How Do You Diagnose Sepsis? What Doctors Look For

Diagnosing sepsis relies on recognizing that an infection has triggered dangerous organ dysfunction, and doctors piece that picture together using a combination of bedside assessment, vital signs, blood tests, and cultures. There is no single test that confirms sepsis the way a throat swab confirms strep. Instead, clinicians work through a layered process, starting with pattern recognition at the bedside and moving quickly into lab work and imaging. The modern definition itself has shifted in recent years, which changes what doctors are actually screening for and how urgently they act.

What Sepsis Actually Means Now

Until 2016, sepsis was broadly defined as infection plus signs of a body-wide inflammatory response, things like fever, elevated heart rate, rapid breathing, and abnormal white blood cell counts. A task force convened by the Society of Critical Care Medicine concluded that this framework was too vague. Many patients with a simple flu or a dental abscess could meet those criteria without being in any real danger. The revised definition, known as Sepsis-3, reframed sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection. Under this definition, the old term “severe sepsis” became unnecessary because organ dysfunction is now baked into what sepsis means in the first place.¹

This shift matters for diagnosis because it moved the goalposts. Doctors are no longer just asking “does this patient have an infection with a fever and fast heart rate?” They are asking “is this patient’s body failing because of an infection?” That is a harder question to answer quickly, but it focuses attention on the patients who are truly at risk of dying.

Bedside Scoring Systems

To identify organ dysfunction formally, the Sepsis-3 framework recommends using the Sequential Organ Failure Assessment, or SOFA score. This tracks six organ systems: breathing, blood clotting, liver function, cardiovascular stability, the nervous system, and kidney output. A rise of two or more points suggests clinically meaningful organ dysfunction. That threshold is associated with an in-hospital death rate above 10%.¹ The SOFA score was originally developed over 25 years ago for intensive care units, and researchers have been working on data-driven updates that may improve its accuracy.² One recent effort tested 56 organ function variables and found that optimized scoring outperformed the current SOFA system across patient databases in three countries.³

Calculating a full SOFA score requires lab results, so it is not always practical the moment a patient walks through the door. For quicker bedside screening, doctors sometimes use a simplified version called qSOFA, which checks just three things: altered mental status, fast breathing, and low blood pressure. If two of those three are present, the screen is positive. The appeal is that you need no blood work at all. The problem is that qSOFA misses a lot of sick patients. In emergency department studies, its sensitivity for detecting sepsis has come in around 28 to 34%, meaning it flags fewer than one in three sepsis cases.⁴⁵

An alternative that has gained traction is the National Early Warning Score (NEWS), which incorporates respiratory rate, oxygen saturation, temperature, blood pressure, heart rate, and level of consciousness. In head-to-head comparisons in emergency departments, NEWS consistently outperforms qSOFA for catching sepsis early. One study found NEWS had a sensitivity around 86% for sepsis detection compared with 34% for qSOFA, though qSOFA had higher specificity, meaning it produced fewer false alarms.⁴ A separate study showed similar patterns, with NEWS achieving an area under the curve of 0.91 for detecting severe sepsis and septic shock, compared with 0.81 for qSOFA.⁵ In practice, many hospitals use NEWS or a similar aggregate vital-sign tool as a first-pass screen, then pursue formal SOFA scoring once labs are back.

Blood Tests That Help Pin Down the Diagnosis

Once infection and organ dysfunction are suspected, blood work fills in critical details. The most common tests doctors order include a complete blood count, a comprehensive metabolic panel to check kidney and liver function, blood lactate, and inflammatory markers like C-reactive protein (CRP) and procalcitonin.

Lactate is one of the most important numbers in sepsis workups. When tissues are not getting enough oxygen or cells are under metabolic stress, lactate rises. Patients with sepsis tend to have meaningfully higher lactate levels than patients without it, and among sepsis patients, those who die have higher lactate readings than those who survive at virtually every time point measured.⁶ Lactate above 2 mmol/L, combined with the need for medication to keep blood pressure up, is part of the formal definition of septic shock.⁷ Doctors often recheck lactate every few hours. A level that stays high or climbs is a red flag that treatment is not working.

