Over the past decade, the global number of deaths as a result of cardiovascular disease (CVD) has risen by 12.5%, and accounts for one-third of all deaths worldwide (Roth et al, 2017). Ischaemic heart disease (IHD) is the largest contributor to CVD mortality (Joseph et al, 2017).
In 2015, CVD was the second leading cause of death in Australia, closely following all types of cancer (Australian Institute of Health and Welfare (AIHW), 2017). IHD was the most common form of CVD and was responsible for 12% of all Australian deaths during 2015. Acute myocardial infarction (AMI) accounted for 43% of these IHD-related deaths (AIHW, 2017).
Relatively few population-based studies have examined the trends and types of AMI; however ST-segment elevation myocardial infarction (STEMI) accounts for 25%–40% of all AMI presentations (Ghimire et al, 2014).
Given the impact of these diseases, there is a strong prehospital focus on early identification, treatment and transport of patients with acute coronary syndrome (ACS). Assessment and identification of acute myocardial conditions is a key part of out-of-hospital care for emergency medical services worldwide, which includes the use of a 12-lead electrocardiograph (ECG) (Morrison et al, 2006).
ECGs are interpreted in a range of ways in the out-of-hospital setting. These include paramedic-based interpretation, the use of a computer algorithm and transmission of the ECG to an on-call emergency physician or cardiologist for direct consultation (Savage et al, 2014; Pilbery et al, 2016).
In Australia, there are a number of emergency medical response services, with a government-appointed provider of ambulance services for each state. Each provider has its own set of guidelines and scope of practice to which paramedics must adhere. In Victoria, Ambulance Victoria (AV) is the sole emergency medical response service and plays a key role in delivery of care to patients with STEMI across the state.
AV operates a two-tiered model of care, with advanced life support (ALS) and mobile intensive care ambulance (MICA) paramedics responding to over 800 000 emergency calls annually (AV, 2019). In 2020–2021, of the 4497 paramedics, 594 (13.2%) were part of the MICA and the remainder with ALS (AV, 2021).
Before 2016, MICA paramedics were the only paramedics in Victoria who had the training and equipment to perform and interpret 12-lead ECGs. MICA paramedics have a higher skill set than ALS paramedics and can perform more advanced procedures. They undertake additional training to be able to make significant clinical decisions and provide clinical support to ambulance crews in the management of more complex patient presentations.
Since 2016, AV has seen the state-wide rollout of Zoll X Series monitors to all crews, including ALS paramedics. As part of this rollout, ALS paramedics completed a basic course in 12-lead ECG use and interpretation. All ALS paramedics attended a training day, in which approximately 4 hours were dedicated to the application, use and importance of 12-lead ECGs. The training session included a refresher module, revisiting the basic structure and function of the heart, the pathophysiology of IHD and ACS, the fundamental components of the 3-lead ECG, taking a systematic approach to analyse a rhythm strip and basic arrhythmia recognition. Also covered were 12-lead placement, acquisition and transmission, lead groupings, systematic interpretation of the ECG and STEMI criteria. Paramedics were educated on the management of these patients, and time was then allocated for them to practise interpreting 10 STEMI ECGs provided.
Although this training is provided, when identifying STEMI, ALS paramedics are currently instructed to rely upon the Zoll Inovise 12L interpretive algorithm; this is a computerised interpretation of the tracing rather than the paramedics' own assessment (Pilbery et al, 2016). If the monitor interprets the ECG and makes a finding of ‘STEMI’ or ‘AMI’, ALS paramedics follow the STEMI Management Clinical Practice Guideline regardless of their individual ability to interpret an ECG (AV, 2019).
The STEMI management pathway stipulates that a MICA crew is requested to continue the management of the patient, and the ECG is transmitted to the receiving hospital via both emergency department and cardiology department notifications. However, if a MICA crew attend, they may use their clinical judgement to override the Zoll's automated STEMI interpretation. Where MICA paramedics are unavailable, hospitals receive pre-notification from ALS paramedics based solely on the Inovise 12L computer algorithm interpretation.
