No abstract

Summary. Backgound: Over-investigation of low-risk patients with suspected pulmonary embolism (PE) represents a growing problem. The combination of gestalt estimate of low suspicion for PE, together with the PE rule-out criteria [PERC()): age < 50 years, pulse < 100 beats min, SaO 2 ‡ 95%, no hemoptysis, no estrogen use, no surgery/trauma requiring hospitalization within 4 weeks, no prior venous thromboembolism (VTE), and no unilateral leg swelling], may reduce speculative testing for PE. We hypothesized that low suspicion and PERC()) would predict a post-test probability of VTE(+) or death below 2.0%. Methods: We enrolled outpatients with suspected PE in 13 emergency departments. Clinicians completed a 72-field, web-based data form at the time of test order. Low suspicion required a gestalt pretest probability estimate of <15%. The main outcome was the composite of image-proven VTE(+) or death from any cause within 45 days. Results: We enrolled 8138 patients, 85% of whom had a chief complaint of either dyspnea or chest pain. Clinicians reported a low suspicion for PE, together with PERC()), in 1666 patients (20%). At initial testing and within 45 days, 561 patients (6.9%, 95% confidence interval 6.5-7.6) were VTE(+), and 56 others died. Among the low suspicion and PERC()) patients, 15 were VTE(+) and one other patient died, yielding a false-negative rate of 16/1666 (1.0%, 0.6-1.6%). As a diagnostic test, low suspicion and PERC()) had a sensitivity of 97.4% (95.8-98.5%) and a specificity of 21.9% (21.0-22.9%). Conclusions:The combination of gestalt estimate of low suspicion for PE and PERC()) reduces the probability of VTE to below 2% in about 20% of outpatients with suspected PE.

No abstract

Objective To derive and validate an objective clinical prediction rule for the presence of uncomplicated ureteral stones in patients eligible for computed tomography (CT). We hypothesized that patients with a high probability of ureteral stones would have a low probability of acutely important alternative findings.Design Retrospective observational derivation cohort; prospective observational validation cohort.Setting Urban tertiary care emergency department and suburban freestanding community emergency department.Participants Adults undergoing non-contrast CT for suspected uncomplicated kidney stone. The derivation cohort comprised a random selection of patients undergoing CT between April 2005 and November 2010 (1040 patients); the validation cohort included consecutive prospectively enrolled patients from May 2011 to January 2013 (491 patients). Main outcome measuresIn the derivation phase a priori factors potentially related to symptomatic ureteral stone were derived from the medical record blinded to the dictated CT report, which was separately categorized by diagnosis. Multivariate logistic regression was used to determine the top five factors associated with ureteral stone and these were assigned integer points to create a scoring system that was stratified into low, moderate, and high probability of ureteral stone. In the prospective phase this score was observationally derived blinded to CT results and compared with the prevalence of ureteral stone and important alternative causes of symptoms. ResultsThe derivation sample included 1040 records, with five factors found to be most predictive of ureteral stone: male sex, short duration of pain, non-black race, presence of nausea or vomiting, and microscopic hematuria, yielding a score of 0-13 (the STONE score). Prospective validation was performed on 491 participants. In the derivation and validation cohorts ureteral stone was present in, respectively, 8.3% and 9.2% of the low probability (score 0-5) group, 51.6% and 51.3% of the moderate probability (score 6-9) group, and 89.6% and 88.6% of the high probability (score 10-13) group. In the high score group, acutely important alternative findings were present in 0.3% of the derivation cohort and 1.6% of the validation cohort. ConclusionsThe STONE score reliably predicts the presence of uncomplicated ureteral stone and lower likelihood of acutely important alternative findings. Incorporation in future investigations may help to limit exposure to radiation and over-utilization of imaging.Trial registrationwww.clinicaltrials.gov NCT01352676.

BackgroundUrinary tract infection (UTI) is a common emergency department (ED) diagnosis with reported high diagnostic error rates. Because a urine culture, part of the gold standard for diagnosis of UTI, is usually not available for 24–48 hours after an ED visit, diagnosis and treatment decisions are based on symptoms, physical findings, and other laboratory results, potentially leading to overutilization, antibiotic resistance, and delayed treatment. Previous research has demonstrated inadequate diagnostic performance for both individual laboratory tests and prediction tools.ObjectiveOur aim, was to train, validate, and compare machine-learning based predictive models for UTI in a large diverse set of ED patients.MethodsSingle-center, multi-site, retrospective cohort analysis of 80,387 adult ED visits with urine culture results and UTI symptoms. We developed models for UTI prediction with six machine learning algorithms using demographic information, vitals, laboratory results, medications, past medical history, chief complaint, and structured historical and physical exam findings. Models were developed with both the full set of 211 variables and a reduced set of 10 variables. UTI predictions were compared between models and to proxies of provider judgment (documentation of UTI diagnosis and antibiotic administration).ResultsThe machine learning models had an area under the curve ranging from 0.826–0.904, with extreme gradient boosting (XGBoost) the top performing algorithm for both full and reduced models. The XGBoost full and reduced models demonstrated greatly improved specificity when compared to the provider judgment proxy of UTI diagnosis OR antibiotic administration with specificity differences of 33.3 (31.3–34.3) and 29.6 (28.5–30.6), while also demonstrating superior sensitivity when compared to documentation of UTI diagnosis with sensitivity differences of 38.7 (38.1–39.4) and 33.2 (32.5–33.9). In the admission and discharge cohorts using the full XGboost model, approximately 1 in 4 patients (4109/15855) would be re-categorized from a false positive to a true negative and approximately 1 in 11 patients (1372/15855) would be re-categorized from a false negative to a true positive.ConclusionThe best performing machine learning algorithm, XGBoost, accurately diagnosed positive urine culture results, and outperformed previously developed models in the literature and several proxies for provider judgment. Future prospective validation is warranted.