Unbiased Analysis of Today's Healthcare Issues

Patient and Physician Risk Preferences

Written By: Jason Shafrin - Feb• 21•17

Physicians are supposed to act as agents for patients, acting in their best interest. But a question is, how well do physician know patient preferences? How well do physician preferences reflect those of patients.

Previous research indicates that for terminal diseases, patients are risk loving. My own recently published research (with co-authors Schwartz, Okoro and Romley) confirms this finding, but adds to the literature by showing that physicians are largely risk neutral or mildly risk averse.

A recent paper by Galizzi et al. (2016), however, reach a different conclusion. The do confirm that physicians are mildly risk averse but they also conclude that patients are also mildly risk averse as well. They find the following:

First, there is a significant difference in time preferences between patients and their matched doctors, with doctors discounting future health gains and financial outcomes less heavily than patients. Second, we find no systematic difference in risk preferences in the healthcare domain between patients and doctors: in our sample both patients and their matched doctors are mildly, but significantly, risk averse. Third, doctors and patients have significantly different risk preferences in the finance domain: whilst doctors are risk averse, patients are risk neutral.

There are a number of potential reasons for the difference in findings. First, the patient populations of interest differed dramatically. Whereas Shafrin et al. examined patients with advanced state melanoma and non-small cell lung cancer, Galizi et al. examined patients and physician preference at outpatient clinics for the following specialties: pathology, cardiology, gynaecology, haematology, surgery, endocrinology, orthopaedics, urology, gastroenterology, nephrology, rheumatology, ophthalmology, and otolaryngology.   Notably, oncology patients were excluded from this study. Second, Shafrin et al. (2017) was conducted in the U.S. and whereas Galizzi et al. (2016) was conducted in Greece. Differences in patient preferences could explain this difference. Third, the methodology for elicitly preferences differed. While both methods compared two treatment options, Galizzi compared two risky treatment options whereas Shafrin et al. compared a risky option against a certainty equivalent option. Further, Galizzi used the Holt and Laury (2002) multiple price list whereas Shafrin et al. used parameter estimation by sequential testing to vary the certainty equivalent value depending on the respondents previous answer.

A useful innovation of the Galizzi paper is that the patients were matched to actual physicians who treated them, whereas the patient and physician selection process was done independently in Shafrin et al.

On the other hand, Shafrin et al. calibrated the “risky” treatment option to real life therapies that were relevant to the specific patient population of interest.

Overall, both studies have useful findings. Galizzi et al. aim to measure risk preferences more generally, across diseases and not calibrated to any one specific treatment. Although Galizzi’s study is broader, it is less tied to real world choices facing patients. The results of Shafrin et al. are more robust for measuring patient and provider preferences across risky oncology treatments, but these should not be extrapolated beyond this specific clinical area.


“You can’t fix by analysis what you bungled by design”

Written By: Jason Shafrin - Feb• 20•17

This is a quote from Light, Singer, and Willet (1990) and is mentioned on a recent commentary by Stephen B. Soumerai  and Ross Koppel in Health Services Research.

The commentary focuses on the use of instrumental variables when conducting health care effectiveness reserach.  They define instrumental variables concisely as:

An instrumental variable (IV) is a variable, generally found in administrative data, that is assumed to randomize a treatment to estimate cause and effect relationships, thus controlling for known and unknown patient characteristics affecting health outcomes. An important assumption is that the IV randomizes treatment but does not directly affect the patient outcome.

There are a number of commonly used instrumental variables that are used to predict use of health care services.  These include a patient’s distance from specific types of providers (e.g., hospitals, specialty providers) and a provider’s average treatment patterns (which generally are correlated with the patient’s likelihood of getting a specific treatment but may not be correlated with the patient’s outcomes except through the treatment received).

The commentary takes aim at the first commonly used IV as applied to access to cardiac catheterization laboratories.

…the central and (we argue, dubious) assumption is that isolated people having a heart attack who live hours away from a hospital with a cath laboratory and are therefore less likely to receive invasive proceduresare identical in their likelihood to survive as those lucky patients living very close to a cath hospital.

The authors claim that patients who live far from a cath hospital may differ from those who live closer.  For instance, maybe the cath hospital is located in an urban core; or in a more affluent part of town.  If this is the case, unobserved differences in patient population correlated may be correlated with both outcomes, the probability of the receive of treatment, and distance from a cath hospital. Clearly one can control for some, but not all of these factors using administrative data.

Another issue is that while IV may help identify a causal effect of treatment, it may not give policymakers a clear pathway to improving outcomes.  Is the answer creating more cath hospitals?  Giving incentives for people to move closer to a cath hospital?

At the same time, while Soumerai and Koppel’s critique has many valid points, their critique is a bit too broad.  IV does have its limitations.  However, identifying a real-world causal effect of a given treatment is a key research question of interest.  How patient outcomes could be improved further is also an important question, but one should not expect a single study to answer all research questions on a single topic and other types of analyses may be preferable to get at the “how” questions rather than at the “what happened” research questions.

Soumerai and Koppel do cite some higher quality IV such as cases where public policies result in lotteries (e.g., the Oregon Medicaid expansion, birth date lotteries), and call for more use of longitudinal (e.g., interrupted time series) rather than a cross-sectional designs.

Although these methods are good to recommend, policy lotteries are not available for all research questions. More broadly, Soumerai and Koppel make a clear point that research design is key and that all analytic approaches should be tested rigorously with sensitivity analyses and falsification checks of assumed causal pathways.


Innovations in Cancer Care: Capturing What Patients Value in the Calculus of Drug Costs

Written By: Jason Shafrin - Feb• 19•17

My current employer, Precision Health Economics, has posted an interesting research brief describing how traditional notions of value may not be capturing the full value patients receive from oncology treatments.

A brief description is below but do check out the full report.

As health care spending continues to rise, payers and providers struggle to accurately measure value to patients, and with health systems all over the world shifting towards paying for value, not just volume, these efforts have never been more important.

Learn about PHE’s groundbreaking studies that have helped define the value of new oncology drugs, and why a variety of global stakeholders have turned to PHE to assist them in defining value in cancer care.

Special Issue on Value Assessment Frameworks

Written By: Jason Shafrin - Feb• 17•17

Value in Health has a Special Issue on Value Assessment Frameworks.   The issue includes one of my papers (co-authored with Taylor T. Schwartz, Tony Okoro, John A. Romley) titled Patient Versus Physician Valuation of Durable Survival Gains: Implications for Value Framework Assessments.  The abstract is below.


Background: Previous research indicates that patients value therapies that provide durable or tail-of-the-curve survival gains, but it is unclear whether physicians share these preferences.

Objective: To compare patient and physician preferences for treatments with a positive probability of durable survival gains relative to those with fixed survival gains.

Methods: Patients with advanced stage melanoma or lung cancer and the oncologists who treated these patients were surveyed. The primary end point was the share of respondents who selected a therapy with a variable survival profile, with some patients experiencing long-term durable survival and others experiencing much shorter survival, compared to a therapy with a fixed survival duration. Parameter estimation by sequential testing was applied to calculate the length of nonvarying survival that would make respondents indifferent between that survival and therapy with durable survival.

Results: The sample comprised 165 patients (lung 1⁄4 84, melanoma 1⁄4 81) and 98 physicians. For lung cancer, 65.5% of patients preferred the therapy with a variable survival profile, compared with 40.8% of physicians (Δ = 24.7%; P < 0.001). For melanoma, these figures were 63.0% for patients and 29.7% for physicians (Δ = 33.3%; P < 0.001). Patients’ indifference point implied that therapies with a variable survival profile are preferred unless the treatment with fixed survival had 13.6 months (melanoma) or 11.6 months (lung) longer mean survival; physicians would prescribe treatments with a fixed survival if the treatment had 7.5 months (melanoma) or 1.0 month (lung) shorter survival than the variable survival profile.

Conclusions: Patients place a high value on therapies that provide a chance of durable or “tail-of-the-curve” survival, whereas physicians do not. Value frameworks should incorporate measures of tail-of-the-curve survival gains into their methodologies.

Friday Links

Written By: Jason Shafrin - Feb• 16•17

The end of “the pill”?

Written By: Jason Shafrin - Feb• 15•17

For years, sexually active women and men that have wanted to avoid pregnancy have used a variety of contraceptives.  From pills to condoms, there are a lot of options out there.  Now add one more to the list, and its not what you might think.

Natural Cycles is an app that helps people identify the time period in which they are likely to be fertile. More about the technology is here.

The mobile application requires the input of basal body temperature recordings and the date of menstruation. Luteinising hormone (LH) test results are optional entry points. The required basal thermometer and the optional LH tests are acquired separately from the application. The users enter their fertility-related data into a device such as a smartphone, tablet or laptop computer. The underlying technology is a statistical algorithm that returns a red (unsafe) or a green (safe) day to the user depending on whether she is considered to be at risk of getting pregnant.

The question is, does it work?  It seems like the answer is yes.  Pharmafile reports:

Natural Cycles has been approved for use in the UK by Tüv Süd, a European regulatory body employed by the Department of Health and acting on behalf of the Medicines and Healthcare products Regulatory Agency (MHRA) following extensive clinical trial testing. Research gathered from the trials placed the app’s effectiveness at 99.5%; by contrast, the NHS places the effectiveness of the contraceptive pill at 99%.

