Provider Directories, Physician data, and Ribbon Health
Get Out-Of-Pocket in your email
Onboarding Clinicians to Tech and Startups
Ribbon Health is an API data platform for comprehensive and real-time information on providers, networks, cost, and quality of care. The company initially started with provider directories and has expanded to add more granular information about providers including where they practice, insurances they accept, patients they see, and more. As the company grows the biggest obstacles they’ll face will be data commoditization and convincing customers that having accurate provider data is a worthwhile investment.
This is a sponsored post - you can read more about my rules/thoughts on sponsored posts here. If you’re interested in having a sponsored post done, email firstname.lastname@example.org.
Company Name - Ribbon Health
Ribbon Health is creating the infrastructure for companies to build healthcare solutions on top of their provider data. Both founders are named Nate which I can assume makes all hands meetings very complicated.
They actually started as care navigation for employers called HealthWiz. But since that name was more appropriate for a urinalysis company, they changed the name and business model. They realized the internal tooling they had built to improve their care navigation system was really valuable and instead decided to turn that into a product. Now they give the greatest gift of all time to other companies: accurate information about providers. And that’s why they call it Ribbon.
Or it’s just SEO-friendly, idk.
The company recently announced their Series B of $43.5M from General Catalyst, Andreessen Horowitz (a16z), BoxGroup, Rock Health, and more.
What pain points do they solve?
I think many of you have probably experienced this pain point at least once. Finding a physician that takes your insurance + knowing they’re the right doctor for your condition.
You want to find a doctor in-network, so you go to the directory of your insurance company or look on a doctor’s website to see if they take your insurance. Ideally you’d like to see a physician that’s an expert in your specific condition, but you know you’re asking too much. You call to make an appointment and find out that the number is no longer in service. Or you actually make the appointment and show up and the office isn’t actually there anymore, they relocated to the other side of town and just haven’t updated their information. And then you go across town to see the doctor and they say “like we told you the last 3 times we won’t surgically create pockets on your body because it would be convenient for the shower”. And you get a $350 bill 5 months later because they’re out-of-network now and didn’t update their page. Hasn’t this all happened to us at least once?
Having up to date provider directories and info about the cost/quality/expertise of the provider is still a massive problem in healthcare today. Honestly providers have never really had to worry about acquiring patients, so they don’t even keep the right phone numbers on their websites let alone any of the other information. The onus is largely on health plans to wrangle this data.
However, every insurer has different formats to capture provider information, different portals they require physicians to go into to update that data, and payers don’t even have good internal processes to make sure that the updated provider directory data they capture is actually getting updated across the board. Considering each provider might be contracted with 15-20+ health plans and physicians might work some days with one health system in a location and other days with a different physician group or facility (e.g. a dialysis center), this gets really complicated really quickly. Each health plan tries calling to get updated information from the providers (which is redundant across the health plans) while they’re also calling the providers about a million other things, too. As you can imagine, everyone gets really frustrated.
The result is bad directories. There are lots of different estimates as to how bad these directories are. CMS did an audit and found close to half are inaccurate. General consensus across studies that look at inaccuracies seems to be that in some capacity they’re inaccurate 25-50% of the time and certain specialties like mental health providers are worse than others. Ribbon says its own customers usually have <50% accuracy in their directories.
Many directories start by pulling from some of the central registries that are supposed to have this information like NPPES or the AMA masterfile. However, a lot of the information is conflicting between them and have relatively low accuracy rates.
Can you imagine how nutty it would be if you used Yelp, showed up at a restaurant, and a quarter of the time the restaurant just didn’t exist there anymore? I would be extremely mildly inconvenienced.
In healthcare the stakes are much higher though since it can be really hard to get appointments, and if you pick wrong you end up with an out-of-network charge. But when it’s really cumbersome for health plans to get up-to-date information from providers and the providers have no real incentive to update it, how do you solve this problem?
But when you step back and really think about it - the fact that we don’t even have the right phone numbers and locations for our providers is really setting the bar low. We have no idea which provider we should go to for our problem and instead we end up defaulting to the brand name institution they’re a part of or get recommendations from friends. If I tear my ACL, I want to see physicians that have seen a ton of patients with torn ACLs. But if I can’t even find a working phone number, how can I expect to find more info about the physician?
What does the company do?
This is where Ribbon Health steps in. Their suite of products is about getting more granular information about physicians, where they practice, the types of patients they see, and their cost/quality.
