How Accurate Is an EKG at Detecting Heart Problems?

An EKG’s accuracy depends heavily on what heart problem you’re trying to find. For detecting a heart attack in progress, a standard 12-lead EKG catches roughly 40% of cases using traditional criteria, though newer reading methods push that closer to 85%. For rhythm problems like atrial fibrillation, accuracy is excellent when the rhythm happens to be abnormal at the moment of the recording, but the test only captures about ten seconds of your heart’s activity, so it routinely misses intermittent problems. For structural changes like a thickened heart wall, the EKG is better at ruling things out than confirming them. The short answer is that an EKG is cheap, fast, and indispensable, but its accuracy swings from outstanding to mediocre depending on the condition in question and the circumstances of the person being tested.

Heart Attacks and Blocked Arteries

When people think about EKG accuracy, they’re often thinking about heart attacks. Here the picture is surprisingly complicated. Emergency departments rely on a set of criteria called STEMI criteria to decide whether a patient is having a serious heart attack caused by a completely blocked coronary artery. The problem is that these criteria miss a lot of cases. In a study of over 800 patients, traditional STEMI criteria caught only about 41% of patients who actually had a blocked artery, while correctly identifying those without a blockage about 94% of the time.¹ That means more than half of patients with a dangerous blockage had an EKG that didn’t meet the classic alarm threshold.

Cardiologists have been working to fix this. A newer approach, sometimes called OMI criteria (for occlusion myocardial infarction), looks at a broader set of EKG patterns beyond the classic ST-elevation cutoffs. In the same study, experienced readers using these broader criteria achieved sensitivity around 80–86% while maintaining specificity above 90%. A separate investigation confirmed that the OMI approach outperformed the traditional method in most accuracy comparisons.² The catch is that reading EKGs this way takes more training, and most emergency departments still default to the older system. So the accuracy of an EKG in a heart attack depends not just on the test itself, but on how skilled the person reading it is and which criteria they apply.

There’s another layer to this. Not all heart attacks produce dramatic EKG changes. A type of heart attack called NSTEMI (non-ST-elevation myocardial infarction) often shows only subtle or no EKG abnormalities. In one small study of NSTEMI patients assessed with EKG, blood tests, and ultrasound, about 18% had no positive findings on any single test, and the EKG alone picked up ischemic changes in only a fraction of cases.³ This is why emergency doctors never rely on the EKG alone when a heart attack is suspected. Blood markers and imaging fill the gaps.

Rhythm Problems and the Ten-Second Window

For rhythm disorders, the EKG has a fundamental limitation that has nothing to do with its electrical sensitivity: it only records about ten seconds of your heart’s activity.⁴ If you have atrial fibrillation that comes and goes, or occasional episodes of rapid heartbeat, the odds of those ten seconds capturing the problem are low. You could feel your heart racing every evening and still walk out of the clinic with a “normal” EKG because the test happened during a calm moment.

This is exactly why longer monitoring exists. A study comparing a standard 24-hour Holter monitor with a 14-day wearable patch found a dramatic difference. The patch caught intermittent arrhythmias in about two-thirds of patients, compared to only 9% with the single-day Holter. For atrial fibrillation specifically, the 14-day patch detected episodes in roughly 22% of patients versus just 3% with the Holter.⁵ The lesson isn’t that the EKG technology is flawed. The electrodes and circuitry work fine. The problem is duration: intermittent conditions need sustained surveillance, and a snapshot test can’t provide that.

When the EKG does catch a rhythm abnormality in real time, it is highly accurate at identifying what the rhythm is. The issue is almost always about missing things that aren’t happening at the moment, not about misreading what’s on the screen.

Thickened Heart Muscle and Structural Changes

Detecting left ventricular hypertrophy, a thickening of the heart’s main pumping chamber, is one of the EKG’s most studied jobs and one of its weakest performances. Doctors use various voltage-based criteria on the EKG to estimate whether the heart wall is abnormally thick, but overall accuracy for the most common criteria lands in the range of 50–62%.⁶ That’s barely better than a coin flip for overall diagnostic accuracy.

The consistent finding across research is that EKG criteria for thickened heart muscle are very specific but not very sensitive. Specificities typically exceed 90%, meaning that when the EKG says the wall is thick, it’s usually right.⁷ But sensitivity can be as low as 35% with some criteria, meaning the test misses most cases. This creates an asymmetry in how the EKG is useful: a positive result means something, but a negative result shouldn’t be reassuring on its own.

