IDENTIFYING PREDICTORS OF PREVENTIVE BEHAVIORS – AN EXPERIMENTAL STUDY

Purpose: Our goal was twofold. Firstly, we wanted to test whether:

1) risk information modeled after the one provided by genetic testing for avoidable diseases,

2) increased effectiveness of prevention,

3) high disease severity

will lead patients to engage in preventive behaviors.

Secondly, we wanted to develop an incentivized experimental design to study decision making in the context of genetic testing hoping to enable a better dialog between health behavior researchers and behavioral economists.

Method: We designed a 2x2x2 full factorial incentized online experiment and varied the following factors: disease severity (high/low), likelihood of developing the disease (high/low) and effectiveness of prevention (high/low). Subjects performed simple memorization tasks to make money and were told that they may lose 45% or 90% of their income at the end of the experiment (disease severity). They were also informed that they belonged to a group with either a high or a low risk of losing money (likelihood of developing the disease). After performing memorization tasks, participants were offer information about their likelihood of losing money, which they either elected to see or not. Finally, we offered subjects an opportunity to decrease their likelihood of losing money by 13 or 26 percentage points (effectiveness of prevention) by paying a percentage of their income.

Result: 390 subjects participated in the experiment. On average 38.52% of participants decided to engage in preventive behaviors (SD= 48.71%). Logistic regression was used to identify predictors of preventive behavior while controlling for locus of control, various risk measures, time preferences, age, sex, education, religion, and income. Data confirmed the first of our hypotheses; being in a high risk group predicts engaging in prevention (odds ratio [95% CI] of 2.00 [1.26, 3.19], p=0.003). Surprisingly, we found effectiveness of prevention not to be a significant predictor (1.30 [0.81, 2.09], p=0.273). Disease severity, while significant (0.34 [0.21, 0.55] p < 0.001) suggests a relation opposite to what we hypothesized. Finally, participants who chose to see their risk information were more likely to buy prevention (2.87 [1.64, 5.01] p < 0.001).

Conclusion: This study was the first step towards using incentivized experimental methods in modeling how risk information influences preventive behaviors. Some of our findings justify closer investigation and further work in this area.