FAIR Health

Today we’re talking about Fair Health and how it’s used at your company. Could you give us a brief overview of the company and your role there?

We are a primary ACA payer. We’ve actually developed all our own tech, including our claims system and backend tools for handling provider data and reaching out to our members. I lead the team that deals with the external side of our health system contracts. We craft new networks, identify new markets, new product lines. Over two years, we identify new markets and contract with major health systems. Another part of our team takes care of these markets long-term - handling provider disagreements, sorting out financial settlements, and even building up partnerships and more in-depth product lines.

Could you provide an overview of how Fair Health works and its application in your workflow?

From what I grasp, Fair Health is a nonprofit. Its main focus seems to be empowering consumers - at least that’s my take, but they’ve also put together and published information on healthcare prices and costs in various markets. On our end as a payer, we’re interested in this information because we use it in dispute settlement with out-of-network providers for what we call post-service contracts. We’ll use Fair Health tools to guide our negotiation strategy when there’s a significant financial dispute.

We also tap into this tool for navigating the diverse ways different states handle arbitration. So, let’s say provider X and payer Y are at odds over a specific payment. There’s a set procedure for arbitration, right? When payer and provider can’t sort it out one-on-one, it leads to either binding arbitration or mediation. Now, this isn’t set in stone, but in many instances, the regulatory entity in a given state looks to Fair Health for guidance on what’s a reasonable offer. So, it’s a resource we work with internally, but it also plays a role in the decisions we have to follow when these cases get escalated.

How long have you been using Fair Health? How does procurement work?

We get access to Fair Health, I believe, through a purchase or some sort of access deal. Occasionally, I’ll hop onto their website to use it directly, but that’s maybe around 20% of the time. What we’ve done is integrate their data into our backend systems. We can plug in a historical claim, and then it spits out various Fair Health rate percentiles, such as the 20th, 40th, and 80th percentiles. This tool crunches the numbers based on a specific CMS locality, geography, or market. It lines up what we actually paid for a past claim against the full charges, and the Fair Health percentile-wise guidelines, which fall between those numbers.

Who, within your organization, uses the product? Is it mainly your team or do other teams have workflows for which they use it too?

Our team definitely uses it for provider negotiations. We also use it to provide guidelines to other teams. Sometimes, we outsource the negotiation to an external vendor, They’ve got access to the same data as us.

I also use it with our legal team, when we need to understand risk. With claims, we submit our best and final offer and the provider does the same. The mediator picks a number and that becomes binding. So, we’ll sometimes try to test the waters and understand what’s within the scope of law, before putting forth our best and final offer.

So, the right way to understand Fair Health is to think of it as a dataset that you build internal tooling and applications around. Is that right?

Yes, it’s primarily used for the data. They scrape data – I think – from Medicare, from private claims from commercial insurers. The first value proposition is the size of the dataset. The second is that these are the numbers that the arbiters refer to. We have no control over it but we have to put ourselves in the mind of the judge to see where we can potentially achieve a better business outcome, so it’s useful from that perspective.

Could you unpack the nitty-gritty of the actual data sitting in that dataset? What kind of data sits in each individual row of the dataset? I’m also curious to get a handle on the dataset’s scope – how extensive is its coverage?

My primary issue with Fair Health’s dataset is that it’s based on providers’ charges. When one thinks about past paid claims, they come from a mixed bag of commercial payers, Medicare, and Medicaid. But since we’re working with data tied to what providers are charging, the numbers get pulled up by the higher end of the spectrum. When we’re looking at that 80th percentile, even though it might look like a fair figure, it can get seriously skewed by those providers who are billing egregious rates.

So, the data isn’t reflecting what’s paid. It’s more of a snapshot of what was charged before any negotiation took place.

Let me give you an example. Imagine that we had a claim for a cardiologist in a state like New York. This provider is in-network at other locations, but was out-of-network for the location where the care activity took place. The facility is in-network but the provider isn’t contracted at that specific facility so counts as out-of-network, and so we’re getting a claim dispute. After adjusting the claim a couple of times, the provider bills us $154K. We have tools we usually use for this sort of thing – based on which, we pay, let’s say, $23K. That’s around 15% of those original charges.

