Thu, 07/12/2018

In my previous blog post, it was determined that data silos in healthcare require utilization of technology with capabilities in three core competencies: Integration, Integrity (Data Quality), and Intelligence.  Since these are not core competencies of the EHR, we should define the problem in more detail before considering further options. 

In this post, you will see three simple data illustrations to help increase our understanding of the issues.  

First, let’s look at some quality, cost, and operations information for patients or members using data from different systems. We’ll include EHRs, materials management, HR/Staffing, and shift/timekeeping systems.  Let’s group the information by gender, which is stored in each of these four systems as follows:

Patient Gender (22)

However, we only really want to group by the following six types:

Patient Gender (6)

This illustrates the need for integration, defined as having a centralized and systematic way of consistently bringing data together for common use throughout an organization. 

If we wanted to use this information for discharge planning and transition to home, then grouping by marital status becomes another opportunity of illustration.  A centralized approach to integration helps us go from this:

Marital Status (37)

To this:

Marital Status (16)

Analysts and query developers save snippets of code to help them properly join these kinds of data together, but it is incredibly time consuming, inefficient, and inconsistent.  Often it results in different answers as code snippets get shared, updated, re-written, or source systems change.

For example, an organization had a data scientist work for weeks trying to integrate data from a few source systems that had hospitals, units, and beds named differently in each database.  By the end of the ordeal, she was exhausted, and when she moved on to another department, her work went with her.

Often, analysts are hired for their critical thinking ability, and have masters’ degrees and/or PhDs.  Unless integration and quality are centrally  and systematically managed, these analysts spend 80% of their time doing basic types of data management, leaving only 20% of their time to develop a report, visualization, or dashboard. This leaves no time at all to actually analyze the results of their work.  

Most health plans and hospital networks use several data sources across their care locations.  Some have single EMRs, some have multiple EMRs, but clinical and financial data needs to be married across all care settings. It is important to also factor in wellness data from devices, lab data, surveys, home health, etc.  In addition to any  inconsistencies, factor in data that are incomplete, missing or outright erroneous and it EXPONENTIALLY increases data challenges for healthcare.

Now that we better understand our problem, we can better discuss strategy and solution.  It is critical to have a platform by which we can validate, harmonize and standardize our data sources before we can meaningfully group them.  For data quality and grouping, it is important to factor in the  1,500,000 types of code in Health Language sets ranging from CPT, ICD and DRG to SNOMED, RxNORM, and LOINC.

Once your data are matched, merged, cleansed and corrected, if you enrich them with the proper code sets and clinical ontologies, you can consistently stratify and aggregate information.  These metadata also give you the ability to implement enterprise data governance, business intelligence, and advanced analytics.   Our Omni-HealthData™ platform does all of this as well as data mastering across 6 key domains and provides access to pre-aggregated health and metric view data marts. 

In the next blog post, we will dive into Data Governance – how to start, organize, and build to sustain!