[Ed. Note: On Friday, May 2 and Saturday, May 3, 2014, the Petrie-Flom Center hosted its 2014 annual conference: “Behavioral Economics, Law, and Health Policy.” This is an installment in our series of live blog posts from the event; video will be available later in the summer on our website.]
Our third panel, moderated by PFC Academic Fellow Matthew Lawrence, addresses the use of behavioral economics techniques to control health care costs. Speakers are Christopher T. Robertson, Brigitte Madrian, Ameet Sarpatwari, Anupam Jena, and Jim Hawkins. (Many projects are co-authored, but I’m only listing the presenters here)
The first speaker is Professor Christopher T. Robertson, coming from Arizona Law but visiting Harvard and the PFC this year, speaking on Cost-Sharing as Choice Architecture. He starts by talking about the cost side of cost sharing, which we know works in reducing consumption from empirical evidence; from the RAND study, it reduced use without reducing health. More recent studies also confirm this. But cost sharing presents four problems:
- Underinsurance relative to ability to pay
- Indiscriminate reductions in health care (more of an ax than a scalpel)
- An unfair tax on sickness (more tentative if we can solve the first two)
- The burden of deciding.
The first three are all soluble (and Robertson is working on solutions to all three of those), but the last is maybe tougher, and relevant to the topic of this conference. One function of cost-sharing is to impose a burden on decision-making.
What are alternative to cost-sharing? One is that someone else rations (but this gives fewer options to the patient); another is no rationing (which has the same number of options, and no price criterion, but you’d get much higher premiums).
Back to the burden of decision-making: it has three problems.
- Choice overload (so many choices!). It can reduce both decision quality and decision satisfaction. Regarding the first, a study by Redelmeier and Shafir found that doctors had only one drug available instead of two, they were more likely to prescribe something, regardless what it was.
- Cost criterion – adding cost adds an incommensurable factor to an already profoundly difficult decision. Mani et al published a study on this – just asking individuals to think about money when making decisions led to poorer quality decisions. Patients agree, and Sommers et all found they generally don’t want to know health cost information.
- Cost sharing can facilitate regret by making things seem like an opportunity cost (what else could you do with the cost-sharing obligation? Go to Tahiti?) Patients face the regret of bad outcomes, the option chosen, or choice process. Bad outcomes can feel even worse because patients take more responsibility.
Still, cost-sharing might be the best of bad outcomes. And at least it preserves choices for patients. So Robertson leaves open the final judgment.
Next up is Professor Brigitte Madrian from down the street at the Kennedy School of Government. Her talk is entitled Active Choice and Health Care Costs: Evidence from Prescription Home Delivery.
She’s presenting data on an active-choice-requiring program by a large U.S. retailer to increase uptake of prescription drug home delivery (which is cheaper and has other advantages). It targeted employees taking a long-term maintenance medication, not already using home delivery, who were contacted to get them to make an active choice between retail pharmacy pick-up vs. home delivery.
They didn’t really force a choice (not really possible) but nagged them, and denied payment after three retail fills without an active choice. Getting insurance coverage after that required making a choice–and the financial incentives then were the same as before the program.
They found that the fraction of all prescriptions for the company (not focused on the targeted employees or targeted prescriptions) found that home delivery went way up to about 17%, after being flat before the program at around 5-6%. For the targeted employees, on the targeted medications, 6% did home delivery before (like the general population); once forced (ish) to make a choice, 40% chose home delivery, 38% chose to stick with the pharmacy, and 22% chose not to choose (and suffered the penalty!). Madrian doesn’t think it was just a coin flip (38% v 40%), because demographic factors predicted choice (older, higher paid workers picked home delivery more).
Unfortunately, medication adherence was lower. Once people were faced with the choice or the penalty, adherence dropped. People deciding not to take the choice? Hard to say – the more individuals would save, the more people picked home delivery (though not dramatically), and more people made more choice. So Madrian thinks the adherence drop was a problem with measure, because people just dropped out of the system and bought cheaper drugs retail. Having a financial incentive appears to really matter.
They calculated about $800,000 in savings per year, and this program has been since rolled out more broadly.
Our third speaker is Ameet Sarpatwari, a research fellow at the Brigham and Harvard Medical School, presenting Behavioral Economics and Physician Prescribing Practices: Legal and Ethical Considerations in the Use of “Nudges” to Promote Generic Drug Use.
First up: how generics aren’t used enough. Health care spending is extraordinarily high, and the ACA doesn’t seem to be making much of a difference. Prescription drug is rising, and a big chunk of spending. Brand-name drugs are far more expensive than generic drugs. It’s important to note two types of generic substitution:
- A-rated interchangeable drugs – bioequivalent; note that no randomized controlled trials have identified clinically significant variations in outcomes between those drugs. Optimizing substitution of these drugs would yield $12 billion annual savings, according to IMS Health. (Maybe we should be more careful for narrow therapeutic index (NTI) drugs, with big dose variation; there’s concern, though it’s not supported by evidence).
- Therapeutic substitution – substituting a brand-name drug with a non-bioequivalent generic drug (similar mechanism of action or clinical effect). These are the large number of “me-too” drugs. But it’s not a bioequivalent drug. This would probably save much more. One study from VA about 4 diabetes drugs suggested $1.4 billion in Medicare spending annual savings.
Sarpatwari then turns to prior interventions. One approach has been academic detailing (where academics tell doctors about the benefits of the approach using a neutral approach). The evidence base is strong – but the effects are fairly small. Another is formulary support (at the point of prescribing, docs are provided with information about national guidelines, contradindications, and co-payment amounts) – a retrospective study found $845,000 in savings per 100,000 patients.
Next is nudges as interventions. At the point of subscribing, states differ on whether they have pro- vs. anti-substitution prescription pads (i.e., where the signature is located, because people default to signing on the right). There’s an 18% difference in the pads. Electronic prescribing (listing generic drugs listed higher in searchers with bold font) resulted in a 20% difference.
Financial incentives are another possibility mentioned [ed – though these don’t seem like nudges to me; they’re just incentives, even if small ones]; they make a difference, but it’s not rigorously evaluated. Sarpatwari asks about the legality of these programs – are these legal under the Anti-kickback statute, which prohibits kickbacks, bribes, and rebates? After the ACA, knowledge isn’t required, and qui tam actions are permitted. But under the regs, this just covers Medicare Part D and Medicate MCOs. [ed 2 – author response via email: “To clarify, we don’t consider financial incentives nudges. We only classify the two point of prescribing interventions (prescription pad alterations and e-prescribing search result modifications) as nudges.”]
Even if doctors can take incentives, should they? Brand-name drugs have problematic incentives (off-label prescribing, rare adverse events). And increased out-of-pocket costs lower adherence – so switching to generics may be ethically preferable for its own reasons.
He concludes that such nudging interventions should be allowed for bioequivalent drugs, maybe less so for NTI drugs and therapeutic substitution. Financial interventions should only be allowed for bioequivalent drugs.
Fourth is Professor Anupam Jena from Harvard Medical and Mass General, speaking on Screening Mammography for Free: Asymmetric Responses to Increases and Decreases in Cost-Sharing for Breast Cancer Screening.
The ACA mandated full coverage of evidence-based preventive services for Medicare beneficiaries beginning in 2011 (we’d like older people to have more preventive care, and cost-sharing decreases that). This piece focuses on breast cancer; it assumes that mammography screening has population benefits or at least remaining agnostic, and that the recommendation rates are correct–or at least, takes that as given.
Jena looked at a Medicare Advantage HMO that voluntarily implemented free screening in 2010, and studied the three years before and after the change, focusing on women over 65 in two groups: a control group with employer-supplemented insurance (full coverage for the whole period) and an individual market group (coverage became free halfway through). He used a difference-in-difference design to observe changes in the difference between the two groups (to control for external changes; lots of other things were going on, including a change in recommendations from semiannual to annual, so overall rates declined dramatically).
Jena found a statistically significant but very small increase in screening (about 1%) when cost-sharing went to 0. The effect was larger (though not statistically significant), and still quite small, among poorer women.
Jena concludes 1) that overall mammography rates are trending down, a bit attenuated by becoming free; 2) incentives along are unlikely to increase screening substantially; and 3) there may be asymmetric price responses, where increases vs. decreases have different effects. (As evidence, co-payments went up between 2006 and 2007 for this same set of individuals, which resulted in a larger decline in screening rates – about 3% measured by difference-in-difference, proportional to the size of the co-payment change).
Our final speaker for the panel is Professor Jim Hawkins from the University of Houston Law Center, presenting a talk entitled Towards Behaviorally Informed Policies for Consumer Credit Decisions in Self-Pay Medical Markets.
Hawkins is talking about self-pay markets – things like fertility treatments and other situations where patients pay for their own treatments. Credit providers often team up with doctors to offer financing. He observes (consistent with the conference in general) that patients deviate from the rational actor model in predictable ways.
In a 2009 study, he found that about 50% of fertility clinics (he looked at basically all of their websites) mentioned credit, almost always through 3rd party lenders. Doctors prefer third parties because they get more business without the risk of default, and don’t have to be debt collectors; they also have limited legal liability (under typical contracts with third parties).
