Active Choice

Active Choice Instead of a Default in Health Policy

Every health programme that asks people to sign up has to decide what happens to those who do nothing. Under opt-in they are left out, under opt-out they are put in, and under active choice each person is asked to say yes or no instead of being assigned an answer. Behavioural economics explains why that rule changes uptake, and evidence from organ donor registration, workplace pensions, influenza vaccination and repeat prescriptions shows the effect can be large, small or even negative depending on how the requirement is built. Economic theory favours it over a default in some conditions, and an illustrative cost-effectiveness comparison of three regimes for a vaccination programme shows how it can be evaluated.

Three ways to structure the same decision

A choice architecture intervention changes how options are presented without removing any of them. For a binary health decision there are three basic structures, which differ in what happens to a person who gives no answer.

In practice the labels blur. Kessler and Roth separate active choice, where people perceive a requirement to choose, from mandated or forced choice, where a person who refuses to answer cannot complete the purpose of the visit, such as receiving a driving licence. Beshears and colleagues add that in most active choice schemes not choosing is still possible, and it triggers what they call an implicit default. Keller and colleagues propose a further variant, enhanced active choice, in which the options are worded to highlight what a person loses by picking the option the communicator does not favour.

Why requiring a decision changes behaviour

Active choice works by removing the routes through which a default operates. Johnson and Goldstein list three. People may read a default as a recommendation from the policymaker, accepting it takes no effort while deciding does, and moving away from the status quo involves a trade-off in which losses loom larger than equivalent gains. Keller and colleagues add present bias and procrastination. Leaving a default usually costs a little effort now for a benefit later, so people who discount the future heavily keep putting the action off.

An active choice regime blocks indefinite delay, because the decision has to be made at a fixed point such as a hire date, a clinic visit or a pharmacy fill. Carroll and colleagues attribute the effect they found to employees having to state their 401(k) decision in their first month instead of being able to delay indefinitely. Keller and colleagues report a second mechanism in a hypothetical flu shot reminder study, where respondents asked to choose actively anticipated more regret from not getting vaccinated than those given opt-in or opt-out formats, and regret aversion partly mediated the effect on compliance.

Evidence from organ donation, pensions, vaccination and prescriptions

The empirical record is mixed, and results depend on the setting, the wording and what happens to people who do not answer. Hypothetical studies report larger effects than the field studies of real health decisions.

Organ donation. In an online experiment with 161 respondents, Johnson and Goldstein found effective consent rates of 42% under opt-in, 82% under opt-out and 79% in a neutral condition that required a choice with no default; only opt-in was significantly lower. They also note that when Virginia used mandated choice, about 24% of the first million Virginians asked said they were undecided. When California switched its licence question from opt-in to active choice in 2011, registration rates were, by Kessler and Roth's estimate, 2.2 to 2.7 percentage points lower than they would otherwise have been, and in a laboratory study using live registry decisions the active choice frame made non-donors 8 to 9 percentage points less likely to register. In a separate hypothetical experiment, subjects were more likely to say next of kin should donate the organs of a person who had simply not opted in than of one who had said no under active choice, which suggests active choice can reduce consent from next of kin.

Pensions. Carroll and colleagues studied a firm where new employees had to return a form within 30 days stating whether they wanted to join the 401(k) plan. Requiring this active decision raised initial enrolment by 28 percentage points compared with standard opt-in, and produced within three months a savings distribution that would have taken three years under opt-in.

Vaccination. In a hypothetical study with 55 employees, Keller and colleagues found that 69% in the two active choice conditions said they would get a flu shot, against 42% under opt-in. Patel and colleagues studied an electronic health record prompt asking physicians and medical assistants to accept or cancel an influenza vaccine order. In an observational comparison of 45,926 patients at three practices, the intervention practice's vaccination rate rose by 6.6 percentage points more than the control practices (a 37.3% relative increase), and more than 99.9% of orders placed resulted in vaccination.

