Nearly 50% of 12,400 physicians were able to take only a partial vacation of up to three weeks in the past 12 months. The statistics come from a recent survey by the American Medical Association (AMA).
5% of the respondents did not take a vacation at all. Doctors call their paid time off (PTO) ”pretend time off” meaning that even on vacation, they have to work. They fill in the data into EHRs and process emails. The absence of sufficient rest and additional working time at home contribute to burnout.
Dealing a lot with medical providers, I would say that introducing electronic health record systems (EHRs) and supplementing them with data analytics and AI powers help medical staff shrink paperwork and increase rapport with patients. As a result, the general productivity of the medical center increases.
Health and tech professionals prioritised face-to-face interactions between a doctor and a patient at the HIMSS conference in Orlando this spring. Traditionally, doctors type the symptoms while talking to a patient. It might be distracting for physicians and patients. As the appointment time is limited, doctors finish their records and transfer them to the EHRs after the shift. They call this “pyjama time”, because it is supposed to be the time to get ready to sleep. AI tools allow doctors to record their conversations with patients. As a result, they do not work extra hours, which leads to less stress and careful diagnoses.
Although physicians already use AI tools for medical diagnosis, I see a tendency for administrative solutions to comprise a more competitive market. Those solutions aid in automating paperwork. They do not influence the decision-making process, when, for example, doctors consult AI tools while compiling the treatment plan. The recent hype around generative AI adds to the AI-powered medical documentation upspring. Current leaders in the domain are:
- Microsoft Nuance
- AWS HealthScribe
- Google Med-PaLM
- Oracle Cerner Enviza
Updating the knowledge base
When a doctor records a conversation with a patient, the system transcribes the recording into text. Then, the large language model (LLM) transforms the text into a health record summary and directs it to the EHR. To generate relevant content, LLM needs regular learning and updating of the input data. For example, OpenAI ChatGPT learns from the requests of its users. In healthcare, sensitive patient data is not accessible to AI algorithms for security reasons. Here, the customized LLM training steps in. At Belitsoft, we have been working with two types of LLM training:
- LLM fine-tuning is adding specific terms and medical glossaries to the general knowledge base. It makes the system more oriented towards the medical domain.
- Retrieval-augmented training deals with dynamic data. It combines the possibilities of a search engine with a content generator.
AI technologies used for medical records
Among the AI technologies currently in demand among medical providers, I would mention speech recognition, optical character recognition (OCR), natural language processing (NLP), and machine learning.
Speech recognition
AI-powered solutions in medical centers use machine learning and NLP. Those technologies transform oral text into written notes. Such solutions usually support multiple languages and several simultaneous speakers. This feature is convenient when doctors share their expertise with colleagues. The system types the information into EHR in standardized terms. It guarantees the clearance of the input data and makes it easy to operate. AI tools help to interview a patient faster, as it takes only around 60 seconds to generate doctor’s notes and instructions.
Machine learning
Such administrative operations as medical billing, insurance claims, and reimbursement procedures take medical experts a lot of time. Medical coders have to classify each medical manipulation according to the International Classification of Diseases. AI tools can store thousands of medical codes and extract them in seconds at the coder’s request. It saves up to 30% of the coder’s working time. When integrated into an inner relational database, AI can examine and suggest codes related to any medical manipulation.
Another domain where machine learning algorithms with generative AI features help medical staff is processing health insurance claims. If performed manually, it takes up to ten working days per client. AI documentation tools accelerate the operation. They make cost calculations and analyze pricing rates with almost instant data completion and verification.
Optical character recognition
OCR helps medical staff to extract text from images and handwritten notes quickly. The system might also indicate missing information. As with speech recognition, OCR helps to avoid entry errors and contributes to compiling a universal database.
Benefits of AI documentation
Saving time
Physicians mention that AI technologies save them much time. They do not have to close the reports and charts at the end of the day. Decreasing the time spent on typing makes it easier to concentrate on a productive interview with patients and minimizes the stress level.
Better clinical diagnosis
AI tools lead to better clinical experience for both patients and doctors. Physicians carefully gather the symptoms and patients are happy with attentive checkup. The app records the conversation. Consequently, doctors do not feel anxiety that they will forget some details while compiling the case history.
Cutting costs
Since AI-powered EHRs perform part of the tasks automatically, medical providers can cut administrative costs. Huge potential also lies in combining AI documentation with telemedicine. According to the statistics, four in ten physicians used telemedicine daily in 2023. Remote appointments are cheaper for patients, as they do not have to get to the hospital. This is especially relevant at the season of high morbidity. At the same time, doctors become more productive as they can reach more patients.
Data accuracy
Machine learning algorithms analyze arrays of medical data and can detect any inconsistencies, missing details, and unclear abbreviations. The system compiles medical histories and provides standardized reports for the EHRs. Further, medical experts can use the data from the EHR to make diagnoses, find appropriate candidates for clinical trials, and conduct research.
Final thoughts: strive for teamwork
I deal a lot with healthcare projects, and I notice that they imply cooperation of efforts between AI tools and humans. It is not a type of software that will operate completely autonomously. Applying AI tools to medical documentation presupposes that doctors will have an eye on medical records, pursuing the so-called “human in the loop” approach. AI generates the data while doctors read it, edit it, and sign off. Software development specialists are working on solutions that assist people in boosting their productivity, but not operate instead of them.
