Clinical Deterioration Tools

[SurfingDoctor]

I was curious people's experience with EMR based warning systems implemented to detect clinical deterioration and trigger alarms/warnings. We are implementing one based on several instances of human error that led to poor outcomes, but the data I've seen isn't very compelling that these systems do much more than increase ICU transfers. Does that fit with people's own experience or have people actually noticed a benefit to the systems being implemented? Just curious.

---

Members don't see this ad.

 

[nikolaite]

My personal experience is in reading charts and records from ER and inpatient setting. From what I’ve seen, once the EMR and BC (“background clinician”) software is installed, there is an interesting statistically significant phenomenon: 1) pt checks by nursing decreases, 2) pt visits by docs decrease, and 3) overall patient condition is recorded to decrease past the point of triggers.

The systems are ignored or bypassed in enough pt ER and PCU/ICU courses as to be glaringly obvious. Care is not coordinated by any one person. No one doc or nurse can answer as to diagnoses or pt course of care.

On top of all that, nurses and docs complain about the burdens involved in interface with the software...the oldies claim everything was easier on paper with in person vitals, etc...the newbs claim they didn’t get into medicine to stare at computers.

I don’t find them compelling either. End rant.

 

[BigRedBeta]

I was curious people's experience with EMR based warning systems implemented to detect clinical deterioration and trigger alarms/warnings. We are implementing one based on several instances of human error that led to poor outcomes, but the data I've seen isn't very compelling that these systems do much more than increase ICU transfers. Does that fit with people's own experience or have people actually noticed a benefit to the systems being implemented? Just curious.

No first hand experience, but my question involves the type of data going in to create the alarm coming out. If it's just an adaptation of VS changes, then it's still going to rely on human interpretation, and I wonder if it's really necessary. If it's something more subtle - eg HR variability monitoring - then that might be cool *if applied to the right patient population*.

I would challenge any system (and probably lose the argument) that is broadly applied without published evidence of validation in the patient population that's going to be monitored. Would also love to talk with the other hospitals directly about their experience....a list of references if you will.

 

[SurfingDoctor]

No first hand experience, but my question involves the type of data going in to create the alarm coming out. If it's just an adaptation of VS changes, then it's still going to rely on human interpretation, and I wonder if it's really necessary. If it's something more subtle - eg HR variability monitoring - then that might be cool *if applied to the right patient population*.

I would challenge any system (and probably lose the argument) that is broadly applied without published evidence of validation in the patient population that's going to be monitored. Would also love to talk with the other hospitals directly about their experience....a list of references if you will.

Ha, funny that you say that. Garbage in, garbage out. The patient who was used as an example and impetus for enacting it was a patient who had no documented BP for like 12 hours. I doubt they tool would have been very useful. But the administration views this as an opportunity to implement the tool instead of dealing with the real issues, ie staffing, education, escalation of concerns, etc.

But maybe there is a system out there that has already deal with and optimized those issues and still found the warning systems useful. I dunno.

 

[nikolaite]

It looks to me like the way these EMR systems work, for example, with telemetry, oximetry, or POCT handhelds, they rely less on human input than passive machine data gathering. The BC software is like an AI system that correlates all passive and human entered data and spits out a clinical score of predicted condition or deterioration. Nurses can also supply numerous important directly gathered data for these systems when they do periodic checks. The BC will automatically flag patient deterioration past some level, but some person still has to scrutinize the EMR data the BC has generated the flag on.

This is not necessarily simple as there are numerous entry points for labs, medication administration and vitals that are often duplicated. This makes a mundane task like writing up an order for discharge medications, for example, take 30 mins or more for three or fewer prescriptions. I saw it take one doctor roughly an hour to decide on how to direct a pt to take his prescriptions on discharge because he could not get anyone on the care team to report, nor find in the EMR, whether the pt had received his statin dose for the day. It took me making a timeline spreadsheet to find that the pt had not in fact been given his dose. I’ve seen it take a consult with billing (who was the fastest at running through the EMR) to discern whether a pt had been given lovenox. Talk about time waste!

Doesn’t sound to me like an aid for better staff morale or pt care, but rather some underhanded aid for the benefit of billing, legal, or insurance.

 

[nikolaite]

Maybe I just have a bad attitude, though.

 

[WheezyBaby]

We use one in our main peds hospital for ward patients. I don't believe any of our adult facilities use one. It's not terrible. I don't know our actual numbers, but anecdotally, I would say somewhere between 1 in 10 or 1 in 20 flag a clinically meaningful change in patient status, and there are probably 5 alerts or so per 24 hours. Not all of these change level of care (most don't) but may otherwise alter care. Our night shift nurses tend to be relatively less experienced, and our overnight physician coverage is largely interns supervised by senior residents. I just see it as a prompt for them to review vitals trends and assess the patient. There aren't so many that it gets to the point of alarm fatigue, and I think it's reasonable to have some back up monitoring. Pretty rudimentary tool currently but big data certainly has capacity to make more robust in the future. I honestly don't recall if ICU routinely shows up to all of these (they do show up to all rapids / codes). I think it would be reasonable to have no ICU team response

 

[chessknt]

My experience with these tools is that they add absolutely nothing to clinical care at the ICU level. Our local group has looked in to using these to identify sepsis patients early to avoid missing the SEP guidelines but it has not been very successful and, as we all know, SEP 1 guideline adherence has nothing to do with improving patient care.

These might have more utility in the floor setting with higher nursing ratios where things can get unnoticed for longer than they do in the ICU but even then data is lacking. It is one of those ideas that seems like it should work but in reality just doesnt. We did a project trying to see how trainees used these types of early chart based alert systems and literally nobody knew where they were in the chart and even after educating them on its location and purpose nobody (and I mean 100%) used them.

 

[WheezyBaby]

My experience with these tools is that they add absolutely nothing to clinical care at the ICU level. Our local group has looked in to using these to identify sepsis patients early to avoid missing the SEP guidelines but it has not been very successful and, as we all know, SEP 1 guideline adherence has nothing to do with improving patient care.

These might have more utility in the floor setting with higher nursing ratios where things can get unnoticed for longer than they do in the ICU but even then data is lacking. It is one of those ideas that seems like it should work but in reality just doesnt. We did a project trying to see how trainees used these types of early chart based alert systems and literally nobody knew where they were in the chart and even after educating them on its location and purpose nobody (and I mean 100%) used them.

None of our ICUs use one. I do think there's more /potential/ for an effective system here due to the huge amount of information to form "data phenotypes" for that a human might not as quickly put together eg Development and validation of a deep-learning-based pediatric early warning system: A single-center study ; uses a limited range of variables but can imagine incorporation of diagnostic code, etco2, ICP monitor trends, etc