Procalcitonin has become a popular marker for distinguishing bacterial infection from other causes of inflammation. A systematic review and meta-analysis found procalcitonin was roughly 88% sensitive and 81% specific for bacterial versus non-infectious inflammation, compared with about 75% sensitivity and 67% specificity for CRP.⁸ In ICU patients, procalcitonin’s diagnostic accuracy for sepsis reached about 75%, with better specificity and positive predictive value than CRP.⁹ Still, neither marker alone can confirm or rule out sepsis, and both can be elevated for reasons that have nothing to do with infection. Procalcitonin is also useful for guiding when to stop antibiotics: if the level drops significantly, it often signals that the infection is under control.

Blood Cultures and the Timing Problem

Identifying which organism is causing sepsis is crucial for choosing the right antibiotic, and blood cultures remain the standard way to do that. A nurse or doctor draws blood, puts it in special bottles, and incubates them to see if bacteria or fungi grow. The catch is that cultures typically take one to three days to yield a result, and their sensitivity drops sharply once antibiotics have been started.

In one prospective study, blood cultures were positive in about half of sepsis patients who had not yet received antibiotics, but only about 28% of those already on treatment.¹⁰ A separate study found that cultures drawn after antibiotics were started had a sensitivity of only about 53% compared to pre-antibiotic cultures.¹¹ This creates a real tension in clinical practice. Guidelines push doctors to start antibiotics fast, ideally within the first hour, because early treatment saves lives. But the faster you start antibiotics, the less likely you are to identify the bug. The standard approach is to draw cultures first, then immediately start broad-spectrum antibiotics without waiting for results. If the cultures later identify a specific organism, doctors narrow the antibiotic choice to match.

Ultrasound at the Bedside

One tool that has quietly become valuable in sepsis workups is point-of-care ultrasound. An emergency physician or intensivist can scan the lungs, heart, abdomen, and urinary tract in roughly ten minutes to look for a source of infection. In a prospective study, ultrasound-guided assessment had about 75% accuracy and 73% sensitivity for identifying the septic source, and it improved the sensitivity of the initial clinical impression by about 25 percentage points.¹² For context, using standard workup alone, fewer than a quarter of sepsis sources were identified within the first hour. Ultrasound got there in under ten minutes.

Beyond finding the infection source, bedside ultrasound helps guide treatment. In patients with septic shock, it can reveal whether the heart is pumping well, whether the lungs are filling with fluid, and whether the patient needs more IV fluids or less. One review found that ultrasound-guided fluid management achieved hemodynamic improvement in over 97% of emergency department patients with septic shock.¹³

Molecular Diagnostics on the Horizon

The slow turnaround of blood cultures has driven interest in molecular tests that can identify pathogens directly from a blood sample, without waiting for anything to grow. One approach uses nanopore-based sequencing to detect pathogen DNA in a matter of hours. In a study of nearly 400 bloodstream infection cases, this technique identified pathogens in about 70% of samples, roughly double the 34% positivity rate of conventional blood cultures, and delivered results in approximately seven hours.¹⁴

Another emerging method targets free-floating DNA from pathogens circulating in the patient’s plasma. A proof-of-concept approach called SUPSETS combined targeted amplification with nanopore sequencing and delivered initial pathogen reads within minutes of starting the sequencing run, with full results in 11 to 13 hours. In its early clinical tests, it matched clinical metagenomics results and outperformed blood culture, while also detecting antibiotic resistance genes.¹⁵ These technologies are not yet standard in most hospitals, but they represent the direction the field is heading.

Researchers are also looking at the patient’s own immune response for diagnostic clues. A four-gene blood test called SeptiCyte Lab measures the expression of immune-related genes to distinguish sepsis from non-infectious inflammation. In validation studies involving hundreds of critically ill patients, it achieved an area under the curve of 0.89 to 0.95 for making that distinction, and its accuracy was not thrown off by how sick the patient was overall.¹⁶ Similar host-gene approaches are being tested in children, where novel transcriptomic signatures can distinguish bacterial from viral infection and identify the dysregulated immune response that leads to organ dysfunction.¹⁷

Why Sepsis Looks Different in Children

Adult sepsis criteria do not translate well to kids. Children have different normal ranges for vital signs at every age, and the organ systems that fail first tend to differ. For decades, pediatric sepsis diagnosis relied on adapted adult criteria that were widely seen as too sensitive, flagging too many children who were not actually in danger. In 2024, a major international task force introduced the Phoenix Sepsis Score, which evaluates four organ systems: respiratory, cardiovascular, coagulation, and neurological. A Phoenix score of 2 or more in a child with suspected infection identifies pediatric sepsis.¹⁸

