Amanda DiTrolio
💡 HTN | Community Brain Trust | 10/17
🧵TOP THREADS OF THE WEEK
In case you missed them, here are highlights of a few interesting conversations from different channels:
Threads included below:
- Assessing healthcare deserts and limited healthcare resources
- The defensibility of AI models: building from scratch vs. using GPT wrappers
- Best practices for outreach to reduce no-show rates
- Billing for remote patient monitoring in FQHCs
- Improving Google Drive system for scaling organizations
1. Assessing healthcare deserts and limited healthcare resources
Q: QQ - How can I take an Excel list of zip codes (or cities/counties) and objectively assess whether that zip code is part of a healthcare desert or has limited healthcare resources / low socioeconomic status?
- For example, would just want a column next the zip code/city/county that says “yes/no” whether it is an area with limited healthcare access by some definition (ie could be as simple as far away from an academic medical center)
- Does anyone know of a list of zips/cities/counties that fall into the category of healthcare desert so I could just plug that in and cross reference?
– Sameer Berry | via #random
Thread Summary: The crew discuss how to objectively assess whether a zip code is part of a healthcare desert or has limited healthcare resources. They share resources such as the HRSA Shortage Area Locator, the Kaiser Family Foundation's Primary Care Shortage Map, and the Rural Health Information Hub's Rural Health Disparities Map.
Top Response:
Dhonam: @Sameer Berry I put your reply to me in chatgpt and it says
Yes, there is an Excel file/database that has a list of zip/county/cities for the HRSA Shortage Area Locator, the Kaiser Family Foundation's Primary Care Shortage Map, and the Rural Health Information Hub's Rural Health Disparities Map. It is called the HRSA Rural Health Data Files.
This Excel file contains a list of all zip codes, counties, and cities in the United States, along with their corresponding HRSA Rural Health Shortage Area (HPSA) status, Kaiser Family Foundation Primary Care Shortage Area (PCSA) status, and Rural Health Information Hub Rural Health Disparity Index (RHDI).
To download the HRSA Rural Health Data Files, go to the following website:
https://data.hrsa.gov/data/download
Click on the link for the Non-Metro Counties (Micropolitan and non-core based counties) and Eligible Census Tracts in Metropolitan Counties Excel file.
This Excel file contains a list of all zip codes, counties, and cities in the United States, along with their corresponding HRSA Rural Health Shortage Area (HPSA) status, Kaiser Family Foundation Primary Care Shortage Area (PCSA) status, and Rural Health Information Hub Rural Health Disparity Index (RHDI).
You can use this Excel file to create a database of zip/county/cities for the HRSA Shortage Area Locator, the Kaiser Family Foundation's Primary Care Shortage Map, and the Rural Health Information Hub's Rural Health Disparities Map. This will allow you to easily search for and identify zip codes, counties, and cities that meet your criteria.
Check out the full HTN Slack convo here!
2. The defensibility of AI models: building from scratch vs. using GPT wrappers
Q: Curious to hear people's thoughts when an AI company just builds their model as a wrapper on GPT and how defensible it is. Is it even possible or worthy to build your own model from scratch?
– Jinghan Hao | via #topic-ai-ml
Thread Summary: HTNers discuss the defensibility of AI models built from scratch versus using GPT wrappers. While some argue that business model defensibility comes from leveraging the model at the right time and using it in novel ways, others believe that differentiation in the market comes from better workflow integration, distribution, and user experience. The conversation also touches on the role of proprietary data in creating distinctive models and the challenges of automating clinical notes using public models.
Top Response:
Vince Hartman: The challenge isn't the direct transcription of a dictation using a transformer model - you are right that those public models are very good and at 99% accuracy. The real challenge is automating the clinical note itself. It's actually REALLY hard to take a transcription and create a really good clinical note from it. You need top NLP scientist to pull that off. And you can't do it at a level that is physician-quality with public models, even GPT-4. The notes will be too templated and have too much hallucinated. You would need proprietary data and methods/workflows to do it. And the physician themselves does an assessment/plan portion in the SOAP note - ideally you want the AI to do portions of the assessment automatically and not require the physician to dictate that portion themselves, which is also really hard. It's just a much harder problem than people realize.
Check out the full HTN Slack convo here!
3. Best practices for outreach to reduce no-show rates
Q: hey everyone-- anyone seen any good research or reports on the protocols for outreach to reduce no-show rates? I know there is a lot of technology out there, but just curious to see the best available stuff out there on outreach protocols -- frequency, medium of outreach, etc.
– Dan Ferris | via #buildersask
Thread Summary: An interesting thread on research and best practices for reducing no-show rates in healthcare appointments. The conversation covers various methods of outreach, including calls, texts, and emails, as well as the importance of easy rescheduling and cultural considerations. Members share their experiences and insights, providing valuable information for improving appointment attendance.
Top Response:
Nick Neral: I used to work for a patient engagement company that sent millions of texts and emails every month. At a high level, you need two different methods of communication (ie text and email). You need one reminder at least a week out to remind patients who may have booked months in advance and totally forgot. Then you need about a 24 hour out and 1 hour out reminder. We developed a tool that would identify when patients scheduled outside their normal times. For example, if a patient was consistently having appointments at 3pm and then they asked for 10am, then that patient potentially needs flagged as a no show risk, among other factors.
