Physicians can perform poorly when patients need special care, study suggests

The study of physician performance is the largest ever to be conducted using actors presenting as patients in doctors' offices. It appears in the July 20 issue of Annals of Internal Medicine and was funded by the Department of Veterans Affairs.

"Physicians did quite well at following guidelines or standard approaches to care, but not so well at figuring out when those approaches were inappropriate because of a particular patient's situation or life context," said Dr. Saul Weiner, associate professor of medicine and pediatrics at UIC and staff physician at the Jesse Brown VA Medical Center, who was lead author of the study.

Weiner said physicians need to understand why a patient is failing, for instance, to control their asthma, rather than just increase the dose of the drugs they prescribe. Specific issues -- such as the lack of health insurance, the need for less costly treatment, or difficulty understanding or following instructions -- must be recognized when making clinical decisions. Inattention to such issues leads to what are called "contextual errors" in patient care.

The study used actors trained to simulate real patients in 400 visits to a wide range of physician practices in Chicago and Milwaukee, including several VA sites. At each clinic, identities were created along with medical records and insurance information for the actor-patients. The doctors had all agreed to participate in the study but were not told which patients were actors.

Unlike real patients, the actors, or "unannounced standardized patients," consistently adhere to a script, enabling researchers to make comparisons of physicians' performance across the visits, said co-author Alan Schwartz, a methodologist and UIC associate professor of clinical decision-making.

Four case scenarios, each representing a common outpatient condition, were developed. Each case had four variants -- uncomplicated, biomedically complex, contextually complex, or both biomedically and contextually complex.

The actors followed scripts that contained hints or "red flags" of significant issues which, if confirmed, would need to be addressed to avoid error. The actors always started with the same two red flags, but were randomly assigned to respond differently based on the variant.

For example, in a case involving a 42-year-old man concerned about worsening asthma, the actor mentioned both a biomedical red flag (coughing at night) and a contextual red flag (losing his job) that suggested acid reflux and loss of health insurance, respectively, as a key part of the problem.

The study looked at whether the physician picked up on the red flags and implemented an appropriate care plan for each of the case variants.

At visits where no modification of customary practice was required, 73 percent of physicians provided error-free care.

But at visits where individualizing care required an alternative to the customary treatment, only 22 percent of physicians provided error-free care during a contextually complicated encounter, 28 percent during a biomedically complicated encounter, and 9 percent during a combined contextually and biomedically complicated encounter.

"To date, measures of doctors' performance have focused on situations where knowledge of the individual patient is ignored," said Weiner. "Under those conditions, physicians did fairly well. But as soon as care required more than following an algorithm -- finding out what's really going on with a patient and acting on that information -- only a minority of physicians got cases right."

The researchers also looked for differences related to physician demographics, training and experience; and for each case they had matched black and white actors randomly assigned to physicians. They also evaluated total time spent with the simulated patients.

"We expected that if physicians had more time with patients, they would be more likely to individualize care," Weiner said. "But what we found was that among those visits where physicians did a great job identifying contextual issues and addressing them, they did not on average spend any more time with patients than the physicians who didn't recognize contextual issues. That was surprising."

The study found that physicians were more likely to respond to the biomedical rather than contextual red flags even when both were equally important to planning appropriate care. "We believe that reflects the way in which physicians are educated," said Weiner. "The lesson here is that there has to be a dramatic change in the way we train physicians."

Other co-authors include Frances Weaver, Julie Goldberg, Dr. Rachel Yudkowsky, Gunjan Sharma, Amy Binns-Calvey, Ben Preyss, Dr. Marilyn Schapira, Dr. Stephen Persell, Dr. Elizabeth Jacobs, and Dr. Richard Abrams.