Unbiased Analysis of Today's Healthcare Issues


Written By: Jason Shafrin - Mar• 15•16

Often times when doing data analysis, you want to find the relationship between two variables.  The first step is typically to plot a scatterplot.  To better understand this relationship, however, it is useful to fit a line to the scatterplot.  Most commonly, this is done with a simple linear regression (i.e., ordinary least squares (OLS) regression).

In many cases, however, the relationship between two variables may not be linear.  Consider the case of y=sin(x) where x ranges from 0 to 2π.  A linear regression would lead to y=β01*x where β01=0. However, this line would poorly fit the data as the sin curve undulates up and down reaching a maximum value at sin(π/2)=1 and a minimum value at sin(3π/2)=-1.

One way to better see this relationship is to fit a LOWESS or LOESS regression, also known as a local regression. The local regression creates smooth curves in your scatterplot by fitting points within local subsets of your data. More inclusive subset definitions (i.e., bandwidth) produces smoother curves, but ones that are less sensitive to local outliers; smaller bandwidth approaches can better capture local deviations, but are also more likely to overfit the data.

What is the difference between LOWESS and LOESS?

A smooth curve through a set of data points obtained with this statistical technique is called aLoess Curve, particularly when each smoothed value is given by a weighted quadratic least squares regression over the span of values of the y-axis scattergram criterion variable. When each smoothed value is given by a weighted linear least squares regression over the span, this is known as a Lowess curve; however, some authorities treat Lowess and Loess as synonyms.

The video below describes how to implement a LOWESS curve in R.  Stata also has a LOWESS command as well.

Preferences for Nonvalvular Atrial Fibrillation Therapies

Written By: Jason Shafrin - Mar• 14•16

Below is an abstract from on of my recent studies on patient and physician preferences for treatments for nonvalvular atrial fibrillation.  The full article is HERE.


The objective of this study was to compare patient and physician preferences for different antithrombotic therapies used to treat nonvalvular atrial fibrillation (NVAF).


Patients diagnosed with NVAF and physicians treating such patients completed 12 discrete choice questions comparing NVAF therapies that varied across five attributes: stroke risk, major bleeding risk, convenience (no regular blood testing/dietary restrictions), dosing frequency, and patients’ out-of-pocket cost. We used a logistic regression to estimate the willingness-to-pay (WTP) value for each attribute.


The 200 physicians surveyed were willing to trade off $38 (95% confidence interval [CI] $22 to $54] in monthly out-of-pocket cost for a 1% (absolute) decrease in stroke risk, $14 (95% CI $8 to $21) for a 1% decrease in major bleeding risk, and $34 (95% CI $9 to $60) for more convenience. The WTP value among 201 patients surveyed was $30 (95% CI $18 to $42) for reduced stroke risk, $16 (95% CI $9 to $24) for reduced bleeding risk, and −$52 (95% CI −$96 to −6) for convenience. The WTP value for convenience among patients using warfarin was $9 (95% CI $1 to $18) for more convenience, whereas patients not currently on warfarin had a WTP value of −$90 (95% CI −$290 to −$79). Both physicians’ and patients’ WTP value for once-daily dosing was not significantly different from zero. On the basis of survey results, 85.0% of the physicians preferred novel oral anticoagulants (NOACs) to warfarin. NOACs (73.0%) were preferred among patients using warfarin, but warfarin (78.2%) was preferred among patients not currently using warfarin. Among NOACs, both patients and physicians preferred apixaban.


Both physicians and patients currently using warfarin preferred NOACs to warfarin. Patients not currently using warfarin preferred warfarin over NOACs because of an apparent preference for regular blood testing/dietary restrictions.

Off label prescribing: Q&A

Written By: Jason Shafrin - Mar• 13•16

The Duke-Margolis Center for Health Policy has a great overview of some of the issues related to off-label prescribing.  Below is a summary of some key points from this article.

What is off-label prescribing?

