As data become more ubiquitous and valued, the prevalent “big data” enterprises have been providing analyses and pattern recognition that previously have been difficult to access. In a recently released article for the Physician Leadership Journal, “Questions to Ask When Considering the Use of Big Data.” Kenneth Poole, Jr., MD, and colleagues offer important considerations for physician leaders contemplating the use of their organization’s data alone or as a contributor to a larger collaborative.
The authors describe how large volumes of patient data can be used to help patients and the broader community; however, like most endeavors in leadership, this area is not without risk. Poole and colleagues propose a series of questions that will help physician leaders assess some of the risks associated with leveraging the large volumes of data our organizations use and determine their organization’s readiness for a collaborative endeavor.
As the authors point out, there are legal, security, ethical, and financial risks to using big data. Often, it is necessary to charter a team of the organization’s experts across these domains to answer the questions that the authors pose in assessing readiness.
As a physician leader, I have had the opportunity to collaborate on big data programming for organizations and communities. In one case, we were able to integrate clinical, genetic, and social data that led us to ask questions and find patterns to help us understand our care, our performance, and the health, well-being, and vulnerabilities of those we served.
I considered these efforts as the biggest strategic planning initiative the organization had undertaken because the data provided us with the power to truly understand the complexion of our community in ways that we had never imagined. We were able to benchmark and drive new programming to address our inadequacies and celebrate and market those areas in which we were high-performing. Of importance, we had partners who shared our vision and value in building a meaningful business case for data sharing — something we knew that we could not do alone. Physician leaders should contemplate big and small data applications to advance the work they do for their patients.
DATA FOUNDATIONS
The ability to manipulate, analyze, and use data to understand metrics and drive outcomes across operational, clinical, and financial domains is a core competency for every physician leader. Early in my career, I was fortunate that my mentors taught me that having a skill set to analyze, interpret, and act upon data was essential for my role as a physician leader.
At first, these skills came from practical discussions as I worked on outcome projects under the guidance of my mentor and the research leadership. Formal education in epidemiology, statistics, database design, and applications solidified my understanding and complemented my skills. Little did I know that a quarter century later, data would become the foundation upon which healthcare would be conducted.
Even more dramatic for me is that those same mentors did not have the luxury of personal computing or the modern tools for large database analysis that are now available on our laptops. My mentors manually collected data on paper-based forms, used computer punch cards to code each data collection sheet, and laboriously analyzed their work using a calculator and all those annoying statistical formulas we all have forgotten. Their work was truly a labor of love. I wonder what tools might be available in the coming decades.
DATA AS CURRENCY
Today, data are not only a foundation of healthcare, they are also, figuratively and literally, the currency of healthcare. The data manipulation process often begins at the keyboard of the frontline physician entering data in an electronic medical record (EMR). Little do physicians realize just how important their work is as they automatically document a patient’s care. It is the collection and documentation of these data through a clinician’s day-to-day clinical activities that ultimately influences what patterns are derived when these data are aggregated and analyzed from the EMR database.
From here, several downstream processes are applied to those data, all with the focus of preparing to submit a bill for services rendered. Data in the medical record are abstracted, coded, classified, and grouped in a variety of ways, including diagnoses, diagnosis-related groups, hierarchical condition categories, and severity of illness, all of which are derived from the clinical data the physician enters into the EMR.
Data validity can be compromised through these truncated classification processes when subtle differences in patient care are grouped together, when non-clinicians make interpretations from the record, and when attempts are made to optimize billing. This matters during analysis because rare events may be lost in these categorical grouping taxonomies.
LITTLE DATA
In contrast to big data, little data, which is much more specific and targeted data, can provide physician leaders with an opportunity to answer a broad range of questions that may have a more immediate impact on care.
Physician leaders at all levels should think of data as the substrate of clinical operations, and they must become adept at using these data to improve how care is delivered. Before the data leave your organization, consider whether you have fully utilized that data to improve the way you provide patient care so that you are not surprised by your performance report from outside of the organization. For starters, consider these questions:
Does our organization have the expertise across leadership to use data to drive decisions?
Do we have access to tools that allow leaders and managers to easily manipulate our data and understand how our core business processes are related to our outcomes?
Are we using the data to improve outcomes in our clinical and non-clinical departments?
Are we prepared to allow others to use our data to evaluate patterns and trends in our care that we may have ignored?
What steps can we take to improve our use of data across the enterprise this year? What investments need to be made, and how will we measure the impact?
LARGE COMPARATIVE DATA EFFORTS
Many states aggregate the volumes of data from hospital encounters and compile them into datasets for further analysis and research. The Agency for Healthcare Research and Quality has further integrated these hospital administrative datasets into the National Inpatient Sample, which provides extensive data on hospital encounters, patient characteristics, organizational structure, and resource utilization associated with each discharge. These datasets can be useful as research and policy tools but are often less reliable for the improvement of patient care.
In addition, many professional organizations, EMR vendors, and contracted entities receive an organization’s data through subscriptions for benchmarking. Often, these groups, for a fee, provide comparative performance across a range of indicators and member institutions for those who subscribe. Benchmarking compares your organization to others that are similar in terms of size, case mix, capability, and organizational structure. For example, you would not want to benchmark outcomes for an illness in your cardiac center to one that performs cardiac surgery and transplantation if you do not offer these services.
Using large databases has its advantages. For example, many of these databases provide broad characterizations of epidemiologic and system-level problems, thereby allowing inferences regarding large groups of patients. In addition, these databases provide important reference points for subsequent and specialized studies on specific patient groups, like the elderly, children, or patients with a specific condition like heart disease or undergoing surgical procedures. Also, the large number of discharges in these databases is especially important for attempting to analyze comparatively rare conditions or events with sufficient statistical power than can be analyzed in single institution studies.
While there are important advantages to participating in some of these efforts, realize that your patients’ data are valuable, and you must ensure that your organization is getting a sufficient return on sharing these data. This return can be measured in clinical or business terms.
For example, are there sufficient comparative reports that provide guidance on improvement efforts you should undertake? Will your team be trained and provided with software that improves their ability to use the outputs? Do you have a team of improvement individuals who can shepherd these projects to improve outcomes? How will your data be protected and used? Will your organization’s performance be rated in aggregate reports, and who will have access to those reports?
CONCLUSION
It seems that everyone is in pursuit of an organization’s data, which have become the commodity for healthcare; the organizations with the largest database of patients and physicians are winners. Physician leaders are uniquely positioned to provide expertise at multiple levels of the conversation regarding data. Poole and colleagues provide an important reference and some questions for those of you considering big data collaborations in which you share your organization’s data.
Physician leaders should ensure that before sharing this valuable commodity, they are getting the most out of their little data for the benefit of their patients, organizations, and communities.
Count your change so that you know beforehand what benefits your organization and community can expect from data-sharing and how those benefits will be derived and measured.