Prescription refills. In two field studies with members of a pharmacy benefit manager, Keller and colleagues found that enhanced active choice raised enrolment in an automatic refill programme from 15.7% to 32% by telephone message and from 12.3% to 21.9% on a website. Disenrolment under enhanced active choice was 22.1% against 21.4% under opt-in by telephone, and 15.2% against 10.2% on the website. Beshears and colleagues studied an employer that made an active choice between home delivery and pharmacy pick-up a condition of the prescription subsidy. Under opt-in only 6% of eligible prescriptions used home delivery; under active choice 42% of eligible employees chose home delivery, 39% chose pharmacy pick-up and 19% made no choice and were assigned the implicit default of pharmacy pick-up without the subsidy.

When active choice beats a default in welfare terms

The economic case for active choice rests on heterogeneity. Carroll and colleagues show in a formal model that active decisions are socially optimal when preferences are highly heterogeneous and people have a strong tendency to procrastinate. If a planner knew each person's optimum, it could set that as the default and spare people the effort of choosing; because the optimum is private, a single default suits some people and harms others. Financial illiteracy, in their model, favours a default. In health, a default suits a decision where almost everyone should take the same action, and active choice suits decisions where the right answer differs between people, such as home delivery of medicines.

Beshears and colleagues add two points that matter for welfare economics. Passive acceptance of a default may not reveal what people actually prefer, so active choice may bring outcomes closer to people's normative preferences, although such schemes rarely work in practice exactly as theory suggests. The implicit default is also a policy lever, because making non-choice less attractive pushes more people to choose, but at the cost of the welfare of those who still do not choose. In their data, people who gained least from the subsidy were more likely to make no choice and so lose it.

Evaluating an active choice intervention

An active choice scheme can be appraised in a cost-effectiveness analysis like any other health intervention. The comparators should normally include opt-in, active choice and, where it is lawful and acceptable, opt-out, because these are mutually exclusive ways of running the same programme. Costs include the prompt or form, staff time and the service taken up. People's time spent deciding is a real cost, and Carroll and colleagues note that an active decision compels people to struggle with a potentially time-consuming decision they may not feel qualified to make.

Registrations, intentions and orders are intermediate outcomes whose link to health depends on follow-through. In the Patel study almost every order led to vaccination, but in the refill and organ donation studies enrolment, disenrolment and later family decisions moved in different directions. Where the target behaviour carries an externality, as with vaccination externalities, the health gain to others belongs in the analysis. Many studies are observational or hypothetical, so a randomised controlled trial gives a firmer effect estimate for a model.

Worked example of an active choice prompt for influenza vaccination

The setting and every number in this example are illustrative. A practice sees 10,000 eligible adults in a season, and the analyst compares three ways of running vaccine offers.

1. Uptake under each regime. Under opt-in, 35% are vaccinated (3,500). An active choice prompt in the record raises this to 41% (4,100), 600 more. Under opt-out, every patient is booked an appointment; 3,000 cancel and 60% of the 7,000 remaining attend, so 4,200 are vaccinated, 700 more than opt-in, and 2,800 booked slots go unused.

2. Costs and effects per additional vaccination. Setting up either regime costs GBP 6,000 a season. Each extra vaccination costs GBP 25 and saves GBP 8 of later treatment, and gains 0.002 QALYs. An unused slot costs GBP 4.

3. Active choice against opt-in. Vaccinations cost 600 × 25 = 15,000 and save 600 × 8 = 4,800, so the net cost is 6,000 + 15,000 − 4,800 = GBP 16,200, and the gain is 600 × 0.002 = 1.2 QALYs, which gives the ratio below.

$$\text{ICER}_{AC} = \frac{\Delta C}{\Delta E} = \frac{16200}{1.2} = 13500$$

where $\Delta C$ is the net cost in GBP and $\Delta E$ the QALYs gained compared with opt-in, so the ICER is GBP 13,500 per QALY.

4. Opt-out against active choice. Vaccinations cost 700 × 25 = 17,500, unused slots 2,800 × 4 = 11,200 and savings are 700 × 8 = 5,600, so opt-out costs 6,000 + 17,500 + 11,200 − 5,600 = GBP 29,100 and gains 700 × 0.002 = 1.4 QALYs. Compared with active choice, it buys 100 more vaccinations for GBP 29,100 − 16,200 = 12,900 and 0.2 QALYs, an incremental ratio of GBP 64,500 per QALY.

