Computational Analysis of Big Pharma

Why would the largest pharmaceutical companies target pediatric emergency care specialists with such intensity?Our initial justifications of these results focused on the nature of the pediatric emergency care specialization.

These physicians care specifically for children in emergency situations.

Because emergency issues are typically time-sensitive, it might be that these medical professionals have complete discretion over what prescription drugs a child takes.

In this, we mean that in the specific scenarios that pediatric emergency care specialists tend to, parents might give complete discretion to the trusted professional without arguing over whether to be prescribed a branded vs.

generic drug.

When Jimmy is SCREAMING because of his broken arm, mom and dad probably aren’t going to argue about the brand-name of a drug; they just want the SCREAMING to stop and their beautiful baby boy to be the image of 5-year-old health again.

If this is the case and pharmaceutical companies understand that, then it would make sense for Big Pharma companies to get pediatric emergency care specialists to prescribe all branded drugs at whatever cost necessary, including making hefty donations.

But why just make unfounded conclusions when you can query the data and make, well, founded conclusions!We then moved to close the loop by directly examining the specific drugs that pediatric emergency care specialists and orthopedic surgeons prescribed.

We did this to investigate whether these substantial donations really did affect the specific drug prescribed by a doctor.

Above, we established that pediatric emergency care specialists were rolling in cash from Big Pharma (okay, yes, the “donations” are labeled as “speaking fees,” and “all-expenses-paid medical conference trips” but just roll with me on this one), now let’s see if that cash paid off.

To investigate this, we:Indexed the Prescriptions data set for pediatric emergency care specialistsFound the resultant set of all Medicare Part D prescriptions made by each pediatric emergency care specialistStep 2: grouping prescriptions by prescriber.

Prescribers are members of specialty most targeted for donations (pediatric emergency care or, separately, orthopedic surgery)3.

Iterate through each physician’s set of prescriptions, flagging a prescription if it was a brand name drug instead of a generic drugStep 3: Branded drugs prescribed by orthopedic surgeons4.

Cross-reference the branded drug in the “Drugs and Manufacturers” dataset to see if manufactured by one of the top-20 companies investigated.

Step 4: Branded drugs prescribed by pediatric emergency care specialists that were manufactured by the companies investigated5.

If the physician prescribed a branded drug, we then queried the Open Payments data set using the physician’s name, to see if that specific physician was given donations by pharmaceutical companies.

By doing this, we would be able to examine, to some degree, the possible influences of pharmaceutical donations on drug prescriptions.

Step 5: Orthopedic surgeons who prescribed branded drugs, those branded drugs were manufactured by the companies in question, and they were donated to by those companies.

Includes value of donation.

6.

Repeat steps 1–5 for the second specialty investigated: orthopedic surgery.

Okay… So what did you find?Our results were surprising to say the least.

After carrying out the above process on both pediatric emergency care specialists and orthopedic surgeons, we found:# of branded pediatric emergency care specialist prescriptions in 2016: 4# of pediatric specialists who prescribed branded drugs and took donations from the drug’s manufacturer (one of the 20 companies investigated): 0# of branded orthopedic surgery prescriptions in 2016: 568# of pediatric specialists who prescribed branded drugs and took donations from the drug’s manufacturer (one of the 20 companies investigated): 4Well… That was disappointing.

Or maybe not, depending on how you see it!.If these findings hold, it appears that pharmaceutical donations are not as effective at influencing physician prescription decisions as hypothesized.

And hey, that would be good news all around.

Our research suggests that in 2016, for the specialties (pediatrics emergency care and orthopedic surgery) that receive the highest average donation value and the highest net donation value from the 20 highest-spending pharmaceutical companies, there is minimal correlation.

In ultra-simple terms it seems like donations did not overtly affect the specific drugs that were prescribed.

Conclusion and CaveatsBut, hold your horses.

CreditThese findings come with some major caveats we would like to address.

At numerous points while conducting the research, we ran into problems and limitations with the data.

First and foremost, the nature of the Prescriptions dataset.

This Prescriptions dataset is limited to Medicare Part D Spending by the federal government for the year 2016.

While the set is massive, it only represents a small fraction of total prescriptions fulfilled in the U.

S.

in 2016.

Medicare is a government program and as a result, its records are free and available to the public.

This is why we used them.

This research project is not definitive simply because it only examined a small sample of the total drug market.

Sampling at a small scale to model large-scale behavior might work with some research but certain characteristics of Medicare further muddied our research waters.

Because Medicare is a government program, it carries with it restrictions and limitations; these limitations become even more refined and restrictive at the “Part D” level.

“Part D” refers to a specific subset of Medicare plans and only covers drugs in six categories: antidepressants, antipsychotics, anticonvulsants, antiretrovirals (AIDS treatment), immunosuppressants, and anticancer.

This immediately poses a problem with our investigation of pediatrics and may explain why we found such a small number of branded, pediatric drug prescriptions.

Furthermore, funding for Medicare is limited by the federal government.

Because the government’s goal is to keep Medicare spending at a minimum while still providing necessary drugs to the public, they may overwhelmingly restrict fulfilled prescriptions to generic drugs and not branded drugs.

This might also explain why we found so few branded drug prescriptions in our research.

Finally, not all doctors accept Medicare.

Doctors have to opt-in to accept Medicare and when, on average, Medicare pays only 80 percent of what private health insurance pays, it seems to be an increasingly less popular option.

Doctors who are already interested in maximizing their profits might not choose to accept Medicare and thus would not be present in our data set.

These same doctors may be the ones who take the most donations from pharmaceutical companies.

In a separate vein, our research was limited by only investigating the top 20 donating pharmaceutical companies.

We made this decision because it made the data easier to manage and work with.

However, it is possible that these companies all, or mostly, focus on a specific sector of the medical market (i.

e.

they make drugs that pediatricians and orthopedic surgeons would never use).

Because we left out dozens of other, smaller companies, we may have missed manufacturers of pediatric/orthopedic drugs.

All these observations point to a single truth that reverberated through our minds at every step of this research process: the perceived pharmaceutical donation problem is exceedingly nuanced and clouded.

At the start of the project, we could never have imagined the number of times we had to consider the actual decisions a pharmaceutical company might make, a physician might make, a pharmacist might make, and a patient might make — all to arrive at meaningful and valid claims about the data we had before us.

The scale of the problem, the legal aspects surrounding public disclosure of certain payments/prescriptions and not others, and working with multiple datasets that each coded its data in a different way and was created/managed by different groups, made this research project challenging… but fun.

But why does investigating any of this actually matter?If you’ve gotten to this point and still haven’t already found your own answer to the question above, here is how we see it, plain and simple:lives could be at stake.

If a doctor prescribes the wrong drug (because of a bias toward Big Pharma) to a patient, diseases could be inadequately managed, symptoms could spread, and treatment could be ineffective.

People could die.

Topics for Further ResearchOur project only scratched the surface of what is possible using the OpenPayments and Medicare Part D datasets.

These datasets are thorough, powerful, and impressive and the opportunities for investigation are limitless.

We discussed investigating the effect of Big Pharma donations on research publications.

For example, if Dave the Doctor was given donations from a pharmaceutical company that produces ExampleDrug A, what are the chances that he will later write/co-author a research article touting the perceived positive effects of ExampleDrug A?.This question is important because thousands of doctors and patients rely on medical journals to understand the benefits/disadvantages of new drugs appearing in the market.

Based on these articles, doctors prescribe specific drugs.

To answer this question, we could use natural language processing to code research articles by tone (positive/in-support or negative/not-in-support) and search the Open Payments Dataset for donations to the author of the article.

Then, we could use linear regression using the sentiment analysis score along with the query into the Open Payments Dataset.

Furthermore, we are interested in using machine learning to predict the projected amount of donations a doctor might receive based on a number of features of the physician (e.

g.

U.

S.

state located, years in practice, etc.

).

We would particularly like to investigate whether the specific form of Big Pharma donations affects prescription bias.

If good ol’ Dave the Doctor receives 150 free lunches every year from FakePharma, Inc.

, is he more likely to prescribe FakePharma, Inc.

’s drugs than if FakePharma, Inc.

was flying him to medical conferences every weekend?All the datasets we used are free to use/download; if you have any ideas on possible further research topics related to them don’t be afraid to dive in!So I made it this far, now what?Great question!.Here are some steps that you can take right now to better protect yourself and your loved ones from potentially biased prescribing.

We implore you to look up your physician on the US government’s Open Payments website or ProPublica’s Dollars for Docs website and think critically about the trust you store in him/her.

We emphatically suggest reading into peer-reviewed medical research studies on the effectiveness of specific drugs.

Don’t just rely on what the TV advertisements tell you or what Karen recommended over deviled eggs the other night.

Do your own research and bring that research to your doctor’s office.

To start, try taking a look at the findings published in the CDC’s Preventing Chronic Disease Journal or other journals like it.

Extra reading for those of you who just can’t get enough1.

Down that rabbit-hole: High drug prices and patent monopolies2.

Big Pharma spends millions on lobbying against lowering prescription drug prices3.

O Canada, you have these problems too!4.

Is Big Pharma contributing to the growing Opioid Crisis?5.

A similar research study! — Pharmaceutical Industry-Sponsored Meals and Physician Prescribing Patterns for Medicare BeneficiariesAnd here is some on-theme music you can listen to while you champion pharmaceutical transparency and fairnessThis project was made with ❤ by me, Sisipho Zinja, and Charles Junior Kwanin.

If you’re interested in replicating our research project to investigate your own burning Big Pharma questions, here’s our GitHub with the code/files!*Karen is a fictional character portrayed in this article for explanatory/descriptive purposes only.. More details

Leave a Reply