Access to this type of large database can also shed new light on the cost-effectiveness of particular health treatments over time. This can be valuable in studies examining rare diseases, specific treatments, or particular patient subgroups, where access to a large overall patient population is necessary to ensure that the sample sizes are sufficient to draw statistically meaningful inferences.
These vast troves of health care information often require a larger scale of raw computing power, such as parallel processing, and new computing approaches that include machine learning procedures and predictive analytics. Health Care Practice Director Paul Greenberg also notes: “Increases in the quantity of available data can change the nature of the analysis required to draw meaningful conclusions. The scope of the data can cross a threshold at which the data become qualitatively quite different from smaller data sets, increasing the importance of human expertise in generating usable results.”
New types of data
As the quantity of available data has increased, information has also grown more diverse, resulting in the proliferation of new types of data that simply did not exist just a few years ago, such as next-generation smart-device biometrics and real-time patient data. Because these data are so new, appreciating the applicability of these information types requires a high degree of industry experience and creativity.
Although some researchers remain uncertain about social media’s compatibility with research requirements, such platforms unquestionably offer new opportunities for health information dissemination that can be leveraged in new ways. These new data could help to track adverse events both from a pharmacovigilance perspective and for drug safety analyses in the context of product fraud litigation. Although many aspects of social media data collection and analysis remain in early stages of development, the use of responsive, real-time information associated with the risk-benefit profiles of drugs is likely to play a larger role in drug safety monitoring over time. This will become particularly important as social media data analysis transitions from straightforward, volume-based assessment to include more sophisticated methodologies that consider cluster effects, herd effects, and source variations across websites to reduce credibility concerns. In a related article, Managing Principal Mei Sheng Duh addresses the use of social media data with respect to pharmacovigilance.
Combined data overlays
The increasing quantity of available data and the promulgation of new data types have also created opportunities to merge multiple, distinct data sets to produce novel outcomes. This additional type of big data could help to clarify potential drivers of prescribing dynamics in the context of various types of disputes. For example, government investigations of alleged manufacturer kickback payments to physicians can often be illuminated by such multilayered data. The insights obtained by overlaying physician prescribing data on speaker honoraria payment histories and event attendance lists can be invaluable in understanding the scope and impact of such events. Although each of these data sets likely was created in isolation, valuable complementarities can be elicited by studying them in combination.
For many years, de-identified medical and pharmacy insurance claims databases have been a rich source of information for health economics and outcomes research. These databases contain information about patient diagnoses and medical procedures rendered or drugs prescribed. Today, there are opportunities to overlay these data over provider clinical records – such as electronic health records (EHR) and laboratory test results – patient and provider surveys for “quality of life” measurements and treatment adherence patterns, and data generated from new biometric recording technologies to provide a comprehensive perspective on disease severity, treatment, and effects. Such insights could not have been easily obtained prior to the advent of the big data era and the cheap computing power upon which it is built.
