Detecting Drug Diversion With Data Analytics–An RxTrace Interview With Analytics Expert Kosta Tzavaras

phiSystems_Logo2This week I have a special treat to share with RxTrace readers.  It is an interview I recently conducted with Kosta Tzavaras, author, publisher, leading data management expert in the pharmaceutical industry and Solutions Manager with Phi Systems in Thousand Oaks, California.  Kosta is a consultant to pharma and biotech manufacturers specializing in data analytics using supply chain and reimbursement data to help detect various kinds of drug diversion.

The original interview was conducted last summer and we updated it last week to reflect the recent passage of the Federal Drug Quality and Security Act.

DirkWelcome Kosta.  Please describe your business…what it is that you do at a high level? 

Kosta:  Phi Systems is focusing in a niche area within the brand protection business in pharmaceuticals dealing with provider type of diversion, that is where healthcare providers are involved.  The two particular areas I’m focusing on are physician office diversion—diversion that is enabled by the actions of certain physicians and their practices—and the second has to do with 340B diversion or PHS (Public Health Services) account diversion.  That is diversion enabled, again, by providers that vary from hospitals to clinics to group practices.  These are facilities that qualify for the lowest pharmaceutical discounted prices in the industry, and have the opportunity to acquire the product at the low price and sell it at a higher price if they choose to violate their agreement with the manufacturer.

DirkWho would be your customers…who is it that wants to employ you to do these kinds of things?

Kosta:  The customer would be a pharmaceutical manufacturer who is marketing a product, the kind of product that is susceptible to diversion.  Not all products are susceptible to diversion because of various reasons:

  1. They may not have enough sales volume to make it worthwhile for somebody to divert…A healthcare provider might not have enough business with that particular drug to make it worthwhile for them to divert it.
  2. Pricing.  The product has to be expensive enough where by arbitraging price, you can actually make some money.  This is likely to happen with the higher cost drugs.
  3. You have to have two markets where there is a significant price differential.  We are talking about the U.S. market, so there is going to be another market out there—whether it is Europe, Canada or third-world markets, etc.—where the price is significantly lower than the U.S. price.

DirkSo the physician office type of diversion you would be detecting would be diversion of drugs that are targeted for somewhere outside the United States but are being sold within the United States improperly.

Kosta:  That’s correct.  Those products were intended for use in other countries and they were improperly acquired by U.S. providers through either international suppliers via a shipping method, or through a secondary supplier here in the United States that acquired/imported the drug from outside the country through various means.

DirkSo this would be similar to the Avastin case (see “How Counterfeit Avastin Penetrated the U.S. Supply Chain”) that most people are familiar with from the last couple of years…

Kosta:   Precisely!  That is the one best known case that involved physician office diversion.  It got a lot of publicity.  The FDA got involved with that case.  They identified the supplier, were able to obtain enough information from the supplier—records, I suppose—they compiled numerous lists of providers that had engaged in diverting Avastin.  And Avastin, I’ll say, is one of the worst case scenarios—the worst nightmare for the manufacturer, because Avastin was not only diverted, but it happened to be counterfeit as well.

A lot of the cases of diversion we see the product is legitimate.  That is, the product is made by its rightful owner but for another country; it is authentic, although it might not have met the cold-chain requirements and supply chain requirements.  The product may have gone bad, but your worst case scenario is a counterfeit product.

DirkYeah, and in the Avastin case, the FDA, as I understood it, found out which providers were acquiring the drug improperly by getting that information directly from the supplier in the U.K., or at least through the U.K. regulatory agency there (the MHRA) rather than through a process like what you are doing.

Kosta:  That is correct.  The FDA had first-hand information through the supplier.  The process that I am utilizing is using third-party data to identify diverters.

DirkSo in the case of a pharmaceutical manufacturer of a block-buster drug that is very high priced—it fits all of the profile that you described earlier—if they were to follow the process that you are describing and find what they believe is diversion, what can they do about that?  I know that’s not your role, but I’m just curious to know if they would employ you for your services, what would they do when you provide them with evidence that indicates that you suspect some diversion.

