Development and Validation of a Prediction Model for Perinatal Arterial Ischemic Stroke in Term Neonates

Ratika Srivastava, Mary Dunbar, Michael Shevell, Maryam Oskoui, Anna Basu, Michael John Rivkin, Eilon Shany, Linda S. De Vries, Deborah Dewey, Nicole Letourneau, Michael D. Hill, Adam Kirton

    Research output: Contribution to journalArticlepeer-review

    1 Scopus citations

    Abstract

    Importance: Perinatal arterial ischemic stroke (PAIS) is a focal brain injury in term neonates that is identified postnatally but is presumed to occur near the time of birth. Many pregnancy, delivery, and fetal factors have been associated with PAIS, but early risk detection is lacking; thus, targeted treatment and prevention efforts are currently limited. Objective: To develop and validate a diagnostic risk prediction model that uses common clinical factors to predict the probability of PAIS in a term neonate. Design, Setting, and Participants: In this diagnostic study, a prediction model was developed using multivariable logistic regression with registry-based case data collected between January 2003, and March 2020, from the Alberta Perinatal Stroke Project, Canadian Cerebral Palsy Registry, International Pediatric Stroke Study, and Alberta Pregnancy Outcomes and Nutrition study. Criteria for inclusion were term birth and no underlying medical conditions associated with stroke diagnosis. Records with more than 20% missing data were excluded. Variable selection was based on peer-reviewed literature. Data were analyzed in September 2021. Exposures: Clinical pregnancy, delivery, and neonatal factors associated with PAIS as common data elements across the 4 registries. Main Outcomes and Measures: The primary outcome was the discriminative accuracy of the model predicting PAIS, measured by the concordance statistic (C statistic). Results: Of 2571 term neonates in the initial analysis (527 [20%] case and 2044 [80%] control individuals; gestational age range, 37-42 weeks), 1389 (54%) were male, with a greater proportion of males among cases compared with controls (318 [60%] vs 1071 [52%]). The final model was developed using 1924 neonates, including 321 cases (17%) and 1603 controls (83%), and 9 clinical factors associated with risk of PAIS in term neonates: maternal age, tobacco exposure, recreational drug exposure, preeclampsia, chorioamnionitis, intrapartum maternal fever, emergency cesarean delivery, low 5-minute Apgar score, and male sex. The model demonstrated good discrimination between cases and controls (C statistic, 0.73; 95% CI, 0.69-0.76) and good model fit (Hosmer-Lemeshow P =.20). Internal validation techniques yielded similar C statistics (0.73 [95% CI, 0.69-0.77] with bootstrap resampling, 10-fold cross-validated area under the curve, 0.72 [bootstrap bias-corrected 95% CI, 0.69-0.76]), as did a sensitivity analysis using cases and controls from Alberta, Canada, only (C statistic, 0.71; 95% CI, 0.65-0.77). Conclusions and Relevance: The findings suggest that clinical variables can be used to develop and internally validate a model to predict the risk of PAIS in term neonates, with good predictive performance and strong internal validity. Identifying neonates with a high probability of PAIS who could then be screened for early diagnosis and treatment may be associated with reductions in lifelong morbidity for affected individuals and their families..

    Original languageEnglish
    Pages (from-to)E2219203
    JournalJAMA network open
    Volume5
    Issue number6
    DOIs
    StatePublished - 1 Jan 2022

    ASJC Scopus subject areas

    • Medicine (all)

    Fingerprint

    Dive into the research topics of 'Development and Validation of a Prediction Model for Perinatal Arterial Ischemic Stroke in Term Neonates'. Together they form a unique fingerprint.

    Cite this