What Are the Duke Criteria for Infective Endocarditis?

The Duke Criteria are a standardized set of clinical, microbiological, and imaging findings that doctors use to diagnose infective endocarditis, a serious infection of the heart’s inner lining or valves. First published in 1994 at Duke University Medical Center, these criteria sort patients into three categories: definite, possible, or rejected for infective endocarditis. The system has been updated twice since then, most recently in 2023, to keep pace with changes in the types of organisms causing the disease and the imaging technology available to detect it.

Why a Scoring System Exists for This Diagnosis

Infective endocarditis is notoriously difficult to pin down. The infection produces a wide range of symptoms, from fevers and fatigue to stroke-like episodes and skin changes, many of which overlap with other illnesses. No single test confirms or rules it out on its own. Before the Duke Criteria existed, clinicians relied on looser case definitions that frequently missed the diagnosis or overcounted it. The Duke system solved this by combining findings from blood cultures, heart imaging, physical examination, and (when available) surgical or autopsy specimens into a structured framework that could be applied consistently across hospitals and research studies.

The underlying disease process helps explain why the criteria look the way they do. Infective endocarditis typically begins when bacteria enter the bloodstream and land on a damaged or abnormal heart valve surface. Valve trauma causes changes in the cells lining the valve, leading to deposits of platelets and a mesh of clotting proteins. Once that sticky surface forms, certain bacteria adhere to it more readily than others, building up a clump of infected material called a vegetation.¹ This injury-clot-infection sequence is the commonly accepted theory of how the disease develops.² Because bacteria are circulating in the blood and a physical mass is growing on the valve, the two pillars of the Duke Criteria are blood culture results and imaging of the heart.

How the Classification Works

The Duke Criteria divide diagnostic evidence into two tiers: major criteria and minor criteria. A patient is classified as having “definite” infective endocarditis if they meet certain combinations of these, or if a pathologist confirms the infection directly from tissue removed during surgery or autopsy. The clinical route to a definite diagnosis requires either two major criteria, one major plus three minor, or five minor criteria. “Possible” endocarditis applies when findings are suggestive but fall short of definite, and “rejected” means the evidence points clearly elsewhere or the patient improved rapidly in a way inconsistent with the disease.

This tiered approach matters because treatment for infective endocarditis is aggressive, often involving weeks of intravenous antibiotics and sometimes open-heart surgery. Clinicians need a system that balances catching real cases against labeling someone with a serious diagnosis they do not actually have.

The Major Criteria

There are two categories of major criteria: microbiological evidence from blood cultures and imaging evidence of heart valve involvement.

Blood Culture Findings

The microbiological major criterion captures the fact that infective endocarditis causes a continuous, low-level stream of bacteria in the blood. Certain organisms are so strongly associated with the disease that growing them from blood cultures carries heavy diagnostic weight. These “typical” organisms include species like Staphylococcus aureus, viridans group streptococci, and enterococci. In the 2023 update, the list of typical organisms was expanded, and pathogens that should be considered typical only when a patient has a prosthetic heart valve or other intracardiac device were added as a separate category.³

A meaningful practical change in the 2023 criteria was the removal of the older requirement that blood cultures be drawn at specific time intervals and from separate needle sticks. The previous rules had stipulated that positive cultures needed to come from draws separated by at least 12 hours or from a majority of four or more separate draws. In practice, these timing requirements were cumbersome and did not always reflect how blood cultures are collected in real emergency settings. The updated criteria dropped those stipulations.³

How central is the blood culture criterion? In one analysis of 241 patients evaluated for the disease, removing the microbiological major criterion would have downgraded roughly a third of definite cases to merely possible.⁴ Blood cultures are not just one piece of the puzzle; they are frequently the piece that clinches the diagnosis.

Imaging Findings

The imaging major criterion traditionally relied on echocardiography, the ultrasound of the heart. Seeing a vegetation, an abscess, or a new dehiscence (separation) of a prosthetic valve on echo counts as a major criterion. Transthoracic echocardiography (the kind done through the chest wall) is usually the first step, but transesophageal echocardiography, where a probe is passed into the esophagus to get a closer view of the valves, is recommended when the initial study is inconclusive, when complications are suspected, or when a patient has a cardiac implant with leads running through the heart.⁵

The same 241-patient analysis found that without the echocardiographic major criterion, 43% of cases would have been reclassified to possible rather than definite.⁴ In other words, nearly half of confirmed diagnoses depend on seeing something abnormal on heart imaging.