Kosta:  The first thing we have to keep in mind is that we are talking about data, the quality of which is very good, however, it is not foolproof.   This is the same data drug manufacturers are using every day for their analytics, for their promotional activities, but there is a margin of error within the data, so once you identify suspect accounts your first responsibility would be validation.  You do that by interviewing the suspect provider, probing about their purchasing practices, to see if they would volunteer any information about their sources of the product.  Experience with previous cases indicates that these providers are forthcoming with information initially, before the get defensive about the whole thing.

And so once you have some evidence, you have a few options in front of you:

  1. Get assurance directly from the provider that they cease the practice of acquiring diverted product
  2. Refer them to the FDA.

It’s all up to the manufacturer what action they’re going to want to take, ranging from amicable agreement with the provider or more extreme measures.

In addition to that, the data will help you identify two categories of accounts:

  1. Suspect:  That is accounts that the data provides enough evidence that suggests that they are engaged in diversion;
  2. High-risk accounts:  There isn’t enough data to indicate that they are engaged in diversion, however, they fit the profile of the typical diverter compiled from a large list of accounts.  So by matching a company’s account universe to that profile you can identify the high-risk accounts.  These are accounts where if diversion were to happen, that is where it would likely happen.  So with this group of accounts, lacking the evidence, the best thing you can do is proactively educate them about the risks of diversion.
  3. In addition, the data also allow you to build the risk profile of a product and estimate its potential exposure to diversion, even estimate the tangible losses.

DirkThe data that you are using you referred to as data that the manufacturer owns, so you are using the manufacturer’s existing data that they have for, as you said, marketing and other things, right?

Kosta:  That is correct.  Most of it is data that manufacturers already subscribe or can subscribe to from third-party sources, and they are using it on a regular basis for their promotional activities.

DirkAbout the data…today you could do this without any serialization data, but we are on the cusp of the time where even more robust data at the unit-level will be available.  Would that help you at all to go any farther than what you have accomplished with data that exists today?

Kosta:  Yes.  The process can be improved in a couple of ways, either with the improvement of the existing data products out there…the ones I was referring to…and in addition the serialization data.  Improved data would bring a new level of analytical precision, identification of more suspects, and improved confidence around the results.

But, Dirk, you know, my opinion is that physician office diversion can be eradicated much before the serialization data comes around.  The FDA has already given enough publicity to the case …it has brought it to the attention of a lot of providers…a lot of providers are already aware of the risks, and, in fact, of the implications.  The Avastin case resulted in the prosecution of a doctor, involving several hundreds of thousands of dollars in fines and a prison term for one of the physicians.  So I’m thinking that with enough attention from the manufacturers and the FDA, the industry in general can get control of the illegal distribution of diverted drugs through physician offices.

But the manufacturers must be willing to talk both about the risks of diversion, and the implications for the providers. In the name of good relations with the customers, manufacturers shy away from talking about the implications.

DirkNow, unlike the diversion we have been talking about so far, the 340B type of diversion is all within the United States, isn’t it?

Kosta:  That is correct.

DirkAnd how does that type of diversion work?

Kosta:  This diversion involves providers that qualify for 340B pricing.  These are typically providers that serve …their business portfolio includes a large number of disadvantaged patients.  HRSA has made provisions for these providers to be able to acquire product at favorable pricing.

DirkFavorable pricing, or favorable reimbursement from Medicare?

Kosta:  It’s not the reimbursement, but the acquisition cost of the drugs.  They can acquire drugs at about 50% of the price of the drugs or less.  This is the deepest discount given to any provider.

Dirk…by the government. And so then the pharmaceutical manufacturer gets reimbursed by HRSA directly?  Or the wholesalers.  Somebody does…

Kosta:  No, they don’t get reimbursed for that…the manufacturer bares the cost of the discounting entirely.  That is the final sale at that discounted price.

DirkOh!  So it seems to me, then, that the manufacturer would not want to sell to 340B providers, because they don’t get reimbursed by the government for that lower price?

Kosta:  Yes, but if they don’t provide the drug at the 340B pricing then they can be disqualified from receiving Medicare reimbursement.  Basically, they are not getting reimbursed for this discount in any way, but the manufacturers can qualify to receive Medicare reimbursement otherwise.

DirkOK.  So you have a doctor, they are receiving the 340B pricing because they have applied and been approved as being a provider for this particular class of patient.

