Can Wearable Devices Detect Sepsis Early?
Your smartwatch counts your steps, tracks your sleep and warns you when your heart rate climbs. Could the same small device also spot sepsis before you feel seriously ill? Researchers are testing that hopeful idea right now. Sepsis moves fast, so a few hours of extra warning could save lives.
In short, wearable devices cannot yet detect sepsis on their own, and no smartwatch sold today can diagnose it. What they can do is watch the vital signs doctors use to spot deterioration, day and night. That steady stream of data, paired with smart software, is where the real promise lies. Our pillar guide, Septosis: Definitive Guide to Understanding, Recognising & Managing Sepsis, explains the condition in more depth.
Please note: this article offers general information only, and it cannot diagnose you or replace medical advice. A wearable is not a medical device, and a normal reading from a watch never rules out sepsis. If you have an infection and feel very unwell, confused, breathless or cold, seek urgent help at once. Call 999 or go to A&E, and say you fear sepsis.
What Wearables Actually Measure
Sepsis is a life-threatening reaction to an infection. The World Health Organization (WHO) describes it as a condition in which the body’s response to an infection injures its own tissues and organs. If doctors miss it early, it can lead to septic shock, organ failure and death. The WHO reports 48.9 million cases and 11 million deaths in a year, about 20 percent of all deaths.
A wearable measures the body from the outside, usually from the wrist, chest or upper arm. Most smartwatches use a light-based method called photoplethysmography, or PPG. A green light shines into the skin, and a sensor measures how much light bounces back as blood pulses underneath. From that signal the watch works out heart rate and heart rate variability, the small changes between beats.
Heart rate is the most useful signal a wearable can offer for sepsis. When infection takes hold, the heart usually beats faster as the body tries to push blood to the organs. A resting heart rate that climbs steadily over hours, especially with a fever, is one of the earliest changes doctors look for. Our guide Early Signs of Sepsis: The First Symptoms You Should Never Ignore lists a fast heartbeat among the first symptoms, with confusion, shivering and breathlessness.
Newer watches and chest patches also track skin temperature, blood oxygen, breathing rate, sleep and movement. A fever, or an unusually low temperature, can be an early sign of infection, and sepsis can make breathing harder. What a wearable cannot measure matters too: it cannot take a blood test, so it cannot check lactate or white blood cells, and it cannot tell which germ is causing the infection. Our article How Is Sepsis Diagnosed? Tests, Scores and What Doctors Look For explains the hospital tests no watch can replace.
How Early-Warning Scores and AI Models Use That Data
A single high heart rate reading means very little on its own. Exercise, stress, a hot bath or coffee can all push it up. The value of a wearable is that it records readings every few minutes, day and night, so software can look for a pattern over time. A heart rate that rises slowly for six hours while you rest is far more worrying than a brief spike during a brisk walk.
Hospitals already turn vital signs into scores. In England, the National Early Warning Score, or NEWS2, is the standard tool for spotting a patient who is getting worse. NHS England endorses NEWS2 and requires it in acute hospitals and ambulances. The score combines breathing rate, temperature, oxygen, blood pressure, pulse rate and consciousness. A score of 5 or more needs urgent review.
Artificial intelligence models go a step further than a fixed score. Instead of adding up points, a model learns from thousands of past hospital records, looking for subtle patterns before sepsis. It might notice heart rate variability falling, breathing rate rising and temperature creeping up, all at once, hours before any single measure looks alarming. Our article Can AI Predict Sepsis Before Symptoms Appear? explains how these models work.
The newest models are small enough to run on the wearable itself, rather than sending every reading to a distant computer. Researchers at ETH Zurich built a model called SepAl that uses only six vital signs a wearable can measure, including heart rate, temperature and movement. Other models focus on heart rate alone. In one recent study, the best model predicted sepsis well one hour ahead, but its accuracy fell when it tried to look four hours ahead.
