Eric Meier

A controlled resuscitation strategy is feasible and safe in hypotensive trauma patients: results of a prospective randomized pilot trial.

BACKGROUND Optimal resuscitation of hypotensive trauma patients has not been defined. This trial was performed to assess the feasibility and safety of controlled resuscitation (CR) versus standard resuscitation (SR) in hypotensive trauma patients. METHODS Patients were enrolled and randomized in the out-of-hospital setting. Nineteen emergency medical services (EMS) systems in the Resuscitation Outcome Consortium participated. Eligible patients had an out-of-hospital systolic blood pressure (SBP) of 90 mm Hg or lower. CR patients received 250 mL of fluid if they had no radial pulse or an SBP lower than 70 mm Hg and additional 250-mL boluses to maintain a radial pulse or an SBP of 70 mm Hg or greater. The SR group patients received 2 L initially and additional fluid as needed to maintain an SBP of 110 mm Hg or greater. The crystalloid protocol was maintained until hemorrhage control or 2 hours after hospital arrival. RESULTS A total of 192 patients were randomized (97 CR and 95 SR). The CR and SR groups were similar at baseline. The mean (SD) crystalloid volume administered during the study period was 1.0 L (1.5) in the CR group and 2.0 L (1.4) in the SR group, a difference of 1.0 L (95% confidence interval [CI], 0.6–1.4). Intensive care unit–free days, ventilator-free days, renal injury, and renal failure did not differ between the groups. At 24 hours after admission, there were 5 deaths (5%) in the CR group and 14 (15%) in the SR group (adjusted odds ratio, 0.39; 95% CI, 0.12–1.26). Among patients with blunt trauma, 24-hour mortality was 3% (CR) and 18% (SR) with an adjusted odds ratio of 0.17 (0.03–0.92). There was no difference among patients with penetrating trauma (9% vs. 9%; adjusted odds ratio, 1.93; 95% CI, 0.19–19.17). CONCLUSION CR is achievable in out-of-hospital and hospital settings and may offer an early survival advantage in blunt trauma. A large-scale, Phase III trial to examine its effects on survival and other clinical outcomes is warranted. LEVEL OF EVIDENCE Therapeutic study, level I.

Interrater agreement in the evaluation of discrepant imaging findings with the Radpeer system

OBJECTIVE The Radpeer system is central to the quality assurance process in many radiology practices. Previous studies have shown poor agreement between physicians in the evaluation of their peers. The purpose of this study was to assess the reliability of the Radpeer scoring system. MATERIALS AND METHODS A sample of 25 discrepant cases was extracted from our quality assurance database. Images were made anonymous; associated reports and identities of interpreting radiologists were removed. Indications for the studies and descriptions of the discrepancies were provided. Twenty-one subspecialist attending radiologists rated the cases using the Radpeer scoring system. Multirater kappa statistics were used to assess interrater agreement, both with the standard scoring system and with dichotomized scores to reflect the practice of further review for cases rated 3 and 4. Subgroup analyses were conducted to assess subspecialist evaluation of cases. RESULTS Interrater agreement was slight to fair compared with that expected by chance. For the group of 21 raters, the kappa values were 0.11 (95% CI, 0.06-0.16) with the standard scoring system and 0.20 (95% CI, 0.13-0.27) with dichotomized scores. There was disagreement about whether a discrepancy had occurred in 20 cases. Subgroup analyses did not reveal significant differences in the degree of interrater agreement. CONCLUSION The identification of discrepant interpretations is valuable for the education of individual radiologists and for larger-scale quality assurance and quality improvement efforts. Our results show that a ratings-based peer review system is unreliable and subjective for the evaluation of discrepant interpretations. Resources should be devoted to developing more robust and objective assessment procedures, particularly those with clear quality improvement goals.

Prehospital traumatic cardiac arrest: Management and outcomes from the resuscitation outcomes consortium epistry-trauma and PROPHET registries.

