Behavioral Economics Interventions to Improve Medical Decision-Making

Abstract

Background

In medicine, a wide gap exists between the medical care that ought to be possible in the light of the current state of medical research and the care that is actually provided. Behavioral biases and noise are two major reasons for this.

Methods

We present the findings of a selective literature review and illustrate how interventions based on behavioral economics can help physicians make better decisions and thereby improve treatment outcomes.

Results

A number of behavioral economics interventions, making use of, for example, default settings, active decision rules, social norms, and self-commitments, may improve physicians’ clinical decision-making. Evidence on long-term effects is, however, mostly lacking.

Conclusion

Despite their apparent potential, the application of behavioral economic interventions to improve medical decision-making is still in its infancy, particularly in Germany.

In medicine, there exists a gap between what ought to be possible and what is actually achieved in care provision. One important reason for this gap is bounded rational behavior which then goes on to exacerbate many health problems such as hypertension, diabetes, and obesity. It is estimated that up to 40% of premature deaths in industrialized countries are attributed to largely preventable behaviors (1). Recommendations from doctors are not followed, alcohol and nicotine are consumed excessively, people have unhealthy diets and too little exercise (1– 3). Currently, a frequently asked question is whether enough people are seeking vaccination against COVID-19 (4, 5).

cme plus

This article has been certified by the North Rhine Academy for Continuing Medical Education. Participation in the CME certification program is possible only over the internet: cme.aerzteblatt.de. The deadline for submission is 22 September 2023.

New treatment methods alone will not bring about a reduction in morbidity and mortality. People can be helped to make and implement decisions that are consistent with their own demands and goals (1, e1). It is often assumed that providing adequate understanding alone could produce good decisions. But this idea rests on an optimistic self-image of the human being, namely the image of a (fully) rational decision-maker who pursues clearly defined goals, rationally processes all available information, and makes decisions under perfect self-control. This self-image is misleading (e2). We know, for example, that smoking, little exercise, poor medication compliance, and using a cellphone while driving can cause us harm, yet we sometimes fail to act accordingly.

Even well-informed doctors are not always immune to such behavioral flaws: A recent empirical study from the United States shows that physicians often do not behave significantly differently when it comes to health than the general population with similar levels of education (6). Guideline-adherence to medication use can also be significantly lower among physicians than among people without medical expertise (7). The reason is our habits as well as our choice architecture (that is, the linguistic, physical, emotional, and social environment in which people make decisions) which make it harder for us to reach good decisions. Along with this, we have systematic biases in information processing and decision-making. This is the point of departure of behavioral economics research.

The image of the rational decision-maker has dominated economics for decades and is the basis of many public health measures (for example, tobacco taxes). In recent years, however, behavioral economics approaches have been increasingly used in health policy to foster healthy lifestylse, for example, with regard to nutrition, to reduce excessive cigarette and alcohol consumption, and to increase vaccination rates (3, e3– e5). So far, these measures have addressed primarily insured persons and patients, but not physicians (8, 9). The relevance of behavioral economics approaches is highlighted by the creation of behavioral insights teams, or nudge units, in the UK and many other countries, in international organizations (for example, World Health Organization), in health insurance companies, and in digital companies involved in the science and practice of behavior change. Nudge units are assigned to important decision-maker entities (for example, the British Cabinet Office, which supports government work).

Yet, when it comes to improving clinical decision-making of physicians, the application of the behavioral economics approach is still in its infancy (10). Every day, physicians make numerous, sometimes complex, decisions about patients’ diagnoses and treatment plans, sometimes with only limited information and under time pressure (11, 12). By necessity, the decision-making process is also guided by heuristics. These are abbreviated cognitive operations to allow quick, workable decisions despite incomplete information (e2). Heuristics are susceptible to cognitive biases that also have an impact on decisions by physicians regarding patient treatment and care (13). Behavioral economic approaches can help shape information flows for physicians and decision-making processes by physicians to improve patient care (14, e6, e7).

bounded rational behavior

Behavioral economics combines concepts from economics, psychology, and computer sciences to unravel decision-making processes and human behavior and explain deviations from a fully rational decision (15, 16). Based on this, decision-making architectures can be created to help reach good decisions (17, 18).

