a b s t r a c t

Study objective: In Rural settings, long distances and transport times pose a challenge for achieving early reperfusion goals in patients with ST-elevation myocardial infarction . This study investigated the association between the method of pre-hospital 12-lead ECG transmission (radio transmission vs. cellular phone transmission) and the success of transmission and legibility of 12-lead ECGs in a rural setting.

Methods: Observational study of pre-hospital 12-lead ECG transmission to the emergency department (ED) in a predominantly rural area. Success of transmission and the legibility of the 12-lead ECG were analyzed to identify barriers to 12-lead ECG transmission and reasons for failed transmission.

Results: Emergency medical services performed ECGs on 1140 patients, 917 of which they attempted to transmit, including 43 cases requiring emergent catheterization. Twelve-lead ECG transmission was successful in 236 (70%) of 337 radio attempts and 441 (76%) of 580 cellular attempts (difference 6.0%, 95%

CI 1.1-12.1). Legibility increased from 164 (49%) of 337 radio attempts to 389 (67%) of 580 cellular attempts

(difference 18.4%, 95% CI 11.8-24.9).

Conclusion: The success of transmission and legibility of 12-lead ECGs was significantly higher with cellular technology by emergency medical service agencies in comparison to radio transmission. In rural settings with lengthy transport times, utilization of cellular technology for transmission of pre-hospital 12-lead ECGs may improve Door-to-balloon times for STEMI patients.

(C) 2013

1. Introduction
   1. Background

In patients with ST elevation myocardial infarction , early reperfusion with percutaneous coronary intervention (PCI) is more effective in improving mortality [1] than thrombolysis [2,3]. A 30- minute delay in reperfusion can increase 1-year mortality by 7.5% [4], and attaining reperfusion should occur within 90 minutes of arrival at the hospital according to the American College of Cardiology/ American Heart Association STEMI guidelines [5]. Among the factors affecting door to balloon times is the early identification of STEMI patients and activation of the catheterization lab while the patient is en route to the hospital [6]. Amit et al showed that pre-hospital trans- mission of Electrocardiograms could reduce door-to-balloon

? Support: This research received no support in the form of equipment, drugs, or grants.

* Corresponding author. University of Vermont Dept of Surgery, 89 Beaumont Ave, Burlington VT 05405, USA. Tel.: +1 802 656 4216; fax: +1 802 656 0860.

E-mail address: [email protected] (K. Freeman).

¹ Formerly with University of Vermont College of Medicine, Burlington, VT.

² Formerly Vermont EMS District 3, Burlington, VT.

time from 94 minutes to 70 minutes [7], even though obtaining a 12- lead ECG in the field increased on-scene time slightly. In one sys- tematic review, an increase of 1.2 minutes on scene had a concomitant reduction in door-to-balloon time of up to 36.1 minutes [8].

Importance

Dhruva et al showed through the STAT-MI trial that a wireless network transmitting ECGs to emergency department (ED) and offsite cardiologists can meet this goal of door to balloon time (D2B) with an improvement in mean door-to-intervention time from 145.6 minutes to 80.1 minutes [9]. The acquisition and transmission of pre-hospital 12-lead ECGs is one of the strategies identified by Bradley et al in their survey of 365 hospitals that have reduced D2B⁷. Activation of the catheterization laboratory while the patient was en route to the hos- pital was shown to improve D2B by 15.4 minutes. The use of pre- hospital ECGs for identification of STEMI has long been recognized as a potential means of reducing the Time to intervention [10-12], but remains an underutilized tool.

Implementing cellular technology from a cardiac monitor to cellular phone (using “Bluetooth” (TM) technology) in ECG transmission has allowed for high-speed transmission of ECGs to both ED

0735-6757/$ – see front matter (C) 2013 http://dx.doi.org/10.1016/j.ajem.2013.07.032

physicians and cardiologists. The ability to transmit 12-lead ECGs in rural settings by emergency medical technicians with cellular technology has not been studied. Werman et al studied the ability of EMTs and EMT-Is to obtain and transmit a 12-lead ECG in rural Ohio, but used satellite phones, not radio or cell phones [13]. Their sam- ple size was also small (89 patients) because it was a feasibility study. Niles et al evaluated their rural medical center’s ability to improve process times, but focused on in-hospital procedures and just STEMIs, not all ECGs [14]. Although they concluded that pre-hospital ECGs reduced the time to the catheterization laboratory, they did not study the means of transmission or the adequacy of the tracings received from the field.

The history of emergency medical services (EMS) is replete with examples of interventions that were adopted on the basis of presumed effectiveness, only to be found later that they lacked this essential quality. This resulted in loss of limited financial resources, significant time, credibility on the part of the public and other health care professionals, and considerable work on the part of field providers and medical directors. Cellular transmission of ECGs is an example of an intervention that fits in this category of untested interventions and deserves evaluation before further implementation takes place.

