Last month we had
yet another example of a failure of health information technology to improve
some facets of care (see our March 2015 Whats New in the Patient Safety World
column CPOE
Fails to Catch Prescribing Errors). We began apologetically noting weve
had so many columns outlining some of the untoward consequences and other problems
with CPOE and healthcare IT in general even though we remain huge supporters of
CPOE and clinical decision support and IT applications in healthcare.
Fortunately this
month we have a health IT winner! Intermountain Medical Center has recently published
a study on the value of a clinical decision support system in reducing
mortality for patients with community-acquired pneumonia (Dean
2015). This is a great example of
what weve long visualized for the utilization of computer support to improve
patient outcomes.
Most physicians in
emergency departments have long used simple decision support tools like the
Pneumonia Severity Index (PSI) or the CURB-65 tool. These have been of some help
in decisions about whether a patient with community-acquired pneumonia might be
treated as an outpatient or admitted to the hospital (and, if admitted, whether
an ICU is indicated). Many hospitals have integrated these tools into their
electronic medical record systems or at least attached links to these tools.
More often emergency physicians simply use these tools on their smartphones. We
find these tools useful but they probably have a limited impact on actual
patient outcomes in the big picture.
Intermountain Healthcare
really took clinical decision support to the next level. The clinical decision
support tool developed at Intermountain lets the computer do work that the busy
clinician often does not or cannot take the time to do. Their clinical decision
support tool has been described previously (Dean 2013).
The IT system culls the patients
medical record and collects and analyzes data on 40 factors. When it calculates
the probability of pneumonia as 40% or higher it notifies the physician. If the
physician confirms the diagnosis of pneumonia the CDS tool provides treatment
recommendations and recommendations about severity that may help with patient
disposition (eg. inpatient, ICU, or outpatient). It
makes recommendations about antibiotics as well. For example, the tool looks for
risk factors for drug-resistant organisms such as a previous hospital admission
or an actual previous culture of a drug-resistant organism. It also looks at
where the patient is coming from, such as a long-term care facility which is
another risk factor for a drug-resistant organism (the CDS tool apparently even
cross references patient addresses with known addresses of long-term care
facilities). Thus, the CDS tool may suggest to the clinician that the patient
is high risk for a drug-resistant organism in cases where the clinician would
not have suspected it. It also provides recommendations about lab tests to
identify responsible organisms.
The currently
published study (Dean
2015) and the related commentary (Intermountain press
release 2015) describe use of the CDS tool and its impact on patient
mortality. The study was not a randomized controlled trial. Rather it was a
prospective study on almost 5000 pneumonia patients, comparing patient outcomes
for 4 emergency departments using the tool vs. 3 emergency departments not
using the tool. There was no statistically
significant difference for the total pneumonia population in severity-adjusted
mortality between intervention and usual care EDs but that included patients
with healthcare-acquired pneumonia as well as community-acquired pneumonia.
When the healthcare-acquired pneumonia patients were excluded, there was a
statistically significant reduction in mortality in the group using the CDS tool
(almost 50% lower mortality).
The CDS tool was
developed at Intermountain over several years following use of paper-based
pneumonia guidelines that had limited impact. The 40 factors utilized include 6
vital sign variables, 6 laboratory values, 25 nursing assessment variables, the
patients age, the patient's chief complaint, and findings extracted from the
chest imaging report using natural language processing. The team chose a
threshold of 40% probability of pneumonia to alert the clinician to balance
usefulness against the risk of alert fatigue (Dean 2013).
Screening tools that emphasize sensitivity generate lots of alerts, many of
which will be false alerts. So they chose to emphasize specificity instead. But
the clinicians also have access to the CDS tool via a desktop icon if they wish
to use it when the alert had not triggered.
Judging from the comments in the press release the CDS tool
is well accepted by the physicians in the emergency departments.
The real value of
computers in medicine is that they can grab data from multiple sources and
assemble that data in one place where complex calculations and rules can be
used to help with diagnosis and management. Few clinicians would be able to
spend the time getting all the data from those multiple sources nor be able to
remember all the rules and calculations that would be needed. This fine
implementation of a clinical decision support system is an outstanding example
of the potential value of such systems and its great to see how it translates
to better patient outcomes.
You can imagine
similar CDS tools to help with other conditions. For example, an antidiarrheal
agent ordered by a clinician for an inpatient might flag the system to activate
a CDS tool that would consider the possibility of a C. diff infection. The tool
could search for evidence of prior C. diff infection on a prior hospital stay
or check to see if the patient had been at other healthcare facilities where
exposure to C. diff is known to occur. Importantly, it can check multiple
sources to see if the patient has received antibiotics, the most common risk
factor for C. diff, or other risk factors such as use of proton pump
inhibitors. Based on the results, the tool could alert the Infection Control
team, suggest contact precautions be implemented, and suggest appropriate
diagnostic testing. If the patient does have a C. diff infection the tool could
also recommend appropriate therapy, including recommendations for recurrent C.
diff infection.
This is really good
work by Intermountain and reinvigorates our faith that clinical decision
support tools will finally make a significant impact on patient safety and
patient outcomes.
See some of our other
Patient Safety Tip of the Week columns dealing with unintended consequences of
technology and other healthcare IT issues:
References:
Dean NC, Jones BE, Jones JP, et al. Impact of an Electronic
Clinical Decision Support Tool for Emergency Department Patients with Pneumonia.
Annals of Emergency Medicine 2015; Published online: February 26, 2015
http://www.annemergmed.com/article/S0196-0644%2815%2900091-8/abstract
Intermountain Medical Center. "New study finds digital
clinical decision support tools save lives of pneumonia patients
." Medical News Today. MediLexicon,
Intl., 12 Mar. 2015. Web. 15 Mar. 2015.
http://www.medicalnewstoday.com/releases/290679.php
Dean NC, Jones BE, Ferraro JP, et al. Performance and
Utilization of an Emergency Department Electronic Screening Tool for Pneumonia.
JAMA Intern Med 2013; 173(8): 699-701
http://archinte.jamanetwork.com/article.aspx?articleid=1669105
Print PDF
version
http://www.patientsafetysolutions.com/