In the development dataset, children who met this threshold had an in-hospital mortality rate of about 7% in higher-resource settings and roughly 29% in lower-resource settings, more than eight times the mortality of children with suspected infection who did not meet the criteria.¹⁹ The score performed better than previous pediatric criteria in terms of positive predictive value while maintaining comparable or higher sensitivity.²⁰ Pediatric septic shock is defined as sepsis with at least one cardiovascular point on the Phoenix score, meaning the child’s circulatory system is failing. In higher-resource settings, children meeting that definition had about 11% mortality; in lower-resource settings, about 34%.¹⁹

Older Adults and Atypical Presentations

At the other end of the age spectrum, sepsis in older adults is notoriously tricky. Age-related changes to the immune system mean that elderly patients often do not mount the classic fever-and-elevated-heart-rate response that younger patients do. Instead, the first sign may be confusion, a fall, sudden fatigue, or simply “not acting right” according to family members. Multiple comorbidities add noise to the picture: a patient with chronic kidney disease already has abnormal kidney labs, so a further change can be subtle. All of these factors contribute to delayed diagnosis in this population.²¹

This means that clinicians caring for older patients need a higher index of suspicion. A slightly altered mental status in an 80-year-old with a urinary tract infection should raise the same alarm bells as a 40-year-old with spiking fevers and a rapid heart rate. Scoring tools that lean heavily on vital-sign thresholds can underperform in this group because the “normal” baseline is already shifted.

Conditions That Look Like Sepsis But Are Not

Part of diagnosing sepsis is ruling out conditions that mimic it. A patient in anaphylaxis can look nearly identical to someone in septic shock: low blood pressure, fast heart rate, flushed skin. Pancreatitis, pulmonary embolism, adrenal crisis, drug overdose or withdrawal, and even spinal cord injuries can all produce vital-sign patterns and lab abnormalities that overlap with sepsis. No single lab test cleanly separates them. Diagnosis usually comes down to combining the clinical history, physical exam findings, and targeted testing to determine whether infection is truly the driver.²² When in doubt, most emergency physicians err on the side of treating for sepsis while simultaneously working up alternatives, because the penalty for missing sepsis is so severe.

Catching Sepsis Before the Hospital

Recognition increasingly starts before a patient reaches the emergency department. Paramedics are trained to screen for sepsis in the field, and studies have looked at how well they do. In one prospective study, paramedic identification of sepsis had a sensitivity of about 73% and specificity of about 79% compared with emergency physician diagnoses.²³ The negative predictive value was roughly 96%, meaning that when paramedics said a patient did not have sepsis, they were almost always right. Pre-hospital sepsis screening gives hospitals advance notice, allowing them to prepare labs, cultures, and antibiotics before the patient arrives.

Which screening tool works best in the field is still debated. A systematic review found that qSOFA’s simplicity makes it practical for paramedics, since it requires no equipment beyond a blood pressure cuff, but its sensitivity remains a weakness.²⁴ Modified versions of SIRS and NEWS that use only prehospital-available data have shown higher sensitivity, sometimes above 96%, though at the cost of more false positives.²⁵ The trade-off between missing cases and over-alerting is a recurring theme across all sepsis screening, whether it happens in an ambulance or a hospital ward.

The One-Hour Bundle and Why Speed Matters

Once sepsis is suspected, most hospitals follow a time-sensitive checklist called the “one-hour bundle.” This includes measuring lactate, drawing blood cultures, starting broad-spectrum antibiotics, giving IV fluids if blood pressure is low, and reassessing. A meta-analysis of studies comparing patients who received this bundle within the first hour against those who did not found that overall mortality was about 21% in the bundle group compared with roughly 25% in the non-bundle group, a statistically significant reduction.²⁶ Those numbers translate to real lives saved at scale. This is why sepsis diagnosis is not purely an academic exercise in scoring and lab interpretation. Every step, from the paramedic’s initial assessment to the first blood draw in the emergency department, feeds into a clock that has already started ticking.

