In Reply We thank the letter writers for their insights. An overarching theme of the letters and public discussion of our article was whether our findings of lower mortality for patients hospitalized with cardiac arrest or high-risk heart failure during national cardiology meetings were causal or confounded by our inability to account for detailed clinical characteristics unavailable in claims data. We do not believe that our findings are due to confounding.
The goal of well-identified natural experiments is to search for sources of patient randomization occurring in real-world settings. In our study, we relied on the assumption that patients do not choose to have acute, high-mortality cardiovascular conditions specifically during cardiology meeting dates. Not only does this make clinical sense, but we demonstrated that on many dimensions (age, sex, race, and 10 different comorbid conditions) patients hospitalized on meeting and nonmeeting dates were statistically identical. Table 1 in our Original Investigation demonstrates a balanced cohort of patients hospitalized with high-risk cardiovascular conditions on meeting dates.1 Due to the balance in Table 1, we believe that patients hospitalized with these specific high-risk conditions on meeting dates were as good as randomized.
Jena AB, Romley J. Changes in Hospitalizations, Treatment Patterns, and Outcomes During Major Cardiovascular Meetings—Reply. JAMA Intern Med. 2015;175(8):1420-1421. doi:10.1001/jamainternmed.2015.1642