Details

Review Date
11/08/2023
Purchase Date
Q3'23
Implementation Time
4-6 weeks
Product Still in Use
Yes
Purchase Amount
N/A
Intent to Renew
N/A
Sourced by

Product Rating

Product Overall
3.5
Use Case Fit
3.0
Ease of Use
4.5
API
N/A
Integrations
N/A
Support
4.5
Value
4.5

About the Reviewer

Purchasing Team
User
Implementation Team
Product Oversight

Reviewer Organization

Specialty Practice
Oncology

Reviewer Tech Stack

Mosaiq

Other Products Considered

N/A

Summary

  • Product Usage: The AI transcription product is used in an oncology practice to replace traditional dictation and improve efficiency, allowing practitioners to spend more time with patients and less time on dictation.

  • Strengths: The AIs ability to learn and improve with each use, increased efficiency, and improved patient interaction are key strengths, leaning towards eliminating the need for traditional transcriptionists.

  • Weaknesses: The tool sometimes pulls in irrelevant information if the patient goes off on a tangent, and has difficulty correctly identifying and spelling names.

  • Overall Judgment: Overall, the product has significantly improved efficiency by reducing the time spent on dictations but would have benefited from comparative analysis with other similar products before purchase.

Review

So today were chatting about DeepScribe and how its used at your company. Before we jump into that, could you give a brief overview of the company and your role there?

We are an oncology practice with a small number of physicians and nurse practitioners who use DeepScribe. We offer consultations for new cancer diagnoses along with radiation treatments and imaging services. We also provide continued follow-up visits to monitor cancer, maintain patient health, and manage treatment side effects and other medical conditions. My business partner and I serve as medical directors, as well as provide direct patient care on a day-to-day basis.

DeepScribe Response 12/06/2023

At DeepScribe, we love to partner with organizations to expand and specialize our product offering. This review is a reflection of what a co-development relationship looks like. In this case, we partnered with this organization to build our Radiation Oncology workflow.

What was the need that drove you to look for a solution like DeepScribe?

Previously, our transcription process relied on old-fashioned dictation without the use of voice-to-text programs. We would dictate using dictaphones or mobile phones and send the recordings to a team of transcriptionists for accurate transcriptions. We started looking for a new solution for three main reasons. First, the classic transcription devices and software we were using were nearing their end of life. Second, the dictation process was becoming increasingly laborious, and we wanted a more streamlined approach to improve efficiency. Last, we wanted to enhance our documentation of the work we do with the patients to enable higher-complexity billing and ensure a higher quality of patient care.

What were the key requirements you used to evaluate DeepScribe and other vendors in the space?

We chose DeepScribe without evaluating any other vendors. I discovered them and thought the product looked interesting, so we decided to look into their solution. They gave us a compelling presentation, and we went with them without considering other products.

Our requirements for DeepScribe were pretty basic. I told them what we were currently doing and wanted to replicate it. I wasnt looking to improve, just to replace our current process with something more efficient. They assured us that not only could they make it more efficient, but also more thorough. They said it would allow us to bill at a higher rate of complexity and provide documentation that would be more helpful in patient care.

How did you find the sales process with DeepScribe?

They presented their product to us, showing examples of how it is used in other practices. However, they hadnt worked with an oncology practice before, so initially they were hesitant to sell to us. But after discussing it further, they became more enthusiastic because it offered them an opportunity to enter the oncology space and expand their market.

They were concerned about underperforming, but I assured them that we would work with them to improve if that happened. They liked that we were willing to partner with them in that regard. In addition, we are a small brick-and-mortar oncology practice and not part of a large health system. We have eight practitioners who use the service, which makes us nimble and able to make quick decisions without many layers of management to navigate. This was attractive to them, since many of their clients are large multispecialty or primary care groups, or health systems that require going through a new technology committee for decision-making.

Did you do any pilots with them before you decided to move forward?

We did a few test examples, but it wasnt an official pilot program. We only did about six cases. Despite it not being perfect, we were pleased with the results. We expected that it would be a gradual process to get to a point where we could trust it to get things right, but we were surprised at how well it worked from the start.

How was the onboarding and setup process?

The implementation process took about four to six weeks from the time we decided to proceed. They had to prepare their systems and gather necessary data before we could go live. During the onboarding process, their key staff, who had previous experience as medical scribes, shadowed us while using their technology. They were able to quickly make any necessary adjustments. After going live, they came back to shadow us again for a short period with a group of patients. Overall, they were pleasant, easy to work with, and caught on quickly.

How are you using DeepScribe in your practice?

At 7 a.m., our schedule automatically gets imported into DeepScribe. Our staff members then make any necessary changes, because our EMR isnt always accurate about assigning tasks. When we enter the patients room, we open the app on our phone or tablet. The patients information is already there, so it only takes a few taps to access it. We start by introducing the situation and stating the patients name, then we begin speaking to the patient. The AI captures the essence of that conversation. Once we’ve completed the exam, we signal the AI to transcribe directly by saying start impression, and indicate when were finished by saying end impression. The final transcribed note is often better than what we actually said. We provide our impressions and recommendations, using specific wording to guide the AI. When were done, we simply hit the end button.