Unfortunately, Hawkins argues, patients probably make mistakes when they consider these arrangements. Product design causes patients to focus on the fixed, short-term costs instead of contingent, long-term costs; 85% agree to a deferred-financing plan, and 25% fail to repay during the promotional period (with major penalties). There may also be a halo effect; patients like their doctor to make treatment decisions, and may not critically evaluate the lender (like you might a car-dealership-associated lender). Finally, the salience of the transaction is decreased; patients tend to sign up in the doctor’s office, rather than shopping around.
But this is bad! Medical credit is expensive, and doctors have a financial interest. Regulatory response has been mixed, focused on mild disclosure.
Hawkins suggests product-use disclosure–i.e., say how frequently people use loans in a good way, and how many people do bad things (renew the loan a bunch of times or default).
Also, doctors could be in a better position to implement protections than lenders. Lenders have an incentive to hide true costs, but doctors still get the same benefits even if they fully disclose.
Finally, we turn to questions!
Nina Kohn from Syracuse University asks about how much we should be concerned with equating cost savings from a nudge to cost-savings to the system overall. (e.g., home-delivery may save plans money, but having patients stay home means they don’t see the pharmacist who often functions as a care coordinator). Brigitte Madrian answers that this is a legitimate concern, and heterogeneity of intervention targets should inform the appropriateness of an intervention. Ameet Sarpatwari adds that looking at specific patient populations for generic drug interventions is key, and Chris Robertson wraps up by agreeing on targeting.
David Korn from HMS says that doctors taking kickbacks is disgraceful, and wants to know why it’s not anti-kickback. Jim Hawkins says that actually, the doctors pay the lender–their financial incentive is to pay less to the lender–but they don’t get money from the lender.
Matt Lawrence invokes moderator’s privilege, asking Jim Hawkins whether credit patients could get on their own is better or worse than what they get from doctors. Relatedly, how do these studies relate to difficulty in determining patients’ interests? Hawkins replies that he assumed that if there are many lenders outside the doctor’s office, some will have better rates than those in the doctor’s offices (even credit cards have better rates); seems basically anecdotally true. Brigitte Madrian and Hawkins have a discussion about whether and where people might get better credit and higher limits.
Leo Beletsky from Northeastern asks Ameet Sarpatwari about the role of patients in generic substitution – do they nudge providers toward brand names, and do brand-name companies nudge patients to nudge providers? Sarpatwari responds that prescribers are certainly influenced by drug company actions. For patient desires, even though the evidence may not be there, we should probably still respect patient wishes. [ed – Why is that different from nudges in general?] [ed 2 – author response via email: “Patient desires matter. We should be careful in overriding them in situations where our evidence base is limited, as in bioequivalent substitution of NTI drugs.”]
PFC’s own Glenn Cohen asks about the role that insurers might play in these spaces – fertility, drugs, and insurance design. Sarpatwari adds that insurers may counteract brand-name companies, but have their own profit motive (though they are likely to include patient health more than drug companies). Hawkins thinks the absence of insurers in fertility makes it a great place for CFPB to be involved – different from so many types of health (price transparency, measurable success rate, &c.). Robertson chimes in – it’s actually a market! Unlike the rest of health care, which is “pharket.”
Rebecca Gauthier, a student at HLS, is unsurprised that mammography rates didn’t change, because the relevant cost barrier is getting to see the doctor in the first place. Anupam Jena agrees – costs of seeing a doctor and waiting in the office are big costs. He raises the flip side; given all those costs, it’s shocking that we ever see such big differences based on costs. In a study of post-heart-attack patients with free or regular-cost drugs, making drugs free only made a 5% difference – and about 50% of people still didn’t take the medication. So many things other than price affect the outcome (which hurts him to say as a Chicago economist). Madrian chimes in – if you want to change behavior, you’ve got to understand why people aren’t doing what you want them to do. Our single-minded focus on cost is tragic.
David Hyman says that costs make a shockingly high difference when it’s salient. And asks Jim Hawkins – isn’t the fact that 25% of people don’t pay within the promotion period nonproblematic? 75% got free credit, for a blended rate of 6-7%, which seems not bad for unsecured credit… Hawkins replies that it’s a good point, which is why his intervention is so minor; it just provides some extra information.
Neil Kalani, a joint HMS/Kennedy School student, asks whether with the rise of HMOs/ACOs, you’ll have patients picking between central rationers, which might be a good middle ground between extremes. Chris Robertson says it’s an open question as to whether patients would prefer for doctors to fight for them, or inform them about what they’re not getting. It’s hard for people to make informed choices about their advisors’ roles.
And that’s a wrap to a fascinating panel (and an exceedingly long post)!
