Marketing Cloud, Trailhead...
March 9, 2020
Q&A: Intelligent Analytics for Personalized Healthcare
By Woodson Martin, EVP & GM, Salesforce AppExchange
Tableau's booth at HIMMS '19
In a world of activity tracking watches and phones, consumers love having personalized insights into their health, from minutes slept to their heartbeat and the number of steps traveled. But healthcare organizations, from hospitals and insurance companies to the manufacturer of your insulin, are still understanding how data and analytics can help transform the way they provide care and medicine. There’s a lot of potential, but a lot of work to be done.
We sat down with Manish Jiandani, Director Product Management Health Cloud, and Dr. Barry Chaiken, clinical lead at Tableau Healthcare, to understand how analytics are changing the game for patients thanks to healthcare organizations’ innovative approaches.
Can you walk us through some of the ways that healthcare organizations are using data and analytics?
MJ: The biggest trend that we're seeing is that healthcare organizations are beginning to think of patients just like retailers think of consumers. The retailer can recommend products to a consumer or enable them to select a particular location to pick up their merchandise. Similarly, healthcare organizations want to give their patients flexibility and care options that work for them — which provider is closest to you and fits your specific needs? Having a lot of data and gathering insights from analytics is at the center of this shift that will enable healthcare organizations to deliver personalized and contextual services to their patients.
BC: In healthcare, there are three different things to keep track of: the provider, the payer, and the medical device and life sciences companies.
Provider organizations are divided into administrative and clinical areas. In the administrative area, there is the greatest use of analytics because providers have deployed ERP (enterprise resource planning) systems for a much longer period of time than we’ve used electronic medical records. This means they’ve collected massive amounts of data using these systems. This data is used for revenue cycle management, staffing, finance, and managing supply chains. For example, during the coronavirus outbreak it is important to ensure that adequate supplies of medications and protective equipment for caregivers are available. It would be unacceptable for these critical supplies to be unavailable during periods of increased demand or disruption of the supply chain.
As for the clinical space, there are challenges around interoperability which hinders the ability to bring data in from multiple, disparate sources. Nevertheless, organizations are beginning to do some really interesting things. One initiative works to generally quantify what is happening in the clinical area — what care patients are receiving based upon set metrics, if processes are out of whack — so they can provide information to the doctors, nurses, and other caregivers, which they can then use to change their behavior or the processes they follow.
In addition, you’ll also see medical device and pharmaceutical companies using analytics in ways similar to how manufacturing companies would use data; where the data is used to maintain quality and reduce costs. Another interesting example is tracking clinical trials — a lot of work goes into recruiting patients and using dashboards and analytics to interact with those patients and their physicians to ensure the trial goes smoothly.
There are clearly many aspects to the healthcare space. You mention the patient level to administrative and organization levels, as well as research and medical devices. Is there a predominant area you’re seeing analytics used?
MJ: We’re seeing use cases across all of these areas. If you look at the clinical level, the organizations are leveraging advanced analytics techniques to predict the efficacy of certain treatments they can offer to patients. On the non-clinical side, healthcare organizations are leveraging analytics to improve operational processes. Some are looking into automating their pre-authorization processes to enable faster service.
Another example could be where healthcare organizations are utilizing analytics to understand the patient population. This is to personalize engagement for consumers, and increase engagement. So let’s say a patient has missed medical appointments in the past. How can I use that knowledge to predict the likelihood of the patient missing the next appointment? If the likelihood is high, then how can the healthcare organization leverage that information to send the patient a reminder or even automatically order an Uber for the patient?
BC: Agreed — we've seen more work done at the organizational level, because that's the easiest place you can do it. When you bring analytics into clinical settings, you need to involve physicians and nurses and train a lot more people to use the tool. The biggest challenge is to promote adoption and embed the analytics within the clinical workflow.
We know healthcare providers, hospitals, and insurance companies have massive amounts of patient data, and that brings implications around security and privacy. How are you seeing healthcare organizations protect patients while also improving healthcare and the patient experience through learnings from data?
MJ: Data security and patient data privacy have always been top of mind at healthcare organizations. After all, healthcare is one of the most regulated industries. It aligns with one of our core foundational principles at Salesforce, which is trust. Salesforce Health Cloud is HIPAA compliant and companies look to us and learn from our best practices to meet all the regulatory compliance requirements for data privacy and security.
How has healthcare’s use of analytics evolved as we are now squarely in the world of predictive analytics and AI? Is access to analytics changing within healthcare organizations?
BC: The most important thing that’s happened is we have so much more data that we didn't have before. And even with the problems of interoperability, we still have so much more data than we ever had before. That's great. We're almost overloaded with the data when you're first starting to think about how to actually use it.
The thing to remember — and this is really important, particularly in clinical care — is that the data set used in machine learning or AI is destined to be biased. All of them will be biased to some degree.
Let me give you a simple example: if the data that's used to identify the dosing levels for a particular medication through AI and machine learning comes from a dataset that is all white male, it's obvious that when you're treating a black female, an Asian female or even an Asian male the recommendations are not going to be accurate for that sub-population.
So the big challenge in the AI & ML world is, can we create a large enough data set of unbiased data? And remember, there are many different biases — selection bias and reporting bias and documentation biases — but can we create a large enough data set of an unbiased or representative data as we possibly can and then be able to apply that into a machine learning environment?
That's stuff that people are focused on now. Clinicians don't talk about AI as meaning artificial intelligence, they talk about AI as augmenting their information; in other words, it gives them one more data point to consider when they're deciding what to do.
Are there any customers you’ve worked with that have succeeded in big changes with the help of analytics?
BC: There's one customer of ours that comes to mind: Texas Children’s Hospital. A physician named Dr. Barbara-Jo Achuff was interested in the children that were being taken care of in the cardiac ICU. She's developed dashboards that closely track the sedation levels received by her patients in that unit. Dr. Achuff uses data from the electronic medical record and uses it to monitor that sedation. With the dashboard, she is able to identify patients who are over-sedated, often long before physicians and nurses historically have been able to identify the problem.
Why is over-sedation important? You might be thinking of people falling asleep or being tired, but that’s not the case. The problem is, over-sedation — meaning receiving too much sedation medication — impacts the cognitive ability of these children when they become adults. So by actively monitoring sedation, Dr. Achuff is able to identify improperly sedated patients earlier than before and adjust the medications used for sedation using guidelines and protocols developed from the data collected by her dashboards.
So it's fair to say that every single day in Dr Achuff’s cardiac ICU, there are patients who are being protected from any type of cognitive loss due to over-sedation. And that's just one example of how these tools can be used. This frees the physician to do other important tasks and assigns this surveillance responsibility to a tireless dashboard tool that acts as a constant watchman protecting the most vulnerable of patients.
And if you can imagine an ICU filled with all of these devices and all these blinking lights, it would be pretty easy to imagine that when someone's really sick something like monitoring sedation levels may not be at the top of the list of things to do. This allows analytics to do something that physicians and nurses struggle to do on their own.
What recommendations do you have for a health organization beginning their analytics journey?
MJ: My recommendation would be to look at opportunities that align with their organization's priorities and capabilities. I would recommend to them to have a bold vision, but start small. Small successes will enable healthcare organizations to build that analytics muscle. Also, don’t be afraid to learn from other organizations’ successes and failures. Lastly, it’s important to just get started — it won’t be perfect at first, but starting is half the battle. Once you get small wins on the board, it will encourage more use and get more teams interested in the benefits that analytics can bring.
BC: I love that question because when people are working on their garden for the first time, they get their shovel and start digging, when what they should do is measure their garden, decide what to plant and where before they start digging holes. If they don’t, they're going to waste a lot of space, plants aren't going to get along with each other, and they're going to have a terrible harvest at the end of the year.
So, the recommendation is that when you're starting on an analytics journey, review your organization’s mission statement and link that to your objectives for the year and beyond. Then link those to what you're trying to accomplish on a department or service line level. At that point you can start to prioritize what things you want to work on; for example, increase revenue or improve patient satisfaction scores. Then you can start to determine what types of analytics you’ll need and how to put them into workflows or processes to influence actions and decisions.
The next step is to empower people and establish a data culture so that people actually use these analytics tools. There is a tremendous opportunity to get analytics in the hands of the experts — whether that is the physician, the pharmaceutical rep, or someone else — so they can use data in what they do every day.