 2013 to follow patients for fixed time point mortality up to two years after their prehospital encounters using social security numbers (SSNs). The Social Security Death Master File is available online and supported by the Global Internet Management Corporation and the National Technical Information Service, an agency of the U.S. Department of Commerce. The Social Security Death Master File is reported to assign unknown status to approximately 10% of submitted records and underestimate deaths by 2% to 4%.¹¹ The pilot datasets were linked, managed, and stored on dedicated research services at the University of Pittsburgh.

Hospital Linkage Algorithm

We used a hierarchical algorithm of commonly available identifiers for the linkage of prehospital records, based on prior experience with EMS data.¹² We did not use probabilistic linkages due to the fidelity of the data and low rate of missing data. We used MS SQL Server 2010 (Microsoft Corporation, Redmond, WA) to complete the linkage using the following variables: first name, last name, SSN, date of birth, date/time of incident, and hospital destination. The algorithm steps were: 1) date of birth, first name, last name, SSN, hospital, date/time of incident ±30 minutes; 2) date of birth, SSN, hospital, date/time of incident ±30 minutes; 3) date of birth, first name, last name, hospital, date/time of incident ±30 minutes; 4) date of birth, last name, hospital, date/time of incident ±30 minutes; and 5) first name, last name, hospital, date/time of incident ±30 minutes. Subsequent refinements relaxed the hospital destination requirement as well as the date-time window to expand to 8 hours. We conducted a manual review by two independent blinded investigators of 100 randomly selected algorithm matches to verify linking accuracy. Records were randomly selected from all matched groups to avoid bias.

Stakeholder Involvement

We partnered with western Pennsylvania stakeholders in EMS medicine to form a Stakeholder Advisory Committee as part of a broader health care organization project funded by the U. S. Health Services and Resources Administration. This federal agency has an interest in improving access to EMS in the United States. The Advisory Committee helps oversee data assembly, offers input into key research questions that would make use of the data infrastructure, and acts as a liaison to policy makers. Committee members came from community advocacy groups, health care providers, and government agencies with an interest in emergency care. Initial partners included the Jewish Healthcare Foundation, the Pittsburgh Regional Health Initiative, the Hospital Council of Western Pennsylvania, the Pennsylvania Emergency Health Services Council, the Pennsylvania Office of Rural Health, the Pennsylvania Department of Public Welfare, and the Pennsylvania Bureau of EMS. The latter played a particularly key role, since this state agency is responsible for the oversight of patient care delivery, as well as policies and regulations that affect EMS agencies in Pennsylvania. Stakeholders met by conference call and in-person to discuss strategies for improving EMS, using an iterative consensus process to identify key data elements, and create a process for developing and prioritizing patient-centered research questions.

Assessment of Patient and Hospital Generalizability

To determine the external validity of research infrastructure, we characterized hospitals using total beds, intensive care unit beds, total admissions, and average daily census. We describe hospital urbanity using the hospitals’ ZIP code population density. We assessed teaching status by membership in the Council of Teaching Hospitals and proportion with resident full-time equivalents . We assessed hospital capability through the presence of adult interventional catheterization, liver transplant, and neurological services as examples of advanced hospital services. We report patient characteristics from the prehospital record to demonstrate the external validity of our cohort, including demographics, initial physiology and vital signs (e.g. heart rate, respiratory rate, Glasgow Coma Scale score, oxygen saturation, systolic blood pressure), and presenting prehospital medical category. We report missing data in both matching and physiologic variables. To illustrate how outcomes may be different using long-term follow-up, we grouped patients into categories that are the subject of recent investigation and have distinct prehospital protocols (e.g. hypotension, chest pain, respiratory distress).^13-16

Data Analysis

We describe the success rate of EMS record matching and categorical data using proportions, and characterize continuous patient and hospital characteristics using means with standard deviations (SD), and medians with interquartile ranges (IQR). We assessed match accuracy using the positive predictive value (PPV) compared to blinded manual reviewers.¹⁷ We illustrate the long-term mortality of EMS conditions using Kaplan Meier curves, comparing among conditions with the log-rank statistic. We set the comparison alpha level at 0.05 and used STATA, version 11.0 (Stata Corp., College Station, TX).

