Delivery of preventive health services for breast cancer control: A longitudinal view of a randomized controlled trial

Health Services Research, Feb, 2002 by Phyllis A. Gimotty, Robert C. Burack, Julie A. George

Randomized controlled trials (RCTs) have become the standard methodology for evaluating intervention strategies used to change the delivery of preventive services. Over the last decade, studies of intervention strategies to promote the use of preventive services have almost exclusively used comparative randomized trials where the effectiveness of the intervention strategy has been determined by increased utilization by the end of the study among those assigned to receive the intervention compared with those who were not. In the case of screening mammography, intervention strategies have focused on physician-specific interventions; including physician education, audit and feedback; and perhaps most effectively, computer-based reminders (McPhee, Bird, Fordham, et al. 1991; McPhee and Detmer 1993; Mandelblatt and Kanetsky 1995; Davis et al. 1995). Meta-analyses of RCTs of physician reminders have demonstrated that the use of reminders is associated with an increase in screening mammography by the end of the inte rvention period, although the magnitude of the effect size varied among studies (Balas, Austin, Mitchell, et al. 1995; Shea, DuMouchel, and Bahamonde 1996; Mandelblatt and Yabroff 1999).

The evaluation of physician mammography reminders in these studies is based on a patient-level outcome, completion of a mammogram during the intervention period, even though the intervention strategy promotes mammography by prompting the physician to refer women when they are due for the procedure. Mammography completion rates provide only partial information about referrals, as some referred women may fail to complete a mammogram, and reluctant women may receive multiple referrals before they succeed. An equally important view is to examine the process of referral as physicians encounter women on an ongoing basis throughout the intervention period. Indeed, the variability in effectiveness among studies characterized in the meta-analyses of studies evaluating physician reminder effectiveness may be due to differences in how physicians use the reminders, in addition to the attitudes and previous mammography experiences of their patients.

Although seldom used, a comparative longitudinal analysis of data from a RCT provides a framework in which to study the process of health service delivery, as well as changes in the delivery of preventive services that result from an intervention. In our previous work, we have used RCTs to evaluate the effectiveness of physician reminders to promote the use of mammography in the primary care setting. (Burack and Gimotty 1997; Burack, Gimotty, George, et al. 1994; Gimotty, Burack, and George 1999). We found that women randomized to physician mammography reminders had a significantly higher mammography completion rate at the end of 1 year of intervention than those who did not receive them in five different primary care clinics. In this article, we reanalyze data from one controlled trial (Burack, Gimotty, George, et al. 1994) and focus on the effect of physician reminders on the process of mammography referral characterized by referrals at primary care visits when women are due for mammography during each mon th of intervention (as opposed to completed mammograms by the end of the intervention). Subsequent to this analysis, we used a retrospective case-control study to examine patient characteristics associated with delayed referral.

CONCEPTUAL FRAMEWORK

The intervention strategy was developed within the context of the Health Belief Model, which postulates that among other factors, cues to action are an important stimulus to health behavior (Becker 1974). Consequently, the role of the physician mammography reminder was to increase the likelihood of a referral for all women seen by their physician while they were due for a mammogram. However, the Health Belief Model, as well as many other health behavior models, focuses on the current event and fails to include the role of previous health behavior outcomes that have accumulated during the past (see Elder, Ayala, and Harris 1999, for a review of health behavior models). Under the assumption that behavioral outcomes are positively associated, past behavior will be predictive of a subsequent behavioral outcome. In addition, patterns defined by previous health behavior outcomes will reflect perceptions of disease susceptibility, disease seriousness, effectiveness of the prevention strategy, and relevant barriers, key variables in the Health Belief Model and other models of health behavior. From this perspective, it is important to incorporate information on past experiences when examining the effectiveness of an intervention to change behavior over time when these data are available.

METHODS

Study Population and Study Design

The women in this study were visitors to a health department primary care clinic in Detroit, MI, between May 1, 1989, and April 30, 1990, and participated in a two-arm multisite RCT of computerized physician mammography reminders who became due for mammography during the intervention trial. Women were 40 years of age or older, had made a clinic visit during the year preceding the intervention trial, and had not had breast cancer. Details of the design of the RCT have been previously described (Burack and Gimotty 1997; Burack, Gimotty, George, et al. 1994).