For example, people tend to behave inconsistently when it comes to time, often attaching disproportionately more importance to the present (present bias) and postponing unpleasant tasks till later (procrastination) (19, 20). An experiment by Read and van Leeuwen shows that 74% of participants ordered healthy food for a meeting planned for the future. On the other hand, the majority of test subjects (70%) preferred to order chocolate for a meeting currently in progress (e8). We are always a “better” person tomorrow—but the problem is: We are not living in the future, but in the here and now.

Individuals are more motivated by avoiding losses than by acquiring equivalent gains (loss aversion), and they overestimate the probability of positive events (optimism bias), while underestimating the probability of negative events (for example, the negative impact of their behavior on their health) (21, 22). Availability heuristic describes how people misperceive the likelihood of an event depending on how easily the idea comes to their mind (23). For example, doctors who have already encountered bacteremia in one patient will also make the same diagnosis more often in the future (24). Status quo bias results in preferring to maintain the current state of affairs so that alternative options are rejected excessively often (25). This list could go on and on. The question is: Is there a remedy for this?

This article provides examples of behavioral economics interventions aimed at improving doctors’ clinical decision-making. The focus is not on treatment and prevention strategies from the patient’s perspective nor on health policy interventions. The reported interventions were compiled using a selective literature review in PubMed (English search terms: “behavioral economics+intervention+physician”, published after 2013) and from studies known to us.

The impact of behavioral economics interventions on physicians’ decision-making

Default setting and active decision control

Decisions are influenced by context, and it is often possible to actively create this context, or choice architecture (e9). Thus, default settings can remain in place until they are actively changed. They are chosen in different ways and so influence behavior. Examples include settings for computer programs, cell phones and data protection laws, regulations covering objection to organ donation, and the automatic enrolment in private retirement schemes, e.g., the US 401(k)-plan (e10– e12). Typically, a distinction is made between a default setting to regulate consent (opt-in) or objection (opt-out). Active choice directly prompts the decision-maker to make a choice without a pre-selected option (e10, e13).

Initial research in a clinical context reveals that changes to their decision-making architecture influence the behavior of physicians (e14). Changing the default setting reduced the excessive prescription of originator drugs instead of generic equivalents. In a quasi-experimental study involving internists and general practitioners, the electronic decision support system linked to electronic health records changed the default setting for physicians from “originator and generic” to “generic”, with the option to change the default setting. This increased prescriptions of equivalent generic beta-blockers, statins, and proton pump inhibitors by an aggregate of 5.4 percentage points as compared with the preintervention period (95% confidence interval: [2.2; 8.7], p <0.001) (26, 27) (Table 1a– Table 1c).

Table 1a. Behavioral economic interventions directed at physician decision-making behavior – Example: Default setting and active choice.

Approach	Intervention	Design and method	Main outcomes
Patel et al., 2014 (26); USA; Clinic of the University of Pennsylvania Health System; Community Health Data Base
default, opt-out	Default setting on the electronic health record is preset to “generics“ for medication prescriptions with the option to disagree (opt-out)	Quasi-experimental study; 9 months; internists and general practitioners (N = 255); comparison: prescription of equivalent generic beta-blockers, statins, and proton pump inhibitors before and after the intervention	Prescription of equivalent generic beta-blockers, statins, and proton pump inhibitors before the intervention: internists 77.3%, general practitioners: 82.7%; after the intervention: internists: 83.1%, general practitioners: 83.0% (p <0.001)
Patel et al., 2016 (27); USA; Outpatient departments of the University of Pennsylvania Health System; Community Health Data Base
default, opt-out	Default setting for medication prescription set to “generics”, option to change this by setting an additional checkmark	Field experiment, difference-in-differences method, 17 months; physicians; all outpatient clinics of the University of Pennsylvania Health System; comparison: prescription of generics for 10 oral drugs before and after the intervention	Prescription of generics for 10 oral drugs before the intervention: 75.3%; after intervention: 98.4% (p <0.001)
Patel et al., 2016 (28); USA; Hospital of the University of Pennsylvania Health System
active choice	Intervention group: Decision whether colonoscopy or mammography should be requested for a patient or not by accepting or rejecting such a default suggestion or not Control group: no active choice intervention decision without default setting	Field experiment, difference-in-differences method; 2 years; physicians from 1 intervention and 2 control internal medicine clinics of the University of Pennsylvania Health System; comparison: number of requests and utilization of colonoscopy or mammography in the intervention group compared with the control group	Colonoscopy, request before intervention: intervention group: 42.4%, control group: 23.4%; after intervention: intervention group: 49.0%, control group: 19.5% Colonoscopy utilization before intervention: intervention group: 16.5%, control group: 10.7%; after intervention: intervention group: 17.1%, control group: 8.0% Mammography, request before intervention: intervention group: 50.5%, control group: 37.7%; after intervention: intervention group: 64.0%, control group: 36.9% Mammography utilization before intervention: intervention group: 38.5%, control group: 26.6%; after intervention: intervention group: 41.2%, control group: 29.5% Intervention group compared with control group: request: colonoscopy: +11.8% (95% confidence interval: [8.0; 15.6], p <0.001), mammography: +12.4% [8.7; 16.2], p <0.001; utilization: colonoscopy: +3.5% ([1.1; 5.9, p = 0.004), mammography: +2.2 %, [–1.0; –5.5], p = 0.18

