Unbiased Analysis of Today's Healthcare Issues

Should pregnant women be included in clinical trials

Written By: Jason Shafrin - Nov• 25•15

Your knee-jerk reaction (and mine) is No!  However, an interesting article from Mosiac makes the case that the argument is not clean cut.

Because it has long been considered unethical to include expectant mothers in clinical trials, scientists simply don’t know whether many common medicines are safe for pregnant women. Of the more than 600 prescription drugs that the US Food and Drug Administration approved between 1980 and 2010, 91 per cent have been so meagrely researched that their safety during pregnancy remains uncertain.

The provide the example of thalidomide as a case study for where running clinical trials on pregnant women could have avoided the serious birth defects that occurred in as many as 12,000 children.

Tuesday Links

Written By: Jason Shafrin - Nov• 23•15

Which value-based payment system is best?

Written By: Jason Shafrin - Nov• 22•15

Clearly, there is no single answer to this question and the answer depends on a number of factors including the market structure, provider responsiveness to intrinsic vs. extrinsic motivation, provider sample size, and the ability to accurately measure quality of care.  Douglas Conrad (2015) uses agency theory to provide an overview of existing value-based payment systems.  The article is a wonderful overview and has a number of useful insights.  For instance, Conrad describes why stable payment arrangements and quality measures are most useful for changing provider behavior:

Microeconomics has established that long-run price elasticities of supply and demand are both greater in absolute value than the respective short-run elasticities. In the short run, certain resources are fixed (e.g., capital inputs) and this dampens supply response to demand price…The Alternative Quality Contract in Massachusetts constitutes a successful example of this principle because of its 5-year duration, the fixity of its global payment method, and the predictable evolution of level of PMPM payment over time (Chernew et al. 2001; Song et al. 2012, 2014). In contrast, payers unexpectedly lowering PMPM payments or unexpectedly raising performance thresholds in future periods, seeking to capture an increased share of savings, will find that they might have shot themselves in the foot.” Short run surprises are likely to discourage future provider participation and to lower provider efficiency incentives. Similar consequences arise from Medicare accountable care organization (ACO) arrangements that inadvertently punish high-performing providers by rewarding improvement in, but not the level of, performance.

For those interested in value-based payment structures, this is a very interesting read.


Are wellness programs legal?

Written By: Jason Shafrin - Nov• 19•15

A post at The Incidental Economist considers the issue:

Workplace wellness programs discriminate. That’s what they do. Employees who adhere to a wellness program pay less for their coverage; those who don’t pay more. Wellness programs thus clash with federal rules that generally require employers to treat their employees even-handedly, regardless of health status.

The ACA, however, does explicitly make an exception for wellness programs.  Nevertheless, the question of whether employers should be in the business of employee health is an important one do consider.

Health Wonk Review: Thanksgiving Edition

Written By: Jason Shafrin - Nov• 19•15

Brad Wright of Wright on Health gives us a “Counting Our Blessings” edition of the Health Wonk Review.  He even offers thumbnails of each post illustrated with classic American thumbnail illustrations.

Why Medicare Advantage is thriving

Written By: Jason Shafrin - Nov• 18•15

In the 1990s, managed care began to take over the health care marketplace. However, backlash against managed care lead to a retrenchment in managed care in the late 1990s.  A paper by Sinaiko and Zeckhauser (2015) notes that:

After the MA-plan payment cuts imposed through the Balanced Budget Act of 1997, HMO availability dropped by nearly 50%: enrollment fell from 16% of the market in 1999 to 12% in 2002.


More recently, the Affordable Care Act also reduced payment to Medicare Advantage (MA) plans.  Nevertheless, MA enrollment has grown over the past 10 years so that by 2015, more than 3 out of every 10 Medicare patients are enrolled in Medicare Advantage.  Why is MA continuing to thrive despite these cuts?

A paper by Sinaiko and Zeckhauser (2015) proposes the following reasons:

  • Moderate cuts: Unlike the payment cuts to plans in the late 1990s, cuts in plan payments legislated in 2010 are less severe and are being phased in over time.   On net, MA plans today receive payments 6% above beneficiaries’ expected FFS cost
  • Patient experience: Medicare beneficiaries today have more experience and comfort with managed care.
  • Quality and Choice: There is vastly improved variety and quality in MA plans.  Since 2003, many plans have offered more expansive physician networks, such as preferred provider organizations (PPOs) and private FFS plans.
  • Switching cost: Cognitive biases in Medicare beneficiary decision making—including the tendency for beneficiaries to stay in their health plans over time—are likely favoring MA plans.  As many Medicare beneficiaries lower premiums above and beyond out-of-pocket dollars, MA plans began offering “offering ‘zero-premium’ plans, while introducing other revenue-enhancing and cost-saving measures, including higher cost sharing and narrower physician networks.”

Will this trend continue?  Only time will tell?


Mid-week Links

Written By: Jason Shafrin - Nov• 17•15

What are regression trees?