Their initial product is building more accurate provider directories that can then be accessed via API and used by other companies in their applications. The directory also maps the relationship between physicians and the healthcare organizations they work at, whether it’s hospitals, facilities like imaging centers, etc. Ribbon Health claims to improve directory accuracy rates for customers by >80%. And they have this very nice, clean documentation page about their API which EVERY COMPANY SHOULD HAVE calm down Nikhil calm down.
The key here is using a combination of data + services + customer workflows to get a much more accurate directory. Ribbon uses 1000+ different data sources (scraping open web, data partnerships, claims data, etc.). Apparently there’s so much conflicting data out there about physicians that Ribbon gets about 50 different phone numbers per physician through the total dataset. Unless you are a character in Breaking Bad you probably don’t have 50 unique phone numbers.
The result is to build a machine learning model that can spit out a confidence score that determines the likelihood a physician’s contact information is correct. Each of these data sources have some clues that help inform whether or not the provider is practicing at a particular address, which phone number the provider can be reached at, etc. There’s a lot of noise in these datasets that need to get parsed through to build the confidence score.
Ribbon does two key things to improve the accuracy. The first is having a two-way data exchange with customers. For example if a company is using Ribbon data to power a downstream referral to a specialist, then when the appointment is confirmed Ribbon also receives data that confirms the specialist is at the location. By having customers contribute data back into the provider directory, it improves the accuracy of the directory for the other customers.
The second thing Ribbon does is use its own call center to reach out to providers to confirm their information. With the confidence score in hand, it makes it much easier to create outreach lists based on priority + missing information. The manually validated data then becomes an additional source for Ribbon’s machine learning models to continually improve how it determines the likelihood a provider is practicing at a given location or can be reached at a given phone number.
And with that, Ribbon has a provider directory! This includes the provider information, location directory, and insurance network information to see which insurers they take at a plan level.
But now that the company has a core product of essentially mapping the universe of healthcare providers and their unique identifiers + hospital affiliations, Ribbon can start stacking other granular data about institutions and the individual physicians.
Ribbon cost and quality - Ribbon connects cost data that shows how much a given provider tends to charge relative to others in the area and gives upfront cost estimates. They can also bring in quality data, patient ratings, and the physician’s Tik Tok account (kidding...I think). They analyze third-party claims data from 300M+ lives, group clinical episodes claims together to collect all of the relevant ones for a specific condition, treatment, or event. Then the metrics are risk-adjusted to account for differences in patient panels then are weighted and combined to form aggregate outcomes quality and cost estimate scores.
Provider Focus Areas - Ribbon recently launched Provider Focus Areas which can help figure out the kinds of patients a given physician sees. Do they see patients of a specific demographic type? Do they see lots of patients with the same condition I have? Etc. They group similar terms to a common ontology (e.g. atopic dermatitis and eczema get mapped to the same concept) and then cluster procedure/diagnosis codes associated with those concepts to spit out a more understandable “focus area” on the front-end.
The key is that it’s very difficult to layer these datasets, cost estimates, quality scores, and focus areas without a clean mapping of physicians. I talked in the past about how different patient data brokers have to create tokens based on different patient attributes to “guess” whether it’s the same patient across datasets. Ribbon has built a similar identity scaffolding but for providers, which makes mapping third-party data to different healthcare organizations and individual physicians much easier.
What is the business model and who is the end user?
The company charges a subscription fee for access to different data modules (i.e. directory only, directory and network, etc). The data access is unlimited (it wouldn’t be great if you ran out of credits right when a patient needed to schedule an appointment).
This has lots of different company use cases:
- Health plans that need to keep up to date provider directories because they’ll get penalized if they don’t + want to guide their patients to more cost-effective care settings
- Primary care providers that need to guide patients to either virtual or in-person care when necessary based on the activity + cost effectiveness + facility (physical therapy, imaging at an independent imaging center, etc.)
- Referral management and care coordination companies that want to send a patient to the right physician that’s an expert in the condition the patient has. And...also that the physician is still practicing.
- Health insurance enrollment companies that want to show patients what kinds of networks the different carriers have
- A dating app exclusively for physicians in-network with Blue Cross plans
- Literally any application that requires looking up a doctor. You can see more of these use cases here.
Interestingly the company has maintained a relatively even 1/3rd revenue split from health plans, providers, and patient facing use cases even as they’ve grown. This might concentrate on a specific use case in the future, but for now the core dataset seems to be useful for a lot of different entities.
You can see all of Ribbon Health’s roles here.
A lot of companies I’ve worked at or worked with have tried to better understand where physicians practice and what those physicians are experts in. It’ll never cease to amaze me that we have a healthcare system where MRI machines use cutting edge physics to see miniscule details inside my body non-invasively, but also when I book the appointment to get one I might show up at a bank accidentally because the imaging center changed locations and forgot to tell anyone.
