Humans, Health Care and Computer Algorithms

One supposed method to improve health care is to have software help people make good decisions and aid them in their health and health care decisions.  I have always been skeptical about the idea that this will either lower costs or improve people’s health.  The most likely users are actually younger, healthier people.  Really sick, expensive patients aren’t even likely to be using a computing device.  And even when people use software, their ability to effectively do so is often questionable.  So research on how people interact with computers and software is pretty interesting, as we see in a new paper from the National Bureau of Economic Research.   (NBER Paper)   This study was a randomized clinical trial in which the researchers provided a decision-making algorithm to older adults to help them choose a Medicare Part D drug benefit plan during the 2017 open enrollment period.  They were exploring how people reacted to personalized information and “expert” recommendations and what differences there were in people who did and did not use the tool or use it effectively.  The study builds on earlier research showing that only around 10% of beneficiaries switch plans in a year, even though switching could generate significant savings, and that people often don’t access or understand information about plans and how that information relates to their personal situation.  These authors sought to see if they could improve both the substantive amount of understanding a consumer had about plan characteristics and improve the consumer’s perception of the utility or value of a plan or features of a plan.

The decision-assistance tool automatically imported information about a beneficiary’s actual prescription drug use; then it created a comparison of likely out-of-pocket costs across different plan options and finally, the program would provide a recommendation on the best plan for the beneficiary.  The control arm of the study did not have access to the software tool.  One intervention arm got the personalized cost information but not a specific recommendation and another arm received the cost data and a specific recommendation for the three “best” plans for the beneficiary.  Those people who had access to the tool in either intervention arm spent more time on the choice of plan, were more satisfied with the process and were more likely to switch to a plan that would save them money.  Use of the software increased overall switching rates by 10 percentage points.  People’s decision-making appeared to be affected both by improved substantive understanding of a plan’s features and better assignment of value to those features.  Older people and those who were less IT savvy were more likely to benefit.  But at the same time, those who used the tool when made available were already more likely to be considering switching plans.  And as I suspected, a critical finding is that the people most likely to benefit from use of the tool, were the least likely to use it.  So a critical task is to figure out what will encourage, or force, consumers to use these decision-making aids that would be beneficial to them and to the system.  But the study clearly suggests that consumer decision-making in health care can be improved by the use of software and information tools.