Written By: Jason Shafrin - Nov• 16•15

Regression trees are a way to partition your explanatory variables to (potentially) better predict an outcome of interest.  Regression trees start with a an outcome (let’s call it y) and a vector of explanatory variables (X).  

Simple Example

For instance, let y be health care spending, X=(X1,X2) where X1 is the patient’s age and X2 is the patient’s gender.

The regression tree will identify a cutoff for both X1 and X2 that will minimize the error from a regression. The error to be minimized could either be the sum of squared errors, the sum of the absolute value of errors, or any other error structure of interest.  For patient age, the regression tree will pick a cut-off (say age 40) that minimizes the error.   In the case of patient gender, this variable is already binary so the regression tree would split the data into males and females.  The regression tree logic will then decide whether the age or gender division best fits the data.

Regression trees are recurssive.  For instance, if it was determined that age minimized the error in the first node, the second set of nodes would look health care spending for <40 and male, <40 and female, ≥40 and male, ≥40 and female.

Visualizing Decision Trees

Luis Torgo’s trial based regression model tutorial gives both a visual and mathematical description of 2 variable regression tree.

regression tree

In practice, however, regressions have numerous variables. Further, more complex regression trees can allow for variable interactions while creating the partitions. Thus, regression trees risk becoming too detailed and losing predictive power. Typically, the user imposes stopping rules, such as a minimum sample size per node, a minimum increase in R-squared for each additional node, or using a validation sample to prune some of the leaf nodes.

Real World Decision Tree Example with Interactions

A paper by Buchner, Wasem and Schillo (2015) uses the regression tree approach and applies it to a risk adjustment model for setting health insurance premiums in Germany. They allow for the risk adjustment to include disease interactions as one would assume that an additional disease acts in a non-linear way with respect to cost.  The authors find that adding interaction terms to the risk adjustment model adds little predictive value.

The resulting risk adjustment formula shows an improvement in the adjusted R2 from 25.43% to 25.81% on the evaluation data set. Predictive ratios are calculated for subgroups affected by the interactions. The R2 improvement detected is only marginal. According to the sample level performance measures used, not involving a considerable number of morbidity interactions forms no relevant loss in accuracy.

Thus, although added detail and model sophistication can be a good thing, one must be sure to closely examine whether the additional complexity adds significant predictive value.   In this case it did not.


Does tying payment to quality improve quality?

Written By: Jason Shafrin - Nov• 16•15

Although the typical economist answer would be yes, in the case of one Medicare Advantage program, the answer is ‘no’.

A paper by Layton and Ryan (2015) [earlier draft] examine the Medicare Advantage Quality Bonus Payment Demonstration (MA QBP) which began in 2012.  In this program:

…plans receive bonus payments based on an overall plan quality rating calculated from a set of clinical quality measures and patient satisfaction ratings. These bonuses are quite large, ranging from 3-10% of plan payments, much larger than the 1-2% of revenue typically at risk in hospital and provider group P4P contracts.

Although their differences-in-differences  (DiD) approach found that quality improved in counties where plans were eligible fore the larger bonus compared to those that were not, bonus counties already had a positive trend towards quality improvement prior to the start of MA QBP, which may invalidate the parallel trends assumption required by the DiD framework. Instead, the authors use a regression discontinuity and propensity score design and find “…a precisely estimated zero effect of bonus size on quality, suggesting that larger bonuses have no effect on health plan quality.”

Why is the reason for this lack of improvement.  The authors give two explanations.  One is that quality improvement has high fixed cost (e.g., training staff) or that effective investment is necessarily “lumpy”, it may not be worth it for plans to invest in high-cost quality improvement efforts.  Second, plans may face short term constraints.  For instance, plans may find it difficult to shift their existing physician network to a a higher quality network, particularly if the current provider network has a multi-year contract.

Regardless of the cause, tying reimbursement to quality by itself may not be enough to move the needle on patient quality of care.


Music therapy

Written By: Jason Shafrin - Nov• 12•15

When we think of medicine we think of pharmaceutical medications, surgery and other interventions.  But can music also be used to heal.  According to the American Music Therapy Association, the answer is yes.  A paper by Gold et al. (2006) finds this is the case for patients with schizophrenia:

In people hospitalised with schizophrenia, adding music therapy to standard care lead to greater improvement in symptoms compared with standard care alone at 12 weeks (change in PANSS [Positive and Negative Syndrome Scale] total score from baseline: 29.00 with music therapy plus standard care vs 22.96 with standard care alone; p = 0.045).

A study by Chang et al. (2008) found that music therapy reduced stress and depression among pregnant women.  Clearly, music therapy will not cure cancer, but adding music could improve patient mental health and indirectly improve receptiveness to more traditional treatments.  One pharmacist agrees with this approach as well. A Carnegie Hall program convinced a 23-year old mother of two with HIV to get back on her medications after 3 years off her meds.

To improve your mental health as we go into the weekend, check out this blues classic from Louis Armstrong.  Enjoy.