I’m actually almost sad that this pain point is still large enough that a company like Ribbon needs to exist, but the reality is it does and Ribbon is taking a more data-driven approach to fixing it. Here are some of the things I like about Ribbon:
Just...generally being an API lol - I’ve talked so many times in the past about why it’s cool that new API-first companies have public and clean documentation that anyone can read (which helps me with research too). Provider directories in particular have generally been maintained in flat files that get sent to whoever uses them. Without access to real-time updates to changes to the directory, the data gets stale quickly and applications that use them suffer. Real-time data is a necessity here, and I’m glad the product is focused around an API.
Data network effects - I generally find the concept of data network effects to be a little hand-wavy and it gets a little hard for me to understand how much models improve with more data or more diverse data. However in Ribbon’s case it’s a bit more obvious - customer’s correcting or confirming the data feeds back to Ribbon’s model and improves the directory for everyone. More users, more chance of mistakes being caught or corrected, better data for all.
Redundant work reduction - The data network effects also have the added benefit of reducing redundant work. Why should providers constantly update their data for each individual payer when some are contracted with 250 different insurance plans? Why should each payer have a call center to confirm the info? This is partly doable because you can set up the exchange of this data via APIs instead of stagnant flat files. By actually getting hooked up into the workflow and having data pass between both customer and Ribbon, the data can get cleaned and refreshed with less actual submission of data in separate places.
If we want to reduce admin costs in healthcare we’re going to need to eliminate redundant, non-competitive tasks between companies.
Provider scoring - I’ve been writing about how we don’t have good quality measures for physicians for about 5 years now. It’s a core tenet of one of my investing theses. We still cannot answer “who is a good doctor for the condition that I have” because historically there hasn’t been a good system to actually map data from individual physicians, to the kinds of patients they see, to the outcomes of those patients. I’m very excited to see how Ribbon’s Provider Focus Area product develops and other cost/quality scores they tie together to create these maps of physicians and healthcare organizations.
As always, here are the questions I’d have for a company like Ribbon as it grows.
The last mile of data - Ribbon’s data accuracy is good, but it’s still not 100%. The core problem is that most providers don’t care about the part of the patient experience that involves navigating to them, so they’re never going to be really proactively making sure their data is up to date. I’d be curious how Ribbon gets from here to 100% accurate provider directory data. It might be that more providers differentiating on consumer experience start becoming more popular (e.g. One Medical, Forward, Tia, etc.). Or it could be that new channel partners and data partners help get their models to higher levels of accuracy.
Data commoditization and competition - One question is whether provider directory data becomes a commodity. In theory as the directories get better in any patient-facing application, the easier it is for a competitor to come replicate the correct data. Plus, new rules are forcing providers to make their directories available via API so they’ll be readily accessible to all. Ribbon knows this which is why the provider directory is the underpinning for their other products that connect different types of data together to provide very different insights into individual providers/physicians (quality scores, insight into their patient panel, etc.
The other related question is whether a company could approach this from a different angle, like third-party data wholesalers/brokers that focus on de-identified patient data. Currently it’s not their focus area, and would probably require a different product build especially around an accessible API that can be embedded in other applications.
Apathy - I think the biggest point of friction is honestly just companies not caring about whether they have a good provider directory, which is sad. Since payers sell to your employer and not the patient, they generally don’t care about your customer experience with bad provider directories. In fact, it looks better if it seems like they have lots of providers in their networks. when they actually don’t, which are sometimes called “ghost networks”.
This is why we try to resort to more penalties for insurers who have out-of-date directories or try to pull the ghost network stunt, but the reality is that the work required to fix these directories + the fact that it looks good for them to have a lot of doctors in-network usually outweighs the probability of getting hit with a penalty or class action lawsuit.
But if the work to get that data was much easier would they be more likely to ingest it? If more companies are trending towards consumer experience as a differentiator will they need to have this data? The answer is probably yes and it’s a bet that Ribbon is making.
All in all I’m happy that someone is tackling this problem because it’s unsexy yet important, so I can relate. But also because our inability to map this out properly causes so many shitty patient experiences downstream like accidentally seeing a physician who’s really out-of-network or getting sent to a different facility that’s way more expensive. Understanding more info about the physicians we see should become table stakes in the future, and hopefully with companies like Ribbon we’ll get there.
Nikhil aka. “damn even physicians are giving me fake phone numbers smh”
If you’re enjoying the newsletter, do me a solid and shoot this over to a friend or healthcare slack channel and tell them to sign up. The line between unemployment and founder of a startup is traction and whether your parents believe you have a job.