Concurrent external validation of bloodstream infection probability models

Rodic, Stefan; Hryciw, Brett N.; Selim, Shehab; Wang, Chu Qi; Lepage-Ratte, Melissa Fay; Goyal, Vineet; Nguyen, Long Hoai; Fergusson, Dean; Walraven, Carl van

doi:10.1016/j.cmi.2022.07.011

Cited by 4 publications

(9 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This can be especially useful in settings where the healthcare system is overloaded, or decisions need to be made very quickly, such as in the Emergency Department. Machine learning-based models allow us to analyze a large amount of data directly extracted from EHR, overcoming the limitations of many published BSI probability models [33].…”

Section: Discussionmentioning

confidence: 99%

A Machine Learning Predictive Model of Bloodstream Infection in Hospitalized Patients

Murri,

De Angelis,

Antenucci

et al. 2024

Diagnostics

View full text Add to dashboard Cite

The aim of the study was to build a machine learning-based predictive model to discriminate between hospitalized patients at low risk and high risk of bloodstream infection (BSI). A Data Mart including all patients hospitalized between January 2016 and December 2019 with suspected BSI was built. Multivariate logistic regression was applied to develop a clinically interpretable machine learning predictive model. The model was trained on 2016–2018 data and tested on 2019 data. A feature selection based on a univariate logistic regression first selected candidate predictors of BSI. A multivariate logistic regression with stepwise feature selection in five-fold cross-validation was applied to express the risk of BSI. A total of 5660 hospitalizations (4026 and 1634 in the training and the validation subsets, respectively) were included. Eleven predictors of BSI were identified. The performance of the model in terms of AUROC was 0.74. Based on the interquartile predicted risk score, 508 (31.1%) patients were defined as being at low risk, 776 (47.5%) at medium risk, and 350 (21.4%) at high risk of BSI. Of them, 14.2% (72/508), 30.8% (239/776), and 64% (224/350) had a BSI, respectively. The performance of the predictive model of BSI is promising. Computational infrastructure and machine learning models can help clinicians identify people at low risk for BSI, ultimately supporting an antibiotic stewardship approach.

show abstract

Section: Discussionmentioning

confidence: 99%

A Machine Learning Predictive Model of Bloodstream Infection in Hospitalized Patients

Murri,

De Angelis,

Antenucci

et al. 2024

Diagnostics

View full text Add to dashboard Cite

show abstract

Section: Discussionunclassified

“…Posiblemente, en un futuro muy cercano, en la atención al paciente con infección grave en los SUH, en el propio triaje se dispondrá de escalas o modelos automáticos o electrónicos que seleccionarán a los pacientes con alta probabilidad de infección bacteriana y bacteriemia, a los que se les aplicará los modelos diagnósticos y predictivos de bacteriemia y gravedadmortalidad, de forma que se pueda optimizar el diagnóstico microbiológico (indicación de obtener HC) y las medidas terapéuticas (antibioterapia adecuada y precoz, control del foco) [ 2 , 19 , 27 - 29 , 77 ].…”

Section: Discussionunclassified

“…Los rendimientos y capacidad predictiva de la PCT para bacteriemia verdadera en los HC obtenidos en los SUH de pacientes con infecciones comunitarias varia desde 0,68 (IC 95%: 0,59-0,77) [50] hasta 0,98 (IC 95%: 0,97-0,99) [43]. Cuando el punto de corte elegido es 0,5 ng/ml de PCT, que se corresponde con el más aceptado en la actualidad por múltiples autores [23][24][25]27,[62][63][64][65][66][75][76][77], el resultado medio obtiene un ABC-atendidos en el SUH con el diagnóstico clínico de infección, la capacidad predictiva de bacteriemia verdadera del biomarcador PCT, considerando de forma global todos los pacientes e independientemente del proceso infeccioso que la hubiera originado. Además, de forma secundaria se ha evaluado el rendimiento predictivo de bacteriemia específicamente cuando el punto de corte de la PCT se establece en 0,5 ng/ml.…”

