Watch videos with subtitles in your language, upload your videos, create your own subtitles! Click here to learn more on "how to Dotsub"


0 (0 Likes / 0 Dislikes)
>>Peter: Healthcare really stands out as the industry [Peter Zemsky | INSEAD, Deputy Dean, Professor of Strategy] where early excitement about the transformative power of AI has proven slow to unlock in practice, despite high-profile projects backed by leading companies like IBM and Google. There are numerous barriers to adoption in this sector, including heavy regulation, the high stakes for patients, and system complexity. Prior digital solutions, like electronic health records, also saw slow adoption before ultimately being embraced. Nonetheless, the long term potential and need for AI is huge. We're going to look here at a case involving a telemedicine provider where AI is already delivering real business value. Along with the barriers from regulation and a resistance to change, a key feature of the industry is a history of escalating costs that have placed really intense pressures on payers, whether they're governments of employers. And this is nowhere truer than in the U.S., where 18% of national output is devoted to healthcare. One prominent trend driven by this cost pressure is the development of telemedicine, where patients can have virtual doctor visits from their phone or PC. Despite interest from payers, adoption by patients has lagged. The value creation for telemedicine is clear. For the patient, it offers the convenience of easy access to physicians for basic consultations anytime and anywhere. For payers, there are several sources of cost savings. There are lower overheads, as teledoctors can work from home on fully digital delivery platforms. The ability to pool doctors nationally provides staffing efficiencies. And finally, online consultations can substitute for expensive emergency room visits. However, this setting also well illustrates the barriers to adoption. Patients have a deeply ingrained expectation to meet their doctor in person. In addition, there are limits to the interventions a doctor can make online. For the providers of telemedicine, this means extra costs for activating patients to use their services. Currently, only a few percent of doctor visits occur online. How can AI help to unlock the value creation potential of telemedicine? One of the leading U.S. telemedicine players, MDLIVE, is actively leveraging AI to enhance their value proposition. Employers and healthcare plans representing over 30 million patients have partnered with MDLIVE, which has a network of over 1,300 care providers. Their key challenge is patient activation. They're delivering around 750,000 online appointments a year, which means most of their eligible patients are not yet active. Here's a marketing graphic highlighting key elements of their value proposition for patients. In healthcare, the satisfaction of payers, though, is just as important to providers. Here are a set of dashboards that MDLIVE provides to its payers. Note the metrics on patient wait times and patient satisfaction, which are typically part of MDLIVE's service level agreements. A key reason for the success of MDLIVE's early AI projects is that they did not seek to replace doctors, but rather to drive doctor efficiency and effectiveness. A great example is their work on staffing analytics. Having the right number of doctors working at any time of day is important to assure low patient wait times and efficient use of provider time. A complication is that doctors need to be licensed in each U.S. state they serve patients, with doctors typically licensed in three or four states. Hence, effective data analytics needs to deliver demand estimates at the state level. While the solution makes use of sophisticated analytics, as often, the biggest challenge was to acquire, clean, and organize the right data for the problem. This included internal data, such as detailed historical records of patient demand by time and state, as well as relevant external data from public and private sources. For example, the flu is a key driver of calls, and hence, they integrated government data on the spread of the flu by age, and Walgreens' data on flu intensity by state. Another key challenge in unlocking the business value from AI is to assure that the insights are integrated into decisions and actions. Here, this means assuring that those scheduling doctors can see where the needs are, and track progress in reducing wait times as they line up additional provider supply. In addition to screens like this one, showing states and times in red, where there's still a lack of capacity, there are further visualization tools that allow the schedules to drill down from this national view to analytics for a particular state. With the right data, the right analytics, and the right visualization tools, MDLIVE was able to significantly improve its metrics. Average wait times decreased by 80%. Patients abandoning calls due to delays dropped by 50%. And doctor utilization increased by over 20%, leading to a surge in their satisfaction. Now, building on the success, MDLIVE took on a more advanced AI application, addressing doctor bedside manner. Interpersonal skills have always been important for general practitioners. And this is especially true with virtual communications where providers are still driving adoption of the service. MDLIVE has high net promoter scores and strong reviews, as illustrated by this data from the site Trustpilot. While most users are satisfied, as with most service businesses, there is variance in experience. As in this review, patients can get frustrated. They feel the doctor is not listening to them or communicating clearly. And here, you see MDLIVE responding to the negative review. What if they could use machine learning to predict when a visit is going poorly, and pursue service recovery before the patient posts a review, or fills out that NPS survey? Possible? Absolutely. As an online platform, MDLIVE has access to a rich set of data related to each call, patient and doctor demographic information, call metrics, including wait times, duration, reason for the call, type of device used to connect. And finally, the outcome of the visit, data on the diagnosis, prescriptions, and the detailed visit notes entered by the physician. Going even further, they're also able to generate automatic transcripts of the calls, and use the latest NLP libraries to generate sentiment analysis. Armed with such rich data sources, and the latest in AI models, MDLIVE was able to achieve highly accurate assessments on patient visit sentiment, 97% of the 50,000 visits they used as training data. Here's a screenshot alerting a call center agent to a potentially unhappy patient, and suggesting they follow up with a call for service recovery. The vision is to continue developing the system to the point where it can deliver real time support to doctors, as in this mock-up here. In summary, we see two well executed use cases for driving business value from AI. On the patient side, we see both improved wait times and the ability to address service issues. With both being important to the critical payers, as well. On the cost side, they're able to deliver greater efficiency and satisfaction of their doctors. And remember that by improving patient satisfaction and word of mouth, they can drive efficiency in their investment in patient activation. In terms of organization, the key lesson for me is integration. First, the integration of diverse data sets needed for sophisticated analytics. But just as importantly, the skillful integration of the insights into easy to use tools, so that decision-makers and front line staff can make use of them.

Video Details

Duration: 9 minutes and 21 seconds
Language: English
License: Dotsub - Standard License
Genre: None
Views: 9
Posted by: csintl on Jun 18, 2019


Caption and Translate

    Sign In/Register for Dotsub above to caption this video.