Goals of this investigation

This report describes the success of transmission of EMS-obtained 12-lead ECGs using radio transmission and cellular technology to a rural tertiary referral center’s ED.

1. Methods
   1. Study design and setting

This observational study began as a quality improvement effort to improve pre-hospital 12-lead ECG transmission to the Communica- tion Center of the ED at the academic medical center associated with the University of Vermont, Burlington, VT, during a phase-in of cellular transmission technology between February 2009 and Decem- ber 2010. This hospital is a tertiary referral center which provides primary cardiovascular catheterization laboratory services to a popu- lation in excess of 300000, performing approximately 65 emergency PCI procedures each year. This study was approved by the University of Vermont’s Committee on Human Research in the Medical Sciences. Consent was waived because the study was an analysis of data already being collected as part of quality improvement efforts and the data was coded to protect confidentiality of subjects. Data and findings are reported in accordance with the STROBE statement [15].

Selection of participants

The EMS District which serves the region surrounding the hos- pital had 165,303 residents living in 909 square miles in 2009 [16]. The area is in a large valley, but terrain is generally mountainous, with numerous “dead spots” for radio or cellular transmission. Of the 14 ambulance services providing 9-1-1 response in the district, all had universal coverage by emergency medical technician-basics (EMT-Bs) and at least some coverage by EMT-intermediates (EMT- Is), but none had paramedics at the beginning of the study period. The largest city in the county (and in the state) has 39000 residents, with 19000 in the next largest community. Although the busiest EMS agencies are in or near the population center of the region, more than half of the services have median transport times that exceed 15 minutes. The hospital associated with the university is the only one in the county and since the next closest hospital is 28 miles (45 km) away, it is the only ED for all but one of the ambulance services participating in this project.

EMT-Is receive approximately 160 hours of training beyond the EMT-B level that covers patient assessment, intravenous access, a supraglottic airway and limited pharmacology. The only cardiac medications EMT-Is can administer are oxygen, aspirin and nitroglyc- erin. ECG monitoring and interpretation are not included in EMT-I training. In September of 2010, 2 agencies began providing paramedic level service.

The EMS protocol requires that a 12-lead pre-hospital ECG be obtained for any patient 40 years of age or older with any of four chief complaints: non-traumatic chest pain, non-traumatic jaw or Arm pain, syncope, or resuscitation after cardiac arrest. Emergency medical technicians were encouraged to obtain ECGs on other patients in whom they thought it was appropriate and to obtain them as soon as possible without delaying transport significantly. Data was collected for all patients who had a 12-lead ECG performed in the field, regard- less of whether it was required. This included patients under the age of 40 and patients who did not have one of the four chief com- plaints for the purpose of assessing the ability to transmit the 12-lead ECG. The project was started as a Quality improvement project under the title “Chest Pain Continuous Quality Improvement” with the intention of enrolling at least 1000 patients. Data collection ceased shortly after this goal was reached.

Methods and measurements

Data collection was performed by academic research associates (RAs),who were trained in data collection procedures and the specific goals of the study. The RAs consisted of undergraduate students and post-baccalaureate students in the university’s pre- medical program, trained in accordance with principles described by Hollander et al [17]. They were available for data collection in the ED from 8:00 a.m. to midnight, 7 days a week, when the university was in session. Gaps in enrollment occurred during exam periods, vaca- tions and summer months. ECGs were interpreted by emergency medicine attending physicians before the patient arrived at the hospital. pre-hospital times obtained from the EMS Patient care report included: dispatch, arrival at the scene, departure from the scene, and arrival at the hospital. Research associates also recorded the patient’s sex, age and disposition after discharge from the ED. The ECG location (at the scene or en route) was determined by the time of departure from the scene.

Intervention

In November 2008, a group of benefactors affiliated with the hospital Department of Cardiology purchased a Zoll E Series monitor for each emergency ambulance in the EMS district. Since the costly equipment needed for cellular transmission was not immediately available, 12-lead ECGs were initially transmitted by means of a Rosetta translation device (General Devices, Inc, Ridgefield, NJ) over a VHF radio frequency, with the exception of one EMS agency, which had cellular technology at the beginning of the study period. Since each service planned to obtain cellular capability, the EMS district medical advisor saw this as an opportunity not only to evaluate EMT- obtained ECGs, but also to observe a “natural experiment” and deter- mine if cellular was truly superior to radio in a mountainous, rural area. Cellular transmission replaced radio transmission as the means of sending ECGs, in part because the radio frequency used for this purpose (155.340 MHz) was not licensed for such use. The frequency was also used for verbal ambulance-to-hospital communication and each kind of transmission interfered with the other. Between February and July of 2010, Bluetooth(R) transmission via cellular telephone was installed in ambulances belonging to the remaining thirteen emergency ambulance services in the district.

