American Statistical Association Discusses Leveraging Big Data

By Megan Williams, contributing writer

The American Statistical Association’s (ASA) Big Data Research and Development Initiative is in its third year of addressing the challenges around Big Data.

The initiative takes a focus on multidisciplinary teams as the most productive approach in achieving its goals. As a result of its efforts, the association has released the paper “Discovery With Data: Leveraging Statistics With Computer Science To Transform Science And Society.” The paper focuses on multiple, statistically-driven areas being transformed by Big Data and puts particular emphasis on healthcare, the challenges the industry faces, and potential results from taking a more scientific approach to the use of data.

The paper highlights methodological advances that are essential for application to the healthcare industry. The final section of the paper focuses on workforce issues and multidisciplinary teams.

Tackling Healthcare Initiatives

A recent McKinsey report estimates that the healthcare industry could stand to save $300 billion to $450 billion in healthcare costs each year through the use of data analytics. Statisticians can play an impactful role in realizing these savings through the improvement of quality and evaluation of care, and personalized disease risk prediction. They can also contribute to the process of addressing the challenges of using analytics in practice.

Quality Of Care

The paper notes that an emphasis on “best possible care” requires evidence-based, comparative evaluation of industry and organizational guidelines, procedures, and protocols associated with each clinical intervention. Many healthcare systems have built the systems needed to collect this data, or are in the process of creating infrastructures that can collect, store, manipulate, and access records of patient contact within the system. The ASA notes five specific areas in which this information can be sure to improve healthcare quality:

• Evaluating existing procedures across a range of domains including efficacy, side-effect burden, cost, and quality of life
• Constructing individualized patient predictions for health outcomes
• Identifying high-risk patients as candidates for a preventative intervention
• Monitoring diseases for outbreak detection and prevention
• Informing the design of and recruitment for clinical trials

The paper also notes, however, that using non-experimental data like EHRs (electronic health records), requires careful, statistical thinking to adjust for elements like sampling bias, treatment bias, non-standardized phenotype and diagnosis definitions, and evolving clinical guidelines and standards of care.

Research

A more statistical approach to research in healthcare holds the potential for many benefits.

If bias is accounted for, large amounts of data can be used for the prediction of adverse events, like stroke and negative drug reactions. As beneficial as prediction can be in the healthcare arena, it becomes particularly challenging when there are a large number of variables. Correction for bias though, can be facilitated through techniques like multilevel regression that allows adjustment for poststratification cells representing small segments of the population in question.

For Your Clients

For more on data warehousing and mining, read “Hospitals Turn Focus To Data Warehousing, Data Mining.”