General Medicine · Medical Literature
A Randomized Trial of Interval Likelihood Ratios for Clinical Decision-Making.
Abstract
Dichotomized reporting of diagnostic test results may oversimplify test characteristics. Interval likelihood ratios (ILRs), test characteristics across a range of values, offer a nuanced alternative, but their impact on decision-making is uncertain. We assessed whether ILRs improve clinical decisions compared to two traditional formats: single-threshold and binary test characteristics. We conducted a multicenter randomized trial of internal medicine residents, attending physicians, and advanced practice providers. Participants were presented with two fictional vignettes (cases involving pulmonary embolism diagnosis and heart failure treatment), each with a clearly correct response centered on a key diagnostic test result. The vignettes were identical except for participants' random assignment to one of three test characteristic presentations: (1) single threshold, (2) binary test characteristics, or (3) ILRs. We compared the correct management decision rates between the ILR and threshold groups and between the ILR and binary groups. Secondary objectives included comparing threshold and binary groups and subgroup analyses by case vignette. The trial enrolled 611 participants. The correct decision rate was 93.4% in the ILR group compared with 82.2% in the threshold group (difference of 11.2 percentage points; 95% confidence interval [CI], 6.9-15.6, P<0.001) and compared with 75.1% in the binary group (difference of 18.3 percentage points; 95% CI, 13.5-23.1, P<0.001). In the pulmonary embolism case, the correct response rate was 91.9% in the ILR group, 70.9% in the threshold group, and 64.7% in the binary group. In the heart failure case, the correct response rate was 94.9% in the ILR group, 93.4% in the threshold group, and 85.6% in the binary group. Presenting diagnostic test information using ILRs improved vignette-based clinical decisions. The magnitude of improvement appeared to vary by case vignette.
Clinical relevance
Does presenting diagnostic test characteristics as interval likelihood ratios (ILRs) — spanning a range of test values — improve correct clinical decision-making among internal medicine clinicians compared to single-threshold or binary test characteristic formats? __SNAPSHOT__ {"studyDesign":"Multicenter randomized controlled trial (three-arm parallel group) using standardized clinical vignettes","setting":"Multiple academic medical centers; institutions included University of Colorado, Weill Cornell Medicine, UCSF, and University of Michigan (United States)","sampleSize":"n=611 participants; approximately 200 per arm (exact per-arm allocation not reported in abstract)","population":"Internal medicine residents, attending physicians, and advanced practice providers across multiple academic centers; each participant completed two fictional clinical vignettes (pulmonary embolism diagnosis and heart failure management)","intervention":"Presentation of diagnostic test characteristics as interval likelihood ratios (ILRs) — test performance data stratified across a continuous range of test result values","comparator":"Two active comparator arms: (1) single-threshold format (e.g., D-dimer <500 = normal); (2) binary test characteristics format (single threshold plus aggregate sensitivity, specificity, positive LR, and negative LR)","primaryOutcome":"Correct clinical management decision rate, compared between ILR vs. threshold group and ILR vs. binary group","followUp":"Not applicable (cross-sectional vignette-based survey design)","fundingSource":"Not reported"}
Key findings
The trial enrolled 611 participants and demonstrated a statistically significant advantage for the ILR presentation format on the primary outcome of correct clinical decision rate. The correct decision rate was 93.4% in the ILR group versus 82.2% in the single-threshold group (absolute difference: 11.2 percentage points; 95% CI, 6.9–15.6; P<0.001) and 75.1% in the binary group (absolute difference: 18.3 percentage points; 95% CI, 13.5–23.1; P<0.001). The NNT to achieve one additional correct decision by switching from single-threshold to ILR format is approximately 9 (calculated); from binary to ILR format, approximately 6 (calculated). In pre-specified subgroup analyses by case vignette, the benefit of ILRs was most pronounced in the pulmonary embolism case: correct response rates were 91.9% (ILR), 70.9% (threshold), and 64.7% (binary). In the heart failure case, correct response rates were 94.9% (ILR), 93.4% (threshold), and 85.6% (binary) — indicating a much smaller ILR-vs-threshold absolute difference of 1.5 percentage points in that vignette. The threshold vs. binary comparison was a secondary objective; specific effect sizes for that comparison are not reported in the available abstract. No patient safety outcomes or adverse events were applicable given the study design.
Practice implications
This trial establishes that presenting diagnostic test characteristics as ILRs is associated with statistically significantly higher rates of correct clinical decision-making among academic internal medicine clinicians compared to both single-threshold and binary formats, with the PE vignette demonstrating the largest effect. However, several important boundaries on inference must be noted. First, the outcome was performance on fictional vignettes with a prespecified 'correct' answer — this is a surrogate for real-world decision-making quality, and whether ILR-driven decisions translate to improved patient outcomes is not established by this study. Second, both vignettes featured clinical scenarios (low-pretest-probability PE with a modestly elevated D-dimer; high-pretest-probability heart failure with a BNP slightly below threshold) specifically designed to highlight the limitations of binary cutoffs — situations where ILRs have the greatest theoretical advantage. The generalizability to decisions where the binary threshold is clinically appropriate, or where pretest probability is unambiguous, is uncertain. Third, the participant pool was limited to internal medicine clinicians at academic medical centers, and generalizability to community settings, other specialties, or less-experienced learners is not established. Fourth, only two clinical scenarios were tested; the heterogeneity of ILR benefit across cases (large in PE, minimal in HF threshold-vs-ILR comparison) suggests that ILR superiority may be context-dependent. The study does not address implementation barriers (e.g., whether laboratory reporting systems currently present ILRs, clinician training requirements, or cognitive burden). Practice change — specifically mandating ILR-format reporting — is premature without confirmatory trials in real-world clinical settings and broader specialty representation. Note: this summary is based on the published abstract only; full-text review may reveal additional methodological details including randomization procedures, allocation concealment, blinding strategy, and funding disclosures. __LIMITATIONS__ Vignette-based surrogate outcome: correct decisions on two fictional cases do not demonstrate improved patient-level outcomes in real clinical practice.,Only two clinical scenarios tested, both specifically chosen to demonstrate the theoretical advantage of ILRs over binary cutoffs; results may not generalize to other diagnostic contexts.,Pronounced heterogeneity of ILR benefit by case (PE vs. heart failure), suggesting effect size is scenario-dependent and an overall pooled estimate may be misleading.,Academic internal medicine clinicians at multiple centers — generalizability to community practice, other specialties, emergency medicine, and primary care is uncertain.,Approximately 36% participation rate (611 of ~1,700 invited, per medscape commentary), raising the risk of selection bias toward clinicians more engaged with evidence-based medicine or diagnostic reasoning.,Blinding of participants to group assignment was not feasible by design; participants knew which format they were viewing.,The 'correct' response was researcher-defined; the study does not adjudicate whether the designated correct answer aligns with current guideline recommendations or whether other responses would be clinically reasonable.,No assessment of cognitive load, time-to-decision, or clinician-reported usability of ILR format.,Funding source not reported in the abstract; potential conflicts of interest cannot be assessed.,No follow-up; the study captures a single time-point decision with no assessment of retention, learning, or behavioral change over time. __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ In this multicenter randomized vignette trial of 611 internal medicine clinicians, presentation of diagnostic test results using interval likelihood ratios significantly improved correct clinical decision rates compared to both single-threshold (93.4% vs. 82.2%; +11.2 pp) and binary formats (93.4% vs. 75.1%; +18.3 pp), with the greatest benefit observed in the pulmonary embolism scenario. These findings are promising and conceptually compelling for educators and test-reporting system designers, but the vignette-based surrogate outcome, narrow clinical context, and academic setting mean that routine ILR reporting cannot yet be mandated based on this evidence alone. Confirmation in real-world clinical settings with patient-level outcomes is needed before practice change.