Outcomes

The primary outcome of successful ECG transmission was defined based on two criteria: if an ECG was both received by the Com- munication Center at the ED and whether it was legible. For purposes of the study, two contiguous illegible leads or significant artifact made the ECG illegible, as judged by the first two authors, who reviewed and discussed each ECG together.

Analysis

Statistical analysis was performed using SAS Version 9.3 (SAS Institute, Inc, Cary, NC). An ordinal logistic regression model using proportional odds was constructed with the outcome variable having three ordered levels: transmission not received (worst), transmission received but illegible (better) and legible transmission received (best). The proportional odds model, if assumptions are met, allows for one odds ratio for each covariate for evaluation from one ordered level of the outcome to the next. For example, the odds ratio for transmission by radio or cell will be the same, whether comparing no transmission received to illegible transmission received or comparing received but illegible transmission to legible transmission. Main effects in the model were: the ECG transmission method (radio or cellular); whether the EMTs were required to obtain an ECG (yes or no); where the ECG was transmitted from (scene or ambulance); and transport interval (continuous). Whether the ECG was required was included as a factor because EMTs might try harder to get an ECG through when they knew they were expected to get one. The last 2 factors, where the ECG was transmitted from (the scene or the ambulance) and the transport interval, were included to account for the expected difficulties of transmission when distance from the hospital was greater. All first order interactions were included as well. Additional covariates were omitted from the model because they were less plausibly influential on the outcome and to ensure there would be enough events per variable for effect estimates to be stable. No attempt was made to eliminate covariates from the model on account of the well documented difficulties of stepwise regression in distinguishing covariates that truly contribute to the model from those that do not [18]. Because 14 EMS agencies participated in the project, clustering by agency was accounted for by using generalized estimating equations. Assumptions for the logistic regression model were evaluated [19], including conformity with a linear gradient for continuous variables. Covariates were also evaluated for collinearity through assessment of correlations of parameter estimates of the fitted model and condition index. A condition index greater than 30 was considered to have severe collinearity. The proportional odds assumption was evaluated by inspection of plots. The score test for proportional odds is known to be anti-conservative, ie, it rejects the hypothesis of proportional odds more than it should, and the use of generalized estimating equations in SAS excludes the use of the LOGISTIC procedure, the only procedure that provides numerical values to evaluate the proportional odds assumption. The goodness- of-fit of the model was evaluated by means of model fit statistics. Standard logistic regression diagnostics were also evaluated. The threshold for statistical significance was P b .05. No attempt was made to validate the model since the purpose of the model was to deter- mine whether changing the method of transmitting ECGs made a difference, not to predict future success.

1. Results
   1. Characteristics of study subjects

Academic associates screened all ED patients for 47% of total time during the 23 months of the study. Although exact numbers are not available, they screened approximately 10000 of the 21169 EMS cases

that could have been potentially screened during the time period of the study (Fig.). Academic associates identified 1140 patients who had a pre-hospital 12-lead ECG. Baseline patient characteristics are displayed in Table 1 and were similar over the course of the study. The most common chief complaints for the study participants were chest pain (53.6%), syncope (13.6%), dyspnea (10.1%), and abdominal pain (13.7%). The mean total pre-hospital time was 36.2 minutes, with much of that time spent on-scene (13.7 minutes) and en route to the hospital (14.1 minutes). The majority (46.7%) of patients were dis- charged home, while 3.5% were transferred to the catheterization laboratory. A 12-lead ECG was performed for 735 of 914 (80.4%) patients who met the inclusion criteria. EMTs performed ECGs on an additional 405 patients.

1. Main results

The success of pre-hospital 12-lead ECG transmission was greater when cellular technology was used, with respect to both reception and legibility (Table 2). A 12-lead ECG was received in the ED for 677 (59.4%) of the study subjects with a significant increase in successful transmission from 236/337 (70.0%) for radio transmission to 441/580 (76.0%) for cellular transmission (difference 6.0%, 95% confidence interval [CI] 1.1%-12.1%). In addition, the legibility of the transmitted pre-hospital 12-lead ECGs improved from 164/337 (48.7%) to 389/580 (67.1%) (difference 18.4%, 95% CI 11.8%-24.9%). The remaining covariates (whether an ECG was required, transport interval and location ECG transmitted from) were not associated with either in- creasing or decreasing odds of success when considered alone. There were, however, 2 significant interactions.

When an ECG was required, the odds of success for cellular transmission more than doubled compared to the odds of success for radio transmission. In this ordinal logistic regression model, there are two definitions of success: reception of a legible ECG (compared to not legible) and reception of an illegible ECG (compared to no reception). Because of this dual definition, the proper interpretation of this model is that when an ECG was required, the odds of receiving a legible ECG were more than twice the odds of receiving an illegible ECG and the odds of receiving an illegible ECG were more than twice the odds of not receiving an ECG at all. Interpretation of the other interaction is analogous, but slightly more complex because transport time is a continuous variable. For each minute that transport interval increased, the odds for successful transmis- sion en route decreased to 0.94 times the odds of successful trans- mission from the scene. Informally, this means that when an EMS incident was farther from the hospital, the odds of a legible ECG were slightly less than the odds of receiving an ECG that was not legible. Similarly, the odds of receiving an ECG that was not legible were slightly less than the odds of not receiving an ECG at all. There were no disagreements regarding whether an ECG was legible.

