One of the first things I learned as a physician-in-training was the value of a complete patient “history.” This usually refers to the story the patient tells the provider about his or her condition, and is obtained primarily through clinical interviewing. Patients want to communicate with healthcare providers. They want to be heard and understood. The patient/provider relationship is paramount in achieving this understanding. Despite its irreplaceable importance, however, there are limitations to the data that can be gathered through a patient interview. It is often challenging to obtain a complete and accurate medical story through that avenue alone. Fortunately, patients speak to us in other ways.
A patient’s medical record can provide additional details and context to the care conversation. The data contained within such a record holds great power and potential. Effectively accessing this data can allow our patients to “speak” to us in a way that isn’t possible through the patient interview alone. It can be challenging to access and synthesize the data contained within a patient’s medical record, however. Electronic Medical Records (EMRs), while of great benefit in many ways, are a frequent source of frustration for patients and providers. Data contained in such platforms is often difficult to access, fractured and fragmented, and overwhelmingly voluminous.
As with the patient interview, however, there is an art and a science to getting the most impactful data out of a medical record. Doing so depends on the integrity and accessibility of the data, and the ability to act upon it in a meaningful way. This highlights the two primary components of effective data science:
1) Data Management – Storing, organizing, standardizing, and making the data consumable.
2) Data Analytics – Processing, interpreting, and acting on the data in a meaningful way.
Leveraging the power of data starts with effective data management. In the clinical interview, the patient needs to be speaking to the physician in a language he or she can understand. Important components of the story should be elicited, identified, and recorded. Similar principles apply to managing data on a much larger scale. Patient data must be standardized and organized in a way that it is understandable and useful. Often, it is necessary to pull data from multiple different sources including clinics, hospitals, labs, etc. Without an effective data management strategy, the data can’t speak to us.
Terms such as “data warehouse” and “data lake” are used to define large data repositories. As with a physical warehouse or lake, finding a specific item or location is impossible without organization and classification. This requires effective data architecture. Data architecture is comprised of standards of data collection, storage, and integration. This is critical to any effective EMR, health data exchange, or care delivery system. It is also critical for the provider caring for an individual patient. If a lab test obtained in a rural clinic two days ago isn’t appropriately tagged as such, it may be invisible to the treating clinician. Leveraging the expertise of skilled Data Architects is crucial in making data accessible.
There are many components of effective data compilation and storage. One of these, is a common data model. An open and portable standard, for example, is the “Observational Medical Outcomes Partnership Common Data Model” (CDM), created by Observational Health Data Sciences and Informatics (OHDSI). The CDM is designed to help identify and evaluate associations between interventions, and outcomes caused by these interventions. Specific patients or patient cohorts (e.g., those taking a certain drug or suffering from a certain disease) may be defined for treatments or outcomes, using clinical events (diagnoses, observations, procedures, etc.). Advanced analytics, such as machine learning (discussed below), depend on such uniformity.
Once data platforms are developed to facilitate data management, data analysis can begin. Every provider has logged into an electronic patient chart to find thousands of data points. Chart notes, lab values, radiology reports, medication lists, and the like, combine to create an overwhelming amount of data. It can be difficult to see the forest for the trees. Even if the data is accurate, complete, and clearly organized, it doesn’t serve its intended function if it can’t be consumed and interpreted in a way that improves patient care.
So how can we as providers gain visibility into the most important data elements? While nothing can replace a discerning clinical eye, I would like to highlight three powerful data analytic tools:
1) Machine Learning – Algorithms and software have been developed to see patterns and opportunities in volumes of data that would be otherwise indecipherable. Any healthcare system collects more data daily than can be effectively analyzed without technological tools. Machine learning takes this to the next level by identifying potential opportunities, prioritizing them, and presenting them to the clinician for further evaluation, interpretation, and intervention. While machine learning does not supplant the expertise of the trained clinician, it can provide knowledge gained from analyzing a larger patient population than any one provider could see in a lifetime. As do clinicians, computer programs learn through experience. This experience can be leveraged to gain insights into an individual patient’s condition.
2) Predictive Analytics – Effective data science goes beyond presenting historical data. Predictive analytic techniques mine a patient’s history to identify future risks, needs, and opportunities. Such tools can help providers know which of their patients are at the highest risks, and what interventions might prevent problems in the future. We as providers already do this every time we treat a patient. We are constantly considering risk factors for disease development or progression, and trying to predict what will happen to our patients after they leave the office, hospital, or clinic. Predictive algorithms can provide tools that enhance and focus our clinical insights, helping us to prevent certain problems before they arise.
3) A Clear and Concise User Interface – Data must be presented in an understandable and actionable format. The key metrics and elements must be visible and obvious so that providers can get the information needed to succinctly treat their patients, instead of poring over spreadsheets. Clinician input is crucial in the development of any such interface.
Having a basic understanding of these and other data-related concepts is becoming increasingly important for healthcare providers. Clinicians need not be data scientists to leverage the power of health data, however. As with any specialized field, expert support is crucial. Data science tools and experts can help clinicians transform data into optimal patient outcomes.
Imagine a data system that integrates information across a wide breadth of sources. The data is stored and organized in a way that makes it easily and uniformly retrievable. Machine learning helps to identify patterns and individual predictors that are meaningful to clinicians and patients. The data is presented in a way where the truly important components are readily identified. Outcomes are predicted according to established probabilities and provided for clinical interpretation and action. Such a system would greatly enhance a provider’s ability to care for patients. While this may sound overly optimistic, we are getting ever closer to this ideal and understanding where we want to go is critical to getting there.
Well-managed, effectively-analyzed data allows us to communicate with our patients on a much deeper level. Without it, we are not getting the whole story.
Dr. Ryan Heyborne is committed to finding solutions to the challenges that abound in the healthcare environment. He currently serves as the Chief Medical Officer for Proskriptive, supporting health data analytic solutions that improve care delivery and patient outcomes. Dr. Heyborne is an emergency physician and remains clinically active. He obtained his MD from the University of Utah and completed Emergency Medicine Residency training at Indiana University. He obtained his MBA in 2015 from Northwest Nazarene University. Dr. Heyborne worked with Blue Cross of Idaho for over four years, serving as the Senior Medical Director.