What the Evidence Shows So Far
The evidence is real, but it is early. Most studies so far are small, were carried out in hospitals, and used research-grade sensors rather than a shop-bought smartwatch. None proves that a consumer watch can detect sepsis at home. What they show is that continuous monitoring can spot trouble hours earlier than routine checks.
One clear example comes from cancer care. At the University of Michigan, researchers gave 62 patients a wearable temperature sensor that took a reading every two minutes. It spotted dangerous fevers about five hours earlier than standard checks, which in hospital usually happen only every four to eight hours. The patients were having stem cell transplants or CAR-T therapy, which leave the immune system very weak.
A study at Mount Sinai in New York followed 30 people receiving CAR-T therapy for multiple myeloma, a type of blood cancer. Among 25 patients who could be fully assessed, wearables picked up 18 of 20 episodes of cytokine release syndrome, a severe inflammatory reaction much like sepsis, a median of seven hours before nurses recognised it. The team stressed the study was small and single-centre, so larger trials are needed.
In Bangladesh, researchers tested a wireless wearable on 100 children with sepsis in intensive care in Dhaka. The device captured more than 99 percent of expected data over 2.2 days on average. A model built only from wearable data flagged worsening sepsis more than 2.5 hours earlier than the clinical team’s notes. A 2025 review of seven studies found wearable algorithms predicted sepsis deterioration 5 to 48 hours before usual recognition.
The SepAl model shows both the promise and the limits in one result. It gave a median warning of 9.8 hours before sepsis onset, using only six wearable signals. Yet its sensitivity was 0.60 and its specificity was 0.48, so it missed many cases and raised many false alarms. A model that cries wolf too often will be ignored, and one that misses cases cannot be trusted.
What Wearables Cannot Do Yet
The limits are not small details. They decide whether the technology is ready for everyday use, and right now it is not. The first limit is accuracy. Wrist sensors estimate heart rate from light bouncing off skin, and that signal is easily disturbed. Movement is the biggest problem: when you walk, type or lift weights, the sensor can lock onto your movement instead of your pulse. Studies find errors during activity are on average about 30 percent higher than at rest.
Fit and skin also affect the reading. A loose strap, cold hands, sweat, a tattoo under the sensor, or darker skin can all weaken the light signal. Some studies find larger errors in people with darker skin, while others find no clear difference, so the picture is unsettled. For fitness, a few beats out matters little, but for medical decisions a small error could hide a real change.
The second limit is that vital signs are not specific to sepsis. A fast heart rate and a fever can mean flu, a chest infection, dehydration, anxiety, or a hard workout. A wearable cannot tell these apart, because it cannot see the infection itself. That is why models trained only on vital signs raise false alarms. At home, a false alarm could cause panic, or teach a person to ignore the next alert, which might be the real one.
The third limit is that most evidence comes from hospital patients, not healthy people at home. Hospital studies use research sensors and patients who are already ill, so results may not carry over to a smartwatch worn at home. No large trial has yet shown that wearables at home reduce sepsis deaths. A watch also needs charging, and data needs a phone or Wi-Fi link to reach a doctor. Our article How AI Helps Doctors Diagnose Infections Faster explains why human judgement stays at the centre of hospital AI.
Who Might Benefit Most
Some groups stand to gain most when the technology matures. The first is people who have recently left hospital after sepsis, surgery or a serious infection. They face a higher risk of a second infection, and they are sent home to watch for warning signs, yet nobody watches with them through the night. A wearable that alerts a care team to a rising temperature could bring them back sooner.
The second group is people with weakened immune systems, including people on cancer treatment, transplant recipients and people taking medicines that suppress immunity. The WHO names weakened immune systems among the higher-risk groups for sepsis. For them, infection can worsen very quickly, and a fever may be the only early sign. That is why the most convincing studies so far focus on cancer patients, and why NHS trial plans prioritise them.
Older adults are a third group. The CDC lists adults aged 65 or older among the higher-risk groups, and older people often live alone or have long-term conditions that make infection harder to spot. Sepsis can cause confusion, so a device that notices a change when the person cannot is genuinely valuable. In places with fewer health services, a low-cost wearable can also fill the gaps between nurse checks.
