Try as we might to avoid it, in the business of healthcare analytics, some jargon is bound to seep into the vernacular. The best we can do is try to limit it in our communications — or, even better, lay out these seemingly complex concepts in practical terms. In doing so, we often find that what might initially read as technospeak actually makes complete sense with a little bit of context.
Case in point, spatiotemporal analysis — specifically, leveraging spatiotemporal data to better understand and improve the health of populations, while reducing healthcare costs.
Stay with us here.
Before we dive into the cosmos that is space-time data analysis in healthcare, let’s start by taking a brief look at the concept of population health. Population health, as defined by Kindig and Stoddart in 2003, refers to “the health outcomes of a group of individuals, including the distributions of such outcomes within the group.” These outcomes are determined by a number of clinical and non-clinical factors, including social determinants of health, or SDoH.
Population health management, or PHM, therefore, deals with understanding and improving the health of a defined population. An effective PHM strategy takes into consideration the wide-ranging factors influencing a person’s health, including SDoH, and utilizes personalized engagement and care management strategies that empower patients to live healthy.
Now, a closer look at SDoH.
The Centers for Disease Control and Prevention (CDC) defines SDoH as the conditions in which people live, learn, work and play. For the purpose of the topic at hand, let’s consider how where a person lives can affect their health.
A metropolitan area that lacks a pedestrian-friendly infrastructure may have a higher obesity rate than, say, a pedestrian-heavy city. Indeed, research has shown a correlation between walkability and obesity rates.
- People living in rural areas or areas with limited transportation options may lack affordable access to healthy food, which has several negative health implications — from a higher prevalence of obesity and chronic disease to developmental and mental health issues in children.
- Living in an area with high air pollution levels has been linked with disproportionately high rates of asthma and other respiratory diseases. A recent study from Harvard even linked long-term pollution exposure to higher COVID-19 mortality rates.
Understanding these geographical considerations, as well as how they may change over time, allows us to support healthcare organizations and their provider networks in delivering more proactive, comprehensive and, thus, more effective, care to patients and patient populations.
Here’s where spatiotemporal analysis comes in.
Spatiotemporal analysis comprises two forms of research: spatial and temporal — i.e., space and time. Spatial analysis generally deals with space in the context of geography, while temporal analysis refers to variance over time. Studied together, spatiotemporal analysis allows researchers to better detect and follow patterns and, as the Columbia Mailman School of Public Health explains, discover emerging trends that could indicate future health risk.
Overlaying spatiotemporal data with population health metrics — as ascertained through claims data — can provide an indirect method for understanding and improving the health of populations.
For example, claims data can identify patients in a certain geographic region who have been receiving medication associated with diabetes and/or high blood pressure for some time. Based on this information, there is a high probability that these patients could soon be prescribed a lipid-lowering agent, such as Lipitor or its generic equivalent atorvastatin. This presents an opportunity to improve health outcomes and curb costs at population scale — leveraging spatiotemporal analysis to identify patients who have yet to experience the hallmark symptoms of disease and intervene before they become chronic and costly conditions.
Here’s how we’re using spatiotemporal techniques to turn data into action.
At HMS, we apply an ensemble machine learning approach (buzzwords to tackle another day, but, in short, a combined decision method that improves algorithm performance) to extract key data insights and observe patterns that can be influenced by spatial and/or temporal effects. We then use these insights to create predictive models that mimic our clients’ systems, taking into consideration seasonality, or cycles that repeat over time.
Based on previously annotated spatiotemporal data, we uncover trends that help healthcare organizations anticipate events, such as when providers will need to scale their operations. With this information, we can then identify and pursue measures that ensure providers are equipped to deliver the highest quality care through the most efficient means possible.
Let us explain.
In our Beyond Buzzwords series, we’re working to decode some of the jargon surrounding healthcare analytics (see previous articles, Applications of Artificial Intelligence in Healthcare and Anomaly Detection in Machine Learning) by demonstrating how we are applying these concepts at HMS to move healthcare forward. Learn more about our analytics-driven population health management and payment accuracy solutions at hms.com.