Introduction

Injury research employs a wide array of approaches including descriptive epidemiology, analytic epidemiology, behavioral research, policy analysis, and program evaluation. For each approach, alternative methods are available to collect information about exposures, outcomes, or both. No method is ideal for all situations, but certain principles apply across studies. The study question, logistical, sampling, and budget issues determine the data collection method. In this chapter, we consider the suitability of various methods for different research questions. We address issues of feasibility, logistics of instrument design, and methods to enhance response rates and data quality.

Sampling

Collecting information from everyone in a large population can be prohibitively expensive. For this reason, investigators will often sample the members ofa study population.

The sample should represent the population to which the results are to be generalized and each person in the population of interest should have a calculable, nonzero probability of being selected. Achieving high response rates is important because of concerns regarding nonresponse bias (Lessler and Kalsbeek, 1992; Groves and Couper, 1998). Researchers often adjust or weight data so that the percent distribution of certain attributes of subjects (e.g., age, race, and sex) is equivalent to that in the population. This will reduce nonresponse bias to the degree that respondents within age–race–sex groups answer similarly to their nonrespondent counterparts.

Simple random sampling from a population has several advantages. First, the sample is easy to select when a list of all population members is available.