Estimating cardiorespiratory fitness in patients with obesity using machine learning

What is it about?

Cardiorespiratory fitness is an important indicator of health, but measuring it directly requires a demanding exercise test using specialized equipment. We investigated whether machine learning could estimate a person’s fitness level, measured as VO2max, using routinely available information such as age, sex, height, weight, waist circumference, blood pressure, and heart rate. Using data from 253 adults with severe obesity, we found that the models could estimate VO2max with an average error of about 11%. The results suggest that machine learning may offer a simpler, more accessible way to assess cardiorespiratory fitness when direct testing is difficult. However, the estimates are not yet as accurate as direct measurement, and larger training datasets, as well as larger studies in independent populations, are needed before the approach can be used in routine clinical practice.

Featured Image

Photo by VO2 Master on Unsplash

Photo by VO2 Master on Unsplash

Why is it important?

Measuring VO2max is an important way of assessing cardiorespiratory fitness, but conventional testing requires specialized equipment, trained personnel, and a demanding maximal exercise test. Our study is unique in focusing specifically on adults with severe obesity, a population in which existing prediction equations may perform less well. We show that machine learning can estimate VO2max from routinely available clinical and anthropometric measurements, achieving an estimation error of 11.3% when additional physiological measurements are included. This could provide an approach that lowers the burden of assessing fitness in clinical settings where direct testing is difficult or unavailable. Importantly, the models also provide an estimate of prediction uncertainty, offering a potential way to identify cases where the estimated fitness level may be less reliable.

Perspectives