Sepsis Subtypes and Personalized Diagnosis

One of the more intriguing directions in sepsis research is the discovery that not all sepsis looks the same at the molecular level. Using gene-expression data from blood samples across three continents, researchers identified four distinct sepsis subtypes. One subtype, characterized by a functioning adaptive immune system, had low mortality. The other three, marked by immune suppression, acute inflammation, or metabolic disruption respectively, all had high clinical severity and worse outcomes.²⁷

This matters for diagnosis because it suggests the question may eventually shift from “does this patient have sepsis?” to “which kind of sepsis does this patient have?” A patient whose immune system is suppressed likely needs a different treatment strategy than one whose innate inflammatory response has gone haywire. For now, subtyping remains a research tool, but it helps explain why some patients deteriorate rapidly while others with similar vital signs respond well to standard treatment. It also adds context to the AI-based early warning systems that hospitals are increasingly adopting. These tools mine electronic health records to flag patients at risk, but systematic reviews have found that integration challenges, alert fatigue, and excessive false positives still limit their real-world impact.²⁸ A subtype-aware AI system could, in theory, generate more meaningful alerts by distinguishing high-risk molecular profiles from low-risk ones, though that remains aspirational for now.

1
JAMA. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3)
2
PubMed Central. The Sequential Organ Failure Assessment (SOFA) Score: has the time come for an update?
3
Critical Care and Resuscitation. A framework and analytical exploration for a data-driven update of the Sequential Organ Failure Assessment (SOFA) score in sepsis
4
PubMed Central. National Early Warning Score (NEWS) Outperforms Quick Sepsis-Related Organ Failure (qSOFA) Score for Early Detection of Sepsis in the Emergency Department
5
PubMed. Comparison of SIRS, qSOFA, and NEWS for the early identification of sepsis in the Emergency Department
6
International Journal of Critical Care and Emergency Medicine. To Determine Blood Lactate Levels in Patients with Sepsis Admitted to a Respiratory Intensive Care Unit and to Correlate with their Hospital Outcomes
7
PubMed Central. New clinical criteria for septic shock: serum lactate level as new emerging vital sign
8
Clinical Infectious Diseases. Serum Procalcitonin and C-Reactive Protein Levels as Markers of Bacterial Infection: A Systematic Review and Meta-analysis
9
PubMed Central. Procalcitonin versus C-reactive protein: Usefulness as biomarker of sepsis in ICU patient
10
Clinical Microbiology and Infection. Impact of antibiotic administration on blood culture positivity at the beginning of sepsis: a prospective clinical cohort study
11
PubMed. Blood Culture Results Before and After Antimicrobial Administration in Patients With Severe Manifestations of Sepsis: A Diagnostic Study
12
PubMed. Accuracy of point of care ultrasound to identify the source of infection in septic patients: a prospective study
13
PubMed Central. Point of Care Ultrasonography for the Septic Patient in the Emergency Department: A Literature Review
14
PubMed. Molecular rapid diagnostic testing for bloodstream infections: Nanopore targeted sequencing with pathogen-specific primers
15
PubMed Central. Suppression PCR-Based Selective Enrichment Sequencing for Pathogen and Antimicrobial Resistance Detection on Cell-Free DNA in Sepsis
16
PLoS Medicine. A Molecular Host Response Assay to Discriminate Between Sepsis and Infection-Negative Systemic Inflammation in Critically Ill Patients
17
PubMed. Host gene expression signatures to identify infection type and organ dysfunction in children evaluated for sepsis: a multicentre cohort study
18
PubMed. International Consensus Criteria for Pediatric Sepsis and Septic Shock
19
JAMA. Development and Validation of the Criteria for Pediatric Sepsis and Septic Shock
20
PubMed Central. Optimizing Pediatric Sepsis Diagnosis: A Narrative Review of Two Decades of Evolution from SIRS to Phoenix Consensus
21
PubMed Central. Challenges with Diagnosing and Managing Sepsis in Older Adults
22
PubMed Central. Clinical Mimics: An Emergency Medicine-Focused Review of Sepsis Mimics
23
PubMed Central. Paramedic Recognition of Sepsis in the Prehospital Setting: A Prospective Observational Study
24
PubMed Central. Screening tools for sepsis identification in paramedicine and other emergency contexts: a rapid systematic review
25
PubMed Central. Diagnostic Accuracy of Simplified Prehospital Sepsis Screening Tools for Identifying Sepsis and Predicting 30-Day Mortality: A Retrospective Diagnostic Accuracy Study
26
PubMed. Association between 1-h bundle and clinical outcomes in patients with sepsis: A systematic review and meta-analysis
27
Communications Medicine. Sepsis endotypes identified by host gene expression across global cohorts
28
PubMed Central. Artificial Intelligence-Based Predictive Modeling for Early Detection of Sepsis in Hospitalized Patients: A Systematic Review and Meta-Analysis