Artificial intelligence (AI)-based solutions have revolutionized our world, using extensive datasets and computational resources to create automatic tools for complex tasks that, until now, have been performed by humans. Massive data is a fundamental aspect of the most powerful AI-based algorithms. However, for AI-based healthcare solutions, there are several socioeconomic, technical/infrastructural, and most importantly, legal restrictions, which limit the large collection and access of biomedical data, especially medical imaging. To overcome this important limitation, several alternative solutions have been suggested, including transfer learning approaches, generation of artificial data, adoption of blockchain technology, and creation of an infrastructure composed of anonymous and abstract data. However, none of these strategies is currently able to completely solve this challenge. The need to build large datasets that can be used to develop healthcare solutions deserves special attention from the scientific community, clinicians, all the healthcare players, engineers, ethicists, legislators, and society in general. This paper offers an overview of the data limitation in medical predictive models; its impact on the development of healthcare solutions; benefits and barriers of sharing data; and finally, suggests future directions to overcome data limitations in the medical field and enable AI to enhance healthcare. This perspective is dedicated to the technical requirements of the learning models, and it explains the limitation that comes from poor and small datasets in the medical domain and the technical options that try or can solve the problem related to the lack of massive healthcare data.

Background: Although electrocardiographic monitoring is valuable for continuous surveillance of intensive care unit (ICU) patients, false alarms are common and have been cited as a cause of alarm fatigue. ANSI/AAMI EC12:2002 states that electrocardiograms (ECGs) should not detect a QRS if the waveform is less than 0.15 mV (1.5 mm) for adult patients, in order to avoid mislabeling P waves or baseline noise as QRSs during complete heart block or asystole. However, ECG software algorithms often use more conservative QRS thresholds, which may result in false-positive asystole alarms in patients with low-amplitude QRS complexes. Objectives: To 1) assess the frequency of low QRS amplitude in a group of ICU patients with one or more false-positive asystole alarms and 2) determine whether low-amplitude QRSs are associated with false-positive asystole alarms during continuous ECG monitoring. Methods: Hospital-acquired standard 12-lead ECGs were examined in a group of 82 ICU patients who had one or more fa...

Background: While there are published studies that have examined premature ventricular complexes (PVCs) among patients with and without cardiac disease, there has not been a comprehensive review of the literature examining the diagnostic and prognostic significance of PVCs. This could help guide both community and hospital-based research and clinical practice. Methods: Scoping review frameworks by Arksey and O’Malley and the Joanna Briggs Institute (JBI) were used. A systematic search of the literature using four databases (CINAHL, Embase, PubMed, and Web of Science) was conducted. The review was prepared adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Extension for Scoping Review (PRISMA-ScR). Results: A total of 71 relevant articles were identified, 66 (93%) were observational, and five (7%) were secondary analyses from randomized clinical trials. Three studies (4%) examined the diagnostic importance of PVC origin (left/right ventricle) and QRS morphology in the diagnosis of acute myocardial ischemia (MI). The majority of the studies examined prognostic outcomes including left ventricular dysfunction, heart failure, arrhythmias, ischemic heart diseases, and mortality by PVCs frequency, burden, and QRS morphology. Conclusions: Very few studies have evaluated the diagnostic significance of PVCs and all are decades old. No hospital setting only studies were identified. Community-based longitudinal studies, which make up most of the literature, show that PVCs are associated with structural and coronary heart disease, lethal arrhythmias, atrial fibrillation, stroke, all-cause and cardiac mortality. However, a causal association between PVCs and these outcomes cannot be established due to the purely observational study designs employed.

Background: False ventricular tachycardia (VT) alarms are common during in‐hospital electrocardiographic (ECG) monitoring. Prior research shows that the majority of false VT can be attributed to algorithm deficiencies.Purpose: The purpose of this study was: (1) to describe the creation of a VT database annotated by ECG experts and (2) to determine true vs. false VT using a new VT algorithm created by our group.Methods: The VT algorithm was processed in 5320 consecutive ICU patients with 572,574 h of ECG and physiologic monitoring. A search algorithm identified potential VT, defined as: heart rate >100 beats/min, QRSs > 120 ms, and change in QRS morphology in >6 consecutive beats compared to the preceding native rhythm. Seven ECG channels, SpO₂, and arterial blood pressure waveforms were processed and loaded into a web‐based annotation software program. Five PhD‐prepared nurse scientists performed the annotations.Results: Of the 5320 ICU patients, 858 (16.13%) had 22,325 VTs. After three levels of iterative annotations, a total of 11,970 (53.62%) were adjudicated as true, 6485 (29.05%) as false, and 3870 (17.33%) were unresolved. The unresolved VTs were concentrated in 17 patients (1.98%). Of the 3870 unresolved VTs, 85.7% (n = 3281) were confounded by ventricular paced rhythm, 10.8% (n = 414) by underlying BBB, and 3.5% (n = 133) had a combination of both.Conclusions: The database described here represents the single largest human‐annotated database to date. The database includes consecutive ICU patients, with true, false, and challenging VTs (unresolved) and could serve as a gold standard database to develop and test new VT algorithms.

