The Triple Helix @ UChicago

Fall 2016

"Is Predictive Analytics the Future of Healthcare?" by Nila Ray

 

In 30 years, physicians may be relying on computers to diagnose you. Predictive analytics is a rapidly growing field that applies the usage of statistical methods, machine learning, and artificial intelligence (AI) to provide predicted outcomes from massive databases of information. With these tools, we can not only minimize costs but also improve the impact of services and the quality of lives. Healthcare, especially, is an area that could utilize predictive analytics to save patients extra visits, lower hospital costs, and improve administration. It is already projected to reveal a booming billion-dollar market within the coming years in the healthcare industry.[1] 

One of the most impressive steps toward predictive analytics was the development of an AI called Watson by IBM. The company originally developed him around 2006 with the intention of creating a machine that could quickly answer any question, like those on Jeopardy. He could take the clues given to Jeopardy contestants and prepare answers faster than human contestants could even buzz in because of his predictive capabilities. IBM’s next goal was to bring Watson into markets, especially to tackle the health industry. There are over 5,000 hospitals[2] in the United States with healthcare data doubling every two years[3]. Much of the data in healthcare is unstructured and invisible to systems, as much as 80%[3]. There are dozens of notes from previous visits, vital signs, illness and wound assessments, pieces of medical and family history, and even daily habits that a patient’s file accumulates over the course of their lifetime, that are often overlooked before diagnosis. During a visit, there is only so much a doctor can review before seeing each individual patient. Analyzing much of this data would allow potential complications to be proactively addressed or a more accurate diagnosis to be given by the provider to avoid multiple visits. With an intelligent system, such as Watson, each patient’s data could be structured, compiled, and analyzed to allow the doctor to recognize the patterns that often remain unseen. However, predictive analytics does not only apply to individualized healthcare. Watson can scavenge through 200 million pages of text in just three seconds.[3] This capability allows analysis of not only individual patient files, but also of the thousands of patients over multiple providers. This data would allow experts to assess population trends and concentrate preventative health education in areas displaying concerning trends. However, this powerful method is not just limited to artificial intelligence. 

Often, the problem in patient care lies in the provider’s diagnoses. For a physician in a hospital, it is crucial to determine how much treatment a patient needs so that the establishment does not waste funds and so that the patient recovers successfully, especially in cases of severe illness. The physician is not able to track each patient’s real-time status in-person to make sure his or her risk of worsening does not increase. There can often be a misrepresentation of a patient’s status and an under- or overtreatment on the physician’s part. However, with the help of a real-time predictive analytics software, such as AWARE, the patient in critical condition can be comprehensively monitored and receive a more precise, timely diagnosis. AWARE is a tool that aids in decision-support with an algorithm that can take real-time and high-value data from electronic medical or health record (EMR or EHR) systems to evaluate a patient’s condition throughout all organ systems.[4] There have been trials showing that the use of AWARE in ICU units to structure clinical data on patients allowed physicians to more effectively deliver treatments.[4] This is just one example of the daily application of predictive algorithms in healthcare. 

The application and foundation of predictive analytics have the ability to be implemented into normal patient care. It is driven on the hundreds of EMR and EHR databases that already hold enough information to predict most patient results. Because algorithms propel the predictions, they can be adjusted to how broad or specific the analysis of data should be. As mentioned before, the purpose is not only to improve individual care, but also overall care for institutions as well. One target area would be to lower costs by assessing treatments given to low- and high-risk patients and to determine preventable readmission rates. By allocating the correct treatments to the patients on a need-based system, the hospital would save funding. A recent report showed Atlantic Health, which oversees five hospitals, cut $70 million in operational costs within three years via predictive analytics.[5] Similarly, by assessing common symptoms among groups of patients—that are more likely to have a stroke, for example—the hospital can determine what the next step is for these patients, and even begin applying targeted preventative care and education.  

Most importantly, the tool can exist, but the providers must make use of them properly. Predictive analytics is not a field that comes without its challenges. The cooperation of those in the healthcare industry and the stability of the tool itself are essential. The influx of patient data will consistently grow, and often, may come with missing gaps. This data must be scalable and appropriately filtered to optimize the analytical results.[6] In addition, the application should be able to fit privacy requirements, such as HIPAA (Health Insurance Portability and Accountability Act), as the current medical record systems do.[6] Health providers and medical staff will need to train toward being able to use the tool with skill and sustainable data management[7], with a collaborative vision to apply the technology to their routine patient care. However, as improvements continue, these challenges can hopefully be overcome for the greater benefit of predictive analytics. With such a tool, the process of treatment in healthcare can become increasingly efficient and expedited – and who knows, it may just be within the next 30 years.

References

[1] Healthcare Finance. 2016. “Healthcare predictive analytics market should hit $19.5 billion by 2025, research shows.” Last modified Nov 28, 2016. http://www.healthcarefinancenews.com/news/healthcare-predictive-analytics-market-should-hit-195-billion-2025-research-shows
[2] American Hospital Association. 2016. “Fast Facts on US Hospitals.” Last modified Jan 2016. http://www.aha.org/research/rc/stat-studies/fast-facts.shtml
[3] IBM. 2016. “IBM Watson Health.” Accessed Nov 14, 2016. https://www.ibm.com/watson/health/ 
[4] Harvard Business Review. 2016. “Making Predictive Analytics a Routine Part of Patient Care.” Last modified April 21, 2016. https://hbr.org/2016/04/making-predictive-analytics-a-routine-part-of-patient-care
[5] Healthcare IT News. 2016. “Predictive analytics help Atlantic Health save $70 million in labor, operational costs.” Last modified Feb 18, 2016. http://www.healthcareitnews.com/news/predictive-analytics-help-atlantic-health-save-70-million-labor-operational-costs
[6] Harbinger Systems. 2015. “Predictive Analytics in Healthcare.” Last modified Dec 24, 2015. http://blog.harbinger-systems.com/2015/12/predictive-analytics-in-healthcare/ 
[7] HealthData Management. 2015. “Top Challenges to Analytics in Healthcare? Not Technology.” Last modified Sept 23, 2015. http://www.healthdatamanagement.com/opinion/top-challenges-to-analytics-in-healthcare-not-technology

 
UChicago Triple Helix