Title: Do Sepsis Alert Systems Work? ... and some thoughts on clinical decision support and AI<br/>
Author: William Teeter<br/>
<a href='mailto:william.teeter@som.umaryland.edu'>[Click to email author]</a><hr/>
Link: <a href='https://umem.org/educational_pearls/4559/'>https://umem.org/educational_pearls/4559/</a><hr/><p><strong>Do Sepsis Alert Systems Work?</strong></p>
<p>Researchers in Korea completed a high quality systematic review and meta-analysis of sepsis alert systems for adult ED patients</p>
<p>Using high quality methods, they identified over 3000 studies with 22 meeting criteria. </p>
<p><em>They found these systems were associated with:</em></p>
<ul>
<li>reduced mortality risk (risk ratio [RR], 0.81; 95% CI, 0.71 to 0.91)</li>
<li>length of hospital stay (standardized mean difference, ?0.15; 95% CI, ?0.20 to ?0.11)</li>
<li>Better adherence to sepsis bundle elements:
<ul>
<li>shorter time to fluid administration (SMD, ?0.42; 95% CI, ?0.52 to ?0.32), </li>
<li>blood culture (SMD, ?0.31; 95% CI, ?0.40 to ?0.21)</li>
<li>antibiotic administration (SMD, ?0.34; 95% CI, ?0.39 to ?0.29)</li>
<li>lactate measurement (SMD, ?0.15; 95% CI, ?0.22 to ?0.08)</li>
</ul>
</li>
</ul>
<p><em>Electronic alerts were further associated with:</em></p>
<ul>
<li>reduced mortality (RR, 0.78; 95% CI, 0.67 to 0.92)</li>
<li>adherence with blood culture guidelines (RR, 1.14; 95% CI, 1.03 to 1.27).</li>
</ul>
<p><strong>Summary (+ a little editorialization)</strong></p>
<p>As annoying as we may find these systems in our daily practice, there is growing evidence that they do provide some benefit with impacts on task saturation and decreasing cognitive load in addition to real patient benefit. While there is also recent evidence that physician gestalt performs well against these systems, there is a suggested benefit in their inclusion in clinical decision making as a safety net or as an “assist”.</p>
<p>The incorporation of rule-based algorithms like these in more advance machine learning methods are covered quite well in a recent opinion piece on “The AI Future of Emergency Medicine”. However, it is important to always know the source of any “algorithm” that you are using, whether rule or mathematically based, given real concerns for bias and error.</p>
<fieldset><legend>References</legend><p><a href="https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2821277">https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2821277</a></p>
<p><a href="https://www.sciencedirect.com/science/article/pii/S019606442400043X">https://www.sciencedirect.com/science/article/pii/S019606442400043X</a></p>
<p><a href="https://www.sciencedirect.com/science/article/abs/pii/S0196064424000994">https://www.sciencedirect.com/science/article/abs/pii/S0196064424000994</a></p>
</fieldset>