RESULTS

Stakeholder Perspectives

The Stakeholder Advisory Committee provided several key insights during the resource process. First, they agreed to the need for a population-based perspective on emergency care research, noting the essential limitation to performing research in a limited geographic area using data from single emergency care providers. Second, they affirmed the importance of long-term, patient-centered outcomes that cut across diseases and provide clinically relevant information back to individual providers. Third, they outlined key strategies for better engaging patients in the research process, including community outreach strategies that highlight not only past users of EMS, but also the “at-risk” population. These may include town hall forums with civic leaders and the use of religious groups/locations that act as natural meeting places, particularly in rural areas.

Accessing EMS records in a regional partnership

We pursued the first two objectives identified by stakeholders (e.g. many EMS agencies linked to long-term outcomes) in a regional partnership, and subsequently identified gaps in policy and barriers to the project (Table 1). First, we learned the creation of the dataset required a time window of approximately 4 months from the EMS encounter. This includes roughly 48 hours for prehospital data entry, 1 month for completion of the hospital record, followed by 2 to 3 months for data extraction and linking. The electronic prehospital records were also housed on local servers in emsCharts and primary data owned by the EMS agencies. Although Pennsylvania statewide EMS policy allows EMS directors to examine prehospital data for quality improvement, it is silent with regards to use of EMS data for research. Thus, we were required to draft data-use agreements for the maintenance of the collaboration beyond this pilot linking study. Such agreements required 6 months of review by attorneys from the health system, university, and EMS directors. No guidance or policy governed access of EMS agency data for patients transported to hospitals other than the hospitals in the study health system – and these records could not be accessed in the pilot study. Finally, the substantial financial resources required to maintain the prehospital linking beyond the pilot matching study were obtained from an investigator-obtained research grant from the National Institutes of Health.

Table 1.

Barriers encountered and potential solutions when accessing EMS records for a regional health system partnership

Observed Barriers	Potential Solutions
1. EMS records may not be electronic or  uploaded into receiving hospital  medical records.	Strive towards NEMSIS compliant, electronic  prehospital records across all EMS agencies  that are automatically uploaded to the  receiving hospital records at completion.
2. Prehospital databases may lack  capability to be sorted, cleaned, and  queried.	Consider proprietary software or local  programmers to develop relational databases  suitable for cleaning, queries, and linking.
3. EMS agencies lack DUAs with  researchers to participate in studies.	Multi-disciplinary process across university,  health system, EMS agency, and legal to  develop local DUAs generalizable for others.
4. Unable to access EMS records for  patients outside of health system  partnership.	Establish DUAs in advance of research  questions, perhaps with regional or statewide  mandates.

EMS = emergency medical services; NEMSIS = National EMS Information System; DUA = data use agreement

Matching to Electronic Medical Records

We found 6,013 EMS records from 33 agencies met eligibility criteria. Of these, 2,800 patients were transported to nine hospitals in the health system, and 3,213 were transported outside of our health system. During the study period, we observed 40,678 patient encounters in the nine-hospital health system, of which 7,595 were emergency department encounters. On average, the nine study hospitals had 232 total beds, 23 ICU beds, and encompass a range of population densities (Table 2). Less than half were teaching hospitals or offered advanced care such as organ transplant or adult cardiac catheterization. The nine study hospitals make up 40% of adult, non-federal facilities in southwestern Pennsylvania (total n = 23), and account for 60% of the hospital beds (2,367 of the 3,940 beds). Among patients (Table 3), we observed a heterogeneous cohort of mostly females (62%) with mean age of 64 years (SD ±22 years). Prehospital respiratory distress was the most common complaint (12%), while prehospital vital signs were largely normal. We found that 317 EMS records (11%) were missing SSNs, and < 1% were missing any of the other matching variables. Linking between EMS records and hospitalizations using SSN alone resulted in an 86% match rate (2,400 of 2,800). We then implemented a seven-step deterministic matching algorithm that required 84 work hours by computer programmers. We observed that 95% of EMS records (n = 2,675) were linked after all steps (Figure 2). Our blinded review revealed only three records to be a false match by either reviewer (positive predictive value [PPV] = 97%); all three false matches occurred when both SSN and date of birth were excluded from the matching step. The reviewed records were similar to the matched cohort with respect to age (60, SD ±24 years vs. 64, SD ±22 years) and hospital admission rate (48% vs. 45%). We also found that unmatched EMS records (n = 125) were dissimilar to matched EMS records (n = 2,675) in terms of age (p < 0.01), but similar in sex, prehospital systolic blood pressure, and respiratory rate (p > 0.05 for all).

