I examine how individuals learn about health risks from household health shocks using US administrative data. When a family member is diagnosed with a chronic condition, relatives increase healthcare spending by 10%, a response that would require price declines as large as 50% to justify on demand alone. I quantify the mechanisms behind these effects, showing they are most consistent with individuals updating their beliefs about health risks. I evaluate the welfare and efficiency implications of this learning using a structural approach. I find that the majority of individuals overreact to diagnoses, overweighting their ex-post risks and offsetting potential welfare gains from informed decision-making.