According to Zoll's published data, the Inovise 12L interpretive algorithm software has the ability to correctly classify an individual as having a STEMI in 89% of cases (sensitivity) (Zoll, 2016). Furthermore, it is reported to correctly classify an individual as not having a STEMI in 100% of cases (specificity) and, 98% of the time, it can detect patients with a positive STEMI ECG who are actually STEMI positive (positive predictive value) (Parikh et al, 2008; Zoll, 2016). These statistics rely on the absence of confounders such as incorrect ECG electrode placement, patient movement, pre-existing cardiac conditions and incorrect age and sex being entered into the monitor (Zoll, 2016).
There are some concerns over reliance on a computer interpretation as the sole method of ECG interpretation. Anecdotally, a cohort of MICA paramedics shared with the lead author that, based on their experience, ‘the Zoll is getting it wrong more often than it is getting it right’. This results in hospital staff having less trust in the ability of paramedics to identify STEMI correctly, and lower paramedic confidence in the Zoll monitor's interpretation.
Additionally, relying on a computer-automated interpretation could lead to ALS paramedics becoming deskilled in ECG analysis. Consensus-based competency standards published by the American College of Cardiology and the American Heart Association suggest a minimum of 500 ECGs should be interpreted during initial training, and at least 100 ECGs yearly to maintain a certain level of competency (Salerno et al, 2003). By using automated interpretation rather than taking a systematic approach to analysing a rhythm strip and clinical decision-making, paramedics could become entirely dependent upon this technology, which could lead to a decline in clinical reasoning skills and patient care.
Assessment of the ACS patient requires a combination of an extensive history taking, a good-quality patient assessment, and a 12-lead ECG to form a complete picture. A computer interpretation alone may not give an accurate account of an evolving myocardial infarction because of changing pathology and/or patient presentation. An ECG may have subtle changes that require serial ECGs, and these changes may indicate that the patient requires percutaneous coronary intervention (PCI) for an evolving infarct.
The main aim of the STEMI system of care is to minimise the time to reperfusion of the myocardium, thereby improving morbidity and mortality rates. Delays within this system of care can negatively impact the effectiveness of the whole system, particularly door-to-balloon time for these time-critical patients. Therefore, the detection/recognition of STEMI by paramedics can have a dramatic effect on patients' long-term health outcomes. For these reasons, it is important to investigate current ECG recognition performance by the Zoll computer algorithm interpretation.
Methods
Research design and setting
This project is a retrospective/descriptive cohort study on the Zoll X-Series Inovise 12L algorithm's diagnostic capability in detecting STEMI patterns on 12-lead ECGs in patients presenting to AV.
When transmitting STEMI ECGs on scene, paramedics are instructed to enter the case number and the first five letters of the patient's surname into the monitor for the purposes of identification. All wirelessly transmitted 12-lead STEMI ECGs are stored by the VA STEMI Quality Initiative (VASQI), along with prehospital data on all paramedic suspected STEMI cases.
For this study, quantitative VASQI data, including patient ECGs, were periodically matched with data from the Victorian Cardiac Outcomes Registry (VCOR), based at Monash University. Hospitals supply VCOR with data on patient diagnoses, interventions provided and discharge status. Hospital diagnoses and management from matched VASQI and VCOR cases were examined to compare the final hospital diagnoses with the Zoll ECG interpretations.
Study population
Included in the study were patients who met each of these criteria:
Ethical approval
Ethical approval for this project was obtained through the Monash University Human Research Ethics Committee (project number 13926), and the AV Research Governance Committee.
To ensure the research process maintained confidentiality, no personal details that could be used to identify patients were included in any data or results. Non-identifiable information relating to patient age, sex and location were included.
Data analysis
Quantitative and demographic data were analysed using Microsoft Excel Version 16.40. Summary statistics including means, standard deviations (SD), and percentages (%) were used to describe patient demographics, clinical characteristics and diagnoses.