Table 1c. Behavioral economic interventions applying to physician decision-making behavior – Example: Self-commitment.

Approach	Intervention	Design and method	Main outcomes
Meeker et al., 2014 (35); USA; Primary care
public commitment	Intervention group: Signed commitment by physicians to reduce inappropriate antibiotic prescription which was displayed in the examination rooms in poster size for patients to see	Randomized controlled study; 9 months baseline, 12 weeks intervention; primary care practitioners (N = 14) Comparison: reduction in inappropriate antibiotic prescribing by an intervention compared with baseline and control group	Antibiotic prescriptions Baseline: intervention group: 43.5 %, control group: 42.8%; intervention period: intervention group: 33.7 %, control group: 52.7% intervention group compared with baseline: –9.8% [0.0; –19.3] control group compared with baseline: 9.9 % [0.0; 20.2] intervention group compared with control group: –19.7% ([–5.8; –33.04], p = 0.02)
Sacarny et al., 2018 (36); USA, Primary care; Choosing Wisely campaign
self-commitment, reminder	All: 2-month control period, signed commitment to adhere to recommendations for ordering services of low medical value; use of post-it reminders of their commitment during the intervention period (1–6 months); follow-up period of 3 months	Stepped-wedge cluster randomized controlled trial; 12 months; primary care physicians (N = 45) Comparison: requests for low-value services during and after the intervention compared with before the intervention	Unnecessary medical treatment requests during the intervention period compared with the control period: low-back pain: –1.2 % [–2.0; –0.5]), p = 0.001; headache: 0.7 % [–0.7; –2.1], p = 0.34; acute sinusitis: –1.4 % [–2.9; –0.1], p = 0.06 Follow-up period compared with control period: low back pain: –0.3% [–1.3; –0.8], p = 0.62; headache: 0.3% [–0.6; –1.2], p = 0.52; acute sinusitis: –2.7% [–6.6; –1.3], p = 0.19

Using mandatory active choice can help improve medical decisions, especially when it comes to the utilization of screening programs. In a randomized controlled trial, physicians were asked to sign into the patient’s electronic health record and decide whether they should be recommended for colonoscopy or mammography. The use of active choice resulted in a significant 12 percentage point increase in the number of requests for these screening tests as compared with a control group with no prompt to decide (colonoscopy: 11.8%, [8.0; 15.6], p <0.001; mammography: 12.4%, [8.7; 16.2], p <0.001). A significant increase in utilization by patients was observed only for colonoscopies (3.5%, [1.1; 5.9], p = 0.004) (28). Similar effects were seen following the introduction of active decision rules for influenza vaccinations, guideline-compliant prescriptions for statins, blood donations, HIV screening, and medication deliveries (e15– e18).

Social norms

An important finding from behavioral economics research is that people evaluate outcomes as gains or losses and express preferences relative to an existing reference point (21). One’s own performance and results are compared with those of others in a group (peer comparison) or with existing social norms (social norm comparison) (e19, e20). Social norms can be understood as standards of behavior within a group signalizing a desirable behavior and on which members of the group can orientate themselves (e21). Behavioral economic interventions incorporate this (29).

