Decision support systems are not new to health care. From prognostic scoring devices to automated second opinion surgery systems, clinical practices have long used informatics to aid in diagnostics and testing. But can a computer-based decision tool aid in end-of-life treatment decisions? David Wendler and colleagues at the National Institutes of Health report that a decision tool called a “population-based treatment indicator” was able to predict the wishes of incapacitated patients better than family or loved ones.
To use the decision tool, a physician enters the incapacitated patient’s personal characteristics and circumstances into a computer, such as well-educated, 70-year-old, male, Native American with severe Alzheimer disease and pneumonia. Based on treatment preferences of similar individuals, the computer analyzes the likelihood that this patient would want antibiotics to treat his pneumonia. In this case, findings show that 90 percent of highly educated Native American men over the age of 50 do not want antibiotics to treat pneumonia when in an advanced Alzheimer’s stage, providing strong evidence that this patient would not want antibiotics either.
When surrogates — or loved ones — attempt to predict patient treatment preferences, studies show an accuracy rate of about 68 percent. Given more defined socio-demographic in the database, Wendler says, a computer could predict incapacitated patient treatment preferences more accurately than family or physicians, who according to studies are poor predictors of patient treatment preferences.
“Of course the question remains: If this decision making tool is used as a supplemental aid, or used in lieu of family members making a decision about treatment, do they feel relieved of some unwanted decision-making burden -- or do they feel cut out of the process?” says Wendler, who says more work needs to be done on the impact such decision making has on surrogates.
Kenneth Goodman, director of the Bioethics Program at University of Miami, says, “A decision support tool like this supports the idea that there is a ‘reasonable person standard’ out there and indicates there may be resources for making inferences about patients’ preferences in the absence of explicit wishes. Is it airtight? How could it be? But surrogates and proxies have difficulty applying general preferences to specific circumstances anyway.”
Want to read more expert articles like this? Send an email to: