FIND · DE-SELECT · CONTROL — a proposed three-step POCUS framework for antibiotic decision-making, checked against the actual literature behind each piece.
The framework's actual insight: POCUS is proposed here not as a diagnostic test with a yes/no answer, but as a tool for narrowing where the infection (or non-infection) actually is.
The pitch is that undifferentiated sepsis presentations — hypotension, tachypnea, fever, elevated lactate — give you a recognizable phenotype without telling you where the problem is, which is exactly the situation that drives broad empiric antibiotics by default. The proposed fix is a three-step bedside sequence: FIND a probable anatomic source, DE-SELECT non-infectious mimics that produce a similar clinical picture, and CONTROL — identify pathology that needs source control (drainage, decompression, surgery) rather than antibiotics alone.
This pillar has the strongest existing evidence base of the three, even though it's cited generically in the post.
The framework lists consolidation, air bronchograms, and pleural effusion as findings pointing toward a pulmonary source. This matches real, well-established diagnostic accuracy data: multiple meta-analyses put lung ultrasound's pooled sensitivity for pneumonia at 92–96% and specificity at 90–96% against chest X-ray/CT reference standards — consistently outperforming chest X-ray alone, particularly in ICU populations (LUS sensitivity 93% vs. CXR 65% in one 2025 Bayesian meta-analysis). The gallbladder criteria (gallstones, wall thickening, pericholecystic fluid) and soft-tissue criteria (fluid collection, cobblestoning, foreign body) are standard, accurate sonographic findings for cholecystitis and skin/soft-tissue infection respectively, though worth noting: gallstones alone are common and often asymptomatic — it's the combination with wall thickening and pericholecystic fluid (or a sonographic Murphy's sign) that actually supports an active biliary source, not stones in isolation.
"B-lines do not equal pneumonia" is correct, well-established teaching, not a novel claim.
Diffuse, bilateral B-lines with no focal consolidation are the classic sonographic pattern of cardiogenic pulmonary edema/interstitial syndrome, not pneumonia — this is core Lichtenstein BLUE-protocol teaching, not a new claim. Pneumonia on lung ultrasound classically shows focal, asymmetric findings: consolidation, air bronchograms, and often a associated effusion. The framework's example (dyspnea + hypoxia + bilateral B-lines + elevated BNP, reduced LV function on cardiac views, no focal consolidation → more consistent with cardiogenic edema than pneumonia) is a textbook-accurate application of that distinction, not a stretch.
The individual pieces are evidence-based. The proposed integration of them into a sequential antibiotic-decision protocol is not.
A search of the antimicrobial-stewardship literature for ED-based interventions turns up active, real programs — but they're built around rapid pathogen panels (multiplex viral PCR, point-of-care urine bacterial load testing), not POCUS-guided anatomic source-finding. Nothing in the current literature tests whether applying FIND/DE-SELECT/CONTROL as a structured sequence measurably changes antibiotic initiation, de-escalation, or patient outcomes. That's a genuine, correctly-labeled gap, not an overstated one.
This is the gap that actually matters. The framework's own best-evidenced pillar (lung/pneumonia) has a documented, large accuracy drop the moment you move from expert operators to the clinicians who'd actually be using this framework at 3am.
A 2024 multicenter diagnostic accuracy study specifically testing newly certified operators using handheld ultrasound for pneumonia found a sensitivity of just 31% (95% CI 26–36) against expert/CT diagnosis — significantly worse than chest X-ray's 66–69% in the same cohort (P<.001). That's a massive gap from the 92–96% sensitivity figures generally cited for lung ultrasound and pneumonia (this audit included), because those pooled meta-analysis numbers are dominated by studies using experienced sonographers or POCUS-fellowship-trained physicians. A separate North American inter-rater reliability study among POCUS-credentialed emergency physicians (not novices) still found only moderate agreement (κ=0.56) for lung POCUS interpretation.
This matters specifically for this framework because the entire pitch is about influencing real-time antibiotic decisions at the point of care — which means the relevant accuracy figure isn't "what an expert sonographer can achieve in a validation study," it's "what the on-shift clinician holding the probe can actually see." The framework's carousel doesn't distinguish these, and the gap between them (31% vs. 92–96% sensitivity for the exact same diagnostic question) is large enough to flip the clinical conclusion: a newly-certified operator's negative lung scan is far less reassuring than the framework implicitly suggests, and treating it as ruling out a pulmonary source could mean withholding antibiotics from someone who actually has pneumonia.
Yes, as a hypothesis for identifying an infection source. Not yet, as a tool you can trust to have actually identified one.
Split the question in two, because the honest answer is different for each half. Is the underlying idea promising? Yes — genuinely. The individual sonographic findings behind FIND, DE-SELECT, and CONTROL are legitimately well-evidenced (92–96% sensitivity/specificity for pneumonia in expert hands, textbook-accurate B-lines-vs-consolidation physiology), the exam is fast, radiation-free, and repeatable, and it's already sitting in most EDs at effectively zero marginal cost to try. That's enough to justify the prospective trial the framework is asking for.
Is it promising as an actual source-identification tool, today, in the hands of the clinician who'd be using it? That's where the promise thins out. The accuracy numbers this framework leans on come from expert operators. The moment you test newly certified operators — the people who'd actually be running FIND at the bedside — pneumonia sensitivity drops to 31%, worse than plain chest X-ray. And identification isn't static: early pneumonia or a deep abscess may not yet be sonographically visible, so a negative FIND doesn't mean "no source," it can just mean "not yet visible." Neither of those caveats is in the original post.
Promising enough to earn the study. Not yet promising enough to trust as an identification tool in a typical operator's hands — the gap between "can identify" in validation studies and "did identify" at 3am with a newly-credentialed operator is exactly what a real trial needs to close before this becomes practice.
This holds up well for what it actually is: a reasonable synthesis of individually well-evidenced POCUS findings (lung consolidation for pneumonia, diffuse B-lines for cardiogenic edema, gallbladder wall thickening for cholecystitis, soft-tissue collections for abscess) into a not-yet-tested three-step decision sequence. Every specific sonographic claim checks out against the literature, and the honest "conceptual, not clinically validated" framing is accurate. The real hole isn't in what's stated — it's in what's assumed: that the accuracy demonstrated by expert-operator studies transfers to the clinicians who'd actually be running this framework, when the newly-certified-operator data shows the opposite (31% sensitivity, worse than plain chest X-ray), and that better source-finding automatically translates into changed prescribing, which no cited or available evidence actually measures. Any future validation study needs to test this with the operators who'd really use it, not the experts who built it, and needs an outcome measure that's actually about antibiotic decisions, not just diagnostic concordance.