Results
Patient characteristics
Of 5269 paramedic-suspected cases of STEMI (based on the Zoll interpretation, paramedic interpretation or both) between 1 January 2018 and 31 December 2019, 765 had both an ECG uploaded to VASQI and VCOR data matched. For these cases, the mean female patient age was 69.9 years, and mean male patient age was 63.3 years (overall mean age of 64.8 years, SD±12.4) (Table 1). Most patients were men (76.2%), with women representing a fewer than a quarter (23.7%) of the cases included.
| Characteristics | n (%) |
|---|---|
| Sex | |
| Male | 583 (76.2) |
| Female | 181 (23.7) |
| Missing/unknown | 1 (0.1) |
| Age | |
| Female, years (mean) | 69.9 |
| Male, years (mean) | 63.3 |
| Mean age, years | 64.8, SD ± 12.4 |
| Area | |
| Metropolitan | 564 (73.7) |
| Rural | 200 (26.1) |
The majority of patients were located within the metropolitan region of Victoria (73.7%). An ALS resource was the initial skill set on scene at 79.2% of paramedic-suspected STEMI cases (Table 2).
| Skill set | n (%) | |
|---|---|---|
| Initial skill set by paramedic type | ||
| ALS | 606 (79.2) | |
| MICA | 123 (16.1) | |
| Missing/unknown | 36 (4.7) | |
| Initial skill set by area | ||
| Metro | ALS | 446 (81.6) |
| MICA | 100 (18.3) | |
| Rural | ALS | 160 (87.4) |
| MICA | 23 (12.6) | |
| Missing/unknown | 36 (4.7) | |
ALS: advanced life support; MICA: mobile intensive care ambulance
Clinical characteristics
Of the 5269 cases of suspected STEMI between January 2018 and December 2019, only 765 (14.5%) could be matched accurately with outcome data and included in the data analysis. Of the 765 cases, 88.9% of these had been correctly identified as STEMI (Table 3). The remaining 10% were categorised as either non-STEMI or unstable angina. No data were available for 1.1%.
| Acute coronary syndrome type | n (%) | |
|---|---|---|
| Paramedic-suspected STEMI | 5,269 | |
| Paramedic-suspected STEMI (ST-elevation myocardial infarction) with matched outcome data | 765 (14.5) | |
| STEMI | 680 (88.9) | |
| Non-STEMI | 73 (9.5) | |
| Unstable Angina | 4 (0.5) | |
| No data available | 8 (1.1) | |
Further to this, data were broken down by type of intervention provided to each patient. This was then compared to ACS type. Where intervention data were available, 60.8% of cases were classified as STEMI, 33.1% non-STEMI and 2% unstable angina (Table 4).
| Acute coronary syndrome type | n (%) | ||||
|---|---|---|---|---|---|
| Type of intervention | Unstable angina | Non-STEMI | STEMI | Blank/no data available | Total |
| Primary PCI for STEMI <12 hours | 0 | 0 | 562 | 0 | 562 (31.9) |
| PCI for STEMI >12 hours (unstable) | 0 | 0 | 38 | 0 | 38 (2.2)ç |
| PCI for STEMI >12 hours (stable) | 0 | 0 | 63 | 0 | 63 (3.6) |
| Stable after full-dose thrombolytics | 0 | 0 | 104 | 0 | 104 (5.9) |
| Unstable after full-dose thrombolytics (non-rescue PCI) | 0 | 0 | 40 | 0 | 40 (2.3)w |
| Rescue PCI (failed thrombolytics) | 0 | 0 | 264 | 0 | 264 (15) |
| PCI post cardiac arrest or cardiogenic shock (non-MI) | 0 | 0 | 0 | 14 | 14 (0.8) |
| PCI for NSTEMI | 0 | 584 | 0 | 0 | 584 (33.1) |
| PCI for unstable angina | 36 | 0 | 0 | 0 | 36 (2) |
| PCI for stable angina | 0 | 0 | 0 | 20 | 20 (1.1) |
| Staged PCI | 0 | 0 | 0 | 11 | 11 (0.6) |
| No angina/angina equivalent | 0 | 0 | 0 | 12 | 12 (0.7) |
| PCI for recent ACS >7 days ago (patient now stable) | 0 | 0 | 0 | 14 | 14 (0.8) |
| Total, n (%) | 36 (2) | 584 (33.1) | 1071(60.8) | 71 (4) | 1762 |
ACS: acute coronary syndrome; MI: myocardial infarction; PCI: percutaneous coronary intervention; STEMI: ST-segment elevation MI; NSTEMI: non-STEMI
Of the 765 cases, 634 (82.9%) had their out-of-hospital 12-lead ECG wirelessly transmitted to the receiving hospital, with 131 (17.1%) either not transmitted or unknown transmission status (Table 5). Possible causes include missing data or transmission failure because of equipment faults or connectivity issues. There were 637 (83.3%) cases where it could be demonstrated that a prehospital notification had been provided to the receiving hospital.