What to Do If Your Watch Flags Something
If your watch alerts you to a high heart rate, a fever or an unusual pattern, treat it as a prompt to check how you feel, not as a diagnosis. First, look at the reading in context. Are you exercising, stressed or hot? Put the watch back on snugly, sit down, and see if the reading settles. A single odd number is often a sensor error, especially if you were moving.
Secondly, check yourself, not just the screen. Do you have a fever, chills or shaking? Are you breathing faster than usual? Do you feel confused, dizzy or extremely tired? Have you passed much less urine than usual? The NHS lists confusion, slurred speech, shivering, muscle pain, difficulty breathing and not passing urine as common sepsis symptoms in adults. If any are present with an infection, you need urgent help.
- Call 999 or go to A&E if you have an infection and feel confused, breathless, very cold, or extremely unwell.
- Tell the medical team about the infection, the watch alert and how long symptoms have lasted.
- Do not wait to see if the reading returns to normal, and do not drive yourself if you feel faint.
- Take your medicines list with you, and mention any condition that weakens your immune system.
- If you feel well and the alert was brief, watch your symptoms for a few hours and call your GP or NHS 111 if unsure.
Frequently Asked Questions
Can my Apple Watch or Fitbit detect sepsis?
No. No consumer smartwatch can detect or diagnose sepsis. A watch can track heart rate, temperature and oxygen, and alert you to an unusual pattern, but those signs have many causes other than sepsis. Only a clinician, using blood tests and a full assessment, can diagnose sepsis. Treat a watch alert as a reason to check how you feel, not as a diagnosis.
How early could a wearable warn about sepsis?
Studies suggest hours, not days. Research models gave warnings from about 2.5 hours to around 10 hours before sepsis was recognised in the usual way, and hospital temperature sensors spotted fevers about five hours earlier than routine checks. Those figures come from small hospital studies, not from what a shop-bought watch will do at home.
What vital signs matter most for sepsis?
Heart rate, temperature, breathing rate and blood oxygen matter most, because they are the measures in the NEWS2 score. Heart rate and temperature are the easiest for a wearable to track, and a rising heart rate with a fever is a common early pattern. Blood pressure and consciousness also matter, but a wrist device cannot measure them reliably.
Are wearable readings accurate enough for medical use?
Not yet for sepsis detection. Heart rate readings are usually reasonable at rest, but errors rise during movement and differ between devices. Temperature from the wrist is less reliable than a thermometer, and wrist oxygen can drift by a few points. Medical-grade chest patches are more accurate than wrist watches, which is why hospitals use them in trials.
Will the NHS give me a wearable to watch for sepsis?
Not as a routine service yet. The NHS has announced plans to trial wearable sensors for patients most at risk, and some hospitals already use them for high-risk groups such as cancer patients on CAR-T therapy. For now, wearables for sepsis are mainly used in trials and specialist care. If you are at high risk, ask your care team if any monitoring programme is available.
Should I buy a smartwatch to protect myself from sepsis?
No, not for that reason alone. A smartwatch can help you notice changes in heart rate or sleep, but it is not a sepsis detector and buying one will not protect you. The best protection is knowing the warning signs, treating infections early, keeping vaccinations up to date and seeking help fast when an infection worsens.
Final Verdict
Wearable devices are a promising tool for sepsis, but they are not a solution yet. They can watch the vital signs that matter, day and night, and feed that data into scores and AI models that spot trouble hours earlier than routine checks. The early studies, from cancer wards to children’s intensive care in Bangladesh, show the idea works in principle.
Yet the limits are clear. Consumer watches are not accurate enough, vital signs alone cannot prove sepsis, false alarms are common, and no large trial has shown wearables at home save lives from sepsis. A watch is a helpful extra pair of eyes, not a doctor. If you have an infection and feel seriously unwell, act on your symptoms, not your screen. Call 999, go to A&E, and say you fear sepsis.