Kosta:  Yeah. Although a 340B provider is usually the outpatient department of a hospital, a clinic, or a group practice owned by a hospital.

DirkEarlier you seemed to imply that they are approved as a 340B provider because they have a certain number of their patients…they may have some patients…are they allowed to have any patients that are not 340B qualified?

Kosta:  Yes.  Once you quality for the 340B pricing then all of your patients are covered.  So the idea is, you’re going to lose some money on the more disadvantaged patients and you’ll average that out with non-disadvantaged, so these discounts help you offset some of your costs.

DirkAnd keeps you in that community where they have the greatest need.

Kosta:  Exactly.

Dirk I see.  So you can detect when this may be happening…that kind of diversion?

Kosta:  That is correct.  The 340B analysis focuses in three areas.

  1. Identifying cases where a provider may be passing the discounted product to an affiliate.  This is the case with hospitals where an outpatient department may be passing the drugs to the inpatient department that doesn’t qualify.
  2. Identify facilities that received the discount outside their eligibility period.  There is an eligibility period—a start date and a termination date—for which the provider is qualified.  And so you want to make sure the discounts fall within the eligibility period.
  3. Identify facilities that received the discount and at the same time received Medicaid reimbursement for the same unit of the product. Providers cannot receive Medicaid reimbursement at the same time they acquire the product at 340B pricing.

DirkCould they load up on drugs while they are in their eligibility period and then dispense them later?

Kosta:  As long as they are qualified they can buy any quantity of the product, but you cannot receive the discount before the effective date or past the termination date.

Dirk OK.  So a provider has to opt for either 340B pricing or Medicaid reimbursements, but not both.  Is that per patient or the entire practice?

Kosta:  It’s the entire practice.

DirkOK, during the eligibility period.

Kosta:  Right.

DirkInteresting.  So in the day coming, when every drug and biologic package has a unique serial number on it, does this problem kind of go away if HRSA would require a serial number of the drug that is being dispensed to each patient, and then perhaps they would have to provide that same serial number back to the manufacturer or the wholesaler or somebody so they can confirm that the pricing was applied properly?

Kosta:  The extent that the serial number can be tied to the facility where the drug was used, then the manufacturer could verify that the product was, indeed, used by an eligible facility.

DirkRight.  That makes sense.  I think the way I described it, it would probably violate the HIPAA regulations.  You can’t take it down to the patient, but at least you can tell that the drug was dispensed during the facility’s eligibility period and the correct discounts were applied.

Kosta:  Right.  Of course, if serialization or pedigree stops with that facility and the product wasn’t used there, but it was passed to a non-qualified patient or another provider and not reported, then you would not be able to detect the diversion.  So in essence, the provider would have to indicate…would have to report the next destination of the product… an unlikely event if the product is knowingly being diverted.

DirkRight, but [to detect that diversion] someone would have to then be monitoring those pedigrees…some responsible organization, [like] the manufacturer, or HRSA, or FDA…someone.

Kosta:  Yes.

DirkWell, in fact, it wouldn’t be the FDA because the FDA isn’t going to regulate that kind of thing, so it would have to be HRSA, if anyone, I think, so then in order for that to happen, HRSA would have to expect to receive pedigrees from the provider showing where they acquired the drug, how the drug got to that provider and that it was dispensed, and should not get a duplicate pedigree from someone else saying they dispensed that same drug in order to get their pricing.

Kosta:  Right.

Dirk So that assumes pedigrees, but is there any value of just having serial numbers and no pedigrees, because now that the federal Drug Quality and Security Act  has passed it will be at least ten years before we actually see some sort of track & trace information or pedigrees [based on unit-level serial numbers] being required.  So is there any value to just serial numbers instead of pedigrees to prevent this or detect it?

Kosta:  Well, serial numbers without the reporting, I suppose, wouldn’t get you any closer to the solution of the problem.  There would have to be validation.

DirkWell, if the practice was required to, at least, include the serial number of each of the drug packages that they dispense under the 340B program to HRSA, even without a pedigree, if multiple providers submitted that identical serial number, especially if they repeatedly did that, that would indicate there is diversion going on, and that one of them may deserve the pricing and the other would not.  So even without the pedigree it seems like that would be of some value as long as HRSA would start to require…

Kosta:  Yeah, except, why would HRSA make this effort to protect the manufacturer?  They don’t have the same vested interests that the manufacturer has to see that this type of diversion doesn’t happen.  So HRSA will provide the framework for protection but the policing falls in the hands of the manufacturer.  It’s their responsibility.

