Guest Article: Crawl, walk, and then run towards analytics and big data in healthcare

I posted an article recently concerning the need to be more practical in the use of data vs. the need to go after the latest buzzwords, i.e. Big Data. Dan Reber posted a great comment on the article that I found enlightening so I reached out to him to expand on his thinking. Dan is in charge of product strategy at Origin Healthcare Solutions for their Business and Clinical Intelligence application (Precision.BI). Given that Dan’s been doing data warehousing and BI in some of the largest university-based medical groups in the country as well as having trained many users on the design of their reports I thought he’d be a great contributor. Here is what Dan had to say about how to crawl, walk, and then run towards data analysis in healthcare:

I recently watched a TED Talks video by Matt Ridley that discussed how present-day ideas evolve from past ideas and how these ideas often combine to form new ones. The video revealed how it takes roughly a million people to create one modest computer mouse – from the engineers and designers, to the oil rig workers who contribute to plastic manufacturing, to the fabricators who put it all together, to the coffee producers who supply the energy for the said engineers, designers, oil rig workers, fabricators, and so on and so on.

This video was an eye-opening example of how technology revolutions are accomplished – through the culmination of ideas.

Healthcare organizations (HCOs) have only recently begun embracing the notion of analyzing the data they’ve been entering into billing and clinical systems for years. They’re doing this using tried and true products and processes that have evolved over the years within other industries. But there’s a problem with this, those products and processes rely on ideas that came before it and had been applied in those industries. In other words, healthcare is starting on step 10 before completing steps 1 through 9.

Technology in healthcare is like the platypus – a piecing together of ideas that have not evolved properly.

I believe there’s a need for evolution in healthcare technology, data analysis in particular, before the latest buzzwords can be implemented. As Shahid has said previously, we need to crawl, walk, and then run. Here are a few significant steps I believe are required for healthcare data analysis to significantly evolve from crawling to walking:

Exception Reports – It’s essential that every organization have standard financial reports to analyze the status of its billing and receivables. These reports are pretty straight forward. However, clinical reporting and quality measures are relatively new and have various requirements that call for strict adherence to procedures to be certain of accuracy.

  • To ensure procedures are followed, exception reports must be created to catch when these procedures are not followed.
  • The most valuable exception reports must capture the “old way of doing things,” to make sure they remain “the old way.”
  • When procedures are changed, new exception reports must be created.

Analytics Department Setup – Many HCOs have their IT department handle all reporting needs. This may work on the financial side (although I advise against it) but it will not work for clinical reporting and quality measures, as the domain knowledge just isn’t there.

Below are three professional positions I believe are essential for a successful, and more importantly accurate, analytics department:

  • Implementation Champion – This position is responsible for the overall direction and vision of the implementation. May oversee projects that have high visibility throughout the organization. The individual is usually the CMO/CMIO or Medical Director and must have the ability to change clinical procedures within the organization. This position can be broken down into Quality Champion and Process Champion
  • Implementation Director – This position is responsible for the day-to-day reporting and operations of the analytics department. The most efficient departments I’ve seen are those with a physician or RN, or both, to lead the operations
  • Business Analyst – This position is responsible for the prioritization of requests and management of larger projects. The individual best-suited for this position is an SME of the EHR, and is usually part of the EHR implementation team.

Data Quality – I have seen many organizations disseminate reports before the data has even been validated. I am an avowed data geek and so I insist that data quality be top priority for any analytics department. Additionally, exception reports must not be confused with data validation. Exception reports do not take the place of validating the data, they simply show when a process is not being followed.

Below are a few important steps to follow for better quality data:

  • Data Comparison – Create an automated process that compares the data in canned reports in the host system, to data within the data warehouse. Only send alerts when the data does not match. Don’t be surprised, though, if the canned reports are incorrect. We have proven many to be inaccurate.
  • Validation Contests – At the beginning of most analytics implementations, many users will say that the data is inaccurate or just doesn’t look right. Why not implement a contest to see who can find the most issues with the data and even offer a bonus to the top three. This will do two things – help locate data issues (and there will be some) and help the users trust the data once said issues are resolved.
  • Spot Checks – Always, always, spot check on each and every report prior to sending out for the first time. Then continue with periodic spot checks on random reports going forward.

Data analysis in healthcare is challenging and when done incorrectly it will be inaccurate. My years of experience within the industry have taught me that the above recommendations will provide a simpler process with – more accurate results, trusted information, and most importantly, will result in better quality patient care.

Healthcare technology needs to first evolve and adapt so that the “big data” revolution can begin.

Newsletter Sign Up

6 thoughts on “Guest Article: Crawl, walk, and then run towards analytics and big data in healthcare

  1. I firmly believe the aspect of healthcare data that most strongly lends itself to the type of analysis leading to the revolution you describe – namely the sensible organization and hierarchy of uniform code sets – is the very thing that most compromises the industry’s ability to truly crawl, walk and then run to where it needs to be in that necessary order. In many ways, the buzz surrounding ICD-10, the fundamental-yet-powerful reporting that can be undertaken based on the myriad CPT/HCPCS codes and the awareness of data relationships ushered in by RAC audits have lulled many health systems into a false sense of confidence regarding their ability to rigorously analyze large data sets, which in turn yields that skipping of steps 1 through 9.

    It will be interesting to see if ICD-10 further perpetuates this collective belief or serves as a wake-up call. My suspicion is that the percolation of the ACO across the healthcare landscape will ultimately serve as the impetus for this recognition since an ACO – in my opinion – must inherently be a data-driven organization.

    Excellent summary – I enjoyed the read and thank you for pulling it together.

    1. Thanks, Daniel. Glad to hear that you enjoyed Dan’s guest posting. I am trying to reach out to anyone with technical experience to help practitioners make the right decision 🙂

      You are quite right that there’s a false sense of confidence in our ability to analyze data sets in healthcare and alot of work is yet to be done.

  2. Good article! I contend that one was to address this issue is to implement disruptive technology that enables people in loosely coupled social networks to share and play with analytic decision models in a way that promotes continuous evolution of such models (including using the methods described above). This means promoting collaboration among people from multiple locations and with different roles, responsibilities and experiences work together—across organizational boundaries—to compare different analytic models and test them for their ability to reflect reality accurately. They may, e.g., manipulate the models to represent different (‘what if”) scenarios, discuss the assumptions and results the models produce, examine the fundamental assumptions built into the models looking for logical flaws and inconsistencies, and debating about the assumptions and practical value of the model. This can emerge useful counterintuitive insights, raise new questions, spark innovation, and create value. In this way, the models would continually improve/evolve, and those that are inadequate would be rejected.

    For more, I wrote about this in a three-part post on my blog at

    1. Steve,

      I agree. What you are describing is a relatively new concept in BI – Collaborative BI. It is so new that it does not even have a full page on wikipedia…all I found was —

      I am a big advocate for “data for the masses” and I believe Collaborative BI is an extension of that ideal. We are actively designing our own Collaborative BI module for our software and I believe it will significantly add to our offerings.

Add Comment