Assessing data quality in a community-based longitudinal study with middle-school adolescents

Background: Longitudinal studies require sustained levels of high quality data to estimate temporal exposure-disease relationships. Adolescents may be particularly susceptible to study fatigue.

Methods: The Rural Air Pollutants and Children’s Health study measured air pollutants (particulate matter <10µm in diameter and hydrogen sulfide) from industrial livestock operations at three North Carolina middle schools while 340 adolescents completed baseline surveys and daily diaries (symptoms, odor reports, and measured lung function) for 3-5 weeks in 2009. We assessed data quality using: 1) percentage of complete daily records (data completeness), 2) log-binomial and linear models assessing associations between baseline and diary responses (internal consistency) and, 3) fixed effects linear and logistic models evaluating day-in-study as a predictor of diary reports and pollutant concentrations (change over time).

Results: Of 5728 diary records, 94.2% were complete. Allergies were associated with asthma (Prevalence ratio=2.7, 95% Confidence Interval: 1.7, 4.3). Current asthma was associated with increased symptom reports (12% for headache, 7% wheeze, 10% backache), as was undiagnosed wheeze (17% for runny nose, 22% headache, 25% sore throat). Peak expiratory flow was 0.10 Liters/minute higher per year of age, 0.18 higher for males, and 0.11 higher per inch of height. Mean livestock odor scored 0-4 was 0.28 higher among students reporting livestock exposure at home. Lung function measures, symptom frequency, and odor scores declined substantially during follow-up, while pollutant concentrations remained variable. 

Conclusion: Data completeness and internal consistency do not translate to sustained data quality. Longitudinal studies with adolescents need short follow-up and efforts to maintain engagement.