If you plug in those numbers into the tools we built on Fair Health Data, they’d tell us that the 80th percentile for this claim is $140K. That’s only 10K less than the original bill. The Fair Health number is based on some charges that are pulling the averages up. The provider whose claim I’m working with is itself a part of that data. So, this number is boosted by some pretty high figures. If you look at the CMS rates, it’s about 24 times the CMS rate. Note that the $23K we paid is already four times what CMS would pay for this procedure.

This is just one example where a state like New York’s laws might bind us, but they’re not payer-friendly. The provider can keep going - either billing a member or pushing the payer into arbitration, in order to get as close to the 80th percentile Fair Health number as possible.

If the provider charge data gets skewed upwards in this manner, what’s the value of being able to use Fair Health? Why not just go straight to CMS rates for a more reasonable assessment?

We use Fair Health to shape our offer strategy, so we can predict where the back-and-forth might end. There’s an escalation pathway, and I wouldn’t like it if it goes in that direction. Say we pay a claim and the providers aren’t happy. They can either bill the member or come to us, and it’s likely that they do both. We’ll make an offer in that situation that’s directionally towards the Fair Health number. It’s not that the CMS number isn’t helpful, but it’s not going to be what is referenced in dispute resolution. The mediator, or the judge-in-theory aren’t using the CMS rate to make a decision. We have to pay attention to that arena we’re in.

Understood. So, the regulators place a greater emphasis on the charge rates rather than the negotiated rates when making their determinations.

Yes, they utilize Fair Health. The process often follows a baseball-style approach, wherein both parties submit their best and final offers, and the victor emerges without an intermediary phase. So, the CMS rates prove helpful, particularly in my experience, for negotiating rates or assessing hospitals within the same geographical vicinity. This allows us to align them for comparison, taking into account differences among facility-specific CMS rates, and region-specific CMS rates. Our task involves scrutinizing these factors to gauge the relative expense of hospitals within the same rating area and geographic location.

Aside from provider charges, what other attributes are present in the dataset? Is it inclusive of provider locality and the specific CPT codes associated with the charges? Am I omitting any elements? Are there any inaccuracies?

There could potentially be various package options, although I’m not entirely aware of all that we’ve procured. Sometimes, I navigate the website’s tool designed for consumers, where one can pretend to be a healthcare consumer and look at their shoppable services and their projected costs. I’ve not personally encountered this data laid out in tabular form, but I presume that such information is accessible to us. For instance, you can explore categories like urgent care or hospital inpatient facility costs. The tool even accommodates inputting a Diagnosis Related Group (DRG), which is helpful. Moreover, they provide figures for Fair Health total treatment costs, so one can input specific procedures like a hip replacement or cataract surgery within a particular locality, and get a sense for costs. While my role doesn’t demand this level of granularity for individual cases, we do utilize the data for historical claims analysis.

Additionally, I’m curious about the extent of coverage within Fair Health’s dataset. Does it encompass all provider charges, or does it pertain to a specific subset?

I’ve observed comprehensive coverage, including areas like behavioral health. Virtually every conceivable type of healthcare service seems to be included. While I don’t possess the exact mechanics, I wonder what the computation process might be. When I input a claim ID into our tool, it promptly generates a figure. My assumption is that the system delves into the claim, scrutinizes its lines, and examines the codes billed for that specific claim. For instance, the example case in New York might entail five claim lines, involving emergency surgery with CPT code G for microsurgery, cranial scan, orbital cranial approach on a skull, and cranial lesion. It’s plausible that the system cross-references each code, retrieves the corresponding Fair Health numbers, and subsequently aggregates them at the 80th percentile.

So, you have the flexibility to adjust the percentile value according to your needs.

Actually, I don’t have that capability at the moment, but it’s a feature we could easily build in the tool we have. Percentiles don’t conform to a straightforward mathematical scaling, although having the option would be helpful. I’ve previously expressed interest in its implementation, as it offers an additional perspective. However, this aspect is currently managed by someone else.

