Social Network Effects of Nonlifesaving Early-Stage Breast Cancer Detection on Mammography Rates

Abstract

We estimated the effect of anecdotes of early-stage, screen-detected cancer for which screening was not lifesaving on the demand for mammography. We constructed an agent-based model of mammography decisions, in which 10|000 agents that represent women aged 40 to 100 were linked together on a social network, which was parameterized with a survey of 716 women conducted through the RAND American Life Panel. Our model represents a population in equilibrium, with demographics reflecting the current US population based on the most recent available census data. The aggregate effect of women learning about one category of cancers—those that would be detected but would not be lethal in the absence of screening—was a 13.8 percentage point increase in annual screening rates. Anecdotes of detection of early-stage cancers relayed through social networks may substantially increase demand for a screening test even when the detection through screening was nonlifesaving.

A General Agent-Based Model of Social Learning

“…One of the most important determinants of the dynamics that the model exhibits is whether each behavior, and , tends to lead to experiences that support the same behavior, or whether they lead to experiences that support switching behaviors. While the models described below are fairly stylized and are, for the most part, simpler than models we would develop for a particular problem (Nowak and Parker, 2014), they can begin to realistically capture the essential dynamics for representative classes of behaviors.…”

“…In other cases, combined information about a behavior and outcome is likely to be transmitted together (Box H). In prior work, we found that most of the time, when women learn about breast cancer diagnoses of others they know, they are aware both of the screening behaviors and prognoses of the women diagnosed (Nowak and Parker, 2014). Finally, sometimes an outcome is more observable than a behavior; for example, if an individual overeats when alone, the likely resulting overweight and obesity are highly observable, whereas the health behavior is not (Box I).…”

“…Subsequently, each individual who is informed about the teal experience informs two others about the black node's experience, and each individual who learns firsthand of the orange experience informs one other individual about the black node's orange experience. As a practical matter, in our breast cancer work, we did not model the effect of secondhand and thirdhand accounts, because it is difficult to gather information about secondhand or more distant accounts in a survey (Nowak and Parker, 2014).…”

“…However, from our survey, we found that many women were still receiving annual screenings, so we decided to model annual screening rather than biannual screening. Nowak and Parker (2014) describes in detail the sensitivity analyses we conducted to test the validity of using parameters we estimated from a statistical model predicting biannual screening to model annual screening. someone who was screened and diagnosed with early-stage breast cancer, screened and diagnosed with late-stage breast cancer, not screened and diagnosed with early-stage breast cancer, and not screened and diagnosed with late-stage breast cancer.…”

“…These are different processes, and they might lead to somewhat different population-level dynamics. Further discussion of limitations of this model can be found in Nowak and Parker (2014).…”