By far, the most often cited reason for which a 12-lead ECG was not transmitted (Table 3) was equipment malfunction (27.3%) while the most common reason for an ECG not being obtained was short transport to the hospital (25.5%). Frequency of equipment malfunc- tion was similar between the two modes of transmission (29 for radio, 32 for cellular) and the primary outcome was not affected by including these cases in the analysis.

The field monitor interpreting the ECG provides one of two noti-

fications when it finds evidence of ST elevation: “** ** ** ** Acute MI **

** ** **” or “** ** ** ** Acute MI Suspected ** ** ** **.” The accuracy of these interpretations, as measured against the emergency physi- cian’s reading and the catheterization laboratory results, is shown in Table 4. When the machine read “Acute MI,” the physician agreed with 16 (67%) of the readings and 13 of the 14 patients who went to the lab had an artery opened. When the machine read “Acute MI Suspected,” the physician agreed with 5 (45%) of the readings and all 4 of those patients who went to the lab had an artery opened.

Fig. Characteristics of ECGs.

There was one case where the physician initially read an “Acute MI” as not a STEMI, but the patient went to the catheterization laboratory approximately 1 hour after arrival. His first ED ECG did not show ST elevation, but a subsequent one did.

Assumptions for the proportional odds logistic regression model were validated. Details are in the Appendix.

1. Discussion

The success of pre-hospital 12-lead ECG transmission as measured by the receipt of an ECG in the ED significantly improved with the implementation of cellular transmission compared to radio transmis- sion. Furthermore, the legibility of the successfully transmitted 12- lead ECGs significantly improved as well.

Inclusion of interaction effects in the statistical model indicated that for longer transports, ECGs were less likely to be successfully transmitted en route than those attempted from the scene, suggesting that a brief delay at the scene may pay off in greater Time Savings in the ED. When an ECG was required, cellular ECGs were more likely to be successful than radio ECGs, suggesting that EMTs may have tried harder to see that the ED received these ECGs.

With the increasing accuracy of the algorithm used by monitors to evaluate ECGs, one alternative is for the EMT to relay the machine’s interpretation to the ED. During this project, when it became clear that some ECGs were not being received, EMS crews were encouraged to do exactly this when they encountered transmission problems. They were also instructed to continue attempts to transmit the ECG itself. Pre-hospital computer-interpreted electrocardiography has been shown to be an effective tool in identifying myocardial infarc- tion, [20,21] although some authors [22,23] have found high false- positive rates (136/529, 26% and 9/23, 39%).

With regard to both obtaining and transmitting ECGs, the most common reason overall was “unknown.” The most common specific reason for failure to obtain an ECG was short transport time (25.5%). This applied primarily to just a few of the EMS agencies that were close to the ED. The question of whether delaying transport when just a few minutes away is appropriate is not clear from the current evidence and may have a system-specific response. The next most common reason cited was performance of a 3-lead ECG (11.9%). This is somewhat puzzling given the emphasis in training on 12-lead ECGs, but has been addressed.

The most common reason specified for failure to transmit an ECG was equipment malfunction (27.3%). Even ECG monitors designed

