Learn how conversational AI helps to build medical chatbots to deliver personalize service at scale and providing health-related information to users.
We have discussed how chatbots and conversational AI can help healthcare institutions in multiple ways before. From patient screening and triaging to symptom checker bots and contact centres to manage enquiries, the use cases in healthcare are plenty.
One crucial use case however that deserves more attention is how chatbots can help clinicians. Even among clinicians, the perception of chatbots is primarily as a tool that can help improve appointment scheduling and enabling patients to locate clinics. In a survey of 100 physicians in the United States, 78 of them cited scheduling doctor appointments as the most useful benefit of chatbots, followed by locating health clinics (71) and providing medical information (71).
But clinicians themselves are also responsible for understanding and staying up to date with the latest medical advancements. They are relied upon to know the best practices and devise the proper clinical workflows and guidelines for the hospital to effectively manage patient care.
As it stands today, they face some common challenges.
Ever expanding medical knowledge
The medical field is one that is rapidly evolving. As new technologies, medications and breakthroughs shed light on diseases, symptoms and treatments, the medical knowledge database also expands.
This expansion has accelerated so much that it has become near impossible for a busy doctor to keep up with the latest diagnostic and treatment approaches. Professional societies, academic medical centres and medical knowledge vendors work to synthesize and publish new knowledge. But the adoption of best practices and application of new knowledge to practice has become a significant challenge.
At the same time, they are expected to be able to synthesize this new knowledge in order to offer evidence-based recommendations. This is where a handy and convenient clinical assistant tool can be used to combine and synthesize the information from the latest and most comprehensive medical databases and calculators such as PubMed, nMD+Calc, QxMD, ToxBase, Wolters Kluwer, UpToDate and others.
Clinical knowledge hard-coded into solutions
While there are technological solutions and tools to help doctors access medical knowledge, these are far from ideal. In these solutions, the clinical knowledge is often hard-coded into the solutions. This means that the knowledge storage and retrieval follows set rules and would need to be updated if new data and learnings became available.
Updating these existing rules and implementing new ones therefore becomes a challenge and takes more time and energy away from clinicians. It would also create significant technical debt and support from technology teams who will need to be frequently involved for updating code and infrastructure. The situation would only get progressively worse over time.
The complexity of translating guidelines into code
To be able to provide evidence-based care using technology, clinicians need to be sure that the guidelines are accurately translated into code. This requires significant commitment from IT staff, which needs to be repeated at each department and organization.
This is a complex process that involves translating complex symptoms, and clinicians treatment recommendations into a clear and unambiguous medical classification code. Hospitals and health insurance firms also want this information for record-keeping purposes, operations and planning as well as for reimbursement of claims.
Moreover, the codes are updated as and when new diagnoses are discovered, based on an International Classification of Disease codes (ICD) from the World Health Organization (WHO). The next iteration of the standard, the ICD-11 which will go into implementation in January 2022 has over 55,000 diagnostic codes. There may also be additional codes based on individual countries’ medical standards.
For the last few decades, this process of medical coding has been done by human medical coders, relying on a “code book” to look up the right code to classify diseases and treatments. Not only is this a slow process, but it is subject to interpretation differences when there are more than one way to code a diagnosis or treatment.
Information is fragmented and often inconsistent
Once the guidelines are translated into code and the solutions implemented, there is still the challenge of keeping the knowledge up to date and consistent. Organizations often struggle to understand what knowledge is embedded where. Clinicians end up spending a considerable amount of time going through multiple sources from intranet, shared drives and hard copy files.
Prioritizing care for critical patients is getting hard
When the number of patients becomes too high, it becomes more important than ever to prioritize care for the most critical patients. This is especially the case when the cases range from mild symptoms to more severe issues which require immediate attention. When the emergency department is overloaded with patient visits, the really serious cases might have to wait and even turned way. Without easy access to patient data and their initial symptoms, it also makes it hard for GPs to recommend the right specialists who may be better suited to provide care straight away.
Chatbots for Doctors to Manage Medical Information Effectively
Healthcare chatbots, especially advanced conversational AI systems that incorporate natural language processing can help clinicians overcome these challenges. Here are some ways they make clinicians’ jobs easier.
Instant access to information
Instead of poring through intranet databases, shared drives and paper documents, clinicians can access the necessary information in a conversation.They can refer to clinical guidelines, use medical calculators and retrieve patient information via a chat interface. Predictive response options can also be programmed based on common queries so that the relevant hospital content is available at their fingertips. All of this helps in reducing their administrative workload, increasing productivity and making evidence-based care recommendations.
Training on customized hospital content
In addition to the medical knowledge that is publicly available, the healthcare organization will have its own specific databases, internal clinical guidelines, protocols, content and conventions. Hospital specialties, list of treatments, amenities and services and the directory of doctors all will be a natural part of information search.
Chatbots make it easier for clinicians to access such information that includes specific terms used by the hospital. For this to happen, the chatbot will need to be trained on customized hospital content. This will also make it possible for clinical departments within hospitals to build customized clinical modules based on their own guidelines and content. As a bonus, these modules can be registered as their own intellectual property.
Integrated data from third party clinical tools and resources
Clinicians use third party resources and tools such as PubMed, QxMD and other drug databases in their work. These often include large information-heavy databases with complex medical jargon and acronyms. For example, these could include lists of illnesses, procedures, symptoms and their layman expressions among others. PubMed for instance, has more than 30 million citations and abstracts dating back to 1966.
Chatbots can be trained on these databases and use language models specific to biomedical natural language processing. This can help clinicians find specific terms within these databases based on their queries. And it’s not just webpages, today’s advanced healthcare chatbots can mine content within longer PDF or Word documents and retrieve the necessary information.
Right siting patients based on their symptoms
With the help of a chatbot, patients can enter their symptoms and be directed to the right healthcare provider to get the appropriate care at the right time and at the right costs. Not only does this lower the administrative and care delivery costs for the healthcare system and better use their front line resources, but it also helps clinicians triage and prioritize urgent cases ahead of less severe ones. This reduces unnecessary visits to the emergency departments and readmission rates while GPs can look at the symptoms and refer the patient to specialists as and when needed.
Remote patient monitoring and follow ups
In reality, the clinician’s work is not done when the patient leaves the hospital. There is often a recovery period where the patient has to follow certain advice, take the necessary medications, temperature readings and monitor their condition. If the situation does not improve, follow ups would be necessary. Such monitoring can be done via common channels like WhatsApp and webchat.
This process of remote patient monitoring can be automated by chatbots, especially if they are integrated with the existing channels and Electronic Medical Record (EMR) systems. Patients can be sent automated reminders for taking readings of their temperature, blood pressure and sugar levels.
This enables patient reported outcomes and clinicians can flag severe cases for follow up while encouraging patients with good readings. This also allows them to scale up their coverage so that they can serve more patients.
As more wearables and IoT devices come to be used, more of such patient monitoring can be done remotely and automated. In the US for instance, various healthcare providers are ramping up remote patient monitoring technologies in the wake of COVID. Mount Sinai Health System in New York uses a remote monitoring platform that allows patients to download an app to track symptoms such as body temperature, cough, breathing levels, and body aches. If any concerning data is seen during the home monitoring, doctors can arrange for chats or emergency care as needed.
Find out how chatbots can help your clinicians with knowledge management and support evidence-based decision making.