Social norms can play an important role in communication with physicians—for example, when they are informed about their own particular position with regard to their prescribing habits relative to physicians from comparable specialties and regions or to a norm that is desirable from a medical perspective (for example, a guideline). In a randomized controlled trial, a letter intervention containing a group comparison (“extremely high antipsychotic prescribing compared to other Maryland physicians”) reduced excess antipsychotic prescribing by approximately 11% of treatment days per prescriber as compared with the prescribing habits of a control group not sent peer comparison letters ([-13.1; -9.2], p <0.001) (30).

Other peer comparisons involve adjustments to prescriptions of antibiotics. In a randomized controlled trial using electronic health records, primary care practitioners were informed by email about their own inappropriate antibiotic prescriptions compared with the top performers (10% of physicians with the fewest inappropriate antibiotic prescriptions). The intervention was able to reduce inappropriate antibiotic prescriptions by 5.2% compared with prescribing habits in a control group ([-6.9; -1.6], p <0.001; and a reduction from 19.9% to 3.7% for peer comparison) (31). Similar effects were found regarding antibiotic prescribing for acute respiratory infections (odds ratio 0.73 [0.53; 0.995], p <0.05) (e21).

A randomized controlled trial in the United Kingdom applied the principle of avoiding “unnecessary antibiotic prescriptions” in an intervention in which the Chief Medical Officer, the UK government’s principal advisor on health issues, sent letters to general practitioners whose antibiotic prescription rate was more than 80% higher than those of all physicians in their local districts. This resulted in a 3.3% decrease in antibiotic prescriptions as compared with the prescribing habits of a control group that did not receive the letter (incidence rate ratio [IRR] 0.967 [0.957; 0.977], p <0.001) (32). A controlled field experiment in Germany showed that expert feedback on treatment duration (reference point) reduced the amount of antibiotics administered by pediatricians for routine cases by 10% (p <0.001) (33).

Self-commitment

Many people intend to eat more healthily, take their medication more regularly, or stop smoking. As the future draws nearer, these intentions sometimes fail to materialize. The reasons for this are time-inconsistent preferences (present bias) and lack of self-control. Binding commitments can help to achieve one’s own goals. Commitment devices employ, for example, loss aversion or safeguarding against being disappointed by others to counteract self-control problems. One of these devices is deposit contracts, whereby people voluntarily deposit money which they can only access again if they achieve a set goal (34). These self-commitments can help people eat more healthily, do more exercise, and stop smoking (e23). Commitments have two features: On the one hand, people willingly enter into commitments (that is, they are aware of possible discrepancies between goals and future behavior), while having to face consequences if goals are not achieved on the other hand. Commitments can close the intention-behavior gap.

There are also studies in the clinical setting on the effect of self-commitment by physicians (e24– e26). A randomized controlled trial examined the effect of public commitment on antibiotic prescriptions. In primary care clinics in the USA, physicians signed letters committing them to a rational prescription of antibiotics which were then displayed as posters in examination rooms for patients to see. This intervention resulted in a significant reduction of inappropriate prescriptions for antibiotics in comparison with prescription habits of the control group (-19.7% [-5.8; -33.04], p = 0.02) (35). It should be noted that when analyzing the effects of behavioral economics interventions, large-scale randomized field studies of antibiotic prescribing cannot usually determine whether individual patients who should have been prescribed an antibiotic in fact failed to receive it. Although such behavior cannot be ruled out, Eilermann et al., in whose study the decisions of the physicians participating in the experiment can be traced for each patient case, show that the reduction in antibiotic administration cannot be attributed to such behavior (33).

Another randomized trial involving primary care clinics in the USA examined precommitments to avoid services of low medical value as part of the Choosing Wisely campaign involving imaging for uncomplicated low back pain, imaging for uncomplicated headache, and unnecessary antibiotics for acute sinusitis. The precommitments were further reinforced by regular reminders. The intervention resulted in a small, but statistically significant, effect on imaging reduction for low back pain (-1.2% [-2.0; -0.5], p = 0.001). However, the effect was no longer evident three months after the intervention (-0.3% [-1.3; 0.8], p = 0.620). There was no significant change in the other low-value services (36). In 2013, the German Society of Internal Medicine also introduced a similar “smart decision-making” initiative in Germany to identify promising evidence-based diagnostic and therapeutic interventions (37). The question of whether the resulting recommendations are effectively applied in patient care cannot yet be conclusively answered (e27). Here, too, behavioral economic findings and experimental studies can utilized as additions to optimize decision-making processes.

