Big data in the clinical laboratory: a win-win situation

Medical decisions are often based on test results from the clinical laboratory. To facilitate the interpretation of these test results, they are compared to the average test results of a group of 'healthy' persons, resulting in a so-called reference interval.

The selection of a valid reference population is not easy. You never know for sure whether the selected group is really healthy and you have to take into account variables such as age, gender, BMI, diet etcetera. Moreover, as each person is unique, reference intervals should be personalized.

Age and gender are known factors that are already taken into account when creating reference intervals. However, the impact of other influencing factors on test results, e.g. renal function, is usually not included in the reference interval. It remains hidden in the pile of data. For the doctor, it is therefore not always easy to interpret test results correctly.
 

Synergy

Collecting additional patient data from various hospital data systems and cleverly combining it with the laboratory's test results can help the doctor.

The search for this synergy is challenging, however, because the application of data analytics within the clinical laboratory is only just emerging. In addition, hospital systems are not designed to use clinical data for the development and implementation of data-driven models.

In her dissertation, Saskia van Loon describes three examples of how to simplify the increasing complexity of interpreting test results from the clinical laboratory by applying data analytical techniques.

Visualization

First, she shows that by visualizing laboratory data in a different way, certain patterns become visible. This makes it immediately clear for a number of tests what is 'normal' and what is 'abnormal' within a certain patient group.

Next, the researcher presents a model that doctors can use when assessing an elevated test result for vitamin B12 deficiency. With this model, doctors can better assess whether the elevated test result is due to vitamin B12 deficiency or impaired renal function. 

Health Index

Finally, Saskia van Loon describes the development of a model for the health status of extremely overweight patients. These patients often have other diseases besides obesity, such as insulin resistance, hypertension or dyslipidemia. The combination of these diseases is called metabolic syndrome and results in an increased risk of cardiovascular disease.

Metabolic syndrome is one of the requirements for extremely obese patients to receive bariatric surgery. However, it is difficult to properly assess a patient's metabolic health status. The researcher therefore developed the metabolic health index ("MHI"), similar to the body mass index (BMI) for weight.

The MHI model combines multiple laboratory test results and expresses metabolic health status in one number. An MHI cutoff value has also been derived, which helps to distinguish between metabolic "healthy" and "unhealthy" patients and thus supports the decision to operate or not. The MHI model was evaluated with data from over 11,000 patients from multiple bariatric centers in the Netherlands.

In practice

The Catharina Hospital in Eindhoven (CZE) is now using both models. The additional information is reported at the same time as the test results from the laboratory and is therefore immediately available to the doctor, without additional costs.

The research presented contributes to the field of laboratory medicine, where more attention is being paid to developing data-driven solutions to clinical needs. For example, the laboratory at Catharina Hospital has also developed the CoLab score, which can be used to quickly screen patients in the emergency room for coronavirus.

Title of dissertation: The search for synergy in laboratory data. Supervisors: Volkher Scharnhorst (CZE, TU/e), Uzay Kaymak (TU/e); Co-supervisor: Arjen-Kars Boer (CZE).