The Impact of Doximity GPT on Physician Workflows

Physician social network app Doximity launched Doximity GPT, what I’ll now refer to as physicians’ copilot.

OpenAI’s Chat GPT launched almost a year ago to-date, democratizing generative AI technology to help us streamline mundane tasks (and more). I covered Chat GPT’s healthcare implications as soon as it launched. Read it here. Note this is the technology Doximity is using behind its new Doximity GPT.

In this article, I’ll dive into Doximity GPT, recap the current generative AI landscape, and explain how I see Doximity GPT playing out from a business and physician perspective.

Doximity GPT

Doximity GPT is a physicians’ copilot, designed to reduce admin burden. It can draft all types of physician letters including insurance appeals, patient education materials, durable medical equipment requests, and referrals. Doximity began experimenting with this new feature earlier this year through DocsGPT.com, with the intention to 'cut the scut' by streamlining the time-consuming admin tasks that physicians hate.

I’ve already experimented with Doximity GPT to write a referral letter, and I’ll confirm it saved me 15 minutes of admin time. Note though, that this feature doesn’t do clinical decision support yet.

Overall, Doximity is focused on improving the productivity of physicians:

• www.BestDoc.ai: a GPT chatbot helping physician users find sub-specialized care in their area. Since Doximity has detailed CVs on over 80% of U.S. physicians, the GPT can sift through all the data to filter physicians in the area based on expertise.

• Hospital Directories: Doximity is launching hospital directories so physicians can call specific labs or hospital departments to get updates on patients. This feature is live at 50 hospitals.

• Hold for Me: Doximity’s launched a hold assistant, allowing physicians to skip the hold music and focus on their work. Once someone picks up, the physician will rejoin the call.

Also, let’s not forget that Doximity acquired physician scheduling platform Amion in 2022 for $54 million—they’re bound to launch some streamlining feature within the scheduling realm.

Streamlining Workflows with AI

Doximity GPT represents the latest GPT launch as the healthcare industry rapidly embraces generative AI. This transformative technology has not only spurred the launch of new AI-backed workflow services (as seen with Doximity GPT) but has also fostered strategic partnerships between tech companies and health systems. Below I recap the most compelling AI-backed clinical and non-clinical workflow services as well as noteworthy partnerships.

Several startups and existing companies are enhancing clinical and non-clinical workflows. New startups like Denyify and dotphrase.ai are testing AI solutions to simplify claims and enhance dictation processes in healthcare. Established startups are categorized into clinical decision-making (Glass Health), copilots (Nabla, Suki, Doximity), and note generation (Microsoft’s Nuance, Ambience). Glass Health's Glass AI and Microsoft’s Nuance Dragon Ambient eXperience (DAX) stand out for their innovative approaches in clinical support and automated documentation.

Strategic partnerships are also shaping the industry:

• Microsoft and Epic are collaborating to develop and integrate generative AI into Epic’s EHRs.

• HCA Healthcare and Augmedix announced a partnership, with HCA investing $12 million in the digital clinical documentation company. Blake Madden covered the details here.

• The University of Kansas Health System and Abridge are partnering to bring generative AI-powered medical documentation to over 1,500 practicing physicians.

• Abridge and Epic announced a partnership to integrate Abridge’s ambient note-taking technology into the EHR. Emory Healthcare will be one of the first to use the Epic-integrated tech over the next three years.

• 3M Health Information Systems and Amazon Web Services (AWS) are working together to develop ambient clinical documentation and virtual assistant solutions using AWS’s generative AI.

• Nabla and Kaiser Permanente are working together to roll out Nabla’s ambient AI clinical scribe for 10,000 KP physicians.

The above developments since ChatGPT's launch highlight the significant momentum in the generative AI healthcare sector. The space is just heating up, too. Generative AI in healthcare is predicted to grow at a compound annual growth rate of 85% through 2027, reaching a total market size of $22 billion—the fastest growth compared to all other industries using generative AI.

Dashevsky’s Dissection

I imagine Doximity GPT being my copilot: up on the screen next to my EHR, ready to take care of my admin tasks: from referral letters to patient education materials. Hours could theoretically be saved from my admin time.

From Doximity’s perspective, they’re about to receive a ton of data on what tasks physicians need the most support with. If thousands of physicians are typing in a similar prompt, such as, “write me a letter of medical necessity for a patient to get a walker,” Doximity’s GPT model may become even stronger. Doximity could even create more features surrounding medical necessity letters specifically for getting patients durable medical equipment!

There’s also great potential for Doximity GPT to tap into other admin areas that eat into physicians’ time such as responding to patient messages, pre-charting, and discharge summaries.

• Patient messages: Doximity GPT could read patient messages or results and produce one-click suggested prompts to respond. Much like how gmail or outlook make suggestions while you type, Doximity GPT could make these suggestions for responding to patient’s requests, questions, or lab results. It could also be used for triaging patient messages, passing urgent messages to clinical staff and non-urgent messages to non-clinical staff. Providence health system just experimented with this triaging using Azure OpenAI Service, leading to a 35% improvement in response time.

• Pre-charting: Doximity GPT could process and analyze clinical notes and results during the interim between visits, producing an interval summary to aid in pre-charting. This would save hours per week of work for physicians. I even tweeted about this fantasy a couple of months ago.

• Discharge Summaries: Doximity GPT could similarly process and analyze an entire hospital course, creating a succinct timeline and summary of the patient’s stay. The discharge summary is the most important document for outpatient providers when they follow up with patients post-hospitalization. However, discharge summaries are often fragmented, with important information left out. It doesn’t help that writing discharge summaries is time-consuming for physicians.

The above features are all fantasy at the moment because the key barrier to implementation is the “‘I’ word”… integration. Doximity GPT is a standalone platform, not integrated into any EHR—a significant barrier to implementation. But I don’t see it being far-fetched for Doximity’s services to integrate into EHRs, given all of their users are using an EHR daily. Integration would just be incredibly expensive.

However, there is an arms race to get generative AI services like Doximity GPT integrated into EHRs. Look no further than my graphic above demonstrating key partnerships between Epic and AI companies like Abridge. Epic themselves are even experimenting with their own GPT for many of the examples I provided above, including patient responses, handoff summaries, and up-to-date insights for physicians.

In summary, Doximity GPT is an innovative AI tool designed to reduce administrative workload for physicians, offering features like medical correspondence drafting and streamlined referral processes. However, integration with existing EHRs is a primary challenge for broader adoption.