DirkI would think that HRSA would take a great interest in stopping providers from participating in illegal activities like diversion, I would assume.  The cost is borne by the manufacturer, it sounds like, if two different doctors claimed that they dispensed under 340B the same drug…I guess that’s not really diversion, it’s just claims fraud.

Kosta:  Right.  Yes.  And I suppose with 340B diversion you don’t have the same associated risks that you have with illegal diversion from importation, etc.  Federal entities wouldn’t be so sensitized to this issue.  It becomes more of a monetary loss for the manufacturer.  It has a different level of risk than the physician office diversion we talked about earlier.  And so HRSA wouldn’t have that same level of interest in dealing with 340B diversion.  They want to protect the manufacturer and they provide the facilities for that, they have the regulations in place for it, but the enforcement of it, they may not take as proactive a position as they would otherwise.

DirkOK.  So, backing up a little bit, forgetting about serialization, you would be using the same data you would be using to watch for diversion from foreign sources as you would for 340B, is that correct?

Kosta:  No, for this type of diversion you rely more on the manufacturer’s own proprietary data…chargeback data, the Medicaid claims data… It’s a little different.  The analytics are a little different than for physician office diversion.


Kosta:  The methodology and the data differ between the two.  340B analysis is more about sales pattern shifts between accounts, eligibility periods and double discounting. A differentiator between the types of diversion is the risks.  In the first case we have the situation where the product can spoil, coming from a foreign market, it can become dangerous to the patient, if it is counterfeit it can endanger the patient’s life, so the risk level is significantly different.  With 340B [diversion], this is product that moves typically through the normal supply chain—the legitimate supply chain—until it reaches that provider who decides to divert, but normally it is product that originates from a legitimate source.  Now, there is the potential of the product not meeting the cold chain requirements perhaps, once it moves from the last provider who is diverting to the next provider or patient, but it’s a lower risk than some product that comes all the way from Europe, for example, where it might be in transit for days.

DirkWhat is the implication for the manufacturer between these two types of diversion?

Kosta:  In the case of physician office diversion the best case scenario is if the diverted product was legitimate and handled safely. In that case, the risk is limited to a financial loss that would be equal to the price difference between the market of product origin and the U.S. market. When the product is counterfeit, the manufacturer incurs a 100% financial loss.  But, the greatest risk to the manufacturer is patient safety. The consequences for that include potential litigation for harming the patient, and harm to the reputation and goodwill of the drug and the company.

For 340B diversion, the product is assumed to be legitimate and safe, in which case, the risk to the manufacturer is purely financial and can be measured based on the difference between the 340B price and the next higher tier of discount the provider would qualify for.

For a diverter to profit, diverted drugs must have a significant price and/or high-volume. So, it’s not a surprise that many of the diverted drugs are biologics, although oral drugs have also been diverted.  One other characteristic of physician office diversion is that the product must be dispensed at the physician office. That’s a given—otherwise, the doctors wouldn’t be able to profit. So it must be [a drug] that qualifies for reimbursement through the medical benefit, a buy-and-bill drug.

Dirk:  How are suspected diverters identified in the data?

Kosta:  Suspects are identified by discrepancies in their purchasing and reimbursement patterns. But because the data for some providers is not perfect, the analysis generates false-negatives. These diverters would have to be identified through other means.  And that is where the profiling comes in.  The idea of the profiling is to provide a safety net for the analysis of those accounts that the data cannot support, and to proactively identify customers that would likely divert given the opportunity (high-risk customers). The profiling is based on the shared characteristics of approximately 500 doctors, most of them identified by the FDA, who are known to be diverters

The profiling is your safety net, where you don’t have enough evidence you’re going for the next best thing.  If you are not a suspect, are you a likely diverter?  And, again, you are not accused of anything, you are just a good target for an educational campaign by the manufacturer to warn you of the risks of diversion.

Dirk In the profiling process you use, are you using it just to narrow down who you would even look for?

