The health industry is moving towards a future of data abundance. But while the world becomes abundant with data, we still face the problem of harnessing it to make health decisions.
- Data is locked away in electronic health record systems
- Drug companies struggle to build dynamic health-economic models from clinical data
- Insurance companies struggle to fund speciality drugs with the highest cost and least predictable quality of life outcomes
I find it useful to frame this as an economics problem. In a marketplace with buyers and sellers, the price of a good is used to coordinate an outcome that raises the status quo for both parties.
Healthcare is an unusual marketplace because someone else pays for goods on behalf of a buyer. In order to make a successful transaction, this entity needs to infer the buyer’s preferences to select the good which best satisfies their needs. In healthcare, buyer preferences are mostly unknown. The only known preference is not being sick. But the dynamics of this model change once treatment begins. Since buying decisions in healthcare happen on a recurring basis over time, a payer can retroactively refine their understanding of the buyer’s preferences once the patient has begun to notice the effects of a drug.
Healthcare expense is independent of outcomes
In industry parlance, the challenge of collecting data to measure treatment is referred to as Outcomes Research. The case below will help explain why this is important.
Consider a patient diagnosed with breast cancer.
- Oncologist creates a prescription for Trastuzumab, a drug (experts would call this a receptor antagonist) which targets the pathway responsible for cancer cell growth, and is administered via IV
Note: the doctor is prescribing Trastuzumab, an HER2 receptor antagonist, but they are not prescribing a drug from a specific manufacturer.
The prescription is sent from the Hospital System to the Pharmacy Benefits Manager (PBM) responsible for the patient’s insurance policy - the PBM is the middleman between the drug and insurance company who negotiates the cost of the drug and selects the specific drug to administer based on a generic prescription.
The Benefits Manager receives a document with the prescription information.
PBM looks up the prescription, Trastuzumab, in their drug formulary table, with the following questions:
- Are there substitute products?
- Are the substitutes cheaper?
- Are the substitute products equally effective?
One dimensional formulary table. It’s one dimensional because the drugs are measured on a single measure of effectiveness (and sorted into cost tiers, with the lowest cost on top)
- The PBM makes their formulary decision, and the selected drug for the Trastuzumab therapy class is sent to the patient’s hospital.
- Patient begins treatment.
Improving the healthcare model
Let’s assume the success rate is high, and the treatment is effective at reducing the spread of cancer. But clearly, this process omits possible valuable clinical data that could result in even higher success rates. The PBM, for example, knows how much a single treatment costs, but they don’t know how many times the treatment needs to be renewed for a patient with 100% compliance (i.e. taking their drugs everyday).
What if the formulary table reflected not only drug cost, but patient adherence to the treatment? Instead of cost based decisions, what if we could make value based decisions?
In this scenario, the formulary table might look a little different, with our new metrics, QALY (quality adjusted life years) and adherence, on the right.
Two-dimensional formulary table: in contrast to our formulary table above, here we measure outcomes in addition to cost, making it a two-dimensional table, with tiers based on a weighted combination of cost and quality.
In contrast to the cost model, our model assesses value through the marginal QALY increase for the patient. When comparing two drugs, the PBM would quantify the incremental cost to obtain an additional unit of a health outcome, like an increase in how many years that patient is expected to live pain-free.
In this scenario, criteria like copayment cost matter less: value is determined based on the ratio of costs to benefits. Thus, our formulary table looks different from the norm. Rather than the lowest-cost drugs being the preferred payer option, we might find that the highest-price drugs are now the most preferred since they have the lowest value-adjusted cost, measured by patient outcomes over the course of the entire disease. 
Everyone saves money
If the pharmaceutical industry moves towards a value-based formulary model, payers will save money.
A 2015 study showed that a formulary decision model based on drug effectiveness (such as QALY increase) and copayer cost resulted in 11% lower pharmacy costs after 12 months. This result was possible because more effective drugs were delivered to patients, and patients were willing to accept a higher copayment for better drugs.
If a patient doesn’t feel like their drugs are useful, they’ll refuse the copayment, or worse, they won’t even pick it up from their pharmacy. The insurance company could easily spend months paying in vein for a drug that a patient never even uses.
As an insurance or drug company, it’s hard to measure whether patients believe that their prescriptions are effective. You could easily figure out how many drugs were prescribed to a patient group or how many prescriptions were renewed. But it’s much harder to assess the subjective value that the individual patient assigns to a drug. Without access to this data, payers lack the information to assess things like net decrease in pain and quality adjusted life years.
If this remains the norm, payers will continue to prioritize drugs at the lowest net cost, while neglecting the net benefit for the patient.
This is a worse outcome for everyone: patients don’t adhere to their prescriptions and stay sick, insurance companies waste money on recurring prescriptions, and drug companies lose access to clinical data that can inform future drug development.
The situation is especially dire for payers: while the drug company will get paid no matter what, the insurance company remains in the dark about which treatments are worth paying for.
Outcomes data matters to drug companies, but it really matters to insurance companies.
It’s clear that there’s a problem which can be solved by aggregating patient data to show real world drug effectiveness. But open questions remain. For example, which therapy classes can outcomes data deliver the most value to?
As we work towards developing a product to help address these issues, it’s worth returning to our economics analogy. In a simple system, a buyer and a seller make a transaction. But when developing a product for a user whose cost is subsidized by a third-party, we need to align our user experience to both entities: the user and the third-party payer.
We’re capable of using our software to deliver better drugs to patients across America. But to achieve this, we need to align ourselves with the gatekeepers who keep the healthcare system functioning. Only by appealing to their interests can we properly achieve our goal of delivering a superior patient experience.
Next week, my goal is to continue looking for folks who work as liaisons to the insurance companies. In industry parlance, these are called “Health Outcome Liaisons,” or sometimes “HEOR Researcher,” and “Medical Science Liaison.” If there’s anyone in your network who might know someone in or around these roles, I’d be grateful for an introduction.
By appealing to the gatekeepers in the value chain, we have the opportunity to make a real impact in the healthcare industry. We just need to figure out how to connect the dots between the buyer and seller.
 In response to a more expensive drug, the plan-sponsor may increase copayment cost. Given the increased effectiveness of the drug, we assume that patients are unlikely to wince at their added cost burden.