How Big Data Shapes Healthcare Delivery
The healthcare industry is always looking for new ways to improve human health on the individual, community, and global scale. Industry experts use cutting-edge technologies and theories to improve the timeliness, accuracy, and accessibility of healthcare delivery and one of the most efficient ways to achieve this goal in the 21st century is with big data.
The potential use case for big data to improve healthcare has not yet been fully realized, but the move towards electronic health records (EHR) was the first big transition for the healthcare industry into the digital world. The Health Information Technology for Economic and Clinical Health Act (HITECH) of 2009 has propelled EHR systems into four out of five hospitals across the country 80 percent of hospitals, and almost as many physician offices.
As the enormous pool of information available from EHRs, wearable technologies, social media, mobile apps, and marketing campaigns continues to grow, new types of data and new intersections between different types of data are discovered, analyzed, and utilized. What’s more, the amount and quality of data available are growing, making results increasingly accurate.
There are many exciting ways in which the healthcare industry is beginning to leverage big data to improve patient outcomes. Read on to learn five ways big data is influencing healthcare delivery.
Using Big Data to Combat Health Mysteries
Sepsis—when the body attacks a bacterial or viral infection with a severe toxic, damaging, or fatal immune response—is an enormous issue in the United States and worldwide. Sepsis is the leading cause of hospitalization in the U.S. each year, coming with an astounding yearly price tag of $27 billion. Sepsis is also the leading cause of hospital readmissions, adding $2 billion to the annual cost of the disease.
According to the Sepsis Alliance, 30 million people are affected by sepsis each year, and eight million die from the disease. In the U.S. alone, 1.7 million people are diagnosed with sepsis each year. Despite being incredibly common, diagnosing sepsis is a challenge for hospitals and healthcare providers because it mimics other conditions, can be caused by a wide range of other infections and can present anywhere in the body.
There is no single blood test to detect sepsis, and for healthcare providers who are juggling heavy workloads, sepsis is easy to miss, misdiagnose, or discover too late. In response to this enormous challenge that impacts the entire healthcare system, researchers across the country are utilizing big data in innovative ways to help healthcare practitioners catch and successfully treat sepsis with higher frequency.
Retrospective Data Analysis for Preventative Healthcare
A second way researchers are using big data for significant impact is by looking retrospectively at discharge data. One example of this is the Sepsis Early Recognition and Response Initiative (SERRI). Researchers in Texas used data gathered by SERRI to analyze the annonymous discharge data of more than 5,000 patients over the course of one year. The results of the study illuminated that hospital staff needed greater awareness of sepsis symptoms, specifically in emergency departments. The study also concluded that misdiagnosis was due to common symptoms, high false positive test results, and a lack of public awareness.
Thanks to the results from the study, the Methodist Hospital Research Institute has been able to identify concrete steps that will lead to better patient outcomes, such as progression prevention, lower rates of organ failure, fewer deaths, shorter hospital stays, and reduced costs.
Machine Learning for Real-Time Care
Researchers at the University of North Carolina, University of Pennsylvania, Carnegie Mellon, and Amara Health Analytics are using big data and machine learning to create real-time warning systems that can assist healthcare professionals in timely diagnosis and effective treatment.
By applying machine learning algorithms—i.e., computer programs that analyze and identify patterns in data and learn from them—to large amounts of EHR data, the systems can understand trends in sepsis. When this understanding is cross-referenced with real-time telemetry data and data from the patient’s course of care, predictive analysis can warn medical professionals that sepsis may be likely.
Amara’s Clinical Vigilance system cross-references 100 different variables and is capable of analyzing structured data (medical codes) and unstructured data (doctor’s notes) to predict if a patient is becoming septic. When sepsis is suspected, the systems alert providers and some offer an operational response to help the patient recover.
Penn’s system helped sepsis mortality rates fall by four percent, and Amara reported that its system could save a 300-bed hospital $2 million by shortening patients’ length of stay. While these systems are being developed specifically to combat sepsis, researchers are hopeful that they can also leverage big data in broader applications to help predict other diseases as well.
More Targeted and Timely Care
Another big way that big data is going to impact healthcare lies in the capacity to pool information about human health from a wide variety of sources, which will help healthcare experts be more precise when diagnosing and treating patients.
In the case of individuals, big data analysis can be utilized to provide more personalized and targeted care. According to NPR, the analysis of medical records, lab sources, data from wearables, health applications, genetics tests, and census information can be used to provide insights about patients that enable providers to come up with the most precise plan of care possible.
The University of Michigan Comprehensive Care Center, in partnership with Thermo Fisher Scientific, has developed a clinical analytics algorithm that detects and marks common genetic anomalies in cancer patients. With the knowledge of genetic variants, providers can determine the best course of treatments for patients right away, saving patients unnecessary tests and treatments.
In another big data experiment run by Banner Health and Philips, patients suffering from chronic conditions like congestive heart failure, chronic obstructive pulmonary disease, and type 2 diabetes were equipped with telemonitoring technology designed to be worn in their homes. Using an algorithm to warn providers when there were signs of trouble, a mobile health team was deployed to the patient’s home to intervene. Not only did patients receive more timely and targeted care in their homes, but hospitalization rates were down 45 percent, care costs diminished by 27 percent, and acute response costs saw a drop of 32 percent. As care becomes more customized, it can also become more cost-effective and comfortable for patients.
Better Public Health Insights
The interconnectedness of anonymized health data has massive implications for public health. For example, Blue Cross Blue Shield and Fuzzy Logix are combining their data to tackle the opioid epidemic. The program analyzed data points, such as frequency and location of doctor visits, and identified more than 700 risk factors that can predict a person’s likelihood of abusing opioids. Although the use of this data falls onto conscientious practitioners, the information holds the capacity to chip away at the current epidemic where, according to the Centers for Disease Control and Prevention (CDC), three dozen Americans die every day from opioid overdose.
The promise of big data to improve public health also lies in connecting data from inside the industry to the world. By looking at public health issues from a global and interdisciplinary perspective, McKinsey & Company predicts savings of $9 billion for public health surveillance alone and $300 billion in savings across the U.S. healthcare system more broadly. On a human level, big data can predict, prevent, detect, and respond to public health crises in an effective and timely manner.
The Limits of Big Data
Big data holds a great deal of potential to improve healthcare delivery in the U.S., but there are still plenty of factors currently in place holding big data back from its highest potential. One of the limits for big data is the variability in the quality of data. The amount of data that EHR systems have been steadily providing since HITECH was implemented is impressive, but the data can be messy.
Based on the EHR systems in place, data can vary widely according to the information that is required to be filed and how it is recorded within institutions. Human error, changes in medical research and medical trends, and the evolution of linguistic standards and operational protocols can limit the amount of data that can be properly analyzed and used in a meaningful way. That said, systems are being designed with this limitation in mind, where the systems analyze dirty data.
The promise of big data is in its capacity to make meaning out of a lot of information, which can be game-changing for the healthcare industry. Currently, the scale of analysis is limited by access and availability. Big data analysis can only output information that is as grand as the amount and scope of data available for analysis.
Presumably, access to and analysis of patient information from around the world would lead to the generation of comprehensive insights that could catalyze improved patient and public health outcomes on a global scale. However, in a landscape where healthcare is big business, and the U.S. healthcare system pays out based on volume, competition between healthcare organizations, research institutions, and individual practitioners can result in refusal or reluctance to share patient data.
Even if competition were not a factor, the proliferation of different EHR systems has resulted in cross-platform communication errors. Because there is no universal language for the collection of EHR data, sharing data often requires the collaboration of third-party vendors, which is costly to providers and ethically obscure when sharing sensitive patient information.
Despite the current limitations, big data analysis has already improved systems, outcomes, and the cost-effectiveness of healthcare. Of the 100 healthcare executives and thought leaders surveyed by Change Healthcare in its seventh annual industry research, two-thirds reported that clinical and data analytics had a positive impact on health outcomes.
There is still room for improvement with big data analytics, and healthcare executives have this improvement in their sight. The Healthcare Executive Group (HCEG) identified clinical and data analytics as the number one concern facing healthcare organizations in 2018. As researchers and data scientists work to overcome some of the current limitations of big data analytics, both healthcare consumers and providers will begin to see more novel and empowering changes fueled by meaningful data analysis.