Differences in demographic, behavioral, and biological variables between those with valid and invalid accelerometry data: implications for generalizability.

BACKGROUND The exclusion of participants with invalid accelerometry data (IAD) may lead to biased results and/or lack of generalizability in large population studies. The purpose of this study was to investigate whether demographic, behavioral, and biological differences occur between those with IAD and valid accelerometry data (VAD) among adults using a representative sample of the civilian noninstitutionalized U.S. population. METHODS Ambulatory participants from NHANES (2003-2004) who were 20-85 years of age were included in the current study and wore an ActiGraph 7164 accelerometer for 7 days. A "valid person" was defined as those with 4 or more days of at least 10+ hrs of monitoring per day. Among adults (20-85 yrs), 3088 participants provided VAD and 987 provided IAD. Demographic, behavioral, and biological information were obtained from the household interview or from data obtained in a mobile examination center. RESULTS Differences were observed in age, BMI, ethnicity, education, smoking status, marital status, use of street drugs, current health status, HDL-cholesterol, C-reactive protein, self-reported vigorous physical activity, and plasma glucose levels between those with VAD and IAD. CONCLUSIONS Investigators should take into consideration the potential cut-off bias in interpreting results based on data that excludes IAD participants.