Success in today’s data-driven healthcare industry will be increasingly defined by leaders who understand data science. This knowledge will be critical as executives build and guide teams toward a harmonious, well-planned vision for healthcare improvement that fully harnesses data’s capabilities.
Up to 30 percent of the world’s warehoused data comes from the healthcare industry. There’s significant opportunity for healthcare improvement in this information cache, including an estimated $300 billion in annual cost savings. But the industry can only welcome these prospects if health systems fully leverage data to identify areas for improvement and promote evidence-based care. Even with this massive data potential, healthcare too often relies on outdated technology. For example, up to 75 percent of medical communication still occurs via fax machine (in an era where automotive companies use data science to add navigation capabilities to cars).
Technology has laid out the opportunities, but, to realize gains in the digital era, healthcare leaders must understand data science and the urgency of investing in data science resources (technology and people).
As a broad term, data science means pulling information out of data, or converting raw data into actionable insights. Data scientists are knowledgeable in their subject matter (e.g., healthcare clinical data) and statistics, and use computer programming skills to tell the computer how to leverage data to derive insights. Data scientists augment traditional data analysis by automating the process of insight delivery through code. This automation can bring efficiency gains and new depths of insight to analytics, and enables real-time predictive analytics by reducing the time it takes to go from data to prediction.
The heart of data science is machine learning models, which are basically statistical models that can be used to extract patterns from data. Data science and machine learning can also be thought of as using the power of modern computing to leverage statistics. Some machine learning models, such as regularized regression and decision trees, lend themselves well to deriving insights and explaining patterns in data (e.g., which clinicians are over-utilizing costly materials). Other machine learning models, such random forests and neural networks (deep learning), are primarily used for prediction (e.g., each patient in a population’s likelihood of readmission after discharge).
Healthcare has long relied on data and data analysis to understand health-related issues and find effective treatments. For example, researchers have used double blind placebo-controlled studies as the foundation of evidence-based medicine. Such studies generate data about the treatment under evaluation and analyze that data to determine whether the treatment is effective, as well as understand its side effects. As a method of generating data and insight, this study process works in a spirit similar to data science, but is costlier and more time consuming.
Today, healthcare needs data to optimize patient outcomes with evidence-based practices more than ever; those insights are waiting to be discovered in data that has already been collected. With data science, the industry can find efficient, cost-effective ways to harness vast amounts of existing healthcare data—to maximize its potential to transform healthcare with faster, more accurate diagnosis and more effective, lower-risk treatment.
Researchers from Stanford University have developed a model that can diagnose irregular heart rhythms (arrhythmias) from single-lead ECG signals better than a cardiologist. Clinicians record more than 300 million ECGs annually, so the data needed for improved arrhythmia diagnosis already exists. With data science, health systems can leverage this information to make more accurate and more efficient diagnoses.
Another group of Stanford researchers has developed a diagnostic model for skin cancers that uses AI to classify images of skin lesions as benign marks or malignant skin cancers. This model, which can classify lesions as accurately as board-certified dermatologists, can potentially save health systems and patients time and cost by transforming the multistep process of diagnosing skin cancer (visual diagnosis, clinical screening, and possible dermoscopic analysis, biopsy, and tissue examination) into a single-step data analysis. While models are not designed to replace clinicians, they can provide valuable diagnostic guidance, making the care process both more efficient and more effective.
Mission Health wanted to improve the accuracy of its readmission risk assessment, so it leveraged machine learning to develop a predictive model based on its own patient population. Mission was using the LACE index to predict risk for readmission, which, while somewhat helpful, was developed using a patient population from Canada that was notably different from Mission’s demographic. With a machine learning model that used its own population, Mission improved its readmission risk prediction to outperform LACE and achieve a readmission rate 1.2 percentage points lower than its top hospital peers.
Data will continue to be a dominant factor in healthcare delivery and outcomes improvement. For organizations to successfully navigate the complexity of a data-driven world and embrace improvement opportunities, healthcare leaders must understand data science; they must become students of data science, understanding how it’s working in other companies and its implications for their health systems. And, if they haven’t already, leaders must start developing data scientist skills on their teams.
Would you like to learn more about this topic? Here are some articles we suggest:
Would you like to use or share these concepts? Download this presentation highlighting the key main points.