Table 1 Characteristics of study subjects
	Radio transmission, no. (%)	Wireless transmission, no. (%)	Not transmitted, no. (%)	Total
Subjects	337	580	223	1140
Demographics
Sex, male	176/337 (52.2)	289/580 (49.8)	121/223 (54.3)	673 (51.0)
Mean age (SD), y	66.1 (17.1) n = 336	66.2 (18.6) n = 580	64.3 (18.0) n = 223	65.9 (18.0), n = 1139
Primary language English	328/336 (97.6)	569/580 (98.1)	218/223 (97.7)	1115/1135 (98.2)
Chief Complaints (n = 337 for radio, 580 for cellular, 223 for not transmitted)
Chest pain	181 (53.7)	272 (46.9)	144 (64.6)	708 (53.6)
Jaw or arm pain	7 (2.1)	10 (1.7)	1 (0.4)	20 (1.5)
Syncope	25 (7.4)	64 (11.0)	27 (12.1)	172 (13.6)
Post- cardiac arrest	3 (0.9)	0	0	3 (0.3)
Dyspnea	46 (13.6)	74 (12.8)	13 (5.8)	133 (10.1)
Diaphoresis	1 (0.3)	4 (0.7)	0	5 (0.4)
Weakness	14 (4.1)	19 (3.3)	3 (1.3)	36 (2.8)
Palpitations	6 (1.8)	8 (1.4)	4 (1.8)	18 (1.4)
Abdominal pain	42 (12.5)	115 (19.8)	24 (10.8)	181 (13.7)
Other	12 (3.6)	14 (2.4)	7 (3.1)	33 (2.5)
Disposition (n = 327 for radio, 570 for cellular, 217 for not transmitted)
Home	143 (43.7)	274 (48.1)	103 (47.5)	520 (46.7)
Hospital	156 (47.7)	252 (44.2)	98 (45.2)	506 (45.4)
Catheterization lab	12 (3.7)	17 (3.0)	10 (4.6)	39 (3.5)
Intensive care unit	9 (2.7)	14 (2.5)	3 (1.4)	26 (2.3)
Death	3 (0.9)	1 (0.2)	0	4 (0.4)
Left against medical advice	1 (0.3)	3 (0.5)	2 (0.9)	6 (0.5)
Other	3 (0.9)	9 (1.6)	1 (0.5)	13 (1.2)
Pre-hospital times, mean (standard deviation), min
Response interval	8.6 (5.2), n = 327	8.7 (5.4), n = 562	8.1 (4.8), n = 214	8.3 (5.2), n = 1103
Scene interval	14.2 (6.2), n = 327	14.0 (6.0), n = 561	13.8 (5.9), n = 214	13.7 (6.0), n = 1102
Transport interval	14.5 (8.8), n = 329	14.8 (9.2), n = 560	14.8 (9.9), n = 216	14.1 (9.2), n = 1105
Total EMS interval	37.3 (13.4), n = 330	37.6 (15.0), n = 561	36.8 (15.2), n = 216	36.2 (14.6), n = 1107

specifically for the challenging pre-hospital environment are subject to mechanical and other failures. Short transport was the second most common reason specified (12.1%), followed closely (10.8%) by “deemed not necessary.” This is another puzzling finding, but may suggest over-reliance on the monitor interpretation of the ECG. Training since the project ended has stressed the importance of transmitting all ECGs.

An uncommon reason for failure to perform (1.7%) or transmit an ECG (2.2%) was “overwhelming tasks.” The rarity with which this was cited suggests that EMTs are able, with training and the support of medical direction, to incorporate this assessment tool into their procedures.

It is possible that 12-lead ECGs obtained for criteria not listed under EMS protocol and not successfully transmitted may have gone unrecognized. Attempts were made to capture data for all patients who had a pre-hospital 12-lead ECG performed. For logistical reasons, no efforts were made to determine how many patients should have received ECGs. Neither the EMS district nor the hospital employed an EMS coordinator with the responsibility of ensuring ECGs were

Table 2

Measures of success in transmitting 12-lead ECGs

Received Legible Legible/received

Transmission method

obtained on appropriate patients or overseeing the daily operation of the system. Future studies with electronic EMS records (instituted after the study concluded) may be helpful in identifying all patients for whom a pre-hospital 12-lead ECG is indicated and obtained.

Limitations of this study include the use of a single hospital and numerous challenges with regard to data collection. Data was collected for only one hospital by student RAs in the ED who may have missed patients as this was the first major project undertaken by the program. Prior studies of potential acute coronary syndrome patients in the ED, however, have shown that data is often collected more accurately by research assistants than clinicians as they are able to focus entirely on the task of data collection [24]. Neither of the two other hospitals in Vermont with cardiac catheterization laboratories performs emergency PCI, so the use of a single hospital is likely to reflect the demographics typical of a rural academic medical center.

A potential source of inconsistency was the lack of synchronization of ECG monitor clocks, dispatch clocks and ED clocks in the EMS district. Times recorded by dispatch may conflict significantly with times recorded by ECG monitors and the ED. Having an academic research associate in the ED at the time of arrival of the patient limited

Table 3

Reason ECG not performed or transmitted

ECG not performed, Performed but not

Radio	236/337 (70.0)	164/337 (48.7)	164/236 (69.5)
Cellular	441/580 (76.0)	389/580 (67.1)	389/441 (88.2)	no. (%), n = 176 transmitted, no.
Difference (mean, 95% CI) ECG required	6.0 (1.1-12.1)	18.4 (11.8-24.9)	18.7 (12.1-25.3)	(%), n = 223
Yes	399/735 (54.3)	318/735 (43.3)	318/399 (79.7)	Short transport	45 (25.5)	27 (12.1)
No	278/405 (68.6)	235/405 (58.0)	235/278 (84.5)	Deemed not necessary	13 (7.4)	24 (10.8)
ECG location				3-lead ECG performed	21 (11.9)
On scene	298/616 (48.4)	242/616 (39.3)	242/298 (81.2)	Equipment not available	6 (3.4)	17 (7.6)
En route	336/457 (73.5)	273/457 (59.7)	273/336 (81.2)	Patient refused or uncooperative	6 (3.4)	0 (0)
Transport Interval				Equipment malfunction	3 (1.7)	61 (27.3)
<=10 min	266/464 (57.3)	218/464 (47.0)	218/266 (81.9)	Overwhelming tasks	3 (1.7)	5 (2.2)
11-20 min	271/427 (63.5)	228/427 (53.4)	228/271 (84.1)	ECG in hand	2 (1.1)	0 (0)
21-30 min	99/162 (61.1)	75/162 (46.3)	75/99 (75.7)	Other	0	15 (6.7)
N 30 min	41/87 (47.1)	32/87 (36.8)	32/41 (78.0)	Unknown	77 (44.0)	74 (33.2)

