We conducted a cross-sectional survey where unique respondents were treated as hypothetical patients. Each “patient” was presented with two scenarios: low risk chest pain and minor traumatic brain injury. Each scenario included three variables (benefit, risk, and cost), and each variable contained four possible values. The benefit of the test was defined as the chance that the patient had a true positive finding on the test requiring medical intervention. The risk of the test was defined as the chance of developing cancer due to ionizing radiation within the next ten years. The cost was an additional out of pocket expense for the test. The survey was pilot tested on medical students and revised based on feedback.
For the benefit and risk variables, the four possible values chosen were 0.1%, 1%, 5%, and 10%. For the cost variable, the four possible values chosen were $0, $100, $500, and $1000. These values were independently randomly distributed amongst respondents, yielding 64 unique scenarios. A subset of the minor traumatic brain injury respondents who had children under the age of 18 were given another scenario, requiring them to make diagnostic testing decisions for their child.
Adults were surveyed using Amazon Mechanical Turk (mTURK). Amazon mTurk is a crowdsourced internet marketplace that enables individuals and business to coordinate use of human intelligence to perform specific tasks. Anyone over the age of 18 with internet access was eligible to participate if they met Amazon’s vetting requirements as a performer of Human Intelligence Tasks. All 1,000 surveys were completed within 1 day of posting. Each respondent has a unique identifier and account with Amazon and was unable to perform the survey more than once. We provided a reimbursement of $1 for survey completion.
Outcome and Explanatory Variables
The primary outcome measured was whether the patient elected to receive testing under varying levels of benefit, risk, and cost. The following demographic information was collected: age, sex, current marital status, kids, level of education, healthcare worker or not, race, ethnicity, history of cancer, diabetes, hypertension, atrial fibrillation, heart attack, and overall self reported health status on a scale of 1-5.
Human Subjects Protection
This study was reviewed by the University of Michigan Institutional Review board and received a determination as exempt survey research.
Introduction: Diagnostic testing is common in the emergency department. The value of some testing is questionable. The purpose of this study was to assess how varying levels of benefit, risk, and costs influenced an individual’s desire to have diagnostic testing.
Methods: A survey through Amazon Mechanical Turk presented hypothetical clinical situations: low risk chest pain and minor traumatic brain injury. Each scenario included three given variables (benefit, risk, and cost), that was independently randomly varied over four possible values (0.1%, 1%, 5%, 10% for benefit and risk and $0, $100, $500, and $1000 for the individual’s personal cost for receiving the test). Benefit was defined as the probability of finding the target disease (traumatic intracranial hemorrhage or acute coronary syndrome).
Results: A total of 1000 unique respondents completed the survey. Increasing benefit from 0.1% to 10%, the percent of respondents who accepted a diagnostic test went from 28.4% to 53.1%. [OR: 3.42 (2.57-4.54)] As risk increased from 0.1% to 10%, this number decreased from 52.5% to 28.5%. [OR: 0.33 (0.25-0.44)] Increasing cost from $0 to $1000 had the greatest change of those accepting the test from 61.1% to 21.4%, respectively. [OR: 0.15 (0.11-0.2)]
Conclusions: The desire for testing was strongly sensitive to the benefits, risks and costs. Many participants wanted a test when there was no added cost, regardless of benefit or risk levels, but far fewer elected to receive the test as cost increased incrementally. This suggests that out of pocket costs may deter patients from undergoing diagnostic testing with low potential benefit.