The authors are almost just doing internet desk research -- counting the number of Covid deaths/cases and just plotting that against the average Vitamin D level in the population of each Indian state in general. Based on a couple thousand healthy patients of the general population. (BTW, I think it's really important that everyone learn to see through the math/symbols/technical jargon when reading such research papers to understand the populations really being studied. All those r and p numbers are meaningless if the underlying patient data are flawed.)
That is really rough. I could not imagine this would be credible if you substituted for example (just to think about it locally with a population you might know), "New York" for "Maharastra" state in the plot and asked if you believed the strength of finding that average Vitamin D level in New York state correlates with Covid death rate. How about for Florida -- would that be believable too? And also, the noticeable points making up the perceived trend are so few, and the points are not even weighted for population size (?)!
Much more work in detail -- at the patient level with Vitamin D measurement -- would have to be done to be credible. I would not act on this information in the present form.