Trends in Sepsis Incidence and Outcome: Potential Implications of Mismatch between Administrative versus True Incidence

Background:

While administrative data suggest increasing incidence of sepsis, discussion remains on the agreement between administrative (billing) data and true disease incidence. This is particularly important as 1) billing data depends on clinical reporting and 2) the recent introduction of a sepsis core measure by the Centers for Medicare & Medicaid Services (CMS) linking sepsis care and outcome to payment. We aimed to compare administrative data to true severe sepsis incidence and outcome trends, rigorously confirmed and monitored at our hospital.

Methods:

We used retrospective data from an 1171-bed academic medical center with a hospital-wide EMR-embedded sepsis screening program with part of the infrastructure a rigorous clinical adjudication process for those screened positive. Data on monthly incidence (per 100 discharges; January 2012-September 2015), mortality and median length of stay (LOS) was plotted for sepsis definitions from administrative data described by Angus, Dombrovskiy, Martin, and Wang, and were compared to clinically adjudicated severe sepsis/septic shock cases (May 2014-March 2016).

Results:

With substantial differences between definitions, the incidence was lowest for clinically confirmed severe sepsis. Incidences from administrative data increased up to 40.4% (Martin; 2.9 to 4.1) while true severe sepsis incidence increased with 26% (1.27 to 1.60). Sepsis mortality rates decreased at rates between -43.3% (Dombrovskiy) and -51.3% (Wang) for administrative data; this decrease was less pronounced for clinically confirmed severe sepsis (-40.7%). Interestingly, median LOS changes were less pronounced using administrative data while a substantial increase was noted for confirmed severe sepsis: median LOS change from 9 to 12 days. 

Conclusion:

These preliminary data provide an important insight into the mismatch between sepsis incidence from administrative data versus true incidence of severe sepsis. While part of the mortality decrease indisputably stems from efforts to screen and standardize care we believe that the mismatch between true and extracted incidence may also partly be due to inaccuracies in coding that may follow from suboptimal clinical reporting. This will affect the sepsis CMS core measure reporting which is dependent on accuracy and timing of provider documentation to the EMR.