Evaluation for an observation effect in a prospective cohort study of oral health outcomes

Evaluation for changes in behavior due to research participants' knowledge that behavior is being observed (also referred to as a Hawthorne effect or reactivity) has received little attention in the dental literature. The Florida Dental Care Study, a prospective, non-randomized, longitudinal study of oral health outcomes, provides some inferential power to evaluate for an effect on dental care use due to participants' knowledge that this behavior was being observed. The purpose of this paper is to document that an observation effect can occur in dental studies, and to estimate its magnitude in four groups that were defined by their typical approach to dental care as stated at baseline: consistent regular attenders (CRAs); inconsistent regular attenders (IRAs); consistent problem-oriented attenders (CPOAs); and inconsistent problem-oriented attenders (IPOAs).

At baseline, 873 respondents with at least one natural tooth and who were 45 years of age or older participated for an interview and clinical dental examination. Respondents were asked about their dental care use in general and check-up use in particular at 6-month intervals over a period of 24 months.

Dental care use in general and check-up use in particular varied across time points and across the four groups of the sample. There was some stimulation in dental care use for the sample overall, but by the 18-to-24-month period, use had returned to baseline levels. In a direction opposite from that hypothesized, results from the CRAs suggested decreased use of dental care over the course of the 24 months of observation. No consistent pattern was evident for the IRAs, CPOAs, or IPOAs.

An observation effect was evident, but was modest in magnitude and differed within and between sub-groups of the sample. While self-selection into dental care user groups is an expected and desirable feature of this design, the size of the user/non-user groups was affected for some subgroups. We conclude that dental care studies with the potential for an observation effect should evaluate for this effect by distinguishing sub-groups of the sample based on their propensity (as stated at baseline) to use dental care. These differential effects across sub-groups should be taken into account as inferences are made.