Table 2.

Hospital Characteristics^*

Characteristic	Summary Statistic *
Total beds, median (IQR)	278 (207-395)
Intensive care unit beds, median (IQR)	23 (14-53)
Member of the Council of Teaching Hospitals, n (%)	3 (43)
Resident FTEs present, n (%) ^	2 (29)
Zip code population density per sq. mile, median (IQR)	5,488 (1,845-8,568)
Hospital services, n (%)
Adult interventional cardiac catheterization	3 (43)
Neurological services	7 (100)
Liver transplant	1 (14)
Total admissions, median (IQR)	17,031 (9,784-20,413)
Inpatient surgical operations, median (IQR)	5,819 (3,015-6,370)
Average daily census, median (IQR)	233 (156–282)

^*Of the 14 hospitals in our health system, nine were matched to prehospital records in our sample, and these collapsed to seven unique IDs in the AHA survey.

^{^}Corresponds to the number and proportion of hospitals that have any FTEs present (i.e. non-zero)

AHA = American Heart Association; FTE = full time equivalent; IQR = interquartile range

Table 3.

Patient Encounter Characteristics

Characteristic	Summary Statistic
Total no.	2,800
Age, yrs, mean (±SD)	64 (±22)
Male sex	1,053 (38)
Examples of dommon prehospital diagnoses
Respiratory distress	340 (12)
Fall	281 (10)
Medical, not otherwise specified	258 (9.2)
Chest pain	192 (6.9)
Abdominal pain	190 (6.8)
Prehospital vital signs
Systolic blood pressure, mean (±SD), mmHg	138 (±29)
<90 mmHg	105 (4)
>180 mmHg	194 (7)
Missing	84 (3)
Respiratory rate, breaths/minute, median (IQR)	18 (18-20)
< 12 breaths/minute	17 (1)
> 36 breaths/minute	39 (1)
Missing	95 (4)
Heart rate, beats/minute, mean (SD)	87 (75-100)
< 60 beats/minute	156 (6)
≥120 beats/minute	205 (7)
Missing	67 (2)
Oxygen saturation, median % (IQR)	97 (95-99)
< 88%	133 (5)
Missing	305 (11)
Glasgow Coma Scale score, median (IQR)	15 (15–15)
Score <8	66 (2)
Score 8 – 14	384 (15)
Missing	608 (21)

Data are reported as n (%) unless otherwise stated

Figure 2.

Proportion of EMS records (N=2,800) deterministically matched to electronic health records at nine hospitals. Each step #2 through #7 matches previously unmatched EMS records.

Long-term Outcomes

To assess patient-level survival, we removed duplicates encounters (277 of 2,675, 10%) from matched records. Of the remaining distinct patients, 2,262 matched to long-term outcome using SSNs (Figure 3). We observed that one in five prehospital patients died within one year of the prehospital encounter, and the survival curves of important prehospital syndromes differed in trajectory (Figure 4). Deaths at 90 days were more common for those with prehospital hypotension (31%) or respiratory distress (20.7%), compared to chest pain (6.3%) or non-specific complaints (9.4%).

Figure 3.

Linkage algorithm and accrual of records.

