Dear Advisor: Why is it so Hard to Integrate with EHRs?

(Or) Why I Don’t Recommend Integrating with an EHR, especially in your MVP

September 2022

Having helped a number of hospitals in CA improve the quality of care by making data-driven decisions in real time for the first time – and having mentored a number of start-ups in the HealthTech space as an advisory board member of CTIP and through other start-up communities, there’s one piece of advice that I keep sharing almost every time: please, please, please don’t center your HealthTech’s product/service strategy around the requirement of integrating with a hospital’s electronic health record (EHR, AKA electronic medical record or EMR). Here’s why.

Before we get started, please note:

  • I’m not affiliated with any of the organizations mentioned, unless specified otherwise.

  • I'll use the terms 3rd-party healthcare data vendor, Cerner, and Epic interchangeably.

Part 1: Why I Don’t Recommend Integrating with an EHR for your MVP?

The reasons are more nuanced than you’d think:

  1. It’s no surprise that as a start-up, when working with a hospital, you'll be considered a 3rd-party vendor. With all the privacy, compliance, and data governance regulations that private companies face, there are even more (understandably) regulations for working with healthcare data, including HIPPA and (potentially) HITECH compliance, as well as certifications for working with PII and research subjects.

      • Even if your cloud platform and team will meet these requirements, it’s up to the hospital’s executive, risk and legal (and possibly other) teams, to decide that they're willing to risk even a minuscule change of a data breach by partnering with you. That’s a high bar to clear!

  2. Suppose that you were able to clear that bar, what you’ll then find is that you’ll need support from not just the hospital’s IT Department for access to the EHR, but also time with a 3rd-party Database Analyst who understands where things are stored in the EHR. Wait, what?!

      • One thing that those that are not in healthcare don’t realize, is that 85% of hospitals use Epic or Cerner (as of 2019) as their database platforms of choice, e.g. rather than maintaining an in-house Postgres/Oracle/<your favorite database here>, 85% of hospitals pay Cerner and Epic to be their database vendor, for collecting, storing and maintaining healthcare-related events. Because Epic and Cerner have created proprietary (e.g. closed source) ways to access and query EHR data, you will need to reserve and pay for the time of a certified Epic/Cerner Analyst (affiliated with that hospital or a Cerner/Epic consultant hired by the hospital directly from Cerner/Epic) to help you understand the fields. Since they support the whole hospital, as you can imagine, they’re very, very busy.

      • Boris Tyukin further points out that each hospital’s implementation of its Cerner (e.g. healthcare data collection platform) will look different at each hospital (!) you talk to.

  3. Say you actually got on the Analysts’ and IT’s calendars. I’m sorry to say that one 30-minute meeting will most likely not answer all of your questions about what’s where and what the variable name 00000123 is (variable name for illustrative purposes only), how often it’s refreshed, etc. And what do you need to do, and whom do you need to meet with, to get usable data access? Maybe it will be another 3rd party vendor, to help you process HL7 or FHIR feeds from monitors.

      • As you can imagine, this may take many, many months. I imagine you're looking to develop and get feedback on your MVP in a much shorter time frame -- and may not have enough runway to wait for months for meetings to get scheduled (!).

      • You may be surprised to learn, that hospital employees, including clinicians, Biostatisticians and Data Scientists (and those in similar roles) also struggle with the same data access issues -- and they work at the hospital (!).

Because this access to healthcare data is so limited (understandably to protect the privacy and security of the patients), and depending on the Master Service Agreement (or similar) that Cerner had at each hospital, helped Oracle decide to buy Cerner for $30 billion, to have access to this data to then be able to build data-driven/AI-driven capabilities on top of it, for example, help hospitals reduce operational costs or for doctors and nurses to have an early-detection system [ref] for when a patient’s health begins to deteriorate.

Part 2: Advice on What to Try Instead

Just as I don’t recommend including AI in your MVP, I don't recommend you try to integrate with an EHR, even if you think it should be core to developing your product offering. Instead, here are some things to try:

  1. First, let’s take a step back and try to clearly understand – and summarize:

      • (1) What customer pain points(s) and outcomes are we trying to solve and move the needle on?

      • (2) What is the current standard of care for this? How much time and money (and headaches) are the patients spending on getting this resolved now?

      • (3) How do we expect to make an impact?

      • (4) How will we align incentives between the solution, customer pain points and product pricing?

      • That is, will we be offering a SaaS platform for early detection of a condition? A platform for a mindset or patient workflow shift that will help patients to do X? Something else entirely?

  2. Tackle the cold-start problem AKA what seems like a catch-22 here: you need healthcare data to see if you can prove out an MVP, but you need an MVP to pitch to hospitals to try to get access to healthcare data. Here’s some things to try:

      • Did you know about the anonymized dataset called MIMIC, with "over 40,000 patients admitted to intensive care units at the Beth Israel Deaconess Medical Center"? Can it help you prove out your MVP without AI?

          • Please note: you may need to complete HIPAA and research subjects training, along with other requirements to be able to access the data.

      • If not, you may be able to get access to healthcare data through partnerships.

          • Do you know about the Cedars-Sinai Accelerator that partners start-ups with clinicians and Cedars-Sinai? Or are you (or someone on your team) a clinician at a hospital that's contributing to Epic Cosmos?

          • If you’re a start-up that focuses on improving quality of care in pediatrics and needs FDA clearance for your device – do you know about The West Coast Consortium for Technology & Innovation in Pediatrics (CTIP, where I’m an advisory board member)?

          • Do you know about these other healthcare-focused accelerators, such as Verizon’s Forward for Good Accelerator (I mentored at Verizon’s Tech for Good Accelerator a few years back) -- or the newly launched Techstars Healthcare (where I'm mentor the LA chapter)?

  3. Instead of focusing on trying to collect all the information about our patients/customers (contrary to my blog post advising you to collect everything from the start), because that will cost a lot of time and money (as we discussed in Part 1 above), start with what’s mission-critical for the MVP.

      • What aspects of that outcome do you currently have information about – with and without – our patient/customer input? What’s absolutely necessary and required for you to know? What’s nice-to-have?

      • Is there a way that you can develop standalone HealthTech product(s) for your patients, that don't require any information from an EHR?

      • If instead, your product(s) is based on user-generated content, how can you incentivize (or bring value to) your patients for the data they can provide for you? e.g. How can you be the next MyFitnessPal for HealthTech? How and how often will you interact with them through your product?

      • Are there other 3rd party data sources you can connect to, that’s not an EHR, that can help you get the information you need? Such as, data from fitness trackers (such as Fitbit) or mobile apps (such as RunKeeper) via APIs? The trick here, as you can imagine, is to think outside-the-box and talk to the patient, to learn about what they’re using and why.

Part 3: Parting Advice

  • Because of the challenges shared with you above, HealthTech is one of the hardest – if not the hardest – industries to break into as a start-up. Once you do, if nothing else, you’ve got a moat.

  • Because clinicians are stretched thin, especially with COVID, they tend to prefer to see the finished full-featured product (e.g. waterfall vs iterative product development) before Beta-testing it. This is especially true for those clinicians working at Intensive Care Units (ICUs): one incorrect decision may stand between a patient stabilizing or dying [ref].

At the end of the day, we’re on the same mission: improving our customers’ and patients’ lives! Hope this advice helps you avoid the most common pitfalls of a HealthTech start-up.

You may also like:

References: