5 ways automation in healthcare saves clinicians’ time

19.06.2020
10 min.
title

Logically, in the times of a pandemic, medical professionals’ main responsibilities are to treat and save lives. However, this is just two of many more responsibilities they have to deal with daily. According to Medscape’s Physician Compensation Review 2019, 74% of doctors invest 10-20+ hours a week in paperwork and administrative tasks:

Hours spent on medical paperwork and administration

With this data in the picture, doctors’ weekly workload climbs up to reach 60+ hours. No wonder medics frequently suffer from burnout. By the way, that’s the case for about 50% of US medical professionals. Curiously, excessive paperwork is cited as the leading cause of burnout (55%) according to the National Physician Burnout & Suicide Report 2020 by Medscape.

So is it possible to lift this burden off doctors’ shoulders? Fortunately, it is. Tedious and repetitive daily tasks may well be entrusted to machines, and healthcare software development is the route to take. Let’s see how it works.

Healthcare automation in action

In fact, healthcare automation is already in full swing. Fact.MR reports that by 2026, healthcare and pharmaceuticals are to become leaders in the business workflow automation market, with over $900 million in revenue.

So what technologies are there to facilitate healthcare automation? Nowadays, the technology stack comprises nature language processing (NLP), machine learning, healthcare BI, and more as the list of technologies applicable to healthcare automation is constantly expanding. So let’s dive in to see how these solutions work to save clinicians’ time and spare them of excessive paperwork.

Natural language processing

Healthcare has one big problem—80% of medical data is unstructured. This includes doctors’ notes, medical images, and various textual information that is valuable for diagnostics and care provision yet is unusable for analytical purposes in its raw form. Moreover, this data can’t even be integrated into patients’ EHRs. That’s where natural language processing helps.

Natural language processing (NLP) is a branch of AI that aims to bridge the gap between humans and machines, helping the latter to understand the former. The technology works in two directions—it can translate unstructured data into a machine-friendly format and retell the machine format in a natural language for humans to use. It also enables doctors to save time on routine tasks, allowing them to dictate information on the go. Then it can be saved in patients’ EHRs or sent to other clinical systems, which solves the issue with interoperability in healthcare, a silo hampering many clinical environments.

But what do doctors feel about these NLP tools?

Allina Health, a Minneapolis hospital, deployed Dragon Medical One, an NLP-driven platform that supports real-time transcription. Dr. David Ingham, Medical Director for Information Services, compares hospital work before and after the adoption:

[Before Dragon] it could be hours before your colleagues can read a note, know what you’re thinking and take action. [Now] I can say, ‘Go to the most recent labs,’ and the computer will navigate there for me. I can say, ‘Order a basic metabolic panel,’ and it’ll tee that up.

Dr. David Ingham, Medical Director for Information Services, Allina Health

What’s more, NLP tools foster not only efficiency and seamless cooperation but also cost reduction. Concord Health, a large New Hampshire clinic, deployed the Dragon NLP tool and reached a 90% adoption among its staff. According to Garvin Eastman, Concord Application Analyst, this helped lower the use of phone-based transcription services by 91% and save $1 million. In addition, the doctors report improvements in the quality of care and doctor-patient relationships. Dr. Ingham notes that in the pre-Dragon era they used to spend more time on typing rather than seeing patients and thinking about their diseases, and now he finds his dictated notes more detailed and understandable to patients.

To deliver value for doctors and patients, though, NLP solutions need a well-developed digital environment. Otherwise, introducing these tools may require additional integration efforts, which often results in lengthy deployment and increased cost.

Meet Dr. Chatbot

This ‘doctor’ is the amalgamation of NLP and machine learning for healthcare. These chatbots can lend doctors a hand in reducing the number of hospital visits, triaging, and acquiring patients.

Let’s add some personality touch to our Dr. Chatbot—meet Grace from Providence St. Joseph Hospital in the State of Washington. This chatbot solves the triaging puzzle with a series of simple yes-no questions and a set of pre-developed scenarios. With this set of tools, Grace decides on the steps to take to offer a patient the most suitable care option. This can be calling the emergency line, scheduling a visit to a particular specialist, or just following simple self-help recommendations at home. Grace even provides an estimate of the costs the person is likely to incur in case of hospitalization or a doctor visit. Impressive? Well, there’s more to it.

Chatbots can easily adapt to the changes in the clinical setting. This is the case with Grace. As the first COVID-19 patients started to arrive to Providence St. Joseph in March 2020, the AI tool got transformed. Since then, it’s been offering coronavirus assessment based on symptoms, geographic location, potential exposure, and more. In the end of the session, the tool provides useful recommendations that are quite doable, which helps reduce the risks for both the person and the healthcare facility:

Grace chatbot, Providence St. Joseph Hospital

All in all, NLP- and ML-driven chatbots have a lot to offer to the healthcare industry, and they are gradually gaining popularity. According to Business Insider Intelligence, about 73% of hospital admin tasks can be automated with chatbots by 2023.

Robotic process automation (RPA) in healthcare

AI-driven RPA tools can capture and reproduce human behavior when dealing with routine repetitive tasks—processing transactions, communicating with other digital systems, working with data, and more. RPA has at least one unbeatable advantage: it can be deployed without disrupting a healthcare facility’s operations, by using the existing digital environment and lowering the risks of human error.

In a healthcare setting, RPA proves valuable for automating several lengthy and effort-intensive processes. Here are the top three candidates for robotic process automation in healthcare:

Patient relationship management. RPA bots can easily schedule appointments matching diagnoses, doctor availability, patient location, and other factors of choice. Moreover, they can manage follow-ups and send patients reminders on the next appointment, prescription drug pick-ups, and more.