Figure 4.

Kaplan Meier curve showing 2-year survival for EMS patients with chest pain (thin dashed), respiratory distress (thick dashed), hypotension (grey), and other diagnoses (black).

DISCUSSION

The large EMS case volume and absence of consumer choice mandates that prehospital care become a subject of rigorous comparative effectiveness research. We show that regional partnering of EMS agencies with an integrated health system is feasible, with infrastructure well suited to answer several key questions in the National EMS Research Agenda.¹⁸ We leveraged patient-level EMS data, electronic medical records at multiple hospitals, long-term outcomes, and community stakeholder perspectives in a model scalable to larger regions and networks.

We learned many lessons in this process. First, we found that overcoming missing and suspect data quality in the EMS record is key to project success. Others EMS systems report that missing data is common (as high as 50% for vital signs),¹² and is closely related to patient acuity.¹⁹ Such missing data may be a crucial barrier to linkages with hospitalizations and long-term outcomes, and may require sophisticated probabilistic linkage approaches or focused quality improvement. We found occasional missing data in our cohort of high- and low-acuity transports, but it did not limit the success of matching. This allowed us to track patients across multiple admissions, exclude duplicates, and avoid an incident-level database with repeat patients. We then could link unique, patient-level data to long-term mortality – a step not feasible with duplicate encounters in a single dataset, and a source of bias when analyzing datasets with duplicates. As described elsewhere, missing data in prehospital physiologic variables or other EMS fields poses a risk of bias during the analysis phase, and deserves dedicated methods such as multiple imputation.²⁰

Second, we demonstrate the feasibility and success of community stakeholder participation in creating a prehospital research infrastructure. Prior examples of stakeholder participation include partnerships established for community consent in emergency research with exception from informed consent under 21 CFR 50.24. This pilot experiment extends community consent to include in-person stakeholder meetings using a format recommended by the Patient-Centered Outcomes Research Institute (PCORI) for “research done differently.”⁸ PCORI requires stakeholder engagement for funding consideration in a process that involves them early, to develop questions relevant to the community, to foster collaboration with and recruitment of patients, and to promote dissemination of research results. Our grass-roots engagement of stakeholders required little funding and is translatable to most communities.

The EMS research partnership required data management support. Hospital-based data managers or honest brokers are often not initially familiar with prehospital records. Larger efforts such as the National Trauma Data Bank or NEMSIS try to overcome these barriers by leveraging federal and non-federal support, whereas future regional efforts may require internal funding, investigator awards, or foundation grants.²¹ Faced with constrained reimbursement,²² few EMS agencies can budget for informatics expertise to bridge their data silos and develop sustainable research collaborations. The problem will be further compounded when extending research partnerships across competing health systems.

We faced regulatory challenges about the ownership and sharing of EMS data for research. These included the establishment of individual data-use agreements between EMS agencies, receiving hospitals, and academic institutions home to researchers. As noted in other regions and clinical trials groups,^4,23 many community EMS agencies lack a supervising institutional review board (IRB) and Federal Wide Assurance (FWA). As established by the Office for Human Research Protections (OHRP), an FWA is an agreement that defines an institution’s obligation to comply with federal regulations governing research. An FWA is required when agencies or investigators receive federal funding for research.²⁴ These regulatory steps may not always be necessary in health services or comparative effectiveness research, as most individual EMS agencies are unlikely to seek direct federal funding. Because the University of Pittsburgh maintains an FWA, additional FWAs were not required for this pilot study. A more important challenge is that smaller, low volume EMS agencies may be the least likely to establish IRB oversight, but whose participation in prehospital research is the most needed.²⁴

Our model for data infrastructure has many implications. As we seek better evidence to improve EMS care, single-agency cohort studies and traditional randomized trials will be unable to fully satisfy the research agenda. Rather, high-quality EMS comparative effectiveness research is now common in other countries with national participation, mandatory data capture, and more sophisticated data infrastructure.^25-28 As the federal PCORI plans a new clinical research network,²⁹ it is important that EMS seek out ways to participate. The inclusion of prehospital data in centralized networks will support key PCORI goals of coverage of large, diverse populations, and the creation of “rapid learning networks.” Alternatively, EMS may be best suited to participate in Agency for Healthcare Research and Quality “distributed research networks” where source data may be analyzed by remote researchers but remain under physical and logistical control of EMS agencies.³⁰ And although 2-year survival is an unlikely primary outcome for EMS research, this pilot study provides “proof of concept” for linkage to later outcomes that are important to patients.

