- From the basic electronic health record to the health information exchange (HIE), clinical decision support (CDS) system, business intelligence ecosystem, and big data analytics dashboard, most health IT infrastructure is geared towards achieving one ultimate goal: providing more sophisticated insights, answers, and suggestions to decision-makers at the point of care.
The healthcare industry has faced any number of well-documented challenges when it comes to piecing together their patchworks of legacy tools, best-of-breed offerings, and multi-vendor products to develop an integrated, interoperable data pipeline, but few challenges are greater than the ones involving the healthcare data warehouse.
It is certainly possible to conduct advanced analytics without a warehouse in place, but this centralized repository for information can be an invaluable asset to the forward-thinking provider – if they understand what they’re getting themselves into.
With several different basic forms and any number of acronyms unfamiliar to those without a degree in data science, exploring the options for data warehousing can be difficult for the uninitiated.
In this primer, HITInfrastructure.com takes you through some of the foundational concepts and key terms surrounding the healthcare data warehouse.
What is the data warehouse, anyway?
The data warehouse is a centralized repository for data that allows organizations to store, integrate, recall, and analyze information. Healthcare organizations may wish to use their warehouses perform clinical analytics using patient data stored in the EHR, or they may try to improve their financial forecasting by diving into business intelligence and revenue cycle analytics using claims and billing codes.
But choosing the right model for your warehouse is essential for developing the capabilities your organization is after. While a data mart model may allow limited analytics on simple, smaller subsets of data, its quick, easy, and inexpensive structure may not be all too useful for multi-faceted predictive population health management.
Providers may wish to opt for an enterprise-wide data model that encompasses multiple subject areas, and gives the organization additional power to match up data sets from all corners of the organization.
A late-binding data model, such as a semantic or graph database structure, is cutting-edge technology that allows users to ask even more complex, free-form queries. These databases do not immediately sort data elements into discrete, immutable categories, but keep information in a fluid form, allowing data scientists to develop new query capabilities on the fly.
While semantic databases offer incredible power for gaining actionable insights into innovative areas of research or patient management, the data curation effort and infrastructure requirements may be too steep for many smaller providers.
Data warehouses can be located within the four walls of the healthcare organization, or it could be located in the cloud. Infrastructure-as-a-service and analytics-as-a-service are two quickly growing markets segments that could reduce the immediate up-front investment burden for providers looking to dabble in big data.
What can a data warehouse help me do?
Healthcare analytics is a nearly limitless field to explore, but it can be broken down into three major areas: descriptive, predictive, and prescriptive analytics.
Descriptive analytics is the story of what has already happened. It can help to chart the growing number of patients who visited the emergency department last year, or track how much money was spent on overtime for nurses within a given six-month period.
Descriptive analytics is rarely real-time, and may be on a delay of several weeks or months, but this type of reporting still affords an incredible amount of insight into the operations of a hospital or the clinical quality of services provided to patients.
As a result of the EHR Incentive Programs, electronic health records can usually provide basic descriptive reporting that can be used for analytics. But you may require a data warehouse of some sort to move into predictive analytics, which demands real-time data manipulation capabilities and a high degree of integration and interoperability between disparate systems, such as bedside monitoring devices and the EHR, or social media streams and geographical information.
Interoperability is one of the major roadblocks of predictive analytics because unstandardized or incompatible data elements cannot be compared and contrasted without being translated into the same type of information first. A data warehouse that normalizes information before it is used for analytics could be the key to solving this fundamental internal problem.
Prescriptive analytics is the ultimate goal of every data warehouse owner, but it is currently beyond the reach of the majority of healthcare organizations. Instead of just describing what has happened or predicting what might happen, prescriptive analytics delivers actionable suggestions about how to avoid a problem all together.
The industry may not be quite there just yet, but it is certainly on its way to leaping past many of the obstacles in its way. A recent report from IDC Health Insights predicts that it may take only a few more years before half of the healthcare system’s big data analytics problems are solved, especially as providers continue to invest in analytics tools to leverage the data they already own.
How do I choose a health IT infrastructure pathway that’s right for my organization?
Picking the right vendors and the best foundational technologies can be a difficult proposition for providers just starting out on the big data analytics journey. While the vast majority of providers now have an EHR in pace, finding complementary technologies that won’t break the bank is a complicated mission.
Healthcare organizations should start the technology selection process by conducting an internal assessment of what tools are already available and what products, staff members, budgets, and timelines are required to fill in the gaps.
Providers should clearly map out what they hope to achieve by implementing a data warehouse. Will your big data analytics focus be primarily clinical? Do you hope to report on business intelligence or financial analytics, as well? Will your project stay limited in scope, or would a more flexible database option prepare you for the unknowns of the future?
Each organization has its own particular needs and vision, but most big data analytics projects will rely on data warehouse capabilities at some point in their development. Choosing a warehouse that is economical but robust and scalable will allow organizations to take part in some of the most exciting innovations in the healthcare data world.