I see. Theoretically, you could construct such functionality using the dataset.

Absolutely. On the website, you can find percentiles such as 40th and 60th. Personally, I find value in the median figures for this context.

Were you actively engaged in the decision-making process for procuring Fair Health’s services? Or did this arrangement come about through prior decisions?

To the best of my knowledge, nearly every payer uses this type of tool to some extent. While I can’t pinpoint our exact adoption date, it would surprise me if we hadn’t been utilizing it for a considerable period. The practice of referencing such data predates my involvement. I’m unsure about what TPAs (third-party administrators) might use it for, but I’m certain that self-insured employers and commercial payers would use it. Medicare could probably do this themselves because they have their own data.

Do you know what the pricing structure looks like?

I don’t have exact figures but I believe that information is publicly available. My guess is that it isn’t volume-based. Since the data is already present and we access it rather than generating it ourselves, it can’t be volume-based. I don’t know what frequency we use for getting new data, but I think it’s yearly. It might be a fixed dataset you buy or subscribe to every year. Fair Health operates as a nonprofit, aiming to assist consumers, although they likely generate revenue from payers who utilize their services.

Are there any other players in the market that offer similar datasets?

I’d compare them to Truven. Truven likely draws from a similar dataset, although it might be somewhat more focused or narrower. While it may not encompass as many data points, it zeroes in on commercial payer data.

These two tools cater to distinct use cases for us. I tend to rely on Fair Health when dealing with arbitration mandates in the northeastern states, because I’m bound by what those states use. As an aside, imagine a hypothetical scenario where I negotiate a $20 million settlement with an out-of-network provider, they would furnish a claim sheet with their calculated Fair Health rates. They say that Fair Health percentiles are higher than their requested amount. However, this serves as a tactical maneuver.

Conversely, when engaging in negotiations for in-network rates, we often turn to Truven’s commercial benchmark data. This dataset doesn’t rely on charges but rather on historical claims, specifically for in-network commercial agreements. It’s a more accurate directional reference encompassing a broader array of healthcare services pertinent to our context, excluding Medicare and encompassing ACA or smaller group plans.

So the Fair Health data set is useful for single case agreements or SCAs because it’s based on provider charges and is what arbitrators are using in making their decisions. Truven tends to be more beneficial for in-network because it captures the final settled price or the negotiated price. Switching gears, is there any sort of customer support or account management included with the dataset?

It’s just the dataset. If we have specific questions on it, we’d likely involve the data science team. They’re the ones equipped to handle that, I think.

Do you know where Fair Health gets their data from?

It’s probably somewhere on their website, though I’m not sure. I would assume they have some sort of partnership with CMS, and states might be required to de-identify and share data with them, since they seem to be using it to develop a healthcare transparency tool for consumers. We might actually be contributing to that effort, but I’m not certain about the specifics.

As we wrap up, what do you like most about the Fair Health dataset?

I like that it’s at least straightforward. It’s really expansive and helps during negotiations, especially when we’re dealing with less favorable terms. It helps us strategize and position ourselves effectively. Plus, the fact that we can seamlessly integrate it into our existing tools is a huge win. Honestly, if we had to manually look up information on their website every single time, I would not use it.

We discussed how it can be a bit frustrating that the dataset leans heavily towards provider charges. But that’s just the nature of their business use case. Is there anything else you dislike about the product?

It’d be really beneficial if they presented not just the provider charges, but also delved into other aspects. For instance, having a breakdown of in-network and out-of-network payments, and analyzing what different payers are contributing, like Medicare, Medicaid, commercial insurers, self-insured employers, and any other relevant categories. And if we could have those same percentile analyses applied to these segments, it would provide a more complete picture, compared to solely focusing on the charges themselves.

For someone who’s delving into their SCA strategy and considering data products or vendors to aid in this journey, what advice would you offer?

I’d suggest keeping a close eye on the state laws governing your operations. Understand your rapport with the relevant state regulator. Grasp the anticipated outcomes and the business dynamics between providers, regulators, and payers in each state. This landscape varies significantly across different states, with a few being payer-friendly.