Off-label prescribing and use can take many forms, such as use of an approved drug for an unapproved clinical indication, use at an unapproved dosage, use as first-line therapy in narrowly defined clinical populations where available treatment options are lacking (e.g., pediatrics or geriatrics), or use in areas where treatment options are limited or have been exhausted (e.g., oncology or rare diseases). Off-label prescribing often involves areas of high unmet medical need, and is increasingly utilized with the goal of individualizing patient treatment based on a broad array of sources of evidence.

Is off-label prescribing harmful to patients?

The answer is, it depends.  Because the level of evidence to support off-label uses is typically lower than on-label uses, the risk of using the treatments is often higher.  In many cases, however, off-label drug use has significant benefits for patients.

Can pharmaceutical firms communicate the benefits of off-label drug use?

In 1982, FDA published a one-page guidance to manufacturers explaining that materials provided to physicians in response to unsolicited [my italics] requests would not be regarded as labeling and therefore could lawfully include information about off-label uses…Another important “safe harbor” policy permitting manufacturers to provide off-label information arises under the “scientific exchange” regulation promulgated around the same time; it applies to communications such as manufacturer publication of original trial results and letters to the editor of scientific publications in defense of public challenges.

Additional guidance is available in the FDA “Good Reprint Practices” document.

What is FDAMA and how does it relate to off-label communication?

The agency’s current stance also has roots in the Food and Drug Administration Modernization Act (FDAMA) passed by Congress in 1997.9 Section 401 of FDAMA enabled the dissemination of peer-reviewed journal articles related to off-label indications with a confirmed commitment to filing a Supplemental New Drug Application (SNDA).

How does FDAMA affect communication of health economic results

Section 114 enabled the sharing of health care economic information with formulary committees, managed care organizations and other large consumers of health products.

What recent court cases have effected off-label communication?

  • United States v. Caronia (2012): Rejected a government effort to criminalize accurate speech about off-label uses.
  • Amarin Pharmaceuticals Inc. v. FDA (2015): Amarin Pharmaceuticals and physician plaintiffs filed a complaint alleging that the firm was being prevented, by FDA regulation, from communicating truthful and non-misleading information about the prescription drug Vascepa. The pharmaceutical company was granted preliminary relief by a district court, which gave the firm permission, based on Caronia, to communicate truthful and nonmisleading speech via specific, court-sanctioned promotional language
  • Pacira Pharmaceuticals Inc. v FDA (2015): Pacira sought to promote its post-surgery analgesic drug, Exparel, for uses that the FDA had alleged were unapproved, although some evidence suggested that the approved labeling could be interpreted to include these uses. The proceedings were settled in December 2015 after the FDA revised the approved labeling to make clear that the contested uses were, in fact, on-label.


Friday Links

Written By: Jason Shafrin - Mar• 10•16

HWR is up at HBB

Written By: Jason Shafrin - Mar• 10•16

David Williams has a freshly posted Health Wonk Review: Tales of the Trump at Health Business Blog.

Cost of Quality Reporting: $15.4 billion

Written By: Jason Shafrin - Mar• 09•16

Medicare aims to move away from fee-for-service reimbursement and towards value-based payment mechanisms based on quality of care.  Although the goal is laudable, there are a number of practical challenges.  First, most care is still provided via fee for service.  In 2013, 95% of all physician office visits were reimbursed using fee-for-service.  Second, collecting quality of care data comes with cost. A paper by Casalino et al. (2016) find that quality reporting is in fact extremely costly.

Each year US physician practices in four common specialties spend, on average, 785 hours per physician and more than $15.4 billion dealing with the reporting of quality measures. While much is to be gained from quality measurement, the current system is unnecessarily costly, and greater effort is needed to standardize measures and make them easier to report.

Clearly, measuring quality of care is important, but Casalino rightly notes that understanding the costs of reporting and identifying strategies to reduce the cost of reporting is vital if value-based purchasing would be able to improve the efficiency of the health care system.


Is Economics a Science?

