Big Data Comes to Healthcare
While the constitutionality and legality of “Obamacare” seems to overwhelm healthcare discussions today, there are life-enhancing changes taking place in the industry that are being driven by big data. The healthcare industry is in the process of migrating from the stone ages of using paper documentation and fax machines to making decisions based upon rules of thumbs and gut feel, to the modern age of leveraging disperse sources of historical data with advanced analytics and new instrumentation approaches to change the cost and effectiveness of the medical care that you and I receive.
A recent article in BusinessWeek titled “This Machine Save Lives, So Why Don’t More Hospitals Use It?” talked about how electronic health records (EHR) are assembling and aggregating patient information from multiple, disparate systems to bring together the information necessary to make data-informed medical and health care decisions. And while there are still significant challenges in accessing all the patient care and lifestyle data across these multiple, disparate applications, the health care treatment benefits are already evident. The article credited Kaiser Permanente’s 24% reduction in heart attacks and 40% reduction in mortality rates due to sepsis in Northern California to their adoption of EHR (though as good data scientists, we know not to confuse correction with causality).
McKinsey Global Institute published the below analysis to help understand the scope of the potential business value of Big Data to the healthcare industry.
Figure 1: Source: McKinsey Global Institute
I remember a meeting with a senior executive at a major insurance firm several years ago, who said “more lives will be saved by the computer than by the microscope.” I guess with the coupling of Big Data and advanced analytics that vision is finally starting to come true.
Healthcare: A Big Data Business Value vs. Challenges Microcosm
Probably no industry represents both the business opportunities and the technology challenges of Big Data better than the healthcare industry. The list of Big Data business opportunities in health care reads like a Big Data wish list:
- Ability to access any data source, no matter where it is located, using new federated query, data discovery and semantic management technologies. This allows health care providers to gain a more timely, more complete understanding of the patient’s current situation so that they can prescribe the appropriate and most effective treatments.
- New instrumentation opportunities to increase the amount and real-time nature of data being captured about patients’ health care. These smart devices include connected in-home diabetes blood monitoring, sensors in toilets, smart tooth brushes, and smart scales that send current patient condition information to the appropriate hospital and doctor data stores.
- Creative uses of image and audio technology to capture even more data such as gleaning changes in a patient’s weight from vacation photos posted on Facebook, of the use of voice recognition technology to monitor the mental health of home care patients via the tone and clarity of daily phone conversations.
- Advanced analytic modeling in areas such as treatment attribution analytics to model patient treatment effectiveness, predictive patient re-admission analytics, and the creation of patient health scores (like the credit scores that are created by financial institutions today).
- In-memory capabilities to facilitate real-time, life-saving decisions at the point of care, especially in high stress, immediate need areas like the emergency room.
- Complex event processing to monitor, analyze, and flag potential health issues on a daily or on-demand basis.
- Decision governance strategies that clearly outline what information hospitals and doctors are allowed to gather and analyze, and how that information can be used to the benefit of the patient (and not necessarily to the benefit of insurance companies or government agencies).
- Real-time monitoring of key patient health care metrics that leverages in-memory computing to more rapidly evaluate incoming patient data streams (from the multitude of new health metrics capturing sensors), flag areas of concern, and score potential health-related issues.
- Uber big data with whole genome sequencing (petabytes+) as the availability and cost of “every day” genome sequencing becomes available to the masses.
Let’s walk through an example coupling some of these different data sources with advanced analytics to help hospitals predict readmissions.
Reducing hospital readmissions has been identified by Congress and President Obama as an area for reducing Medicare spending. In 2010, President Obama signed into law comprehensive health care reform legislation, the Patient Protection and Affordable Care Act. Health reform legislation allows the federal Centers for Medicare & Medicaid Services (CMS) to withhold a portion of Medicare payments to hospitals that have excessive readmissions, starting with up to 1 percent in fiscal year 2013 and rising to 3 percent in 2015 in penalties.
What if you knew, due to access to the “right” data sources coupled with advanced analytics, which patients are most likely to be readmitted? If your analytics could provide these likelihoods or scores, then hospitals would manage patients much differently. During the initial admission, hospitals would identify patients who will benefit from a longer length of stay, versus working to get these patients out of the hospital as quickly as possible. Then, post discharge, hospitals would apply analytics and more focused patient management systems in order to:
- Improve the design and management of effective hospital discharge planning
- Identify patients that would benefit from more frequent follow up
- Identify patients who should be directed to alternative care settings such as Long Term Care
Big Data Analytics Process For Hospital Readmissions
The first step would be to assemble a patient’s demographics, history, diagnosis, and treatment and post discharge data, and then calculate the patient’s likelihood of having a related and unplanned readmission within 7, 15, and 30 days (see chart below).
Data that could be used to feed this process include:
- Patient demographics
- Physician demographics
- Admission clinical features
- Comorbidity features
- Hospital course features
- Post-discharge features
- Patient medical history features
- Family history features
- Lifestyle features
- Last blood work features
Then the analysis and model building process would take place to look across all of these data sources to identify those combinations of metrics and dimensions that are the best predictors of readmission. We could look at the predictability of age and gender of readmissions (see chart below).
Next we could examine the potential effect of cholesterol and heart rate on readmissions predictability (see chart below).
The end goal is to come up with a simple yet actionable model that helps the hospital administrators to predict readmissions cases so that they can take preventive action instead of the traditional more expense reactive actions (see chart below).
I know that this topic raises all sorts of privacy and “right of choice” issues. One just needs to pick up a local paper to see the battleground issues and posturing that is going on today between Democrats and Republicans around Obamacare. However, the benefits are indisputable – the intersection of Big Data and advanced analytics in healthcare holds the promise of more cost efficient, more effective patient care.
Be sure to join EMC on October 16-17 at the Strata Rx Conference. EMC will have a booth there, you will be able to speak to industry experts, as well as see demos from Greenplum and VMware. I will be joining a panel of experts to discuss Big Fast Data in Health Sciences, as well as delivering a keynote. Don’t miss out!
 BusinessWeek, “This Machine Saves Lives, So Why Don’t More Hospitals Use It,” June 25-July 1, 2012
 Special thanks to Dr. Pedro Desouza for his advanced analytics work on readmissions