Hello Nicole,

In a case like this, starting with a process of elimination can be helpful. Using medical records is a perfectly valid source of information, and interviews were unnecessary for what the researchers were tracking, so both A and B are not correct answers.
From there you can compare options C and D. Both regard the number of participants in some way. In this case you may not be sure if the number of participants in the sample is sufficient, or if the number of cases excluded is significant. However, exclusions are always worth a closer look when it comes to limitations in design. The total sample of 5880 cases were taken randomly. The 567 excluded cases were the result of hospitals refusing to participate or having improper records, which could skew the randomization of the sample and introduce systemic error if there are confounding variables associated with the hospitals that refused participation. For example, if those hospitals that did not participate engage in systemic discrimination and didn't want it exposed, a 10% exclusion of such cases could significantly skew results and as such could pose a methodological limitation.

Nicole