Background: Ventricular tachycardia (V-tach) is the most common lethal arrhythmia, yet 90% of alarms are false and contribute to alarm fatigue. We hypothesize that some true V-tach also causes alarm fatigue because current criteria are too sensitive (i.e., ≥6 beats ≥100 beats/min [bpm]). Purpose: This study was designed to determine (1) the proportion of clinically actionable true V-tach events; (2) whether true actionable versus nonactionable V-tach differs in terms of heart rate and/or duration (seconds); and (3) if actionable V-tach is associated with adverse outcomes. Methods: This was a secondary analysis in 460 intensive care unit (ICU) patients. Electronic health records were examined to determine if a V-tach event was actionable or nonactionable. Actionable V-tach was defined if a clinical action(s) was taken within 15 minutes of its occurrence (i.e., new and/or change of medication, defibrillation, and/or laboratory test). Maximal heart rate and duration for each V-tach event were measured from bedside monitor electrocardiography. Adverse patient outcomes included a code blue event and/or death. Results: In 460 ICU patients, 50 (11%) had 151 true V-tach events (range 1–20). Of the 50 patients, 40 (80%) had only nonactionable V-tach (97 events); 3 (6%) had both actionable and nonactionable V-tach (32 events); and 7 patients (14%) had only actionable V-tach (23 events). There were differences in duration comparing actionable versus nonactionable V-tach (mean 56.19 ± 116.87 seconds vs. 4.28 ± 4.09 seconds; P = 0.001) and maximal heart rate (188.81 ± 116.83 bpm vs. 150.79 ± 28.26 bpm; P = 0.001). Of the 50 patients, 3 (6%) had a code blue, 2 died, and all were in the actionable V-tach group. Conclusions: In our sample, <1% experienced a code blue following true V-tach. Heart rate and duration for actionable V-tach were much faster and longer than that for nonactionable V-tach. Current default settings typically used for electrocardiographic monitoring (i.e., ≥6 beats ≥100 bpm) appear to be too conservative and can lead to crisis/red level nuisance alarms that contribute to alarm fatigue. A prospective study designed to test whether adjusting default settings to these higher levels is safe for patients is needed.

Background: Treatment for unstable angina (UA) or non–ST-elevation myocardial infarction (NSTEMI) is aimed at plaque stabilization to prevent infarction. Two treatment strategies are (1) invasive (ie, cardiac catheterization laboratory 24 hours for recurrent symptoms). However, it is not known if the frequency of transient myocardial ischemia (TMI) or complications during hospitalization varies by treatment. Purpose: We aimed to (1) examine occurrence of TMI in UA/NSTEMI, (2) compare frequency of TMI by treatment pathway, and (3) determine predictors of in-hospital complications (ie, death, myocardial infarction [MI], pulmonary edema, shock, dysrhythmia with intervention). Methods: Hospitalized patients with coronary artery disease (ie, history of MI, percutaneous coronary intervention/stent, coronary artery bypass graft, >50% lesion via angiogram, or positive troponin) were recruited, and 12-lead electrocardiogram Holter initiated. Clinicians, blinded to Holter data, decided treatment strategy; offline analysis was done after discharge. Transient myocardial ischemia was defined as more than 1-mm ST segment ↑ or ↓, in more than 1 electrocardiographic lead, more than 1 minute. Results: Of 291 patients, 91% were white, 66% were male, 44% had prior MI, and 59% had prior percutaneous coronary intervention/stent or coronary artery bypass graft. Treatment pathway was early in 123 (42%) and selective in 168 (58%). Forty-nine (17%) had TMI: 19 (15%) early invasive, 30 (18%) selective (P = .637). Acute MI after admission was higher in patients with TMI regardless of treatment strategy (early: no TMI 4% vs yes TMI 21%; P = .020; selective: no TMI 1% vs yes TMI 13%; P = .0004). Predictors of major in-hospital complication were TMI (odds ratio, 9.9; 95% confidence interval, 3.84–25.78) and early invasive treatment (odds ratio 3.5; 95% confidence interval, 1.23–10.20). Conclusions: In UA/NSTEMI patients treated with contemporary therapies, TMI is not uncommon. The presence of TMI and early invasive treatment are predictors of major in-hospital complications.