Early sepsis detection software has a role in CDI/coding efforts, but not a substitute for trained professionals

By Brian Murphy

An interesting piece up on the blog of Epic: Reducing mortality and saving money by identifying signs of sepsis sooner.

(Note: This is not a commercial for the EHR giant, for which I hold no financial interest).

Per the article, after implementing v. 2 of Epic’s Early Detection of Sepsis cognitive computing model, Saint Luke’s Health System in Kansas City, Missouri, reduced order-to-antibiotic turnaround time by 32%, sepsis mortality index by 16%, and improved CMS bundle compliance for sepsis by 48%.

That’s pretty impressive. See link to full article below.

I find this topic fascinating, as we’re starting to see additional early detection tools in play beyond just sepsis. Remote patient monitoring reportedly reduced 30-day readmission rates at another organization by 50% (see link below).

Keep in mind these are commercial products, but the fact that they are case studies of real organizations adds weight.

Reading these, I started asking how CDI and coding professionals integrating with these technologies in their day-to-day work, and how accurate they find them. And asked a poll on LinkedIn, which yielded some interesting results:

Over the years I’ve heard widely varying reports on early sepsis detection software. 

Some have told me they use these tools extensively and with great success, others have reported huge numbers of false positives, or generalized provider “alert fatigue” that resulted in shutting them off altogether.

But I am encouraged that these tools could reduce provider fatigue and aid coding and CDI professionals at the same time. Again from the article, St. Luke’s decided that when the model identifies a patient as potentially having sepsis, nurses in the ED don’t need to consult with the ED provider before initiating treatment. “Nurses are the ones to say, “The patient needs to go down this pathway. Let’s draw some labs, insert an IV, and increase monitoring.” The predictive model streamlines their workflow and empowers them to initiate treatment quickly.”

As coders cannot code from nursing documentation (save in some limited instances I won’t get into here), this type of scenario won’t help them too much. But, a CDI professional could put it on his/her radar for query. And the additional documented workups and monitoring could help a provider render a definitive diagnosis in shorter time.

In theory these tools should be getting increasingly sensitive and commonplace. 

The results of the poll are interesting. With only 8% of respondents describing their organization’s early sepsis detection software as “very accurate,” no one should be taking their output to the bank. Sepsis can be notoriously tricky to diagnose in its early stages, but clearly the accuracy of these machines is in need of work.

But, with the largest majority (58%) describing them as “somewhat accurate,” early sepsis detection software plays a role and should not be discounted, either.

Bottom line: Put a good CDI or clinically savvy coder in the mix with the right software solution, and it’s a powerful combination of getting sepsis detected early, diagnosed, treated, documented, and coded. Machines alone aren’t enough.

As one of my commenters said:

While accuracy of detection drives better care there are still many documentation barriers. Carrying the diagnosis of sepsis through to the discharge summary is still our biggest documentation barrier. Way too many times just the localized infection makes it to the DC summary and we are left chasing our tails in coding due to problem list “problems” and provider handoffs that don’t keep pulling forward the dx.

I would love to hear any of your additional thoughts/comments on the subject.

References

• EpicShare: Reducing Mortality and Saving Money by Identifying Signs of Sepsis Sooner: https://www.epicshare.org/perspectives/streamlining-sepsis-prevention-and-care
• HealthcareIT News: RPM reduces 30-day readmission rates by 50% at Lee Health: https://www.healthcareitnews.com/news/rpm-reduces-30-day-readmission-rates-50-lee-health

If you need help with your sepsis claims, a CDI chart review, or a coding audit, please reach out to Norwood at consulting@norwood.com.