Outlook

Behavioral economics can help close the wide gap between what ought to be medically possible and what is actually provided by offering treatment for behavioral errors and biases. People fail their own goals and process information in an inadequate and biased manner—but they do not simply behave irrationally or chaotically. Human behavior satisfies its own systematic and predictable rules. This creates opportunities for interventions to present information and shape decision architecture which can help physicians and patients make better decisions. A recent and important example involves the contributions of behavioral economics towards public health measures to increase COVID-19 vaccination rates (e28– e30).

Prevailing inconsistences of medical judgement are not only driven by biases but also by noise, the unsystematic variability within decision-making (38). Different experts may evaluate the same facts very differently. The creation of effective decision hygiene is recommended to reduce the effects of noise (38). Multidisciplinary tumor boards have been shown to be helpful in evaluating surgical interventions in cancer patients, often leading to revision of initial treatment decisions (38, e31). In addition, approaches already successful in quality circles (e32, e33) could be further enhanced by behavioral economics approaches.

Ethical standards must not be ignored in all these types of interventions. Consideration must be given to possible objections to behavioral interventions, including any constraints on core moral values—such as freedom, autonomy, and respect. Whatever the case, behavioral interventions should be made transparent, goals must be clearly defined, and the freedom of action of the target groups should not generally be restricted (e34, 39, 40). This is especially true as technological advances and greater data availability continually increase the effectiveness of behavior change interventions (e35). At the same time, however, it would be negligent in many cases to fail to review and continuously evaluate choice architectures to determine whether they create disincentives or encourage misconduct (e34). Finally, it is vital to consider long-term effects, as well as possibly different effects, of behavioral economic measures on different groups of people, something which is hardly to be found in the literature to date (e36, e37).

Behavioral economics, in combination with experimental methodology, provides both a conceptual framework and a practical toolbox. Theory-driven randomized laboratory and field experiments provide useful insights into the causal mechanisms of interventions (33, e38– e41). Insights into biases and preferences also contribute to the development of patient-centered prevention and treatment strategies.

With the increasing use of algorithms and artificial intelligence, behavioral science findings on the interdependence of information design processes and decision environments in clinical decision-making are becoming more important and digital interventions more effective (10, e35). Multidisciplinary collaboration between medical experts, behavioral economist, and data/computer scientists is already quite common practice in several corporations and organizations, with the recent collaboration in a clinical context (for example, Penn Medicine Nudge Unit) (e42). In addition, there already exists a great, and so far, largely untapped, potential to jointly develop innovative “therapeutic measures” against irrational behavior for the benefit of patients.

Table 1b. Behavioral economic interventions directed at physician decision-making behavior—Example: Social norms.

