Why physicians struggle to stop prescribing, even to the detriment of patients

In 2021, researchers Gabrielle Adams and Leidy Klotz published a landmark paper titled “People Systematically Overlook Subtractive Changes.” Across eight experiments, the result was remarkably consistent: When people are asked to improve something, they instinctively add something. They rarely consider taking something away, even when subtraction is the better solution.

The researchers called this phenomenon subtraction neglect.

This tendency is neither laziness nor ignorance. It is a deeply embedded cognitive default that becomes even more pronounced under time pressure and cognitive load. This raises an important question: What happens when a profession built around thousands of daily decisions operates under the exact conditions that amplify subtraction neglect?

The answer may help explain one of health care’s most persistent challenges: overprescription.

When human bias meets a health care system

Overprescription is often framed as a failure of incentives, pharmaceutical marketing, fragmented care or patient expectations. And those factors certainly matter. But they share a common feature: They all reinforce a preexisting cognitive tendency toward adding things.

When patients present with symptoms, clinicians diagnose and treat. When new evidence emerges, guidelines expand. When side effects appear, additional therapies are frequently introduced to manage them. Every individual decision is grounded in rational clinical judgment. Yet over time, medication lists grow longer, complexity increases and opportunities to remove therapies receive far less attention.

The result is not simply overprescription. It is the absence of a systematic review for what no longer needs to be prescribed.

Subtraction neglect helps explain why.

Why medicine is engineered for addition

The health care system did not intentionally create overprescription. It evolved in a way that makes additive decisions easier, faster and more visible than subtractive ones.

Medical education teaches clinicians what to start, disease by disease. There is no equivalent curriculum devoted to what should be stopped. Clinical guidelines are overwhelmingly additive. Specialists optimize treatment within their area of expertise, at times (by no fault of their own), without full visibility into the broader medication burden carried by the patient.

The clinical encounter reinforces the same pattern. Writing a prescription results in an immediate, tangible intervention. Comprehensive medication review requires considerably more time and a high-fidelity medication reconciliation list, and its benefits are harder to see, both in the moment and over time. When a medication is added, the effect is visible and attributable. When one is removed, success looks like nothing happened: an adverse event that never arises, the hospitalization that never occurs. The data leave little doubt that the harms prevented are real, but no one can point to them. Prevention leaves no fingerprints.

Liability concerns push in the same direction. Starting a guideline-supported medication is straightforward to defend. Discontinuing a therapy, even appropriately, can feel riskier if an adverse outcome occurs later.

Patient expectations add another layer. Direct-to-consumer pharmaceutical advertising has conditioned many patients to ask about specific medications. Few arrive asking whether something can be safely removed.

Taken together, these forces create a system optimized for adding things.

Subtraction neglect supplies the cognitive bias. Health care infrastructure amplifies it.

What the numbers say

The scale is measured in lives, not just dollars:

• Among the leading causes of death: Medication-related harm is estimated to rank among the top causes of death in the United States, with 250,000 to 300,000 deaths a year attributed to adverse drug events. Overprescription is a major, preventable driver of that toll.
• $528 billion a year: That is the cost of nonoptimized medication therapy, approximately 16% of all U.S. health care spending (Annals of Pharmacotherapy, 2018).
• Much of it is preventable: Roughly half of adverse drug events trace to medication errors, meaning they are avoidable, not inevitable.
• Approximately 7.3 billion: That is the number of doses of potentially inappropriate medications dispensed to individuals 65 years and older in the U.S. in a single year (Medicare Part D, Beers Criteria).
• One in three: Roughly 34% of individuals 65 years and older on Medicare take at least one inappropriate medication; the percentage climbs above 75% once a patient reaches polypharmacy.
• An 8.2% to 17.1% increase: Between 1999 and 2018, the percentage of adults in the U.S. experiencing polypharmacy more than doubled. Medication side effects cause 750 senior citizens to be hospitalized every day.

The burden extends beyond cost. Approximately half of patients do not take medications as prescribed, and adherence declines as medication complexity increases. Individuals taking six or more medications daily are more than twice as likely to be nonadherent as those taking three or fewer.

At the same time, deprescribing has demonstrated measurable benefits. Across 231 clinical trials involving more than 44,000 patients, structured deprescribing interventions reduced inappropriate medication use, reduced adverse drug events and improved outcomes. Comprehensive medication review has been associated with a 26% reduction in all-cause mortality.

Every prescription represented in those statistics likely made sense when it was first written. And that is precisely why the problem persists.

The issue is rarely a single bad decision. It is the cumulative effect of hundreds of reasonable decisions that were never systematically revisited.

The deprescribing gap

Appropriate polypharmacy and accumulated prescribing are not the same thing.

A patient taking multiple medications may be receiving exactly the right therapies for their condition. Another patient may be taking medications that were appropriate years ago but no longer serve a meaningful purpose. Both scenarios can look identical on a medication list.

The distinction is critical. Yet, few workflows consistently force the question: What can safely come off the list?

Most health care systems are designed to identify treatment opportunities. Very few are designed to identify opportunities for reduction.

Subtraction neglect is no longer simply a cognitive bias at that point. It becomes embedded in operational practice.

Subtractive medication review is structurally different

Most pharmacy interventions introduce additional complexity. Prior authorizations, formulary restrictions, utilization management programs, member outreach campaigns, discount card searches and patient-facing applications all require someone to do more work.

Subtractive medication review approaches the problem differently. Its objective is not to create another barrier or another decision point. Its purpose is to identify opportunities that already exist but remain hidden inside fragmented clinical and pharmacy data.

Technology is particularly well-suited for that task. Large-scale analytics can evaluate formulary design, medication history, age, comorbidities, interaction risk, prescribing patterns and cost alternatives simultaneously. The technology performs the data-intensive work. The physician performs the clinical work.

The division of labor is intentional:

• Technology identifies opportunities.
• Doctors evaluate whether those opportunities make sense for a specific patient.

What the physician sees

The most effective systems do not issue directives. They surface questions such as:

• Is this medication still appropriate given the patient’s age, comorbidities and current clinical status?
• Was this therapy started for a condition that has since resolved?
• Is one medication being used primarily to manage the side effects of another?
• Is there duplication or therapeutic overlap?
• Is there a clinically equivalent alternative that would reduce financial burden?
• Is there a clinically significant interaction hidden within the patient’s broader medication profile?

These questions sound simple, yet they are precisely the questions that subtraction neglect makes easiest to overlook.

The Adams and Klotz research demonstrated that people become more effective problem solvers when prompted to consider subtraction. The same principle applies in medicine.

When physicians are given visibility into opportunities for simplification, safer and more efficient treatment pathways often emerge.

The bottom line

Overprescription is the predictable result of a universal human cognitive bias operating inside a health care system optimized for addition. Clearly, prescriptions are written for good clinical reasons. The challenge arises when those decisions accumulate faster than they are revisited.

Physicians remain the appropriate decision makers. They understand the patient, weigh competing risks and own the clinical outcome. The opportunity for technology is not to replace that judgment.

The opportunity behind new technology is to surface the signal that cognitive bias, time pressure and system complexity make difficult to see.

Sometimes the most important clinical question is not what should be added. It is what can finally be taken away.

Siva Mohan, M.D., an Atlanta-based interventional cardiologist and health care entrepreneur, is the co-founder and president of RazorMetrics. He also practices at Emory University School of Medicine and has held leadership roles with ePatientFinder and Health Point Partners. With a strong track record of raising capital, advising venture firms, and serving on boards for Emory Healthcare and the American Heart Association, Mohan combines clinical expertise and business acumen to drive health care innovation.