Kosta:  Yes, exactly.  Depending on the drug, the pharmaceutical manufacturer’s customer-base could be very large, and that means a lot of resources to do an educational program.  The program can be as simple as an email or a letter, or a face-to-face conversation with your sales rep.  It’s up to the company how they choose to deliver that message.  Companies do both, but it’s different when you have a ratio of 10 high-risk accounts per rep and when you have 50 accounts per rep.  And typically a rep has 200 accounts to target, so if you treated everybody as a high-risk account and everybody is subject to this educational program, you need a lot of resources for that.  Whereas if we took out those accounts that are not likely to divert, then we go to where better than 90% of the problem is.

The profiling analyzed characteristics both from the doctor standpoint and the practice standpoint. The key theme that emerged from the analysis is that diversion happens mainly where the provider has the discretion on the “expense” and the “revenue” side of the business, being able to choose where to source the drugs, and being able to extract the profits. And this is mostly possible in small, independent practices.  Themes also emerged around age, gender, and cultural characteristics of the doctors. The results of the profiling are made available through PSI (Pharmaceutical Security Institute) to its members.

DirkAnd it’s not just doctors, I assume, it would be anyone who is in control of managing the acquisition [and payment of drugs].

Kosta:  Exactly, and it’s worth mentioning that not every physician practice is owned by a physician.  Some practices are owned by individuals who are not care providers themselves, and whose sole motive would be profiting. And we found a few of those, actually.  These practices are run by individuals who may own or manage them with a lot of discretion where to buy drugs and the reimbursement as well.  We found some practices where the office manager was in control of purchasing a diverted drug.

Subsequently, when I looked at the FDA list of Botox diverters, I found facilities that aren’t even medical offices, per se. They are pseudo-medical, cosmetic facilities with an affiliation to a doctor, perhaps.  And this happens a lot when you have a dermatological kind of product involved.

DirkThe last time we met, you told me about an experiment you did where you used the [initial] list of [88] doctors [who were found to have acquired improperly imported Avastin] that the FDA published to check your methods to see if you would detect the same level of diversion.

Kosta:  Yes. When I did the analysis for a couple of oncology products, the list of “suspects” for those two products, included some of the same doctors that appeared in the [FDA’s] Avastin list. This was a sort of validation that the data and the analysis work. Naturally, if a doctor has chosen to go rogue and is able to profit from one diverted drug, he or she may try to profit from another drug in the same therapeutic category as well, to the extent that it is available for purchase…

DirkThe same kind of doctors, or the same doctors?

Kosta:  The same doctors!  A few of the same names appeared in both the FDA list and the list of “suspects” for the products that I was analyzing.

DirkThat’s incredible and it really shows how analytics applied to existing data can be used to expose behaviors that would otherwise remain hidden.  Hopefully we’ll see more of this kind monitoring in the future.  Thank you for providing this glimpse into your work.

If you would like to learn more about Kosta’s work, check out his website at


5 thoughts on “Detecting Drug Diversion With Data Analytics–An RxTrace Interview With Analytics Expert Kosta Tzavaras”

  1. Dirk:
    Excellent interview with Mr. Kosta and the write-up provided excellent information in easy understandable approach. Nicely done.

    1. Marv,
      Thanks. Kosta is a very interesting person to chat with. The interview-type essay was an idea that I have been thinking about for quite a while now and I thought he would be the perfect person to start with. I have another interview recorded with another expert in a different field. I hope to have that one ready to publish sometime soon and I hope to create others whenever I can.


  2. Hello Dirk. How does big data come into play from the social networks? Social networks provide a substantial conduit for illegal drug trafficking.

    1. Ishmael,
      Thanks for your comment. The sale of prescription drugs on social networks is not legal so the kind of data analytics this essay discusses do not apply to that type of activity. Once again, there is a big difference between technologies and approaches that are designed to keep the legitimate supply chain from being invaded by criminals, and those that are aimed at detecting and cleaning up channels outside of the legitimate one. Rarely do technologies aimed at the former work with the latter and vice versa.

  3. Dirk,
    A very interesting discussion that has very similar “grey-market” parallels in the electronics industry. By chance, have you had any discussions with Kosta regarding the implications to the problems and/or his product offerings he anticipates with the advent of unit level traceability?

Comments are closed.