Table 4

Outcome of patients with regard to field ECG interpretation by ECG monitor and physician

Outcome Field ECG monitor reading

MI MI Suspected

Total 24 11

Physician read as STEMI 16 (67%) 5 (45%)

Cath lab	14 (58%)	4 (36%)	delayed reception of some cellular transmissions by several minutes.
Artery opened	13 (54%)	4 (36%)	This was not known until one month before the end of the study, so

transmission are appropriate. Our large sample size (the largest reported to date for this kind of study) may also mitigate some of this concern.

Two incidental findings occurred that merit reporting. Although radio transmissions were essentially instantaneous when they got through, cellular transmissions had to be relayed through an inter- mediary of the monitor manufacturer to convert the signal. This

Physician read as not STEMI	8 (33%)	6 (55%)
Cath lab	1	0

Artery opened 1 0

the number of discrepancies in times, but future research in this system will need to address this issue.

Transmission for more distant EMS agencies was sometimes limited by cell phone reception, but they were encouraged to re- attempt transmission upon obtaining a signal. Although more than half of radio transmissions were not legible, even with the improved performance of cell transmission, more than 30% did not get through to the ED, a reflection of the relative paucity of cell transmission towers in less populated areas. For purposes of immediate patient treatment and disposition, the legibility of the 12-lead ECG was determined by the physician in the ED. In future studies, a more standardized system for determining the legibility of the 12-lead ECG might be of benefit. The qualitative nature of this reading is, however, representative of emergency medicine practice.

Gaps in coverage occurred at several points because of student schedules: between midnight and 0800, between semesters, and during semester breaks. Although the subjects of this study cons- titute a convenience sample for this reason, our subjects are largely representative of the kinds of cases where ECG acquisition and

we were not able to document its frequency or impact.

Additionally, the emergency physician agreed on presence of a STEMI in 67% of the MI-classified ECGs and 45% of the MI suspected ECGs. The vast majority of those patients went to the catheterization lab and almost all of them had an artery opened. What is surprising is the high false-positive rate of both groups (46% for MI and 64% for suspected MI) when the monitor interpretation is applied solely, exceeding the results reported in 2009 from Orange County, California²². Their larger sample size allowed for meaningful con- fidence intervals to be calculated, a step that would have little value in this study with only 18 patients who had a coronary occlusion treated in the catheter laboratory. This does, however, strongly suggest the need for a qualified person to interpret the ECG before activating the catheter laboratory.

This study’s external validity may be limited because of system- specific characteristics, eg, mountainous terrain, limited number of towers for cellular transmission, small population, availability of a single hospital destination, and low-volume ambulance services staffed primarily by volunteers, but these features are common in many areas outside urban and suburban population clusters.

Like any observational study, factors outside the control of the investigators cannot be ruled out as a cause of any changes observed. Although no modifications of EMS providers’ Scope of practice

Table 5

Results of logistic regression analysis

Covariate	Odds ratio	95% confidence interval	P	? Coefficient	SE	Interpretation of OR
Intercept 1	NA	NA	0.01	0.8600	0.3512
Intercept 2	NA	NA	b 0.0001	1.5430	0.4128
Main effects
ECG transmission method [transmission]	2.88	1.81 – 4.58	b 0.0001	1.0565	0.2370	Using cell instead of radio almost triples the odds of receiving
						a transmission (vs not receiving it) and almost triples the odds
						of receiving a legible transmission (vs. receiving an
						illegible transmission)
Whether the EMTs were required to obtain	1.19	0.64-2.20	0.46	0.1712	0.3155	No effect when considered alone
an ECG [required] Location ECG transmitted from [location]	0.98	0.51-1.86	0.95	-0.0208	0.3288	No effect when considered alone
Transport interval (min)	1.01	0.98 – 1.05	0.46	0.0133	0.0178	No effect when considered alone
Interactions
Transport interval X transmission	1.00	0.98-1.03	0.90	0.0016	0.0133	No interaction
Transport interval X location	0.94	0.92-0.96	b 0.0001	-0.0629	0.0124	When transport interval increases by 1 minute, the ratio of the
						odds of success for en route transmission to the odds of success
						for scene transmission decreases by 6%. Success refers to
						both reception of a 12 lead (compared to not receiving it) and
						receiving a legible 12-lead (compared to receiving an illegible
						12-lead).
Transport interval X required	1.00	0.98-1.02	0.95	0.0008	0.0114	No interaction
Required X transmission	2.25	1.59-3.21	b 0.0001	0.8137	0.1793	When an ECG was required, the ratio of the odds of success for
						cellular transmission to the odds of success for radio
						transmission more than doubled compared to when an ECG was
						not required. Success refers to both reception of a 12 lead
						(compared to not receiving it) and receiving a legible 12-lead
						(compared to receiving an illegible 12-lead).
Required X location	0.72	0.44-1.17	0.19	-0.3242	0.2469	No interaction
Transmission X location	1.37	0.80-2.33	0.25	0.3144	0.2722	No interaction
Overall model evaluation	QIC	QICu
Quasi-likelihood information criteria	1518.4	1502.3