The Minor Criteria

Minor criteria capture findings that are consistent with endocarditis but not specific enough on their own to confirm it. Each one nudges the probability upward. The traditional list includes:

Predisposing condition: A heart condition known to increase the risk of endocarditis, such as mitral valve prolapse, a bicuspid aortic valve, a prosthetic valve, or a history of previous endocarditis. A systematic review found that mitral valve prolapse raised the odds of developing infective endocarditis roughly three- to eightfold, while having a bicuspid aortic valve carried about a sixfold increase.⁶ Injection drug use also counts as a predisposing factor because it introduces bacteria directly into the bloodstream.
Fever: A temperature at or above 38°C (100.4°F).
Vascular phenomena: Signs that infected material has broken off and traveled through the bloodstream, such as embolic strokes, septic lung infarcts, or certain hemorrhagic skin findings.
Immunologic phenomena: Immune-mediated complications like glomerulonephritis (a type of kidney inflammation), certain retinal spots called Roth spots, or rheumatoid factor. Routine testing for these immune markers in patients suspected of endocarditis helped reclassify more patients from possible to definite in one study.⁷
Microbiological evidence not meeting major criteria: A single positive blood culture with a typical organism, or serologic evidence of active infection with a known endocarditis pathogen, when the findings do not quite meet the threshold for a major criterion.

No single minor criterion is enough to diagnose the disease, but stacking several together, especially alongside a major criterion, can push a case into the definite category.

What Changed in the 2023 Update

The original 1994 Duke Criteria were modified once in 2000 and then substantially revised in 2023 by an international panel. The revision was prompted by the fact that the microbiology, epidemiology, diagnostics, and treatment of infective endocarditis had all changed substantially over the intervening decades.³ The biggest shifts involved microbiology and imaging.

On the microbiology side, the expanded list of typical organisms and the dropped timing requirements for blood cultures (described above) reflected how endocarditis-causing bacteria have shifted over the years. Staphylococcus aureus has overtaken streptococci as the leading cause in many countries, and healthcare-associated infections have become more common. The criteria needed to account for organisms that are clearly endocarditis pathogens in patients with prosthetic material but would be considered contaminants in other settings.

On the imaging side, the 2023 update formally incorporated advanced imaging modalities that did not exist or were not widely available when the original criteria were written. Cardiac CT scanning can now visualize abscesses, pseudoaneurysms, and valve damage with high resolution, providing an alternative when echocardiography is inconclusive.⁸ PET/CT scanning, which detects areas of abnormal metabolic activity by using a radioactive glucose tracer, has proven particularly useful for prosthetic valve infections where echocardiography struggles. In one study, PET/CT accurately reclassified patients from the ambiguous “possible” category into either definite or rejected, with sensitivity around 75–87% and specificity around 86–92% depending on whether the infection involved a native or prosthetic valve.⁹ Recognizing these tools as major imaging criteria was one of the most impactful changes in the 2023 revision.

How Well Do the Updated Criteria Perform

An external validation study compared the 2023 Duke-ISCVID criteria head-to-head against both the older modified Duke criteria and the 2023 European Society of Cardiology criteria. The updated Duke-ISCVID system was more sensitive than the older versions, correctly identifying about 84% of true endocarditis cases compared to roughly 75% for the modified Duke criteria and 80% for the 2015 European criteria, without a meaningful loss in specificity. When compared to the 2023 European criteria, the Duke-ISCVID system had similar sensitivity but was substantially more specific (94% versus 82%), meaning it was better at correctly ruling out people who did not have the disease.¹⁰ The biggest gains came from the new microbiological and imaging modifications.

That said, about 16% of true cases still slip through the criteria. This is where clinical judgment fills the gap. The Duke Criteria were always intended as a standardized framework to improve consistency, not as a replacement for a clinician’s overall assessment. A patient who does not technically meet criteria but whose clinical picture screams endocarditis will still be treated.

The Prosthetic Valve Problem

Prosthetic valve endocarditis has always been a diagnostic headache. Mechanical and bioprosthetic valves create acoustic artifacts on echocardiography, making vegetations harder to spot. The organisms involved can be different from those causing native valve infections, and the clinical presentation is often more subtle. One study evaluating the modified Duke criteria specifically in prosthetic valve cases found that how you handle the “possible” category dramatically affects performance: treating possible cases as positive yielded high sensitivity (96%) but mediocre specificity (60%), while treating them as negative flipped the balance to 74% sensitivity and 98% specificity.¹¹

This highlights a real clinical tension: the “possible” bucket, while useful for research categorization, leaves clinicians in limbo. The 2023 updates help somewhat by giving cardiac CT and PET/CT formal roles, since these tools are especially good at detecting infection around prosthetic material. But prosthetic valve endocarditis remains the scenario where the Duke Criteria are most likely to leave you uncertain. Cardiac device infections involving pacemaker and defibrillator leads pose a similar challenge, and the modified Duke criteria are routinely used for those cases as well.¹²

Children and the Duke Criteria

Infective endocarditis in children is less common than in adults, and most of the research underlying the Duke Criteria was conducted in adult populations. The criteria have been generalized to children with acknowledged limitations.¹³ When researchers specifically tested the modified Duke criteria in a pediatric population, sensitivity was about 88%, meaning the system correctly flagged most true cases. But 12% of children with confirmed endocarditis still failed to meet the threshold for a definite classification.¹⁴

One reason for the gap is that children with endocarditis are more likely to have congenital heart disease as their predisposing condition rather than the degenerative valve disease or injection drug use seen in adults. The spectrum of organisms can also differ. Children with congenital heart defects who have undergone surgical repair are at particular risk, and the echocardiographic findings can be complicated by the abnormal anatomy. Pediatric cardiologists tend to use the Duke Criteria as a starting framework while leaning more heavily on the overall clinical picture and imaging than the strict algorithmic score.

