What I like about the CDC ’ s Get Ahead of Sepsis campaign is that you ’ ll see this approach to electronic detection for sepsis gives a glimpse of insight into relative strengths and weaknesses of our current reporting program versus a full array of hospitalacquired nosocomial infections that are out there .”
— Michael Klompas MD ,
MPH , FIDSA , FSHEA
data sources and documentation practices ( such as EHRs )? I ’ d argue that there are some significant difficulties with our current surveillance targets of CAUTI , CLABSI and SSI . There are key definitions for each and applying these definitions are challenging because they are complicated , labor Intensive , subjective and in many cases , the definitions themselves are non-specific . It ’ s hard , and there are some of the questions that come up routinely in our hospital and I ’ m sure they sound familiar to you : Is this patient ’ s fever due to infection or some other cause ? Is this bacteremia due to abdominal infection or line-infection ? Is this coag-negative Staph a contaminant or a true pathogen ? Is this positive C . diff test colonization or infection ? Was this present on admission or was it hospital-acquired ? While we might have some answers to these questions , it is an internal struggle for us to try to apply these definitions . This is brought home by studies that compare efforts by infection preventionists to computerize algorithms .”
Klompas pointed to a study from 2010 in which infection preventionists ’ CLABSI surveillance was compared to the same surveillance being conducted by a computer algorithm . “ The algorithm found far more cases than the infection preventionist , and if hospitals are ranked by their number of infections , there is an inversion in terms of the lowest number of infections found by IPs had the highest infections found by computer algorithm and vice versa . This raises questions about the generalizability , the credibility of infection surveillance conducted by manual methodologies . The other factor is the infections we focus our efforts on are now rare – CLABSI was found to be 0.13 events per 1,000 patient days , CAUTIs 0.15 , MRSA bacteremia 0.05 , C . diff LabID was 0.37 ; the most common of these infections is colon SSI and that is only 2 out of every 100 procedures , so we are focusing a lot of effort on very few rare events .”
Klompas continued , “ And they are becoming rarer .” He pointed to the CDC ’ s point prevalence study in 2011 versus 2015 where a 40 percent decrease in SSI , 40 percent decrease in CAUTI , and 40 percent decrease in CLABSI were seen . “ Meaning that we are focusing more and more effort on an ever-diminishing , ever-vanishing set of outcomes that might be reaching their natural nadir . And finally , they miss the most common HAIs – those same point-prevalence studies told us that the most common HAI is hospital-acquired pneumonia , and of course that is not in our primary list of infections we commonly survey for and report . They also vary widely in their impact , some being quite severe , and some being measured by their costs to prevent , such as CLABSI at $ 45,814 per event , going down to CAUTI at a trivial $ 896 per event , and of course , missing out on VAP , which is one of the most impactful conditions , yet not under our current surveillance framework .”
He added , “ The other factor that is becoming increasingly important is that they miss non-device-associated infections . Yet we are focusing on CAUTI and
CLABSI and we are going to miss those non-device infections . So , what ’ s the alternative ? Where else can you possibly go ? I ’ d argue that the solution in 2021 and beyond is to leverage the breadth and detail of electronic health data to improve the completeness , timeliness and significance of surveillance . The CDC has taken some substantial steps in this direction , and the easiest example is the LabID events , which are based purely upon electronic discernable events in a given timeframe in a given location ; the limitation of that is that they are pathogen-specific and so for each potential pathogen of interest , you must have an additional LabID event defined and programmed .
The other foray that the CDC has made into more automatable surveillance using more subjective electronic data is ventilator-associated events , Klompas noted , pointing to the VAE calculator from the CDC , which he said “ realizes the objective of making surveillance more objective , dependable and efficient by taking advantage of data in the electronic health record . It can be done on the population level or the individual level by using the calculator .”
Another example Klompas pointed to is hospital-onset bacteremia . “ This is not routine yet but I think the data behind it are quite promising ,” he said . “ There ’ s a study from 2016 in which the authors said instead of focusing on CLABSI , why don ’ t we look at all bacteremias beyond 48 hours after admission . The rationale being catch any hospital-onset infection , not just CLABSI . By applying that methodology , they were able to demonstrate that HOB is far , far more common compared to CLABSIs . It allowed for more discrimination between different kinds of hospital ICUs versus using CLABSI rates .”
Another example of a condition that is not yet part of routine electronic reporting is sepsis , Klompas emphasized . “ What I like about the CDC ’ s Get Ahead of Sepsis campaign is that you ’ ll see this approach to electronic detection for sepsis gives a glimpse of insight into relative strengths and weaknesses of our current reporting program versus a full array of hospital-acquired nosocomial infections that are out there .”
There are 1.7 million cases of sepsis among adults per year in the U . S ., accounting for 20 percent of ICU admissions in North America , and representing 37 percent to 56 percent of hospital deaths .
“ About 20 percent of cases are hospital-acquired and therefore you would say these are nosocomial events , and the question is , how well do we detect those and track those with our current limited set of reportables ?” Klompas asked in his Decennial Day presentation . “ One way to try to determine this would be to say , let ’ s use diagnosis codes ; the challenge is when you try to get a diagnosis code , there are lots of different ways to define sepsis , and using diagnosis codes is inherently flawed . If you are trying to find a hospital-onset event , a diagnosis code does not differentiate and so it is an imperfect strategy , and secondly , when you are looking for organ dysfunction