N = 841. Coding of categorical variables: ECG legible: 1 = legible, 2 = not legible, 3 = not received; ECG required: 0 = no, 1 = yes; ECG transmission method: 1 = radio, 2 = cell phone; ECG location: 1 = transmitted before departure from scene, 2 = transmitted at or after departure from scene. NA, not applicable.

occurred during the study, 16 EMT-Intermediates were enrolled in a local paramedic course which ended just before completion of the study. Data collection began several months after the EMS agencies started to use their monitors, allowing the EMTs to become familiar with the equipment and procedures, but a “practice effect” may also have occurred as EMTs gained more experience with ECGs and felt more comfortable not transmitting those interpreted by the monitor as perfectly normal, although they were discouraged from doing so.

The barriers to implementing cellular technology in rural settings include the time and costs required for the installation of the units and the education of EMTs. The care of myocardial infarction in rural settings will continue to remain a challenge given both longer pre- hospital times and fewer advanced life support providers trained to read 12-lead ECGs in rural settings. The early provision of a pre- hospital 12-lead ECG to emergency physicians through the use of cellular technology represents a significant advancement in the treatment of acute myocardial infarction by permitting earlier identification of STEMI patients and quicker delivery of definitive care to this subset of patients. The combination of a prehospital ECG and a qualified clinician to interpret it reduces the number of false- positive activations of the catheter laboratory. More widespread implementation of this technology in the future, especially in rural settings, could have significant effects on the mortality and morbidity of myocardial infarction.

Appendix. Details on logistic regression model checking Assumptions for the proportional odds logistic regression model

were validated. The number of events per variable was 60 (240 ECGs not received/4 variables), well in excess of the recommended mini- mum number [25] of 10. A plot of transport time, the only continuous independent variable, against probability of successful transmission was approximately linear, with a longer transport time associated with a slightly increased chance of successful transmission. Evalua- tion of collinearity yielded no condition index greater than 11.7. Goodness-of-fit statistics, odds ratios and 95% confidence intervals are displayed in Table 5. SAS does not provide typical goodness-of-fit statistics when generalized estimating equations are used, but does provide quasi-likelihood information criteria for model fit. Inspection of logistic regression diagnostic plots of ??², ?D and confidence interval displacement against both predicted probability and leverage revealed 17 points that might have undue influence on the values of the coefficients. All the cases involved long transports (range 28-52 minutes). In 12 cases, events occurred that were contrary to expec- tation: in six cases an ECG transmitted by radio was legible and in six a cell phone transmission resulted in an illegible or no ECG. Although these events were contrary to expectations, all of these occurrences were plausible and may have reflected how distance is sometimes not as important as location, especially the location of cell phone towers. Additionally, they constitute less than 2% of the data points. Hence, they were not omitted from the dataset.

The authors acknowledge the work of the emergency medical

technicians and the EMS agencies in Vermont District 3. The cardiology department at Fletcher Allen Health Care generously raised the funding for the ECG monitors. Wendy James, MD in Emergency Medicine was instrumental in this activity. No other outside funding was obtained.

