It’s not uncommon for clinicians and health care executives to see one another as adversaries. It’s understandable when you consider the differences in training, schedules, and languages. But what these groups share is meaningful commitment to better outcomes for patients. Each side faces challenges, and I believe finding a common language would help all of us build a better, more effective health care system.
As clinicians, how do we know which of our treatments work best for patients? The “therapeutic illusion,” as it’s called, is the idea that we believe our interventions are more effective than they actually are in practice. In a classic example, a 1978 study found that for outpatients without definite evidence of illness, the outcomes of treatment versus no treatment were equivalent. In other words, there was no value received from the treatment, despite the best intentions of the physician (and the patient). Perhaps that example is too dramatic, possibly even absurd. But the idea that clinicians overestimate the impact of their treatments is not new. A recent New England Journal of Medicine article contained examples of overconfidence in treatment of back pain and cancer chemotherapy.
In the parlance of evidence-based medicine, this bias is an underestimation of the number-needed-to-treat. The number-needed-to-treat is an estimate of the effectiveness of a therapy — how many patients must be treated in order to achieve the desired outcome for a single patient. Said differently: for every treatment, some patients benefit, some are harmed, and some are unaffected. For example, the number-needed-to-treat for taking a daily aspirin to prevent a first heart attack or stroke within one year is 1,667. This means that for every 1,667th person who takes the drug, 1,666 either have no benefit or are harmed. Only one person avoids a heart attack or stroke. Here are a few more examples:
The number-needed-to-treat can be extended to topics in the realm of public policy and delivery system design. Let’s apply the concept to non-clinical leaders: how do we know if our systems strategies improve care for our population? A notable study in the American Journal of Public Health found that adults without insurance had measurably higher mortality rates than those with insurance. The number-needed-to-treat in this study to prevent death by providing adults insurance is 333 — for every 333 adults who receive health insurance, a life is saved. While very meaningful methodological and statistical caveats abound in this example, the logic alone suggests there may be as much value in delivery systems design as there is in the treatments those systems are designed to deliver. This feels like prime opportunity for collaboration.
For policymakers and leaders of health care organizations, these ideas should bring with them a sense of obligation: mistakes made in the boardroom can have the same consequences as mistakes in the exam room. If these number-needed-to-treat data come as a surprise, that suggests we are at risk of underestimating the value of delivery systems design. The therapeutic illusion can apply both ways. We seem to overestimate the effect of medical decisions, yet we may also underestimate the impact of system-wide design choices.