Septic Shock: Diagnostic Disagreement and ICU Mortality

Key Takeaways

• At eight acute care hospitals, patients with septic shock had documented early disagreement over shock etiology in a minority of cases.
• Disagreement was significantly more frequent at the academic center than at community sites.
• Blood-culture status significantly modified the disagreement–ICU mortality association, although the culture-positive subgroup odds estimate did not reach conventional statistical significance.
• Among culture-positive patients, treatment receipt rates did not differ by disagreement status, and treatment timing did not statistically mediate the mortality association.

Identifying sepsis as the cause of shock precedes early antibiotic, fluid, and vasopressor treatment. During the first hours in an intensive care unit (ICU), clinicians may document different etiologies for the same patient's shock. Whether those differences relate to ICU death or the timing of treatment is the question investigators examined.

Investigators retrospectively evaluated 2,233 patients meeting Sepsis-3 organ-dysfunction criteria at eight acute care hospitals during 2018–2023. Regular expressions identified clinician-documented shock etiologies from the first 48 ICU hours.

Investigators compared etiologies using cosine similarity and classified scores below 1 as disagreement. The multicenter cohort had no assigned intervention. Investigators assessed ICU mortality using inverse-probability-of-treatment-weighted multivariable logistic regression. Fine–Gray models examined administration of antibiotics, bolus fluids, and vasopressors; mediation analyses examined whether treatment timing accounted for the mortality association.

Documented disagreement occurred in 213 patients (9.5%). It was significantly more frequent at the academic center (13.3%) than at community sites (7.2%).

Blood-culture positivity significantly modified the adjusted association between documented disagreement and ICU mortality: interaction odds ratio (OR) 2.37 (95% confidence interval (CI) 1.03–5.45; p=0.042). This interaction describes a difference in the association by culture status, not the mortality association within culture-positive patients alone.

Within the culture-positive subgroup, documented disagreement had an adjusted ICU mortality OR of 1.99 (95% CI 0.98–4.01; p=0.056). The OR estimate did not meet conventional statistical significance despite the significant interaction. In a separate adjusted linear-probability analysis, the authors reported a 16.9-percentage-point higher mortality risk (95% CI 1.1–32.7; p=0.036). Among culture-positive patients, rates of receiving antibiotics, fluids, or vasopressors did not differ by disagreement status. Timing of those treatments did not statistically mediate the mortality association in that subgroup.

Disagreement was derived from retrospective clinician documentation and a cosine-similarity threshold, not prospectively adjudicated diagnostic error. The authors' culture-positive subgroup estimates differed in statistical significance: the adjusted odds-ratio estimate did not meet the conventional threshold, while the adjusted absolute-risk estimate did. Because the analysis was observational, the association does not establish that disagreement caused ICU death. The absence of detected timing mediation does not establish identical treatment ordering times or show that timing is immaterial.

The investigators characterized early documented disagreement over shock etiology as a potential prognostic marker among patients with culture-positive septic shock.

Clinician Questions

Which factors were included in the adjusted ICU mortality analysis of diagnostic disagreement in septic shock?

For patients with septic shock, investigators used inverse-probability-of-treatment-weighted multivariable logistic regression adjusting for demographics, comorbidities, Sequential Organ Failure Assessment (SOFA) score, blood-culture positivity, and ICU specialty.

Were positive blood cultures required for inclusion in the septic shock diagnostic-disagreement cohort?

Patients entered the multicenter septic shock cohort by meeting Sepsis-3 organ-dysfunction criteria, not by having positive blood cultures. Investigators subsequently evaluated blood-culture positivity as an effect modifier; culture-positive patients constituted a subgroup.

Recommended Reading

• For more on prognosis in septic shock: PSP Shows Higher Prognostic AUC Than CRP and PCT in Sepsis