5. Sensitivity to the effect size. If active choice adds only 300 vaccinations, the net cost is 6,000 + 7,500 − 2,400 = GBP 11,100 for 0.6 QALYs, or GBP 18,500 per QALY. At an illustrative threshold of GBP 20,000 per QALY, the prompt needs at least 261 extra vaccinations, as the inequality shows.

$$6000 + 17N \le 20000 \times 0.002N = 40N \quad \Rightarrow \quad N \ge \frac{6000}{23} = 260.9$$

where $N$ is the number of additional vaccinations and 17 is the net cost in GBP of each one (25 − 8).

6. Interpretation. At these illustrative values, active choice is the efficient regime. Opt-out vaccinates slightly more people, but wasted appointments make its extra gain expensive. Because the fixed cost is spread over the extra vaccinations, the ICER equals 3,000,000/N + 8,500 GBP per QALY, so it can never fall below GBP 8,500 per QALY, however large the effect.

Boundaries between active choice, defaults, nudges and incentives

Active choice sits between a default and a mandate that requires one particular action. It keeps every option open, like a default, but removes the passive route, so the opportunity cost of a person's time and attention moves from the few who opt out to everyone who must answer. Enhanced active choice is active choice plus framing, which favours one answer through wording, so it is closer to persuasion than a neutral question.

Once refusing to choose carries a real penalty, as with the lost prescription subsidy in the Beshears study, the scheme also works as a financial incentive. For vaccination, active choice is one option in wider vaccination policy alongside reminders, incentives and mandates; for repeat prescriptions Keller and colleagues present it as a route to better medication adherence.

Where active choice disappoints, and common misreadings

Active choice is often presented as a neutral middle path, but its design determines its effect. Four errors are common.

• Assuming active choice always beats opt-in. California's organ donor switch reduced registrations, and Kessler and Roth found no gain in a study of real registrations.
• Reading intentions as behaviour. Several high figures come from hypothetical choices; the field effects in the vaccination and refill studies are smaller.
• Ignoring non-choosers. Some people do not answer even when required; how they are treated is an implicit default with welfare effects, in the Beshears study mostly on people who gained least from the subsidy.
• Treating the prompt as free. Time spent deciding and staff time are costs, and where most people should make the same choice a well-chosen default can do better.

Sources

• Beshears J, Choi JJ, Laibson D, Madrian BC. Active choice, implicit defaults, and the incentive to choose. Organizational Behavior and Human Decision Processes. 2021;163:6-16. doi:10.1016/j.obhdp.2019.02.001. Author manuscript: https://pmc.ncbi.nlm.nih.gov/articles/PMC8112790/
• Carroll GD, Choi JJ, Laibson D, Madrian BC, Metrick A. Optimal defaults and active decisions. Quarterly Journal of Economics. 2009;124(4):1639-1674. doi:10.1162/qjec.2009.124.4.1639. Working paper version: NBER Working Paper 11074, https://www.nber.org/papers/w11074
• Johnson EJ, Goldstein D. Do defaults save lives? Science. 2003;302(5649):1338-1339. doi:10.1126/science.1091721.
• Keller PA, Harlam B, Loewenstein G, Volpp KG. Enhanced active choice: a new method to motivate behavior change. Journal of Consumer Psychology. 2011;21(4):376-383. doi:10.1016/j.jcps.2011.06.003.
• Kessler JB, Roth AE. Don't take 'no' for an answer: an experiment with actual organ donor registrations. NBER Working Paper 20378. 2014. https://www.nber.org/papers/w20378
• Patel MS, Volpp KG, Small DS, Wynne C, Zhu J, Yang L, Honeywell S Jr, Day SC. Using active choice within the electronic health record to increase influenza vaccination rates. Journal of General Internal Medicine. 2017;32(7):790-795. doi:10.1007/s11606-017-4046-6.
• The Organ Donation (Deemed Consent) Act 2019 (Commencement No. 3) Regulations 2020, SI 2020/520. https://www.legislation.gov.uk/uksi/2020/520/made