Michael A. Simon, a principal data scientist at Arcadia Healthcare Solutions, argues that everyone needs an analytics solution for different reasons. The issue is that your analytics solution may be holding you back – such that it will leave you without some of the most important features you’re looking for. In this post, Michael describes the three crucial properties to embrace if you are thinking about introducing a new analytics platform into your health IT environment, or you’re wondering whether your current system is really bringing you all the power you thought it would.

Everyone needs an analytics solution today!

Your reasons may vary. It may be the federal mandates, or the proliferation of incentives (and penalties) dependent on a growing number of distinct measure sets, or the increasing complexity of managing an aging population with chronic diseases. Or maybe you just don’t want to be left out of the Big Data Boom.

No matter the specific reason, it is prudent to seek out a powerful and effective analytics solution. The market agrees, as thousands of companies pour in to what will be a $20 billion industry by 2020 (IQ4I Research & Consultancy).

The catch is that your analytics solution may be holding you back. Despite all the potential and big promises, many analytics systems will leave you without some of the most important features you’re looking for – the ability to understand your population’s health state, to identify points of risk and mounting utilization, or even the ability to simply have confidence in the results.

That’s why, if you’re thinking about introducing a new analytics platform into your health IT environment, or you’re wondering whether your current system is really bringing you all the power you thought it would, you should consider whether it embraces these three crucial properties:

  1. Don’t Separate; Integrate Your Data

If your system only offers you access to claims data, you get a view of the world as it was a month or two ago. Or three. Or more. Claims data are easy – structured, clean and orderly –  but claims data also lack many crucial elements like patient vitals, important details from encounter notes, and activities that don’t typically get a notation in a claim. Wondering whether the patient smokes, and whether he or she is being counseled to stop? Claims won’t tell you.

On the other hand, an EHR will tell you those things, and much more. It is a real-time, clinically rich and hugely powerful dataset.  Unfortunately, if you are only using EHR data, you are forced to operate solely within the boundaries of that EHR. Absent is the vast domain of events that occurred outside that realm, like urgent care or ED visits. How many of our patients actually went to a specialist for that eye exam? And how many are currently taking anti-psychotics but haven’t been seen in months? The EHR doesn’t know.

But a well-integrated system does. An integrated analytics platform brings together both clinical and claims data to create a more up-to-date view of the world, both within the ambulatory center and out into a world of disconnected inpatient and specialist facilities. An integrated system offers greater completeness. It combines knowledge of an unexpected urgent care visit with the patient’s vitals for the months and years preceding it and offers considerable intelligence at every level of the healthcare world, whether it’s a care team discussing an overall population health strategy or a physician preparing to see a longtime patient after an unexpected event. And the person is at the center: Any integrated analytics platform needs a powerful Master Person Index, so there’s never any doubt of whom you’re dealing with.

  1. Connect Completely and Confidently

If data are the fuel that fires up your analytic platform, what’s your pipeline? All of the fancy analytics, performance measurement and predictive scores go to waste if they’re not fed a high-quality, clinically rich dataset. Does your platform use a condensed extract like the Continuity of Care Document (CCD)? Specifications like CCD offer a compact and convenient way to exchange information between defined roles for explicit purposes, but they also impose severe limitations. Depending on vendor or configuration, the CCD could exclude many of the data elements your analytics system needs in order to answer your questions, resulting in your taking a severe hit in the data-quality department.

Speaking of data quality, does your analytics platform ensure the accuracy of the data it takes in? Of course, it may never be able to tell you that the blood pressure entered should have been 130/70 and not 130/60. But what if it’s 40/70, or 12.6/70, or Yes/70? And what about making sure structured information are in a useful structure? If there are 71 different definitions of “smoker” in the system, analysis could get a little hampered.

What you need is a data pipeline, or connector, that is both complete and confidence-inspiring. A connector should have access to at least a Minimally Viable Dataset that supports basic reporting functions, population-level health analysis, and patient-level care management. Your connector should be transparent in how it handles the data-scrubbing process, and, like any good relationship, should be able to provide feedback on what your data look like and when there are deviations from your typical, “normal” mode of operations. And don’t forget the mappings: Your data should be mapped to appropriate standards (LOINC, CPT, NDC/RxNorm) once they pass through the connector, but there should also be knowledge about local customs and workflows and the ability to map those to understood clinical concepts so that any user can sit down at your analytic platform and spend time learning about people and health and not about how specific items are coded.

Michael Simon, Ph.D., is a principal data scientist at Arcadia Healthcare Solutions, with experience in data analytics, process efficiency, and experimental design. Since joining Arcadia in 2011, Michael has focused on demonstrating and enhancing the value of client data through exploratory data analysis and advanced statistical methodologies, and on the analysis of healthcare data quality. Michael is a former National Science Foundation AAAS policy fellow and earned his doctorate in biology from Tufts University in Medford, Mass. He also holds a Bachelor of Arts degree in economics and a Bachelor of Science in electrical engineering from Rice University in Houston, Texas.

Part II of this article can be found here.