Blood Culture-Negative Endocarditis

One of the most frustrating scenarios in cardiology and infectious disease is the patient whose clinical picture strongly suggests endocarditis but whose blood cultures come back negative. This happens more often than you might expect. In the 241-patient analysis mentioned earlier, blood culture-negative cases accounted for over 70% of the cohort.⁴ That particular study was conducted in a setting where prior antibiotic use and fastidious organisms (bacteria that do not grow easily in standard lab conditions) contributed heavily to negative cultures, but even in well-resourced hospitals, culture-negative endocarditis accounts for a meaningful minority of cases.

When blood cultures are negative, the microbiological major criterion evaporates, and the diagnosis rests almost entirely on imaging and minor criteria. This is where advanced serologic testing for organisms like Coxiella burnetii (the cause of Q fever endocarditis) or Bartonella species becomes critical. The 2023 criteria expanded the serologic and molecular diagnostic options that can satisfy the microbiological criterion, partly in recognition of the fact that traditional culture-based diagnostics miss a substantial proportion of cases.

When the Duke Criteria Point the Wrong Way

False positives and false negatives both happen. False positives are relatively uncommon because the criteria are designed to be specific, but they can occur when a patient has a different condition that mimics endocarditis findings. Non-bacterial thrombotic endocarditis (sterile vegetations that form on valves in patients with cancer or autoimmune diseases) can produce echocardiographic findings identical to infected vegetations. A patient with bacteremia from another source, like a urinary tract infection, combined with a preexisting valve abnormality seen on echo could technically meet Duke Criteria without actually having endocarditis.

False negatives are a bigger concern. Early-stage disease, right-sided endocarditis (which tends to produce smaller vegetations that are harder to see), and infections caused by atypical or slow-growing organisms all reduce the criteria’s sensitivity. Patients who have received antibiotics before blood cultures are drawn are especially likely to be missed. The 2023 update improved overall sensitivity, but no version of the criteria has eliminated this blind spot entirely. Clinicians who strongly suspect endocarditis despite an initial classification of “rejected” will often repeat imaging and cultures over the following days rather than accepting the initial negative result at face value.

Where Advanced Imaging Fits In

The formal inclusion of cardiac CT and PET/CT in the 2023 criteria reflects a broader trend in how endocarditis is diagnosed at major medical centers. Cardiac CT excels at revealing structural complications like abscesses, pseudoaneurysms, and fistulas that echocardiography can miss, particularly around prosthetic valves where ultrasound shadows are a problem.⁸ PET/CT offers a fundamentally different kind of information: instead of showing anatomy, it shows metabolic activity, highlighting areas of active infection or inflammation that may not yet have produced visible structural damage.

These tools have their own limitations. PET/CT can produce false-positive results in patients who have recently undergone cardiac surgery, because post-surgical inflammation lights up on the scan in a way that can mimic infection. Cardiac CT requires intravenous contrast and radiation exposure, which are not trivial considerations. Neither modality is universally available, and in hospitals without access to them, echocardiography and blood cultures remain the diagnostic backbone. The Duke Criteria are designed to work with whatever imaging is available, but having access to the full toolkit clearly improves diagnostic accuracy.

How Immunologic Testing Adds Value

The immunologic phenomena minor criterion is sometimes overlooked in clinical practice because it feels old-fashioned. Roth spots are rare, and checking rheumatoid factor in a febrile patient is not always top of mind. But systematic testing for immunologic markers in patients under evaluation for endocarditis appears to have genuine diagnostic value. A study that implemented routine testing for these phenomena found that it helped reclassify more patients into the definite category, particularly in cases that were hovering at the possible-definite boundary.⁷ For clinicians who have a patient stuck in diagnostic limbo, checking these markers systematically rather than selectively can tip the scales.

Glomerulonephritis, one of the immunologic phenomena, deserves a specific mention because it is often detected incidentally through routine blood and urine tests rather than through targeted evaluation. A patient being worked up for possible endocarditis who shows unexplained kidney inflammation with blood in the urine has just gained a minor criterion that might not have been recognized if the connection to endocarditis was not considered. The criteria work best when the clinician is actively looking for each component rather than passively waiting for findings to appear.