| Communication, treatment and cardiac arrest | n (%) |
|---|---|
| Prehospital wireless ECG transmission | |
| Yes | 634 (82.9) |
| No | 50 (6.5) |
| Missing/unknown | 81 (10.6) |
| Prehospital notification | |
| Yes | 637 (83.3) |
| No | 9 (1.2) |
| Missing/unknown | 119 (15.6) |
| Intravenous fibrinolysis | |
| Yes | 77 (10.1) |
| No | 685 (89.5) |
| Missing/unknown | 3 (0.4) |
| Out-of-hospital cardiac arrest | |
| Yes | 72 (9.4) |
| No | 693 (90.6) |
There were 77 patients of 765 (10.1%) who had an ECG uploaded in VASQI and received an intravenous fibrinolytic (tenecteplase) from AV paramedics (Table 5). The decision to administer this medication is made through consultation with a cardiologist, and their assessment of the 12-lead ECG and patient's clinical presentation; this only occurs in regional and rural areas of Victoria. Neither the Inovise 12L algorithm nor the paramedic make the diagnosis.
A total of 72 (9.4%) patients experienced an out-of-hospital cardiac arrest (OHCA) (Table 5). The data did not indicate whether the OHCA occurred before or after 12-lead ECG capture.
Discussion
This study examines the patient characteristics, demographics and performance of the Zoll monitor Inovise 12L interpretive algorithm within the STEMI system of care.
The results demonstrate that, within AV, the Zoll monitor Inovise 12L interpretive algorithm correctly identifies STEMI patients in 88.9% of cases. However, the matched cases represent only a small portion of the overall number of paramedic-suspected STEMI cases attended.
A potential reason for only 14.5% of paramedic-suspected STEMI cases being matched include hospitals not providing all of their suspected STEMI cases. This could be because of poor data collection systems or because a portion of suspected STEMI cases end up being treated as a different condition to a potential ACS, so the outcome data is never provided to VCOR. It is likely that cases that are high-acuity ACS, such as STEMIs and non-STEMIs, are more actively reported to VCOR because of positive patient outcomes from the STEMI system of care.
Therefore, cases that are identified by the Zoll monitor as STEMI but are dismissed by hospital cardiology teams immediately based on the prehospital ECG transmission could slip between the cracks, and are never included in outcome data. At face value, the statistics demonstrate that a high portion of patients are correctly identified as STEMI; however, with such a low number of matched cases, no definitive answers can be drawn about STEMI recognition within AV.
Further to this, the false positives that the Inovise 12L algorithm produces are sometimes not transmitted, as MICA arrive on scene and conclude that the interpretation is incorrect before the ECG is transmitted. These ECGs are then never included in the paramedic-suspected STEMI VASQI database so cannot be followed up. This may explain why anecdotal evidence provided by MICA paramedics indicates that they have negative views regarding the use and effectiveness of the Zoll monitor ECG interpretation.
Nonetheless, comparing Zoll's sensitivity of 89% for STEMI identification using the Inovise 12L interpretive algorithm with this study's finding of 88.9% positive STEMI identification supports the sensitivity statement from Zoll. Unfortunately, there is no way to ascertain the specificity of the Zoll monitor for STEMI, as ECGs that are STEMI negative are not transmitted so the data are not kept or stored.
