Use different national surveys for the numerator and denominator in estimating the incidence rates and the confidence intervals of a variety of injuries

Background/Purpose

A common epidemiological research problem is that desired information often resides in different surveys.  Few national data sources record injury case and exposure at the same time, making estimation of incidence rates difficult. Among the few which document both such as the National Health Interview Survey (NHIS), the available information (e.g. injury and time exposed) is constrained by the questionnaires’ content.  We sought to overcome these limitations by developing a procedure to use different national surveys for the numerator and denominator in estimating incidence rates and confidence intervals (C.I.) for a variety of injuries.

Methods

The 2010 National Health Interview Survey (NHIS) was used for the injury outcomes, and the 2010 American Time Use Survey (ATUS) for time of exposure. We used work-related and sport-related injuries as examples respectively. The incidence rate was estimated by dividing the number of population-level injuries (from NHIS) by population-level exposure (ATUS); the population totals were calculated with appropriate sampling weights. The standard error, and hence confidence intervals, were calculated using Taylor Linearization (Delta method), which took the complex survey structures into account.

Results/Outcomes

In 2010, the incidence rate for work-related injuries was 1.44 per 100,000 hours (C.I. = (1.10, 1.77)). For sport-related injuries, the incidence rate (and C.I.) was 13.36 per 100,000 hours (C.I. = (11.38, 15.38)).

Conclusions

This procedure makes it possible to estimate the incidence rates and 95% confidence intervals even when the numerator and denominator come from different data sources, offering an alternative approach for injury surveillance and research when data availability is constrained.