Subscribe to DSC Newsletter

Data science has become such a part of everyday life that it is easy to lose sight of how this technology can assist entire industries in very significant ways. Healthcare is a prime example. When applied strategically, data science has the potential to profoundly influence healthcare delivery from both an operations and clinical perspective.

Clinical Applications of Data Science 

Diagnostic applications, such as reading imaging scans or pathology reports, are probably one of the most intuitively obvious places where we would expect to see data science deployed in the near to medium term. This solution can quickly provide answers to diagnostic questions or treatment options based on very large datasets culled from hundreds of thousands of similar cases across all of the institutions that agree to pool their data. Advanced data science also can discover counterintuitive patterns and sources of causality that are impossible for humans to do effectively. Solutions like this and predictive analytics could dramatically streamline the diagnostic and treatment processes to consistently yield better outcomes that increasingly become less dependent on the skills of the individual physician.

While data science could provide substantial benefits to both patients and doctors in this scenario, it is important to remember that every encounter between a provider and a patient is a unique, one-off event that is never identical to any other encounter — even if the same patient and the same provider are involved. Hence, relying on only analytics to get it 100% right every single time is not realistic. It is much more likely that we will rely on the expert judgment of a clinician for the “last mile” decision that needs to be made. In some sense, this is much like self-driving cars are likely to depend on an alert driver ready to take over the wheel for quite a few more years.

The role of algorithms is best deployed as augmented intelligence that enables doctors and nurses to perform their best by providing them with timely, data-driven recommendations that they can accept, reject or modify based on their personal expertise and their judgment of the situation context at that moment in time.

Operational Applications of Data Science 

Hospital operations are complex, highly variable and deeply interconnected systems — this is why it is operationally very challenging to simultaneously deliver high utilization of assets, low wait times for patients, and a large number of available slots for patients seeking an appointment over the next few days. The day-to-day operation requires hundreds of decisions being made by staff members at all levels of the organization on a daily basis — some of these could be to decide on the perfect slot to squeeze in an add-on appointment or which patients need to be called in order to have their appointments rescheduled. Unfortunately, the reality of each day rarely goes exactly according to plan, and as a result, the front line is forced to rely on a static dashboard, half-formed predictions, or gut feel to make these decisions. While each decision in isolation may not seem like a “make-or-break” choice, collectively, patient flow can be adversely impacted to a significant extent. Advanced data science has the potential to provide the front line with the real-time wisdom to improve the speed and the quality of the hundreds of decisions they make each day in order to improve the flow of patients through the various clinical services involved in delivering appropriate care.

It’s worth breaking this down because quite a lot goes into hospital and health system capacity management that most people never see. To improve the patient flow using data science and predictive analytics, there must be detailed simulation models based on the historical volume, type of appointment, patient arrivals, and distribution of time durations for each service performed. There are also operational constraints that are based on the number of staff, their availability, the skills they possess, as well as the specific equipment and rooms required. All of this can then be factored as constraints into optimization algorithms that recommend the ideal sequencing of appointment slots for individual patients, which may change dynamically. 

Many data analytics systems focus on “descriptive analytics” which answer the question “what happened?”  Leveraging data science allows the analytics to move toward “prescriptive analytics” which make recommendations about “what should happen.”

It is still almost impossible for the final appointment schedule to perfectly match the optimized recommendations, but this is where data science is most helpful. Comparing what actually happened after the fact to what the models had predicted at the start of the day creates the opportunity for learning the recommendations and tweaking them in subtle ways that continuously move the performance closer to the optimal frontier.

Again, the intent is never to replace the decision-making of the front line — this expertise, context and judgment are invaluable. Instead, the intent is to leverage the data and continuous learning to help the front line consistently make the most informed decisions possible across a wide variety of situations based on all of the data and accumulated wisdom of the optimization algorithms.

Patient care and operations are two very different, everyday examples of ways data science can disrupt the healthcare industry. There are countless more examples, yet hopefully, these demonstrate the opportunity to use data to improve the experience in large and small ways.

The Cost of Change

In order for an entire industry to change, however, the financial return on investment has to be extremely tangible and measurable. For example, infusion centers have seen an increase in the effective capacity of their infusion treatments by as much as 15-20% with the same levels of staff and hours of operation as well as the same (or reduced) wait times for patients. Optimization software deployed at Stanford Health Care, UCHealth, Memorial Sloan Kettering and NewYork-Presbyterian has been able to deliver and sustain these results over multiple consecutive years. These results are possible because the solution is able to draw from the historical dataset and to continue learning on a daily basis to ensure that the recommendations are timely, accurate and effective.

This is a big deal on multiple levels. First and foremost, it helps with patient access. Rather than expecting patients to wait several weeks or months for an appointment, which can exacerbate a patient’s condition as well as place undue stress on a patient anxious for treatment, it is likely that appointment slots will become available over the next few days or week. Cancellations and rescheduling are common issues. With good predictive software, however, centers can better plan ahead. As a result, more patients can be seen quickly, and patient wait times can be reduced by 30-50% once they set foot into the center.

The bottom line is this: Optimization models that learn and adapt can help improve the effective capacity of the many expensive assets in a health system that include operating rooms, inpatient beds, and imaging equipment, to name just a few. Here’s where it gets really exciting to those obsessed with cost. The aggregate value of all of these assets across the 5,000 hospitals in the United States is well over $2 trillion. Improving the utilization by even 5% will unlock over $100 billion of value per year and can have a significant impact on the cost of care.

So, yes, data science will significantly impact the future of healthcare because it has the potential to improve patient care, achieve better outcomes, and enhance the experience while better utilizing the resources on hand.  As more solutions are adopted and put into play, expect a whole new world of opportunities to open up that will transform the healthcare industry for the better.

Views: 809

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Videos

  • Add Videos
  • View All

© 2019   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service