Using Big Data in Clinical Nursing Research

By Nancy M. Albert, PhD, CCNS, CHFN, CCRN, NE-BC, FAHA, FCCM, FHFSA, FAAN

Real-world research generally refers to clinical research that uses real-world data as evidence. These data is observational; patients are not randomized into groups. Rather, all possible cases that meet inclusion criteria are used to answer research questions. Most often, data used in real-world clinical research come from big data, which is data that comes from large data sets, such as hospital electronic medical records and registries.

In hospitals and healthcare centers, big data is all around us. At our healthcare system, one of our favorite sources of big data is our billing database. It contains many patient characteristics and also many outcomes variables, such as hospital length of stay, intensive care unit stay and number of days, discharge disposition and 30-day rehospitalizations. Other big data resources are the medication administration database and long-standing registries that contain multiple variables surrounding surgical or medical procedures.

The beauty of big data is that it represents a high volume of a variety of day-to-day data that can be generated or retrieved quickly to provide insights to common problems and issues, including patient quality and safety. Once analyzed, teams can interpret findings in a way to make better future decisions or to implement new or altered interventions. Big data is easily retrievable, usually represents a lot of data and can be managed in Excel files.

In addition, many organizations with a research or quality focus have a program of big data collection and maintenance, especially organizations that are the source of Centers for Medicare and Medicaid Services (CMS) clinically based quality data. It is important for nurses to ask questions, both within their healthcare system and outside it, about the availability of big data and contact persons.

But there are some questions that must be answered before planning to use big data as research data. For variables being considered or used, questions are:

• Has each variable been defined in writing? And if definitions have changed over time, is that information available?
• Is the amount of data retrieved feasible? It is possible that the data may overwhelm the software tools used to capture and analyze data.
• Has data been recorded consistently? For example, is height data recorded in centimeters or inches?
• What is the missing data rate? Variables with greater than 25% missing data should not be used.
• How was the “quality” of data assured (if quality was even considered)? Were data collectors hired for the purpose of collecting registry data? Were data entered into the database as part of usual care by multiple clinicians, or were they entered by only a few people who were focused on high quality? Has anyone assessed the quality of data entered into a database, compared to the source data?

For outcome variables, it is up to nurse researchers to ensure that the research sample size represents the optimal size to answer the research questions. In addition, questions to ask are:

• Do available data match the definition of data needed to answer research questions? It is possible that data may be too ambiguous. Know what is available before getting started!
• Were data collected consistently over the period of time needed? Were compliance standards in place? Outcomes data need to be rigorous for all the time periods involved.

For some variables, it is easy to tell when data are inaccurate. For example, it should be a red flag to see more than a few patients over the age of 100 years being listed in an open heart surgery database; thus, nurse researchers would need to assess each case to be sure that the patient’s age was accurately recorded. It should be rare to see an adult female’s weight at less than 38 kg, even if the patient is short. It is likely that someone accidently made an inaccurate or transposed keystroke when entering data.

Not all research questions can be answered using big data. But, findings of big data analyses can be important to customer service, to operational efficiency, and to decisions that could affect future productivity and clinical outcomes. Further, research findings from big data can be used to ensure optimal delivery of medical and nursing care. With regard to ensuring high-quality patient care delivery, big data findings provide evidence of conformity to optimal medical therapy, and longitudinal analyses can show trends over time that allow sites to assess maintenance or progression of high quality.