It would be hard to overstate how complicated the nervous and immune systems are. On top of the biology, the provision of medical care is also absurdly complex: there are all sorts of biases in how patients are treated, some of which turn into feedback loops. An early heart transplant might stave off vascular dementia, while a patient with severe dementia might not be eligible for transplants at all.
It is certainly possible for a new person to notice something that people already working in the field have missed. At the same time, domain knowledge is worth a lot, especially when trying to figure out if a relationship seen in messy, observational data is real. I'd be skeptical either way, but collaborating with a more established team would help a little bit.
I don't think that software, in general, has quite so many foot-guns laying around. However, people do give similar advice for cryptography and (sometimes) machine learning because those fields do offer a number of subtle ways to really foul things up.
It would be hard to overstate how complicated the nervous and immune systems are. On top of the biology, the provision of medical care is also absurdly complex: there are all sorts of biases in how patients are treated, some of which turn into feedback loops. An early heart transplant might stave off vascular dementia, while a patient with severe dementia might not be eligible for transplants at all.
It is certainly possible for a new person to notice something that people already working in the field have missed. At the same time, domain knowledge is worth a lot, especially when trying to figure out if a relationship seen in messy, observational data is real. I'd be skeptical either way, but collaborating with a more established team would help a little bit.
I don't think that software, in general, has quite so many foot-guns laying around. However, people do give similar advice for cryptography and (sometimes) machine learning because those fields do offer a number of subtle ways to really foul things up.