Background: Practice-based research networks (PBRNs) replicating the National Ambulatory Medical Care Survey (NAMCS) must sample more than 1 year to account for presumed seasonal variation in illnesses. This study evaluated the effects of seasonality on diagnoses within NAMCS family physician data.
Methods: Using combined data from the 1995-1998 NAMCS, diagnostic clusters that accounted for more than 1% of total visits were analyzed for seasonality. Seasons were coded categorically as dummy variables with summer as the reference category. A logistic regression was performed with each diagnosis as an outcome on the full data. To examine the ability of alternative sampling strategies to replicate the full year of data, a simulation study was carried out drawing 50 samples of 1,000 visits each for winter-summer and spring-fall sampling periods.
Results: We found 23 diagnostic clusters that had a frequency more than 1%, of which 10 had seasonal variations (P < or = .001), primarily between winter and summer. If sampling were restricted to spring, the diagnostic clusters of pregnancy and coronary artery disease would account for less than 1% of visits. All other diagnostic clusters, though changing rank order, would account for more than 1% if sampled in a single quarter. In the simulated sampling strategy, visit prevalence dropped below 1% for at least 1 diagnosis in 24 of 50 samples in spring-fall compared with 20 of 50 samples for winter-summer (P > .20).
Conclusions: There is little seasonal variation in the 23 diagnoses that occur in more than 1% of visits to family physicians. There is, however, important seasonal variation in the rank order of these diagnoses. A sampling strategy that uses any quarter of the year but spring (March, April, May) could be used to understand what diagnoses are frequently seen within a PBRN.