I am, however, a bit skeptical of these results.  Although they appear impressive, one of the key studies is a non-randomized study.  Patients who are more motivated are more likely to sign up for Natural Cycles; but they are also the same people who would be more likely to regularly take their pill.  Also, discontinuation rates of health care apps are often high and this study was no exception; about one-third of individuals discontinued use of the app within 6 months.

While Natural Cycles may be a good fit for motivated individuals or those sensitive to contraception side effects, it is unlikely to be a cure-all.

Did the ACA cause industry consolidation?

Written By: Jason Shafrin - Feb• 14•17

The Affordable Care Act (ACA)–among other things–mandated a number of reforms to the Medicare reimbursement system.  For instance, the ACA created Accountable Care Organizations (ACOs) and bundled payment initiatives were initiatives.  If Medicare started paying providers more based on quality and total cost of care across all provider settings, one would hypothesize that industry consolidation would accelerate.

An article by Neprash, Chernew and McWilliams, (2017) however, argues that the data does not support this conclusion.  Although industry consolidation has increased, it likely is not due to the ACA.  They write:

Drawing on data from a number of sources from 2008 onward, we examined the relationship between Medicares accountable care organization (ACO) programs and provider consolidation. We found that consolidation was under way in the period 200810, before the Affordable Care Act (ACA) established the ACO programs. While the number of hospital mergers and the size of specialty-oriented physician groups increased after the ACA was passed, we found minimal evidence that consolidation was associated with ACO penetration at the market level or with physiciansparticipation in ACOs within markets. We conclude that payment reform has been associated with little acceleration in consolidation in addition to trends already under way, but there is evidence of potential defensive consolidation in response to new payment models.

Despite the observation that the ACA may not have caused provider consolidation, a trend towards more provider market power and less competition is worrying for consumers and should not be ignored.


Need health care data?

Written By: Jason Shafrin - Feb• 13•17

The International Society For Pharmacoeconomics and Outcomes Research (ISPOR) has put together a database of databases from around the world. The ISPOR International Digest of Databases has goal is to organize global health care databases into searchable models.  Their working group on this topic is chaired by Carl Asche and Elisabeth Oehrlein.  If you have any databases you suggest that should be added to ISPOR’s International Digest of Databases, you can email the working group with suggestions using this form.

While I have not reviewed the quality of this database in detail, the opportunity to have a single searchable database to identify high value data around the world would be extremely valuable for researchers. The University of Michigan’s ICPSR attempts to the same but focuses on social science more broadly.

Regardless, providing researchers with more information and access on health care tools and data will hopefully greatly improve the quality of social science and health care research in the future.

Smarter deductibles?

Written By: Jason Shafrin - Feb• 12•17

Are high deductible health plans a good thing?  Republicans typically argue yes as they say that increased cost sharing reduces moral hazard.  That is, when people have to pay for medical care out of pocket, they don’t ask for unnecessary care or use care more frugally.  Democrats typically argue that increased cost sharing reduces demand for effective health care, decreases medication adherence, and may even increase long-term cost if foregoing effective care results in higher downstream medical cost.

There is evidence for both arguments.  The RAND Health Insurance Experiment showed that people consume less health care when they have higher out-of-pocket cost.  A recent study by Wharam et al. (2017) found that low-income patients with high-deductible health plans did not use necessary health care and this resulted in a higher number of emergency department visits.

A commentary by Fendrick and Chernew (2017) argues for a “Precision Benefit Design”.  They propose a value-based insurance design (V-BID) which  would lower cost-sharing for high-value services and increase cost-sharing for low-value services.  They state that “In a nuanced design, cost sharing for eye examinations would be substantially lower for those with diabetes than for those without.”  The authors argue for pre-deductible coverage of high value services even for patients who have HDHP.   In fact, law has already been passed to facilitate this type of plan.

To better enable the synergies between value-based payment models and benefit designs, in July 2016, the bipartisan H.R. 5652 “Access to Better Care Act of 2016 ”was introduced. This bill givesHSA-qualified high-deductible health plans the ability to provide coverage for services that manage chronic disease prior to meeting the plan deductible.

While V-BID are good in theory, there are some issues in practice.  First, it is more complicated and expensive for plans to administer as deductibles will vary within a plan type depending on a person’s condition.  Second, it will be more difficult for patients to price compare across plans if deductibles and copayments routinely change each year depending on their health condition.  Third, V-BID will likely result in more paperwork and appeals regarding what constitutes high value care and who (and when) people are eligible for a cost sharing waiver.

Despite these challenges, reducing cost sharing for high-value treatments is an appealing approach.  The question to be answered will be whether these logistical challenges can be overcome.


Friday Links

Written By: Jason Shafrin - Feb• 10•17