References

1. McNamara RL, Wang Y, Herrin J, Curtis JP, Bradley EH, Magid DJ, et al. Effect of door-to-balloon time on mortality in patients with ST-segment elevation myocardial infarction. J Am Coll Cardiol 2006;47:2180-6.
2. Andersen H, Nielsen TT, Rasmussen K, Thuesen L, Kelbaek H, Thayssen P, et al. A comparison of coronary angioplasty with Fibrinolytic therapy in acute myocardial infarction. N Engl J Med 2003;349:733-42.
3. Keeley EC, Boura JA, Grines CL. primary angioplasty versus intravenous thrombolytic therapy for acute myocardial infarction: a quantitative review of 23 randomised trials. Lancet 2003;361:13-20.
4. De Luca G, Suryapranata H, Ottervanger JP, Antman EM. Time delay to treatment and mortality in primary angioplasty for acute myocardial infarction: every minute of delay counts. Circ 2004;109:1223-5.
5. Antman EM, Anbe DT, Armstrong PW, Bates ER, Green LA, Hand M, et al. ACC/AHA guidelines for the management of patients with ST-elevation myocardial infarction- executive summary. A report of the American College of Cardiolo- gy/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 1999 Guidelines for the Management of Patients with Acute Myocardial Infarction). J Am Coll Cardiol 2004;44:671-719.
6. Bradley EH, Herrin J, Wang Y, Barton BA, Webster TR, Mattera JA, et al. Strategies for reducing the door-to-balloon time in acute myocardial infarction. N Engl J Med 2006;355:2308-20.
7. Amit G, Cafri C, Gilutz H, Wolak A, Ilia R, Zahger D. Benefit of direct ambulance to coronary care unit admission of Acute myocardial infarction patients undergoing Primary percutaneous intervention. Int J Cardiol 2007;119:355-8.
8. Morrison LJ, Brooks S, Sawaksky B, McDonald A, Verbeek PR. Prehospital 12-lead electrocardiography impact on acute myocardial infarction treatment times and mortality: a systematic review. Acad Emerg Med 2006;13:84-9.
9. Dhruva VN, Abdelhadi SI, Anis A, Gluckman W, Hom D, Dougan W, et al. ST- segment analysis using wireless technology in acute myocardial infarction (STAT- MI) trial. J Am Coll Cardiol 2007;50:509-13.
10. Foster DB, Dufendach JH, Barkdoll CM, Mitchell BK. Prehospital recognition of AMI using independent nurse/paramedic 12-lead ECG evaluation: impact on in- hospital times to thrombolysis in a rural community hospital. Am J Emerg Med 1994;12:25-31.
11. Kereiakes DJ, Gibler WB, Martin LH, Pieper KS, Anderson LC. Relative importance of emergency medical system transport and the prehospital electrocardiogram on reducing hospital time delay to therapy for acute myocardial infarction: a preliminary report from the Cincinnati Heart Project. Am Heart J 1992;123: 835-40.
12. Karagounis L, Ispen SK, Jessop MR, Gilmore KM, Valenti DA, Clawson JJ, et al. Impact of field-transmitted electrocardiography on time to in-hospital thrombo- lytic therapy in acute myocardial infarction. Am J Cardiol 1990;66:786-91.
13. Werman HA, Newland R, Cotton B. Transmission of 12-lead electrocardiographic tracings by emergency medical technician-basics and emergency medical technician-intermediates: a feasibility study. Am J Emerg Med 2011;29:437-40.
14. Niles NW, Conley SM, Yang RC, Vanichakarn P, Anderson TA, Butterly JR, et al, Dartmouth-Hitchcock ST Elevation Myocardial Infarction Process Upgrade Project (STEPUP) Group. Primary percutaneous coronary intervention for patients pre- senting with ST-segment elevation myocardial infarction: process improvement in a rural ST-segment elevation myocardial infarction receiving center. Prog Cardiovasc Dis 2010;53:202-9.
15. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. STROBE Initiative. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)statement: guidelines for reporting observational studies. BMJ 2007;335(7624):806-8.
16. U.S. Census Bureau, Population Division. Table 5. Annual estimates of the resi- dent population for minor civil divisions in Vermont, listed alphabetically within county: April 1, 2000 to July 1, 2009 (SUB-EST2009-05-50). Release Date:

June 2010.

1. Hollander JE, Singer AJ. An innovative strategy for conducting clinical research: the academic associate program. Acad Emerg Med 2002;9(2):134-7.
2. Derksen S, Keselman HJ. Backward, forward and stepwise automated subset selection algorithms: frequency of obtaining authentic and noise variables. Br J Math Stat Psychol 1992;45:265-82.
3. Hosmer DW, Lemeshow S. Applied logistic regression. . 2nd ed.New York: John Wiley & Sons; 2000 .
4. Patel RJ, Vilke GM, Chan TC. The prehospital electrocardiogram. J Emerg Med 2001;21:35-9.
5. Aufderheide TP, Rowlandson I, Lawrence SW, Kuhn EM, Selker HP. Test of the Acute cardiac ischemia time-insensitive predictive instrument (ACI-TIPI) for Prehospital use. Ann Emerg Med 1996;27:193-8.
6. Swan PY, Nighswonger B, Boswell GL, Stratton SJ. Factors associated with false- positive emergency medical services triage for percutaneous coronary interven- tion. West J Emerg Med 2009;10(4):208-12.
7. Youngquist ST, Shah AP, Niemann JT, Kaji AH, French WJ. A comparison of door-to- balloon times and false-positive activation between emergency department and out-of-hospital activation of the coronary catheterization team. Acad Emerg Med 2008;15:784-7.
8. Cruz CO, Meshberg EB, Shofer FS, McCusker CM, Chang AM, Hollander JE. Interrater reliability and accuracy of clinicians and trained research assistants performing Prospective data collection in emergency department patients with potential acute coronary syndrome. Ann Emerg Med 2009;54:9-11.
9. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;49:1373-9.