Healthcare cycle improvements. The amount of data that providers collect from their customers daily is enormous, from personal information to doctors’ notes, appointment summaries, prescription details, etc. For humans, managing this pile of information is time-consuming and error-prone. For RPA systems, that’s virtually effortless. These tools can generate valuable insights from available data to help medical professionals improve diagnosing and deliver personalized treatment plans for better patient outcomes.

Claim management. This is yet another lengthy process requiring vigilant attention and concentration, which are not always there due to the human factor. With the unbiased RPA, claim management runs seamlessly and the human factor is eliminated.

RPA is a powerful technology that can be used for automating various healthcare workflows, from onboarding and first-appointment scheduling to the payment. According to ResearchAndMarkets’ RPA in Healthcare – Global Forecast by 2025, RPA in healthcare will grow at a CAGR of 20%.

While the technology is promising, before deployment you’ll still need to get prepared to prevent potential issues. Some RPA deployments end up being inefficient, and the key reason for this is a hectic implementation without due research and the outline of the problems to solve. Besides, successful RPA implementation requires advanced staff training, detailed use cases, and the analysis of the most suitable solutions on the market, including the option of custom development. Evidently, rushing is not the best strategy here. It’s better to take time and make grounded decisions in cooperation with key stakeholders and technical specialists.

Computer vision in diagnostics

Computer vision is yet another alliance between ML and data science, which proves beneficial for a number of medical areas, from precision medicine to diagnostics.

Computer vision tools use deep learning (DL) and convolutional neural networks (CNNs) to mimic humans in image classification and recognition with the accuracy of up to 90%. In some cases, machine-powered recognition outdoes doctors. This is the case with the new AI-driven algorithm for the lung cancer detection by Google.

Google’s lung cancer detection AI

When precisely calibrated, medical image analysis greatly facilitates diagnostics and allows doctors to come up with effective treatment decisions without investing much time. In diagnostics, computer vision tools allow doctors to detect the problem early, reduce false positives and false negatives as well as to craft most suitable treatment plans.

In some cases, AI algorithms can even help practitioners personalize treatment. For example, the Cleveland Clinic, Ohio, deployed an AI algorithm that was trained to combine patient records, health history, and CT scans. Having analyzed the data, the smart tool builds mathematical models to evaluate how well patients would react to certain treatments. Here’s how it works:

A deep learning framework for individualising radiotherapy dose

Doctors then weigh the results and offer the most suitable scientifically grounded treatment. And yet, here comes an issue to consider. Surprisingly, the two key players responsible for computer vision—data scientists and clinicians—don’t have a common language for productive communication. Consequently, we have some knowledge gaps that may hurt diagnostics. At the moment, medical professionals have little training in ML and data science to understand how their smart tools work. This calls for building up productive partnerships between clinical and technology experts to ensure informed decision making at both ends.

A treasure chest: predictive analytics

Clinicians face risks on a daily basis. They have to make decisions on suitable treatments, the signs of recovery or exacerbation, the need for hospitalization, and more. Quite often they rely not only on clinical data but also on their gut feeling, which is not a universally reliable strategy. Besides, in some cases this intuitive decision-making is not applicable at all. For instance, how can doctors intuitively predict the spikes in demand caused by a pandemic to provide for efficient resource utilization?

Luckily, nowadays doctors have a trustworthy assistant—predictive modeling in healthcare. Predictive analytics is a unique field located at the crossroads of statistics, data mining, and machine learning. Data scientists study hospital workflows and other data sets in question to formulate a problem and develop suitable predictive models for solving it. After model assessment and validation, the tool is released to work its magic—to look into a mass of historical data to detect dependencies and make predictions about future events.

Predictive analytics proves efficient for personalized health management. Analyzing patients’ medical histories with the help of this type of analytics, doctors get a clear visualization of the likelihood of complications with risk scores. Then they can assess the number of at-risk patients and prepare for their potential hospital stays in time. Besides, analyzing patient data (patient-generated health data (PGHD), claims, lab tests, etc.) with predictive algorithms may help providers design personalized disease management strategies aimed at improving the quality of life and reducing costs.

Predictive analytics also helps resolve health challenges on a larger scale—in communities. Thanks to advanced digitalization, predictive analytics can plunge into large health-related datasets to evaluate population health and detect risk-prone groups of people in need of special protection schemes. Cohort health analytics is of special significance for epidemiology, as it can provide timely actionable insights on the disease outbreak prevention and management, identifying population segments at risk.

Given all the pros above, it’s not surprising that predictive analytics adoption in healthcare is on the rise. According to Allied Market Research, it is to hit $8.464 million in the global healthcare market by 2025 at a CAGR of 21.2%.

Automation in healthcare: trust or caution?

With all these AI-powered automation solutions, there comes a legitimate question: is there a chance that healthcare professionals will be ousted by smart machines? Keep calm and automate—this is highly unlikely.

However smart, machines can only perform repetitive routine tasks; creative problems that doctors and nurses solve every day are out of their reach. Still, with routine tasks, machines are far more efficient than humans: they work faster and don’t make mistakes associated with fatigue and lack of concentration. So why not let them bring these benefits to your clinic?

In fact, automation doesn’t rob healthcare professionals of their work. It spares them the time they can dedicate to solving complex professional problems and to self-education in order to deliver top-quality care and improve patient outcomes. This is how automation works as a trustworthy partner to healthcare institutions.

Tags: