Test vs. diagnostic: why sepsis tests don't need to be perfect to have an impact

Medicine is often framed in absolutes. A patient either has a disease or does not. A diagnosis is either correct or incorrect. In reality, clinical decision-making is far more complex. Every day, clinicians work with incomplete information, building a picture of a patient’s condition using symptoms, medical history, physical examination findings, laboratory results, imaging, and other data.

This reality contrasts with how new diagnostic tests are evaluated. In areas such as sepsis, where delays in treatment can have serious consequences, diagnostics are often expected to perform at near-perfect levels before they are considered useful.

Dr. Ephraim Tsalik, Vice President and Chief Scientific Officer for Infectious Disease and Acute Care, argues that this expectation may be limiting innovation. Instead of focusing on perfection, he suggests that the industry should prioritize additive diagnostic value and clinical impact.

The problem may begin with the word 'diagnostic'

According to Tsalik, one reason for the industry's expectations may be surprisingly simple.

"Some of it is just semantics," he says. "The word diagnostic implies that you're making a diagnosis."

That assumption creates a powerful expectation. If a test is described as a diagnostic, many clinicians naturally expect the result itself to be diagnostic.

"And therein lies a lot of the expectation. If I'm running a diagnostic test, as opposed to any other kind of a test, then the result I get should be diagnostic of the condition," Tsalik explains. "And so when it fails to be diagnostic, then people criticize the test as failing to achieve what it was set out to do."

Yet he argues that this does not reflect how medicine actually works.

"There are some tests that are used as diagnostics, but even the best diagnostic tests have limitations," he says. "And if we think about all the other tests that we regularly use, it's remarkable how often they are not intended to be diagnostic nor are they used that way by clinicians."

Instead, clinicians routinely combine information from multiple sources.

"They are just individual pieces of information that add to the collective story combined with the patient's history, the examination, the epidemiology," Tsalik says. "And all of those things ultimately serve to inform the diagnosis but not necessarily make the diagnosis."

In other words, many tests contribute incrementally rather than delivering certainty.

Why new tests face a higher bar

Established laboratory tests are deeply embedded in clinical workflows. Their strengths and limitations are well understood, and clinicians are comfortable interpreting their results.

New diagnostics, however, must earn that same level of trust.

“It is the difference between what we already know and what is unfamiliar,” Tsalik says. “New tests must overcome that lack of familiarity to find their place in the clinical process.”

Adoption depends on more than analytical performance. Hospitals and laboratories must consider workflow integration, staffing, infrastructure, and cost. Questions such as who will run the test, where the instrument will be located, and whether it delivers sufficient value all influence uptake.

These practical barriers can slow the adoption of innovative diagnostic technologies, even when they offer meaningful clinical benefits.

Bayesian reasoning in clinical decision-making

Tsalik’s perspective is grounded in Bayesian reasoning, which describes how clinicians update their thinking as new information becomes available.

“Bayesian reasoning is a formal term for what clinicians intuitively do,” he says. “Every time you get a new piece of information, you interpret it in the context of what you already know.”

For example, a symptom such as leg swelling could indicate a range of conditions. Its significance changes depending on the patient’s history, including factors such as heart failure, lymphedema, or prior blood clots.

Clinicians continuously revise their diagnostic hypotheses as additional information is gathered through questioning, examination, and testing.

“With each new piece of information, the probability of different diagnoses shifts,” Tsalik explains. “Some possibilities become more likely, others less likely, until you arrive at a working diagnosis.”

This process highlights a key point. No single test operates in isolation, and there is no such thing as a perfect diagnostic test.

“Every result has to be interpreted in context,” Tsalik says. “There is no test that can give you a diagnosis in a vacuum.”

Sepsis diagnostics and the challenge of sensitivity

Sepsis presents a particularly difficult case. The condition is associated with high mortality, and delays in treatment can worsen outcomes. As a result, clinicians are understandably cautious about relying on tests that could miss cases.

Tsalik points to an example of a test with 90 percent sensitivity. By most standards, this represents strong performance. However, clinicians often focus on the remaining 10 percent of cases that could be missed, which is true if nothing but this test is used to make the diagnosis.

This creates an environment where anything less than perfect seems inadequate. However, this perspective overlooks the broader clinical context.

If current approaches to identifying sepsis misses some cases, a new test that improves overall detection, even incrementally, could still deliver significant value. The question should not be whether a test is perfect, but whether it improves decision-making and patient outcomes.

Rethinking how diagnostic tests are evaluated

For decades, biomarker development in sepsis has focused on performance metrics such as sensitivity, specificity, positive predictive value, and negative predictive value. While these measures remain important, they do not fully capture how tests are used in real-world clinical settings.

Tsalik believes that the emphasis should shift toward clinical utility.

“We should be asking what impact the test has,” he says. “Does it change decisions? Does it improve outcomes?”

A diagnostic test does not need to be flawless to be valuable. It only needs to provide incremental, measurable improvement over existing approaches.

“A test only needs to make the diagnostic process better in a meaningful way,” Tsalik explains. “It does not need to be perfect.”

The risk of demanding perfection

Setting unrealistic expectations for diagnostic performance may have unintended consequences. If new tests are required to meet unattainable standards, potentially useful innovations may never reach clinical practice.

"The perfectionist mindset sets a threshold that cannot be met,” Tsalik says. “And in doing so, it can prevent the adoption of new diagnostics.”

This is particularly relevant in sepsis, where improved early detection could help clinicians initiate treatment sooner and improve patient outcomes.

Ultimately, the future of sepsis diagnostics may depend on a shift in mindset. Instead of searching for a single perfect test, the focus should be on identifying tools that enhance clinical decision-making.

Even small improvements can make a meaningful difference.

Lucy Lawrence, Science Editor for SelectScience was speaking with Dr. Ephraim Tsalik, Vice President and Chief Scientific Officer for Infectious Disease and Acute Care.