Section: Comparación Y Rendimientos Diagnósticos De La Pro-unclassified

“…Aunque se han publicado otras revisiones sistemáticas y metaanálisis previos sobre la precisión diagnóstica de la PCT para predecir bacteriemia verdadera [22,24], ninguna publicación de forma parcial o completa ha evaluado su rendimiento en pacientes del SUH con procesos comunitarios desde el año 2014 [24]. De ahí, que esta revisión trata de enfatizar en la importancia de estos resultados para normalizar el uso de la PCT como herramienta de ayuda en los SUH, tal y como se usa en servicios de medicina crítica [23][24][25][62][63][64][65][66][75][76][77][78] En el año 2007, Jones et al [22], en un metaanálisis que incluía solo 4 artículos (de los 17), entre sus estudios, con pacientes de los SUH, mostraba globalmente un ABC-COR de 0,84 (IC 95%: 0,75-0,90), una sensibilidad del 76% (IC 95%:66-84) y una especificidad del 70% (IC 95%:60-79) [22]. Esta revisión sistemática y metanálisis mostraba dudas sobre la utilidad generalizada de la PCT para la precisión diagnóstica de bacteriemia verdadera, pero se debe tener en cuenta las limitaciones de dicha revisión que, incluso en alguno de sus artículos, no En esta línea, en los estudios incluidos en nuestra revisión a partir del año 2015 [26,41,43,49,50] con 7.734 pacientes de los SUH, mostramos un ABC-COR media estimada incluso mayor (0,85) que la de Hoeber et al [24].…”

Section: Comparación Y Rendimientos Diagnósticos De La Pro-unclassified

See 2 more Smart Citations

Diagnostic accuracy of procalcitonin for bacteremia in the emergency department: a systematic review

Julián-Jiménez,

Eduardo García,

Merinos-Sánchez

et al. 2023

Rev Esp Quimioter

View full text Add to dashboard Cite

Introduction. Obtaining blood cultures (HC) is performed in 15% of the patients treated with suspicion of infection in the Hospital Emergency Services (ED) with a variable diagnostic yield (2-20%). The 30-day mortality of patients with bacteremia is two or three times higher than the rest with the same process. Procalcitonin (PCT) is a biomarker that has been used as a tool to help predict bacteremia in HEDs. The main objective of this systematic review is to investigate the diagnostic accuracy of PCT in predicting true bacteraemia in adult patients treated with clinical suspicion of infection in the ED, as well as to identify a specific PCT value as the most relevant from the clinical decision diagnostic point of view that can be recommended for decision making Method. A systematic review was performed following the PRISMA guidelines in the PubMed, Web of Science, EMBASE, Lilacs, Cochrane, Epistemonikos, Tripdatabase and ClinicalTrials.gov databases from January 2010 to May 31, 2023 without language restrictions and using a combination of MESH terms: “Bacteremia/ Bacteraemia/ Blood Stream Infection”, “Procalcitonin”, “Emergencies/ Emergency/ Emergency Department” and “Adults”. Observational cohort studies and partially an systematic review were included. No meta-analysis techniques were performed, but the results were compared narratively. Results. A total of 1,372 articles were identified, of which 20 that met the inclusion criteria were finally analyzed. The included studies represent a total of 18,120 processed HC with 2,877 bacteraemias (15.88%). Ten studies were rated as high, 9 moderate and 1 low quality. The AUC-COR of all the studies ranges from 0.68 (95% CI: 0.59-0.77) to 0.98 (95% CI: 0.97-0.99). The PCT value >0.5 ng/ml is the most widely used and proposed in up to ten of the works included in this systematic review, whose estimated mean yield is an AUC-COR of 0.833. If only the results of the 6 high-quality studies using a cut-off point (PC) >0.5 ng/mL PCT are taken into account, the estimated mean AUC-COR result is 0.89 with Se of 77.6% and It is 78%. Conclusions. PCT has a considerable diagnostic accuracy of bacteraemia in patients treated in EDs for different infectious processes. The CP>0.5 ng/ml has been positioned as the most suitable for predicting the existence of bacteraemia and can be used to reasonably rule it out.

show abstract

Weighing the odds of bloodstream infection. What is the perfect model to predict this risk?

Bailey

Al-Hasan

2023

Clinical Microbiology and Infection

View full text Add to dashboard Cite

Concurrent external validation of bloodstream infection probability models

Cited by 4 publications

References 30 publications

A Machine Learning Predictive Model of Bloodstream Infection in Hospitalized Patients

A Machine Learning Predictive Model of Bloodstream Infection in Hospitalized Patients

Diagnostic accuracy of procalcitonin for bacteremia in the emergency department: a systematic review

Weighing the odds of bloodstream infection. What is the perfect model to predict this risk?

Contact Info

Product

Resources

About