BACKGROUND: Evaluation of incident outcomes requires excluding patients with prevalent disease. The appropriate length of the lookback period to define prevalent disease may vary depending on the outcome; chronic conditions may require all available data, but acute or recurrent conditions may require shorter periods. Objectives: To inform decisions about the length of prevaccination lookback periods, we investigated daily incidence of diagnoses of outcomes relative to vaccination.
METHODS: Patients with end-stage renal disease receiving hemodialysis were identified in the United States Renal Data System (2010-2016) on the date of influenza vaccination. For each adverse reaction (anaphylaxis, angioedema, seizure, encephalopathy, urticaria/hives, rash, pain in limb, cellulitis, myalgia/myositis, fever, nausea and vomiting, and syncope), we plotted daily incidence rates of recorded diagnoses relative to the vaccination date, from 180 days before to 180 days after vaccination. Multiple diagnoses per patient were permitted.
RESULTS: Among 520,876 vaccinations from 216,843 unique patients, daily incidences of many outcomes exhibited clear periodicity, frequently corresponding to regular health care interactions such as monthly comprehensive physician encounters or hemodialysis sessions every 2-3 days. Daily incidence rates of the following outcomes at least doubled in recurrent, 30-day spikes before and after vaccination: anaphylaxis, cellulitis, myalgia/myositis, nausea/vomiting, and fever. For example, daily incidence rates of fever were approximately 140-150 cases/10,000 patients for most days but were elevated to approximately 250-350 cases/10,000 patients in narrow spikes around days 30, 60, 90, 120, and 150 before and after vaccination. Both syncope and seizure outcomes exhibited weekly spikes in incidence; pain in limb, rash, diarrhea, and angioedema exhibited spikes occurring every 2-3 days. Plots of outcomes requiring hospitalization (e.g., encephalopathy, secondary outcomes of hospitalized fever, nausea/vomiting, diarrhea) did not demonstrate the same periodicity observed in nonhospitalized events.
CONCLUSIONS: Diagnoses for recurrent, minor, or sustained conditions may be routinely recorded at regularly occurring health care encounters. The duration of baseline washout periods should account for diagnostic patterns in special populations, such as end-stage renal disease, with regular patterns of health care utilization (e.g., extending washout periods to > 30 days to capture the previous month’s comprehensive physician visit).