The situation gets more complicated when a patient also has a conduction abnormality called left bundle branch block, which distorts the EKG pattern. A meta-analysis of 12 studies found that most standard criteria performed poorly in this group. One lesser-known criterion achieved sensitivity of 86%, but the others were substantially worse.⁸ In practice, when thickened heart muscle is genuinely suspected, an echocardiogram or cardiac MRI is used to confirm it because the EKG simply isn’t reliable enough on its own.

The EKG’s role in primary care reflects this reality. A normal EKG result, combined with a normal blood test for a marker called BNP, is useful for ruling out significant heart dysfunction when clinical suspicion is moderate.⁹ The EKG works best as a gatekeeper: if it’s normal, you probably don’t need an expensive imaging study right away.

Inherited Electrical Disorders

Some heart conditions are purely electrical and show up only on an EKG, not on imaging. Long QT syndrome is one of the most important examples, a genetic condition that can cause dangerous arrhythmias and sudden death. You’d think the EKG would excel here since a prolonged QT interval is the defining feature. But roughly half of patients with genetically confirmed long QT syndrome have a normal or borderline-normal QT interval on a resting EKG.¹⁰ The interval fluctuates, and a single resting recording can easily catch it in its normal range.

This is one of the more sobering limitations of the standard EKG. A condition that is defined by an electrical abnormality is still missed half the time by the test specifically designed to measure electrical activity. It underscores why family history, genetic testing, and repeated or prolonged recordings all play a role in diagnosing these conditions.

How Body Type, Age, and Sex Affect Accuracy

Your body doesn’t just passively sit between the heart and the electrodes. It actively changes what those electrodes pick up. Obesity is the most studied confounder. In patients with high blood pressure and a thickened heart wall, obesity significantly reduced the EKG’s ability to detect that thickening by dampening the electrical signals that reach the skin. Sensitivity dropped from 56% in non-obese patients to just 24% in obese patients using the best-performing criterion for each group.¹¹ Obesity also increased false positives through structural changes in the heart that mimic hypertrophy on the EKG without truly being hypertrophy. So the test gets less accurate in both directions for larger patients.

Sex and body composition influence baseline EKG readings as well. Men and women show consistent differences in several key intervals on the EKG, and these differences are large enough to matter clinically.¹² Race also affects normal values. A study of 2,400 healthy children in North America found that African-American children had significantly higher R-wave amplitudes than white children, and that the QT interval differed by both sex and race from early childhood onward.¹³ If a doctor is comparing your EKG to reference ranges built from a population that doesn’t look like you, the result could be misleadingly abnormal or falsely reassuring.

EKGs in Children

Pediatric EKGs are a distinct challenge. A child’s heart is structurally and electrically different from an adult’s, and those differences change rapidly as the child grows. Normal heart rate, PR interval, and wave amplitudes all shift with age, sometimes substantially over just a few years. Data from school-aged children in a sports screening program found that sex-specific differences in heart rate and PR interval begin emerging in older children, suggesting that age- and sex-matched reference values are clinically important for accurate interpretation.¹⁴

The risk of applying adult criteria to pediatric EKGs is real. Patterns that look alarming in a 50-year-old can be entirely normal in a 7-year-old. At the same time, subtle abnormalities in children that would mean nothing in an adult could signal a serious inherited condition. This is one area where the accuracy of the EKG is inseparable from the expertise of the reader.

Computer Reads Versus Human Reads

Every modern EKG machine spits out an automated interpretation, and clinicians rely on these more than they probably should. A recent analysis found that about 39% of EKGs were interpreted incorrectly by the machine’s algorithm. Among those misread, about 58% were false negatives (real problems missed) and 57% were false positives (normal hearts flagged as abnormal). The diagnosis of ischemia, arguably the most important call the algorithm makes, was falsely positive in about 16% of cases and falsely negative in about 22%.¹⁵ Older patients and those with higher body mass were more likely to receive incorrect automated readings.

For atrial fibrillation specifically, a study comparing computerized EKG analysis with cardiologist interpretation found that the computer’s sensitivity was 78% versus 87% for the cardiologist, and its positive predictive value was only 86% compared to 98%.¹⁶ Among the EKGs the computer misclassified, more than 40% of those errors went unrecognized by the treating team, leading to delayed or inappropriate treatment in a meaningful number of patients. The practical takeaway: the printout at the top of your EKG is a suggestion, not a diagnosis. It needs a trained human to confirm or override it.