Beyond research, our model infrastructure has applications to other aspects of the EMS system. A link to patient-level long-term data will facilitate quality improvement for EMS personnel, agencies, and regions. Mature data systems that extend beyond the EMS encounter will also grow real-time performance improvement, benchmarking, and quality initiatives across participating agencies.^31,32 The data infrastructure may even grow the development and testing of financial incentives and reimbursement models that promote incentivized EMS care.²² And as other fields embrace fixed time point mortality as primary outcomes,^33,34 EMS will be similarly positioned to look beyond process measures or other surrogates.

LIMITATIONS

We recognize limitations to our current model. First, we designed our approach in western Pennsylvania specific to our EMS and hospital attributes. Although the elements of this process are generalizable to other regions, the model will require adaptations when scaled up. For example, the missing data in both prehospital records and electronic medical records may result in differential matching rates when applied in other settings.^20,35 Similarly, communities have different historical interaction and investment with EMS agencies, such that stakeholder participation will vary. In fact, our initial stakeholder engagement effort could be expanded in the future to include more direct providers of clinical prehospital care, and more rigorous qualitative methods to derive consensus (e.g. Delphi methods). The project also took substantial time commitment for regulatory approvals, data extraction and linking, and maintenance of databases. Under-resourced EMS agencies may not have the academic or financial support to engage in this time-intensive process. Finally, the nature of hospital locations, provider networks, and cooperation will affect any approach to establish an EMS research infrastructure linked to more than one or many distinct hospital systems.

Several policy changes may facilitate the broader adoption of similar infrastructure for EMS comparative effectiveness research. First, we endorse legislative agendas that address the privacy concerns imposed by HIPAA and FWA requirements for EMS research.³⁶ One approach is to mandate the collection of electronic patient care reports and identifiable data at the state level, with the facility to link these records to relevant patient outcomes. This process is inconsistently accomplished in only some states, and identifiers are often not available to researchers;³⁷ a legislative approach could help reduce community EMS agencies’ burden to obtain and maintain individual data use agreements and FWAs. And as shown in the Resuscitation Outcomes Consortium,²³ such parent agreements could reduce the extensive time required for regulatory preparation.

Second, there is a clear need for funding to support EMS research. Funding limitations are the main reason for inadequate data infrastructure, as found by the National Association of State EMS Officials.⁵ No one single federal agency funds EMS research, although the newly formed National Institutes of Health Office of Emergency Care Research will help coordinate acute care efforts in the future.³⁸ Currently, most EMS research endorsed by this office focuses on cardiac arrest, trauma, and stroke – conditions afflicting a minority of patients receiving prehospital care.³ Only three of 1,981 NIH mentored career development awards (K08, K23) and one institutional research training program (NHLBI K12) in 2013 address prehospital topics, a barrier for the next generation of EMS investigators.³⁹

Finally, a robust infrastructure with efficient capture of patient-centered outcomes provides the framework for value-based purchasing in prehospital care.²² As U.S. health care seeks to incentive higher quality care, EMS agencies will be further motivated to contribute complete and accurate data for linking to performance metrics. To date, a comprehensive value assessment of EMS care does not exist, yet millions depend on it “being there” in times of crisis.

CONCLUSIONS

We describe an infrastructure for EMS comparative effectiveness research based on combining stakeholder recommendations and unique data systems. We note the existing challenges and solutions needed to integrate data silos in the emergency care system and demonstrate the potential to capture long-term, patient-centered outcomes for EMS.