Using Big Data for Health Outcomes

July 26, 2017 | By Eric Chen

The healthcare industry produces and makes use of a wide range of data to achieve its clinical and operational objectives, from insurance claims to electronic health records (EHRs), from laboratory reports to radiology images, and much more. The successful use of this data for population health management (PHM) relies on a comprehensive approach, as data aggregation in healthcare ultimately seeks to improve health outcomes for patients by providing physicians with a better picture of the context in which disease manifests.

Claims data offer a good starting point, due to its standardized nature, structured format, completeness, and availability. It includes demographics, diagnoses, and the dates and costs of service. EHR data then introduce the non-billable aspects of care, such as vital signs, allergies, and immunization records. However, it presents a larger hurdle for data aggregation efforts due to the high number of free-text fields. Ultimately, the next step to provide a more comprehensive view of patient health will be to incorporate social determinants of health, such as income, access to transportation, and education, as well as patient-generated data from surveys and Internet of Things devices.

With the high-volume of unstructured data, successful aggregation will need to not only statistically normalize it, but also account for potential duplication, standardize information from different sources, and remain flexible to allow for the entry of new data streams. Once these hurdles are overcome, however, data can be stored in a way that is not compartmentalized by department, facilitating cross-organizational communication and new applications for clinical trials, epidemic forecasting, and disease management.

Aggregation can be divided into two realms. Patient-facing aggregation involves the compilation of an individual patient’s health data, such as vital signs, radiology reports, and lab results, into a single dashboard like Apple’s HealthKit, which could then feed into the patient’s EHR. Doing so empowers the patient with access to their own information, which itself can produce better health outcomes. Alternatively, population-facing aggregation involves the development of databases with many patients’ information. This can then provide insights into population-level trends and, when leveraged correctly, lead to optimal interventions for different patients.

Attempts to develop these multi-patient databases are becoming more feasible with the growth of social media, Internet of Things devices, and cloud computing. These new platforms for healthcare data, however, also cause a new set of concerns about privacy because people are generally unaware of the types of data that they are leaving behind that are not covered by the Health Insurance Portability and Accountability Act (HIPAA), such as information from fitness trackers. Although these concerns are valid, the narrative surrounding big data analytics in healthcare should focus more on the data’s social value and addressing the technological impediments, including the standardization of data from different Internet of Things devices.

As a whole, the healthcare industry is moving towards patient-centered, data-driven, coordinated care. Entirely doing so will require a combination of descriptive, predictive, and prescriptive analytics. Descriptive analytics can be used as a starting point to benchmark patient outcomes and understand population health trends. In order to move into predictive and ultimately prescriptive analytics, however, real-time integration will also need to be built in as a capability of the data aggregation system.

Once achieved, data aggregation will equip clinicians with information that can be used to quickly identify potential complications for high-risk intensive care unit (ICU) or surgical patients. They will be able to not only derive population averages, but also provide specific, actionable health advice based on a patient’s age, sex, health status, fitness level, geographic location, and more. Furthermore, the timeliness of this advice will result in cost savings by allowing clinicians to act on a preventative basis.

Achieving meaningful population-facing health data aggregation will require solutions to a variety of technical issues such as late-binding aggregation, information filtering, and real-time integration. Nevertheless, it will provide an opportunity for physicians to make more informed health decisions, facilitating the medical paradigm shift towards prevention, early intervention, and optimal management.