Blood tests are currently a one-size-fits-all solution – machine learning can determine exactly what is actually “normal” for every patient

If a health care provider has ever ordered you to have a blood test, there may be probability they’ve done one complete blood countor CBC. One of the The commonest blood tests Billions of CBC tests are performed worldwide every year to diagnose diseases and monitor patients' health.

But despite the test's ubiquity, the best way doctors interpret and use it within the clinic is usually lower than ideal. Currently, blood test values ​​are based on uniform reference intervals that don’t take individual differences under consideration.

I’m a mathematician on the University of Washington School of Medicine, and my team is researching ways to make improvements using computational tools clinical blood test. In order to develop higher methods for recording individual patient definitions of “normal” laboratory values, my colleagues and I within the Higgins Laboratory studied at Harvard Medical School 20 years of blood tests of tens of hundreds of patients on the East and West Coasts.

In our newly published study, we used machine learning to discover healthy people Blood count areas for individual patients and predict their risk of future disease.

Clinical tests and complete blood count

Many people commonly view clinical tests as purely diagnostic tests. For example, a COVID-19 or pregnancy test is either positive or negative and provides details about whether you’ve gotten a certain disease. However, most tests don’t work this fashion. Instead, they measure a biological characteristic of your body regulates constantly up and down stay inside certain limits.

Your complete blood count can be a continuum. The CBC test creates an in depth profile of your blood cells – equivalent to what number of red blood cells, platelets and white blood cells are in your blood. These markers are used day by day in just about all areas of medication.

For example, hemoglobin is an iron-containing protein that permits your red blood cells to hold oxygen. If your hemoglobin level is low, it could indicate that you simply are Iron deficiency.

Platelets are cells that help form blood clots and stop bleeding. If yours The platelet count is lowThis may mean that you’ve gotten internal bleeding and your body is using platelets to form blood clots to plug the wound.

White blood cells are a part of your immune system. If your white blood cell count is high, it could mean you’ve gotten an infection and your body is producing more of those cells to fight it off.

Normal ranges and reference intervals

But all of this raises the query: What actually counts as too high or too low in a blood test?

Traditionally, doctors determine so-called reference intervals by conducting a blood test on a variety of healthy people. Typically they take the center 95% of those healthy values ​​and call them “normal,” with anything above or below being too low or too high. These normal ranges are used almost in every single place in medicine.

However, reference intervals face a significant challenge: what’s normal for you might not be normal for another person.

Almost all blood count markers are hereditary, i.e. yours Genetics and environment To a big extent, determine what the healthy value for every marker can be for you.

For example, on the population level a normal platelet count is between 150 and 400 billion cells per liter of blood. However, your body will want to maintain a platelet count of 200 – a price called your set point. This means your normal range may only be 150 to 250.

Differences between a patient's true normal range and the population-based reference interval may cause problems for doctors. They could also be less prone to diagnose a disease in case your set point is way from a limit. Conversely, in case your setpoint is just too near a limit, they could perform unnecessary testing.

Define what’s normal for you

Fortunately, many patients have blood work done yearly Routine examinations. Using machine learning models, my team and I were in a position to make estimates Blood count goal values for over 50,000 patients based on their clinic visit history. This allowed us to look at how the body regulates these set points and to check whether we are able to develop higher ways to personalize laboratory test values.

Over several a long time now we have found that the person normal ranges are approx 3 times smaller than on the population level. For example, while the “normal” range for white blood cell count is around 4.0 to 11.0 billion cells per liter of blood, now we have found that the majority people's individual ranges are much narrower, closer to 4.5 to 7 or 7.5 to 10. When We used these goal values ​​to interpret latest test results. They have helped improve the diagnosis of diseases equivalent to iron deficiency, chronic kidney disease and hypothyroidism. We were in a position to detect when an individual's result was outside their smaller personal range, potentially indicating an issue, even when the result was inside the conventional range for the final population.

The set points themselves were strong Indicators of future risks to develop an illness. For example, patients with high white blood cells were more prone to develop type 2 diabetes in the long run. Additionally, they were almost twice as prone to die from any cause as similar patients with low white blood cell counts. Other blood count markers were also strong predictors of future morbidity and mortality risk.

In the long run, physicians may have the opportunity to make use of thresholds to enhance disease screening and interpretation of recent test results. This is an exciting avenue for personalized medicine: defining what exactly “healthy” means to you based on your individual medical history.