For the provider, its really simple. We dont have to take any notes, although we can if we want to. In the past, we used to dictate the whole letter, either with the patient in the room or after they left. Now everything is done through ambient listening. We receive an initial transcribed note, which is then reformatted and structured by a transcriptionist. If the note is already perfect, it can be sent out, but our transcriptionists still review the note. They insert any information from our system that wasn’t automatically included and ensure that names are spelled correctly and sentences are appropriate. Once its reviewed, they send it to us for approval.

Currently, we still need our own staff transcriptionists, but were hoping that the AI will improve enough to eliminate the need for that additional oversight. Were making progress in that direction. Over the past month, the AI has improved significantly. Every week it gets better. Previously, our nurse practitioners would review and scrub the transcription, but now they tell me theres nothing to scrub. They used to have to fix any errors before forwarding it to our inbox. Now, they simply pass it along because theres nothing to correct.

When DeepScribe was selling you on their product, did they indicate that you would still have to have a transcriptionist or scribe in the loop?

DeepScribe promised that their software would eliminate the need for transcriptionists and save us money. We chose to keep our transcriptionists while we’re implementing the new technology because we wanted to ensure the accuracy of the transcriptions and avoid the need to rehire if the software didnt work out. Were on track to eventually eliminate the need for transcriptionists, but were not there yet. We currently have eight different users who are using DeepScribe as intended. However, some other practitioners are not as comfortable with the software and have been turned off by early transcription mistakes. They continue to dictate their notes as usual, without using the ambient visit capture feature. We believe they may become more comfortable with the software as it improves.

Once the transcription is completed, the document is created in our EMR on our letterhead and queued for our approval. Once approved, they’re ready to be sent out to other physicians or facilities. Currently, we’re using DeepScribe for follow-up notes and history and physical examination notes, but we have plans to expand its use to procedure notes as well. We currently send the transcripts to our transcriptionists, and they enter them into our templates.

Does it do any of the medical coding for you?

The system was meant to provide ICD-10 codes, but it hasnt fully delivered on that promise. We need to push them on this issue. Currently, we only input the primary ICD-10 code with no secondary coding, but DeepScribe should be able to provide that coding for visits with higher levels of complexity.

What would you characterize as some of the strengths and weaknesses of DeepScribe?

The biggest strength of the AI is that it learns and improves with each use. It allows me to spend more time with patients and less time on dictation. It also has the unexpected benefit of improving patient interaction, as I provide a summary and explanation of their background and history during the visit that the patient can hear and contribute to, and the AIs ability to understand what the patient is here for is further enhanced when we vocalize it. We also vocalize the exam findings for the AI to capture while we’re examining the patient, where we used to have to write or dictate them. I appreciate that it allows me to add the impression and recommendations in front of the patient, reinforcing our plan. There’s still room for improvement, but we’re learning how to interact with the AI effectively.

As far as weaknesses, it may pull up random information if the patient goes off on a tangent. I may delete these random thoughts, but sometimes I leave them as a reminder for when I interact with the patient on the next visit. Additionally, names can be challenging for the AI, although it may be able to learn to spell common names in the future.

How reliable and stable has DeepScribe been?

The system is highly reliable and stable. There was a crash today, but that was the first time it has happened to us, and it seems to have fixed itself.

How has the support and account management been with DeepScribe?

Our administrator has regular meetings with DeepScribe and our transcriptionists to ensure smooth operations. They’re pretty responsive.

Do you feel like you made the right decision in going with DeepScribe?

In hindsight, it would have been better to assess multiple products, but I wasnt aware of any alternatives at the time. If I were to do it again, I would definitely research more.

The biggest improvement for me has been efficiency. Previously, I would have a busy day and then be faced with a huge stack of charts to dictate. It was a challenge to remember all the details for each patient. But now I can stay on schedule throughout the day by putting away each chart after leaving the room. So the primary goal for me was to improve efficiency, and it has definitely been achieved. We are still working through some secondary issues, though.

For providers who used to dictate on the fly, there was no worry at the end of the day because they would simply dictate before leaving the exam room. The transcriptionists knew them well, their style, and the common things they would say, so dictating the whole thing took only a minute or so. But DeepScribe is saving them that amount of time. Initially, I felt like we were doing more work, since I spent more time reviewing everything and making adjustments. But now that Ive gained trust in the product, I can sign off on things without spending too much time on the back end.

Are there any areas you’d like to see growth in with DeepScribe?

We plan for the AI to eventually replace dictation for creating procedure notes and other internal notes, but it may take a year to fully implement that change. I’d like to see the AI integrated into the other processes that are currently done by our transcriptionists. We would like the AI to generate treatment notes and automatically import information from our EMR. Currently, our transcriptionists manually input necessary information into records and templates. The transcriptionists also handle procedure notes, on-treatment notes, and some letters of medical necessity, although we can use speech-to-text technology for the latter. The transcriptionist workload has decreased, but we’d like to get to the point that AI can handle all of those duties.

What advice do you have for someone who is making a decision about this type of product right now?

Initially, our people were frustrated and wanted to abandon the system. The transcriptionists found it a waste of time. However, after discussing their concerns and explaining how the AI-generated transcripts could save time by providing already organized text, they became more receptive. The initial pushback was primarily due the learning curve. Any potential buyers should be prepared for pushback from their own staff.

DeepScribe Response 12/06/2023

At DeepScribe, we love to partner with organizations to expand and specialize our product offering. This review is a reflection of what a co-development relationship looks like. In this case, we partnered with this organization to build our Radiation Oncology workflow.