As others typically note, the barrier to rescheduling/cancel/confirm has to be near zero. If a patient has to call or login to a portal with a password to cancel/reschedule then you’re going to experience no shows.
Check out the full HTN Slack convo here!
4. Billing for remote patient monitoring in FQHCs
Q: Hi All, wondering if and how anyone is billing for RPM in an FQHC. I keep meeting roadblocks trying to get out billers and coders to look at this for CGMs and being told we can’t do this at an FQHC because of how we code/ bill. Anyone out there have experience with this?
– Kayce Amelia Sol, CDCES | via #topic-remote-patient-monitoring
Thread Summary: HTNers discuss the challenges of billing for Remote Patient Monitoring (RPM) in Federally Qualified Health Centers (FQHCs). It is mentioned that FQHCs are currently not allowed to bill for RPM due to coding and billing restrictions. However, there is hope for change as the proposed 2024 Medicare Physician Fee Schedule suggests that FQHCs may be able to bill for RPM in the future. The final rule is expected to be issued in the coming months.
Top Response:
Matt Fisher: As far as I’m aware FQHCs aren’t allowed to bill for RPM because those codes aren’t recognized in the FQHC fee schedule. An FQHC can bill the CCM codes, but the rules are a little different and the reimbursement rate is lower I believe. For the why, it’s tied into what the base FQIC bundled rate includes. I seem to recall CMS having a document that explains what an FQHC can bill.
Yes, your comment is correct the proposed 2024 Medicare Physician Fee Schedule is suggesting that FQHCs would be able to bill RPM and RTM. However, that rule hasn’t been finalized yet, so not in stone as of now that it will actually happen. The final rule is often issued in the mid to late November timeframe.
Check out the full HTN Slack convo here!
5. Improving Google Drive system for scaling organizations
Q: Hey! Awell is scaling the # employees and our Google Drive is falling apart (aka it’s poorly set up). Anyone here that was able to build a good Google Drive system (that got stress tested when scaling) at their org and is willing to chat?
– Rik Renard | via #buildersask
Thread Summary: The group discuss strategies for building a robust Google Drive system that can handle the demands of scaling organizations. The main challenge identified is poor discoverability, and members share their experiences and solutions, including using folders, migrating to other platforms like Coda or Notion, and implementing enterprise search tools like Glean.
Top Response:
Brandon Ballinger: @Rik Renard - poor discoverability has been a longstanding problem, unfortunately. We tried a few strategies at Brex with mixed success:
1. Aggressively use folders. You’ll have to train/remind people at first, but this at least ensures there’s some information hierarchy, and therefore a way to find things.
2. Port documents to a different system, like Coda or Notion. This obviously requires a lot of upfront effort and you may end up with fragmentation if (when) the migration is incomplete. The nice payoff here is that those platforms don’t just solve discoverability, they’re more advanced as a whole. For example, Coda lets you create docs that are “programmable.”
Check out the full HTN Slack convo here!
🤖HTN KNOWLEDGE BOT
If you have your own question(s) to ask, don’t forget that a good place to start is our HTN Knowledge Bot. It’s our smart search tool that makes it easier to access the wisdom shared within the HTN powered by ChatGPT. You can log in and use it on the website (here) or see how to use it directly in Slack here.
Check out the example ask below!
Member Question:
What are good resources to collect SDOH data?
Knowledge Bot Answer:
There are several resources available to collect Social Determinants of Health (SDOH) data. Here are a few:
1. The Accountable Health Communities program by CMS: CMS has released a fact sheet on this program, which aims to screen populations for SDOH needs. It provides insights into the program's impact and the social needs identified. You can find more information in this HTN Weekly Health Tech Reads article.
2. NowPow: NowPow is a comprehensive offering in the SDOH space. It provides a platform for support staff to identify specific needs and access resources across various areas. You can learn more about NowPow in this Health Tech Nerds Slack conversation.
3. Data-driven clustering and geographic SDOH data: Some organizations use data-driven clustering techniques, such as random forests, on historical claims data to identify distinct clinical clusters. They then introduce geographic SDOH data at the zip or census tract level to identify correlations with avoidable hospitalizations and ED visits. This approach can help identify social determinants correlated with health outcomes. You can find more information in this Health Tech Nerds Slack conversation.
4. Community health worker programs: Engaging community health workers can be an effective way to collect SDOH data. These workers can connect with patients, identify their social needs, and provide appropriate support and resources. You can read about a pilot partnership between a clinic and BCBS MN that focused on connecting patients with community health workers in this Yahoo News article.
These resources provide different approaches and platforms to collect SDOH data. It's important to consider your specific needs and goals when choosing the most suitable resource for your organization.
✨HIDDEN GEMS
Here we highlight helpful resources from across the community:
- 6 Proven Strategies for Getting Patients to Engage With Your Digital Health Product Using Behavioral Science via Kristen Berman – A helpful article on best practices to improve patient engagement strategies from the team at Irrational Labs.
- Provider Data: How it Can Make or Break Your Value-based Care Strategy by Joe Mercado – A solid write up unpacking the impact of provider data on VBC strategy.
- Modern Care Delivery Vendor Library via Betty Chang – A super cool library of 150+ digital health point solutions cataloged by condition categories and customer channels.