Predictive analytics, it is being said, is going to redefine how health care is delivered. It will predict occurrences of critical illnesses and probability of readmissions in the future. Other sectors such as food and beverage, publications and entertainment have already reaped benefits from using predictive analytics — there is no reason health care cannot do the same.

However, the definition and scope of predictive analytics needs to first be understood purely in the context of health care. The one-size-fits-all model is not going to work. It is also important that the infrastructure for delivering analytics is provided and it is able to deliver the required information to the health care professionals in the right format. To deliver the proper and proactive health care, health care professionals need to be given the right context and metadata. So, while predictive analytics is good for health care, it must first be customized and the right data in the right format must be delivered. (To learn about big data's role in health care, see Will Big Data Revolutionize Health Care?)

Webcast: Details Matter: Getting It Right With High-Powered Analytics - Sign Up Here

What Is Predictive Analytics?

Predictive analytics is a branch of advanced analytics that provides predictions of certain events based on historical data, data patterns and other inputs. Proactive steps can be taken to address the requirements arising out of the predictions. To make the predictions, predictive analytics leverages techniques used in other branches such as data mining, artificial intelligence, modeling, machine learning and statistics, and it integrates information technology, management and modeling business processes. The predictions can be used to identify risks and opportunities in the future. Predictive analytics can help business organizations to achieve a lot of things. A few examples include:

  • Identifying hidden associations and patterns
  • Improving customer retention
  • Reducing risk to minimize loss and exposure
  • Improving customer satisfaction

There are a lot of real-life examples of how businesses have benefited from the use of predictive analytics. Accenture conducted a survey to find out how different businesses have benefited from using predictive analytics. Some of the findings are:

  • Best Buy discovered that less than 7% of its customers contributed to 43% of its sales. It then segmented its customers logically and redesigned its stores and in-store experience to reflect the buying habits of specific customer groups.
  • Olive Garden, an American casual dining restaurant, uses data to design and redesign its menu. That way, it has been able to cut down on food wastage significantly.

Predictive analytics is being applied to a lot of domains such as health care, customer relationship management (CRM), fraud detection and risk management. Predictive analytics is also frequently being combined with prescriptive analytics. Prescriptive analytics in this context means that not only are predictions made regarding certain events, but also definite steps are given that must be taken to handle the situation. These steps will be provided by the analytics engine itself. (Learn more about fraud detection with Machine Learning & Hadoop in Next-Generation Fraud Detection.)

Predictive Analytics in the Context of Health Care

Theoretically, predictive analytics has a big role in improving health care. Although it is still a new entrant in health care management and its scope is still being worked out, predictive analytics can analyze historical patient data and provide predictions for things like illness risks, probability score of heart attacks and asthmatic attacks based on patient profile, and probability of readmissions.

The human brain cannot deeply analyze more than six to eight variables at a time to properly profile a problem. But, the algorithm of a predictive model can analyze hundreds of variables at a time to create an accurate profile of a medical problem. Based on the profile, accurate diagnosis and risk predictions, if any, can be made.

Predictive modeling can help control costs related to medical care. In the U.S., one in five Medicare patients are readmitted to the hospital within 30 days of discharge, which results in an expense of $17 billion a year.

Role of Predictive Analytics in Health Care

Simply put, predictive analytics can play an important part in delivering health care that is based on prevention and proactive actions. Health care costs account for a substantial chunk of the GDP in the U.S. Health care costs in the U.S. are 17.6 percent of the GDP, $600 billion more than the ideal health care expenditure considering the population, per capita income and economy of the U.S. Readmissions and critical illnesses such as heart attacks, stroke, diabetes and kidney problems account for a big chunk of the health care costs. A certain percentage of the hospital readmissions and critical illness recurrences could be prevented or managed better with the help of data pattern analysis and the analytics thereafter.

Case Studies

The case studies described below set an example of how predictive analytics could transform the health care industry.

Case Study 1: Steadman Hawkins Clinic Improving Profitability

The Requirement

The main objective of the Steadman Hawkins Clinic of the Carolinas was to improve profitability by optimally combining the services of available physicians with the ancillary services such as pharmacy, phlebotomy and physical therapy. The clinic wanted to optimize the availability depending on the peak demand hours or seasons so that no opportunity to earn revenue was lost. For example, asthmatic attacks could peak during the transition period between two seasons. That is when pulmonologists and other technicians are in demand the most.

The Action

Steadman Hawkins Clinic partnered with River Logic, an analytics firm that implemented predictive analytics. All data points and constraints were taken to determine the optimal way to design the clinic facility and the staff required. Some of the steps that were taken include:

  • Illness patterns of different patients and their correlations with seasons. For example, onset of asthma intensified in winter.
  • Illness profile of patients and the type of investigations they ordered historically
  • Availability of resources to fulfill patient needs and missed requests, if any

The data sets from the above actions provided comprehensive materials for analytics.

The Results

The Steadman Hawkins Clinic was able to increase their net profitability by $20 million a year. They were also able to improve the accuracy of their financial predictions from 30 to 32 percent.

Case Study 2: Unnamed Clinic Improving Profitability

The Requirement

The clinic wanted to both improve services to the patients and improve their profitability by optimally using their resources which includes staff, facilities and instruments.

The Action

The clinic collected copious data on different variables such as type of care needed by patients, staff profile and qualification, patient profile, quality of services delivered such as response time, outcome, patient experience and wait time for patients. Based on the data collected, predictive analytics was put to use. They expected concrete analytics and course of actions to put in use.

The Result

Though the clinic is still in the process of implementing policies based on their predictive analytics, there are signs that they are on course to achieve at least 10 percent higher profitability than before.

Important Points to Remember

It is not that implementing predictive analytics will start doing wonders right away. The results depend on the approach. First, the industry needs to determine what predictive analytics mean in its context and then specify its scope. Also, the health care industry needs to remember the following lessons from other industries:

  • The amount of insights is not directly proportionate to the amount of data. You are not going to get more insights just by increasing data collection.
  • Insights do not necessarily provide value. You need to first customize the insights in your context so that it becomes useful.
  • Implementation of predictive analytics is going to be a big challenge. You need to embrace the right technologies and deliver insights to the healthcare professional in the right format.


Predictive analytics needs to be merged with prescriptive analytics in order to deliver the right results, because the industry needs not only the predictions but also a course of action. While the concept seems to be rewarding in the end, businesses need to make the right investments and be patient with the results if they hope to reap the benefits.