Combination of procalcitonin and C-reactive protein levels in the early diagnosis of bacterial co-infections in children with H1N1 influenza.

OBJECTIVE: This study evaluated the diagnostic value of measuring the levels of procalcitonin (PCT) and C-reactive protein (CRP) to differentiate children co-infected with H1N1 influenza and bacteria from children infected with H1N1 influenza alone. METHODS: Consecutive patients (children aged < 5 years) with laboratory-confirmed H1N1 influenza who were hospitalized or received outpatient care from a tertiary-care hospital in Canton, China, between January 1, 2012, and September 1, 2017, were included in the present study. Laboratory results, including serum PCT and CRP levels, white blood cell (WBC) counts, and bacterial cultures, were analyzed. The predictive value of the combination of biomarkers versus any of the biomarkers alone for diagnosing bacterial co-infections was evaluated using logistic regression analyses. RESULTS: Significantly higher PCT (1.46 vs 0.21 ng/mL, P < 0.001) and CRP (19.20 vs 5.10 mg/dL, P < 0.001) levels were detected in the bacterial co-infection group than in the H1N1 infection-alone group. Using PCT or CRP levels alone, the areas under the curves (AUCs) for predicting bacterial co-infections were 0.801 (95% CI, 0.772-0.855) and 0.762 (95% CI, 0.722-0.803), respectively. Using a combination of PCT and CRP, the logistic regression-based model, Logit(P) = -1.912 + 0.546 PCT + 0.087 CRP, showed significantly greater accuracy (AUC: 0.893, 95% CI: 0.842-0.934) than did the other three biomarkers. CONCLUSIONS: The combination of PCT and CRP levels could provide a useful method of distinguishing bacterial co-infections from an H1N1 influenza infection alone in children during the early disease phase. After further validation, the flexible model derived here could assist clinicians in decision-making processes.

Evaluation and accurate diagnoses of pediatric diseases using artificial intelligence.

Artificial intelligence (AI)-based methods have emerged as powerful tools to transform medical care. Although machine learning classifiers (MLCs) have already demonstrated strong performance in image-based diagnoses, analysis of diverse and massive electronic health record (EHR) data remains challenging. Here, we show that MLCs can query EHRs in a manner similar to the hypothetico-deductive reasoning used by physicians and unearth associations that previous statistical methods have not found. Our model applies an automated natural language processing system using deep learning techniques to extract clinically relevant information from EHRs. In total, 101.6 million data points from 1,362,559 pediatric patient visits presenting to a major referral center were analyzed to train and validate the framework. Our model demonstrates high diagnostic accuracy across multiple organ systems and is comparable to experienced pediatricians in diagnosing common childhood diseases. Our study provides a proof of concept for implementing an AI-based system as a means to aid physicians in tackling large amounts of data, augmenting diagnostic evaluations, and to provide clinical decision support in cases of diagnostic uncertainty or complexity. Although this impact may be most evident in areas where healthcare providers are in relative shortage, the benefits of such an AI system are likely to be universal.