This paper has some red flags and it would be great if someone who has experience with this domain provides more light.

1) The original paper is at https://www.mdpi.com/1660-4601/17/20/7594

2) The paper was written based on the findings of the software https://wittyfit.com/ The sample is not random in any sense as it looks like its based on 187 people who are using this software. Correlation != Causation.