Healthcare analytics is the application of statistical tools and methods to healthcare data. It enables analysis of the patterns of past events so they can be better understood and help inform future decision-making.
It’s a hot topic these days, and rightly so, since it’s increasingly evident that healthcare is an information-intensive industry. Fueling the demand for healthcare analytics are drivers such as meaningful use metrics, quality outcomes reporting and understanding population health patterns across accountable care organizations.
In the past, this kind of analytics was hurt by the lack of sufficient data, other than claims data. But as an increasingly large number of healthcare provider organizations adopt electronic health records and continue to make major investments in other IT systems, data capture is no longer as much of an issue.
However, the obvious primary function of most health IT systems is to automate clinical and business transactional workflows. The use of this data for aggregate analysis and predictive modeling remains, at best, a secondary concern for many vendors. So while the sheer amount of data being captured electronically is growing exponentially, accessing this data effectively and combining it with data from disparate systems in order to see the full picture remains an elusive goal.
Laying the Foundation
Despite the claims of many vendors of business Intelligence systems, the key to establishing a solid foundation for healthcare analytics lies not in the technology itself, but rather in how data is managed across the organization. Successful analytic initiatives begin with an intentional strategy – typically referred to as data governance – that oversees how data is managed. Data governance treats data as a critical asset to the organization, and it comprises several areas. These include:
- Data standards: These set specific definitions for data and how it will be structured, including the use of standardized vocabularies (such as LOINC and SNOMED) where appropriate. Governance around data standards also impacts front-end system design, for example, defining the permissible values that can be selected from a drop-down menu.
- Business processes: These affect how data is captured and treated. How race and ethnicity data is captured during a registration process, for example, can impact the accuracy of subsequent demographic analyses.
- Data quality: This refers to the tools and processes that ensure data are correct, complete and valid. A thorough data quality review process includes analysis of how the data is being captured, transformed and normalized, and presented in the analytic tools. It is also an ongoing process to ensure the validity of data as upstream systems and business rules change over time.
SEE ALSO Increasing Importance of Data Analytics
Maturity Models in Healthcare Analytics
Most healthcare organizations today perform some form of retrospective data analysis. By reviewing what happened in the past, analysts can identify trends around inefficiencies in operational departments, areas needing improvement in patient satisfaction, or variability in clinical outcomes for disease populations. Typically this will result in some intervention based on the data, and then it is resampled to determine whether there has been an improvement. While this type of analysis can help identify patterns, decision-making is still based on historical information, somewhat like driving while looking in the rearview mirror.
The next level of maturity in healthcare analytics is the use of predictive modeling for forecasting. Predictive models are developed from the analysis of historical data combined with statistical methods, such as regression. After the model is created, analysts then can use new data to determine the likelihood of a future outcome. Depending on the complexity and validity of the model, predictive analytics can be very useful in decision-making, such as resource adjustments or patient compliance with certain types of therapy.
Further along on the healthcare analytic maturity continuum is simulation and scenario planning. Requiring fairly advanced rules engine development and operations analysis, these systems enable “what if” type scenarios in which users can adjust variables within the system (for example, emergency department staff) to see how outcomes (such as patient wait times) are affected.
SEE ALSO Big Data for the Small Practice
Future Trends in Healthcare Analytics
Despite wider adoption of electronic health records (EHRs), a large proportion of healthcare data remains unstructured, and that hampers refinement in healthcare delivery. For example, knowing a patient’s response to treatment, and aggregating this information across hundreds or thousands of patients is critical to doing comparative effectiveness analyses. Yet this data is often captured in the free text or dictated portion of the clinical note. Advances in areas such as natural language processing and various text analytics methods are beginning to demonstrate some effectiveness in understanding context and extracting unstructured data so it can be analyzed.
“Big Data” is a term that is being hyped these days by database and analytics vendors. It refers to datasets that are so large (think terabytes and exabytes) that more traditional methods of capture, storage, sharing and analysis are unwieldy. While this is rarely a current problem in even large healthcare organizations with EHRs, it will increasingly become a real issue with the advent of affordable whole genome sequencing and other genomic data sets used for personalized medicine.
Big Data will require new technologies and deep analytic talent in the near future. The McKinsey Global Institute (MGI) has described Big Data as the next frontier for innovation, competition and productivity. In examining the healthcare sector, MGI estimates that the effective analysis of Big Data would create more than $300 billion in value each year.
Mark Hulse is vice president and chief information officer at Moffitt Cancer Center and Research Institute, Tampa, Fla.