The Athlete Paradox

Athletes pose a unique problem for EKG accuracy because regular intense training reshapes the heart in ways that mimic disease on an EKG. A study of trained military sportsmen found that 81% had significant EKG changes. About two-thirds of those changes were considered normal training adaptations, while a third were classified as genuinely abnormal.¹⁷ Separating the two categories is difficult, and the consequences of getting it wrong run in both directions: you can unnecessarily sideline a healthy athlete or clear someone who has a dangerous condition.

The variability in how different doctors read the same athlete’s EKG makes this worse. Research on inter-reader agreement in young athlete screening found that interpretation and the resulting cascade of follow-up tests were highly dependent on the individual physician, even among experienced cardiologists.¹⁸ Two equally qualified doctors looking at the same EKG might reach different conclusions about whether further testing is needed, which has real financial and psychological consequences for the athlete.

Smartwatches and AI-Assisted EKGs

Consumer smartwatches now offer single-lead EKG recordings, and for atrial fibrillation, they perform better than many people expect. A meta-analysis of 11 studies covering over 4,200 participants found that the Apple Watch had pooled sensitivity of about 95% and specificity of about 95% for detecting atrial fibrillation.¹⁹ A separate emergency department study found that the watch’s ability to distinguish normal from abnormal rhythms in a single lead was comparable to the standard clinical EKG.²⁰

Smartwatches that use pulse-wave sensing rather than electrical leads showed even higher pooled sensitivity and specificity for atrial fibrillation in a meta-analysis, reaching about 97% and 97% respectively.²¹ These devices aren’t replacements for a 12-lead EKG, which views the heart from multiple angles. But for catching intermittent atrial fibrillation over days and weeks of wear, they have an obvious advantage over a ten-second clinic recording.

On the artificial intelligence front, researchers are training algorithms to extract information from EKGs that human readers can’t see. A systematic review of externally validated AI models found that AI could detect reduced heart pump function from a standard EKG with promising accuracy, even in people with no symptoms.²² This is a condition the standard EKG was never designed to identify, which makes it a genuinely new capability rather than an incremental improvement. The technology is still being validated, but it could eventually turn the humble EKG from a simple rhythm and voltage detector into a broader screening tool for conditions currently requiring expensive imaging.

When the EKG Needs Backup

No experienced clinician treats the EKG as a standalone verdict. For suspected heart attacks, combining the EKG with blood troponin levels and bedside ultrasound raises diagnostic accuracy well beyond what any single test achieves. In one study of cardiac arrest patients, combining ST elevation on the EKG with ultrasound wall-motion abnormalities and troponin elevation predicted coronary artery disease with 94% sensitivity.²³ The specificity of that combination was poor, meaning it flagged some people unnecessarily, but in an emergency setting, catching almost everyone with a real blockage matters more than avoiding extra tests.

For the specific subset of NSTEMI heart attacks, which by definition lack the dramatic EKG changes of a STEMI, the EKG alone is even less reliable. In a feasibility study that combined prehospital EKG, point-of-care troponin, and portable ultrasound, the ultrasound was the sole positive finding in about 14% of confirmed NSTEMI cases.³ Without it, those patients would have been missed entirely in the prehospital phase.

Why the EKG Persists Despite Its Gaps

Given all its blind spots, you might wonder why the EKG remains the most ordered cardiac test in the world. The answer is practical. It costs almost nothing, takes less than a minute to perform, involves no radiation or contrast dye, can be done at the bedside or in a moving ambulance, and delivers results instantly. No other cardiac test comes close on all those dimensions simultaneously. An echocardiogram is more informative but requires equipment, expertise, and time. A cardiac MRI is more precise but costs thousands of dollars and isn’t available in an ambulance. Blood tests are powerful but take minutes to hours to process.

The EKG’s real value lies not in definitively diagnosing every condition, but in quickly sorting patients into risk categories. A normal EKG in someone with chest pain doesn’t rule out a heart attack, but it changes the probability enough to guide the next steps. An abnormal EKG in someone who feels fine may be the first clue that something warrants further investigation. One evidence review of the resting EKG as a screening tool found that its sensitivity for predicting future cardiac events was too low for mass screening of healthy people, largely because the conditions it detects are uncommon in the general population.²⁴ But in the right clinical context, where a patient has symptoms or risk factors that raise the baseline probability of disease, the EKG’s modest sensitivity becomes much more useful because the test has more disease to find.