In addition, the results of this study indicate that a large portion of paramedic-suspected STEMI cases are treated at hospital for some type of occlusive cardiac event, whether that is STEMI or non-STEMI. Therefore, regardless of final diagnosis, patients are still being transported to the most appropriate facility that can provide them with PCI or thrombolytics (Table 4). In this way, AV is providing patients with best care in order to achieve high-quality outcomes and, in turn, prevent secondary patient transfers. However, as mentioned above, the small number of matched cases gives only a small insight into the overall performance of the Victorian STEMI system of care.
The vast majority of patients included in the study were located within the metropolitan region of Victoria and were predominately men. This is consistent with current data that demonstrate that men are more likely than women to experience an AMI (Boyer et al, 2012). Further to this, the average age of patients included in the study aligns with published statistics that show men are typically younger than women when they experience their first AMI (Boyer et al, 2012). While the mean age of 64.8 years is slightly below (by 2.2 years) the 67.0 years reported by Boyer et al (2012), it is still similar.
Rates of ECG teletransmission for the data set was 82.9%. However 17.1% of cases had missing data, which significantly alters the overall transmission rate. Recent studies have demonstrated 12-lead ECG hospital transmission rates of up to 93%, indicating that there is some work for AV to do in this area (Sowizdraniuk et al, 2019). Given that ECG teletransmission, in conjunction with prehospital notification, allows receiving cardiology teams to activate the catheterisation laboratory, the rate of 82.9% should be improved upon.
Regarding prehospital notification by treating paramedics on scene, evidence shows this leads to a significant reduction in door-to-balloon time by as much as 35.2 minutes (Sowizdraniuk et al, 2019). This is the key outcome within the STEMI system of care, with the ultimate goal being a door-to-balloon time of <90 minutes (Boyer et al, 2012). A prehospital notification had been provided to the receiving hospital for 83.2% of cases in this study, with 1.2% showing no prehospital notification and a further 15.6% with no data available. Prehospital notification needs to increase for suspected STEMI cases to achieve best possible patient care and outcomes.
For STEMI patients living rurally, where timely access to a PCI-capable facility is not feasible, fibrinolysis is administered by paramedics. The decision to administer this medication is made through consultation with a cardiologist, and their assessment of the 12-lead ECG and their patient's clinical presentation; this occurs only in regional and rural Victoria. Therefore, the STEMI diagnosis is made by neither the Zoll monitor nor the treating paramedic. Fibrinolysis in this patient cohort is important, as it is the most viable option for myocardial reperfusion given the geographical barriers to definitive care (Danchin et al, 2004).
The incidence of OHCA in this study was 9.4%, which is higher than published rates of 5.6% (Danchin et al, 2004). Patients who have a sudden cardiac arrest after STEMI in the out-of-hospital setting have a mortality rate 10 times higher than those with STEMI without OHCA (Karam et al, 2019).
Finally, an ALS resource was the initial skill set on scene at 79.2% of paramedic-suspected STEMI cases. This highlights the importance of ALS STEMI education. As ALS paramedics are typically on scene first at these cases, recognition of STEMI on 12-lead ECGs is vital for best patient care and outcomes.
Limitations
There are a number of limitations to this study. First, only ECGs that have ‘STEMI’ or ‘acute MI’ interpretive statements that have been wirelessly transmitted to the receiving hospital are available on Zoll Online (web-based data system for storing wirelessly transmitted ECGs and rewatching cases in real time/vital sign survey trends); no other ECGs are transmitted. Second, any ECGs that display this interpretive statement that have been deemed inaccurate/incorrect by MICA on scene will not have been transmitted so were not included in the study. Third, as mentioned above, a data matching rate of only 14.5% is fairly low, and definitive conclusions cannot be drawn. For these reasons, the exact accuracy of the Zoll monitor cannot be determined.
Conclusions
Based on the data available, the authors are unable to say whether the diagnostic capability of the Zoll Inovise 12L interpretive algorithm in the out-of-hospital setting is safe and feasible.
Because of limited data-matching with patient outcomes in hospital, no hard conclusions can be drawn. Furthermore, there is no way to ascertain how many false positives the Zoll interpretation algorithm generated. Further investigation is required to truly assess the diagnostic capability of the Zoll Inovise 12L interpretive algorithm itself.