How do physician payments affect drug prescriptions?

San Diego, California
MedicineData Science
$60
Raised
1%
Ended on 6/07/17
Campaign Ended
  • $60
    pledged
  • 1%
    funded
  • Finished
    on 6/07/17

About This Project

Health care companies enter into financial relationships with doctors, hospitals, and other health care providers. The Sunshine Act requires them to disclose these relationships. A previous analysis uncovered a correlation between payments to doctors and the likelihood of prescribing a brand name drug. We'll repeat this analysis and release the full data to the public. Additionally, we will reanalyze the data and attempt to identify factors that can help explain the observed effects.

Ask the Scientists

Join The Discussion

What is the context of this research?

Health care companies frequently enter into financial relationships with doctors, hospitals, and other health care providers. These relationship are often in the form of speaking fees, research, gifts, or meals. Doctors also can invest in the products of pharmaceutical products, and potentially receive a share of the profits in procedures or drugs they prescribe. The Open Payments data from the Centers for Medicare and Medicaid Services was recently released online, and discloses these financial relationships to the public.

What is the significance of this project?

ProPublica revealed that doctors who receive payments are, on average, more likely to prescribe a higher percentage of brand-name drugs. This analysis showed a correlation, however, it doesn't prove causality. We will dig into this data more deeply and investigate possible explanations for the observed effects.

We will additionally integrate other external data sources, including data on hospital of care and physician quality of care, in order to investigate the correlation between brand name prescriptions, doctor payments, and general health outcomes of the patients.

What are the goals of the project?

We need to redo the ProPublica analysis, because it was never released to the public. Once this is done, we will release it under an open source license. Several data fields need to be normalize and standardized in order to allow us to cross-reference this data across databases.

We will build a robust statistical model using many additional factors that weren't considered by the ProPublica analysis including hospital affiliation, drug class and usage, and the patients diseases they are being treated for.

Budget

Please wait...

All members of the team have full time jobs, without this money team members will not be able to reliably allocate personal time to this project in a consistent manner.


We already have all the computational, storage and associated resources required to complete the analysis so no additional expenses will be required to complete the project. Additionally, all the data has already been secured, so no datasets required for the analysis require a subscription or additional expense to acquire.

Endorsed by

I have worked with Greg and Max for two years and they have the technical and scientific ability to make this exciting project a reality.

Flag iconProject Timeline

We expect it'll take about a month to understand and properly normalize the Open Payments data. With the data available, two months should provide enough buffer to properly replicate the ProPublica analysis. We expect additional time will be needed for this as everyone on the team intimately understands the difficulties involved in replicating previous research. The additional two months to do our own analysis should be more than sufficient.

May 08, 2017

Project Launched

Jul 01, 2017

Download and normalize Medicare Prescribing data

Jul 01, 2017

Download and normalize OpenPayments data

Sep 01, 2017

Replicate ProPublica Analysis

Oct 01, 2017

Download, normalize, and integrate health outcome data

Meet the Team

Max Nanis
Max Nanis
Computational Biologist
Gregory Stupp
Gregory Stupp
Bioinformaticist
Tiffany Comandatore
Tiffany Comandatore
Business Analyst, Statistician

Affiliates

ID Analytics, Qualcomm
View Profile

Team Bio

Max and Greg have worked together for over 2 years in the department of Molecular and Experimental Medicine department at The Scripps Research Institute. Outside of the laboratory, we’ve collaborated on award winning projects in the past, most notably winning the 2015 Big Data Category at CISCO Live by building a tool to predict death dates using DNA methylation data. Paired with Tiffany's complementary skills, we'll be able to uncover insights and the stories the data is telling.

Max Nanis

Max Nanis is a programmer and computational biologist. He studied biology, chemistry and sculpture at Bennington College. Max logged onto the Internet when he was three years old and has never signed off.

First joining The Scripps Research Institute (TSRI) in 2012, Max worked with Dr. Arthur J. Olson at the Molecular Graphics Laboratory developing OpenGL tools for protein docking simulations. At the Su Lab, he’s a senior computational biologist and is the lead developer of Mark2Cure, a citizen science tool that empowers anyone who can read to make biomedical research literature more useful for researchers. As part of Mark2Cure, Max develops language processing tools, builds user contribution analysis pipelines, and disseminates user collected data openly for researchers to leverage. Previous work at TSRI included software development for GeneWiki, an effort to provide a well written and informative Wikipedia article for every notable human gene.

As the founder of General Research Lab, a chemically oriented creative software and animation studio, he helps companies develop new ventures and solve hard problems. Taking on anything from database optimization, to product driven biological analysis, he’s found ways to convey technically difficult concepts to anyone. His work has been covered by The New Yorker, The Association for Computing Machinery, Bloomberg Businessweek, The New York Times, and others

Gregory Stupp

Gregory Stupp is a biochemist turned bioinformatician. He obtained his Ph.D. in Biomedical Sciences in the College of Medicine Interdisciplinary Program at the University of Florida, Gainesville. There, he studied the ways in which nematodes can detoxify their environment and fight off toxins produced by pathogenic bacteria they often encounter in the wild. Another part of his research involved creating tools to analyze the large amounts of data generated by measuring the compounds these nematodes produce when they are challenged. He joined the Su Lab as a Research Associate in September of 2014. His work focuses on understanding the role of the human microbiome in inflammatory bowel diseases. Additionally, he contributes toward biomedical ontologies, crowdsourcing, and improving science through better data standardization and integration, using Wikidata as a knowledgebase. In his spare time, he contemplates the intricacies of computer-based learning while silently meditating to the gustatory glories of exquisitely assembled entrees comprised of Coryphaena hippurus flesh, a mixture of flour and water, and spices/oils of unknown origins.

Tiffany Comandatore

Tiffany is currently working as a Business Analyst and has previously worked as a Statistician at Qualcomm in the corporate legal department. She has her M.S. in Statistics from San Diego State University. Her research has revolved around patents (is royalty stacking really happening? What are software patents? How to value patents?) and patent litigation (what role are standard essential patents playing in litigation? The effect of eBay vs. MercExchange case on practicing entities and non practicing entities). Unlike Greg, in her spare time she enjoys the beach, hiking, camping, and cycling.

Additional Information

The purpose of looking at other factors that may affect brand name prescribing is that we can identify specific conditions in which this effect may be observed more or less strongly. These conditions may help explain why this effect exists.

For example, one possible scenario is that a significant correlation between industry payments and brand name prescriptions is observed in oncologists, but not in internal medicine doctors. Or maybe this correlation only exists in doctors receiving over $50k, but is not seen when considering all other doctors. We would like to investigate possibilities such as this, which would better inform the public as to the underlying factors that influence the relationship between payments and prescriptions.

In addition, by stratifying treatment centers, it allows us to associate OpenPayments data with the Hospital Compare dataset. It includes 57 measure of performance across 7 categories: mortality, safety of care, readmission, patient experience, effectiveness of care, timeliness of care, and efficient use of medical imaging. Until there are meaningful ways to quantify quality of care given by a physician, we will entertain these associations for potential insights between prescribing behaviors and treatment outcomes.

References


ProPublica


Project Backers

  • 2Backers
  • 1%Funded
  • $60Total Donations
  • $30.00Average Donation
Please wait...