Written By: Jason Shafrin - Mar• 07•16

That is the question addressed by in a paper by Colin Camerer and co-authors in Science. The authors repeated 18 economics experiments conducted in a laboratory setting. These articles were published in leading economics journals including the American Economic Review and the Quarterly Journal of Economics between 2011 and 2014.

As The Economist reports, they found that

For 11 of the 18 papers (ie, 61% of them) Dr Camerer and his colleagues found a broadly similar effect to whatever the original authors had reported. That is below the 92% replication rate they would have expected had all the original studies been as statistically robust as the authors claimed—but by the standards of medicine, psychology and genetics it is still impressive.

One theory put forward by Dr Camerer and his colleagues to explain this superior hit rate is that economics may still benefit from the zeal of the newly converted. They point out that, when the field was in its infancy, experimental economists were keen that others should adopt their methods. To that end, they persuaded economics journals to devote far more space to printing information about methods, including explicit instructions and raw data sets, than sciences journals normally would.

Maybe Economics is a science after all.


Cost effectiveness analysis Q&A

Written By: Jason Shafrin - Mar• 06•16

What is cost effectiveness analysis or CEA?  

One definition is that CEAs–at least in the field of health care–measure the difference or ratio between cost of care and the benefits of care for a given intervention compared to an alternative treatment strategy.  The intervention could be a new surgical procedure, a drug, a behavior modification program or any other intervention of interest.

How frequently are CEAs conducted?

Fairly commonly.  In the past, CEAs were conducted primarily in Europe for health technology assessment (HTA) organizations such as the the National Institute for Health and Care Excellence (NICE) in the UK.  However, with the advent of new initiatives to measure care value by CMS, the advent of value-based insurance design initiatives by private and public payers, and the development of independent institutions that measure cost effectiveness, CEAs are on the rise.

Are there historical data on CEAs?

In fact there are.  The Center for the Evalutation of Value and Risk in Health (CEVR) at Tufts University in fact has an entire registry of CEA studies.   CEVR was founded by Peter Neumann and Joshua Cohen and has two databases:  Cost-effectiveness Analysis Registry and the National Coverage Determinations Database.  The CEVR CEA database has been used in a number of publications.

Can I get access to the CEA registry?

The CEVR website does allow to search for CEA articles that are in their registry here.  The CEA registry FAQs are also helpful.

Friday Links

Written By: Jason Shafrin - Mar• 03•16

Core Quality Measures

Written By: Jason Shafrin - Mar• 03•16

One challenge providers have faced in the past is that quality measure reporting has been complex.  Medicare may ask for quality measures with one definition, commercial payers may define quality a second way, and Medicaid may ask for a third definition of quality.  Keeping track of these definitions and recording quality measures distracts providers from actually providing quality care to their patients.

With this issue in mind, CMS and a trade organization known as America’s Health Insurance Plans (AHIP) have create seven sets of core measures that public and private payers will use in a consistent measures.  As AJMC reports, there are  7 sets of core quality measures.   The seven sets of core measures fall into the following groups:

  • Accountable Care Organizations (ACOs), Patient Centered Medical Homes (PCMH), and Primary Care
  • Cardiology
  • Gastroenterology
  • HIV and Hepatitis C
  • Medical Oncology
  • Obstetrics and Gynecology
  • Orthopedics

Why is this coordination needed?  A CMS statement says:

“In the U.S. Health care system, where we are moving to measure and pay for quality, patients and care providers deserve a uniform approach to measure quality,” said CMS Acting Administrator Andy Slavitt. “This agreement today will reduce unnecessary burden for physicians and accelerate the country’s movement to better quality.”

Centralizing quality measures has the advantage of simplifying quality reporting for providers.  However, CMS and AHIP will need to approve changes to these core measures.  As new medical techniques and treatments become available, will the core measures be flexible to adapt to changing practice patterns?  My guess is ‘no’.  Thus, CMS should carefully consider how best to track quality while maintaining flexibility to adjust the definition of “quality” based on how the state of the science evolves over time.