Approach	Intervention	Design and method	Main outcomes
Sacarny et al., 2018 (30); USA; Federal State of Maryland; Primary care
peer comparison, social norms	Intervention group: information letter about antipsychotics prescription rate of own practice in comparison with other practices (peer comparison). Control group: placebo letter discussing a different, unrelated medical topic	Randomized controlled study; 2 years; general practitioners and internists (N = 5055); comparison: prescribing of antipsychotics between intervention and control group	Days of antipsychotic treatment after intervention: intervention group: 2456 days, control group: 2864 days of antipsychotic treatment of the intervention group compared with the control group after intervention: –11.1% (95% confidence interval [–13.1; –9.2], p <0.001)
Meeker et al., 2016 (31); USA; Boston and Los Angeles; Primary care
peer comparison, social norms	Peer comparison: monthly feedback on practice’s antibiotic prescribing rate compared with other practices Control group: no intervention	Cluster randomized controlled study; 18 months; primary care physicians (N =248) Comparison: inappropriate antibiotic prescribing in the intervention groups compared with the control group	Inappropriate antibiotic prescriptions before versus after intervention (peer comparison): control group: 24.1% versus 13.1%: 19.9% versus 3.7 % (p <0.001); comparison with the control group: –5.2% ([–6.9; –1.6], p <0.001)
Hallsworth et al., 2016 (32); United Kingdom; Primary care
social norms, feedback	Intervention group: information letter addressed to physician or patient with feedback on high antibiotic prescribing rate in the practice Control group: no feedback	Randomized controlled study (2 x 2 factorial design); 8 months; general practitioner practices (N = 1581) Comparison: antibiotic prescribing in the intervention and control group	Antibiotic prescribing after intervention in intervention group compared with the control group: –3.3% (IRR 0.967 [0.957; 0.977], p <0.0001)
Eilermann et al., 2019 (33); Germany; Pediatric care
social norms, feedback	Intervention group: Antibiotic prescribing in hypothetical cases in three stages 1 st Stage: no feedback 2 nd Stage: expert feedback announced beforehand, then provided 3 rd Stage: see stage 2 Control group: no feedback	Controlled field experiment (framed field experiment); pediatricians (N = 73) Comparison: antibiotic prescribing (for hypothetical patients) before and after expert feedback compared with control group before and after the intervention	Average recommended length of antibiotic therapy in the intervention group: Stage 1: 7.98 days [7.42; 8.53]; Stage 2: 7.83 days [7.31; 8.35]; Stage 3: 7.23 days [6.93; 7.53]; change in recommended length of antibiotic therapy in Stage 2 compared with Stage 1: intervention group (−0.15; SD 0.63 [–0.34; –0.06]) versus control group (–0.06, SD 0.25 [–0.28; –0.16], p = 0.577). Stage 3 compared with Stage 2: intervention group (−0.60, SD 0.97 [–0.91; –0.29]) versus control group (−0.06, SD 0.25 [–0.15; – 0.03], p = 0.000)

IRR, incidence rate ratio; SD, standard deviation

Questions on the article in issue 38/2022:

Behavioral Economics Interventions to Improve Medical Decision-Making

The submission deadline is 22 September 2023. Only one answer is possible per question. Please select the answer that is most appropriate.

Question 1

What percentage of premature deaths in industrialized

countries are estimated to be due to preventable behaviors?

1. approx. 7%

2. approx. 15%

3. approx. 30%

4. approx. 40%

5. approx. 65%

Question 2

What is the term used to describe teams involved in the science and practice of behavior change in various healthcare organizations?

1. push units

2. nudge units

3. shove units

4. kick units

5. prod units

Question 3

When do heuristics significantly affect decision-making?

1. When complete information and consultation with colleagues are available

2. At multidisciplinary boards and case conferences

3. When planning elective surgery

4. When treating chronic diseases

5. When under time pressure with incomplete information

Question 4

What is meant by “present bias” when referring to time-inconsistent behavior?

1. The present matters unduly more than the future

2. The past matters more than the present

3. The past matters more than the future

4. The future matters more than the present

5. Unpleasant things are done as soon as possible

Question 5

Which term describes preference for the current state of affairs while frequently rejecting alternative options?

1. loss aversion

2. present bias

3. status quo bias

4. availability bias

5. laziness bias

Question 6

How much was the reduction in “inappropriate prescriptions for antibiotics” reported in a US study using a self-commitment intervention for “rational prescription of antibiotics”?

1. approx. 10%

2. approx. 20%

3. approx. 30%

4. approx. 40%

5. approx. 50%

Question 7

Which term describes the option of rejection in connection with behavioral economics interventions?

1. opt-in

2. active choice

3. objection

4. opt-out

5. protest

Question 8

In 2013, the German Society of Internal Medicine introduced an initiative in support of evidence-based decisions to improve patient care. What is this initiative called?

1. smart decision-making

2. clever decision-making

3. quick decision-making

4. safe decision-making

5. better decision-making

Question 9

What is the name given to the phenomenon whereby a physician, having previously made a particular diagnosis in one patient, will reach the same diagnosis more often in the future?

1. performance heuristic

2. memory heuristic

3. availability heuristic

4. recall heuristic

5. probability heuristic

Question 10

A randomized controlled trial in the USA examined the impact of mandatory decision-making (in the intervention group) with regard to offering screening examinations. Which screening examinations were recommended significantly more often as a result of the intervention and were indeed utilized significantly more often by patients?

1. mammography

2. PSA test

3. colonoscopy

4. Pap smear

5. skin cancer screening