We examine textualfeedback using tens of thousands of patient comments and their correspondingnumerical ratings of medical care providers. Patient comments incrementallyexplain care providers’ current overall numerical ratings beyond granularsubcomponent ratings. Moreover, care providers appear to use the information inwritten comments to improve performance in the future; after controlling fornumerical ratings, care providers’ future performance improves when theyreceive textual feedback that is more negative, specific, or salient, althoughspecific comments appear to have a greater impact when they are not negative. Weexamine nuance in these average results by tying the content of comments todirectly related outcomes. Comments about wait times are associated with longerfuture wait times when the feedback is not negative, specific, orsalient, perhaps because care providers have already maximized performance on thisdimension. Further, care providers who receive patient comments about medical mistakesexperience a reduction in future mortality rates regardless of the manner inwhich the feedback is given, suggesting stylistic attributes of comments do notmatter when they include information that is inherently negative, salient, andspecific. Our results suggest that patient comments provide valuable feedbackto care providers, even though patients have a disadvantage in terms of medicalknowledge and expertise.