BACKGROUND Traumatic arrests have historically had poor survival rates. Identifying salvageable patients and ideal management is challenging. We aimed to (1) describe the management and outcomes of prehospital traumatic arrests; (2) determine regional variation in survival; and (3) identify Advanced Life Support (ALS) procedures associated with survival. METHODS This was a secondary analysis of cases from the Resuscitation Outcomes Consortium Epistry-Trauma and Prospective Observational Prehospital and Hospital Registry for Trauma (PROPHET) registries. Patients were included if they had a blunt or penetrating injury and received cardiopulmonary resuscitation. Logistic regression analyses were used to determine the association between ALS procedures and survival. RESULTS We included 2,300 patients who were predominately young (Epistry mean [SD], 39 [20] years; PROPHET mean [SD], 40 [19] years), males (79%), injured by blunt trauma (Epistry, 68%; PROPHET, 67%), and treated by ALS paramedics (Epistry, 93%; PROPHET, 98%). A total of 145 patients (6.3%) survived to hospital discharge. More patients with blunt (Epistry, 8.3%; PROPHET, 6.5%) vs. penetrating injuries (Epistry, 4.6%; PROPHET, 2.7%) survived. Most survivors (81%) had vitals on emergency medical services arrival. Rates of survival varied significantly between the 12 study sites (p = 0.048) in the Epistry but not PROPHET (p = 0.14) registries. Patients in the PROPHET registry who received a supraglottic airway insertion or intubation experienced decreased odds of survival (adjusted OR, 0.27; 95% confidence interval, 0.08–0.93; and 0.37; 95% confidence interval, 0.17–0.78, respectively) compared to those receiving bag-mask ventilation. No other procedures were associated with survival. CONCLUSIONS Survival from traumatic arrest may be higher than expected, particularly in blunt trauma and patients with vitals on emergency medical services arrival. Although limited by confounding and statistical power, no ALS procedures were associated with increased odds of survival. LEVEL OF EVIDENCE Prognostic study, level IV.

Understanding traumatic shock: out-of-hospital hypotension with and without other physiologic compromise.

BACKGROUND Among trauma patients with out-of-hospital hypotension, we evaluated the predictive value of systolic blood pressure (SBP) with and without other physiologic compromise for identifying trauma patients requiring early critical resources. METHODS This was a secondary analysis of a prospective cohort of injured patients 13 years or older with out-of-hospital hypotension (SBP ⩽ 90 mm Hg) who were transported by 114 emergency medical service agencies to 56 Level I and II trauma centers in 11 regions of the United States and Canada from January 1, 2010, through June 30, 2011. The primary outcome was early critical resource use, defined as blood transfusion of 6 U or greater, major nonorthopedic surgery, interventional radiology, or death within 24 hours. RESULTS Of 3,337 injured patients with out-of-hospital hypotension, 1,094 (33%) required early critical resources and 1,334 (40%) had serious injury (Injury Severity Score [ISS] ≥ 16). Patients with isolated hypotension required less early critical resources (14% vs. 52%), had less serious injury (20% vs. 61%), and had lower mortality (24 hours, 1% vs. 26%; in-hospital, 3% vs. 34%). The standardized probability of requiring early critical resources was lowest among patients with blunt injury and isolated moderate hypotension (0.12; 95% confidence interval, 0.09–0.15) and steadily increased with additional physiologic compromise, more severe hypotension, and penetrating injury (0.94; 95% confidence interval, 0.90–0.98). CONCLUSION A minority of trauma patients with isolated out-of-hospital hypotension require early critical resuscitation resources. However, hypotension accompanied by additional physiologic compromise or penetrating injury markedly increases the probability of requiring time-sensitive interventions. LEVEL OF EVIDENCE Prognostic study, level II.

Comparison of Natural Language Processing Rules-based and Machine-learning Systems to Identify Lumbar Spine Imaging Findings Related to Low Back Pain

RATIONALE AND OBJECTIVES To evaluate a natural language processing (NLP) system built with open-source tools for identification of lumbar spine imaging findings related to low back pain on magnetic resonance and x-ray radiology reports from four health systems. MATERIALS AND METHODS We used a limited data set (de-identified except for dates) sampled from lumbar spine imaging reports of a prospectively assembled cohort of adults. From N = 178,333 reports, we randomly selected N = 871 to form a reference-standard dataset, consisting of N = 413 x-ray reports and N = 458 MR reports. Using standardized criteria, four spine experts annotated the presence of 26 findings, where 71 reports were annotated by all four experts and 800 were each annotated by two experts. We calculated inter-rater agreement and finding prevalence from annotated data. We randomly split the annotated data into development (80%) and testing (20%) sets. We developed an NLP system from both rule-based and machine-learned models. We validated the system using accuracy metrics such as sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). RESULTS The multirater annotated dataset achieved inter-rater agreement of Cohens kappa > 0.60 (substantial agreement) for 25 of 26 findings, with finding prevalence ranging from 3% to 89%. In the testing sample, rule-based and machine-learned predictions both had comparable average specificity (0.97 and 0.95, respectively). The machine-learned approach had a higher average sensitivity (0.94, compared to 0.83 for rules-based), and a higher overall AUC (0.98, compared to 0.90 for rules-based). CONCLUSIONS Our NLP system performed well in identifying the 26 lumbar spine findings, as benchmarked by reference-standard annotation by medical experts. Machine-learned models provided substantial gains in model sensitivity with slight loss of specificity, and overall higher AUC.