The proper answer is not much. Any given report is actually one eyeball and that is never really good enough. You want comparable work by other eyeballs that also allow you to have a chance of check king the underlying data relied on.
It is all a work in progress and great care must be applied when using new information Are you really replicating? In medicine, that really matters.
I have often noted that several great mines have been located on highways driven by every geologist in the country. Remember those eyeballs. Are we paying him to have a positive statement regarding muddy facts?.
i would like to be more positive but it really does not work that way.
- Mechanical compression was not better than manual compressions for CPR. (JAMA. 2014;311:53)
- Early and aggressive methods for care of patients with sepsis (severe infection) were no better than usual care. (JAMA. 2017;318:1233)
- The REACT-2 trial found that routine use of an immediate total-body CT did not impact mortality or benefit compared with conventional imaging and selective CT scanning in patients with severe trauma. (Lancet. 2016;388:673)
- Platelet transfusion after acute hemorrhagic stroke was found by the 2015 PATCH study to worsen survival in the platelet transfusion group (68%) compared with the standard care group (77%). (Lancet. 2016;387:2605)
- The study should be a human study, or, if it is an animal study, the limits of such a study should be declared.
- The study should follow basic rules about how to determine causation, and avoid the trap of claiming that "association" or "coincidence" is proof of causation.
- The study should avoid surveys and questionnaires as a source of "evidence" since recall bias is always a problem in survey or response studies.
- The study should always be measured in terms of the magnitude of the "effect," and the rule is that magnitude of effect should be "robust" — at least 2 or 3 times the increase in effect over the baseline.
- The study should establish a mechanism to explain the causal effect asserted — for example, ice cream consumption is associated with an increase in drowning deaths, but it is not a cause of those deaths.
- Although I could argue that peer review and publication are not a good standard for reliability, the source of the research and the reputation of that source as well as the reputation of the journal the research was published in is often worth something. How much it is worth is the question.