Browsing by Author "Foraker, Randi E."

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Development and external validation of the KIIDS-TBI tool for managing children with mild traumatic brain injury and intracranial injuries
    (Wiley, 2021-12) Greenberg, Jacob K.; Ahluwalia, Ranbir; Hill, Madelyn; Johnson, Gabbie; Hale, Andrew T.; Belal, Ahmed; Baygani, Shawyon; Olsen, Margaret A.; Foraker, Randi E.; Carpenter, Carpenter; Yan, Yan; Ackerman, Laurie; Noje, Corina; Jackson, Eric; Burns, Erin; Sayama, Christina M.; Selden, Nathan R.; Vachhrajani, Shobhan; Shannon, Chevis N.; Kuppermann, Nathan; Limbrick, David D., Jr.; Neurological Surgery, School of Medicine
    Background Clinical decision support (CDS) may improve the postneuroimaging management of children with mild traumatic brain injuries (mTBI) and intracranial injuries. While the CHIIDA score has been proposed for this purpose, a more sensitive risk model may have broader use. Consequently, this study's objectives were to: (1) develop a new risk model with improved sensitivity compared to the CHIIDA model and (2) externally validate the new model and CHIIDA model in a multicenter data set. Methods We analyzed children ≤18 years old with mTBI and intracranial injuries included in the PECARN head injury data set (2004–2006). We used binary recursive partitioning to predict the composite outcome of neurosurgical intervention, intubation for > 24 h due to TBI, or death due to TBI. The new model was externally validated in a separate data set that included children treated at any one of six centers from 2006 to 2019. Results Based on 839 patients from the PECARN data set, a new risk model, the KIIDS-TBI model, was developed that incorporated imaging (e.g., midline shift) and clinical (e.g., Glasgow Coma Scale score) findings. Based on the model-predicted probability of the composite outcome, three cutoffs were evaluated to classify patients as “high risk” for level of care decisions. In the external validation data set consisting of 1,630 patients, the most conservative cutoff (i.e., any predictor present) identified 119 of 119 children with the composite outcome (sensitivity = 100%), but had the lowest specificity (26.3%). The other two decision-making cutoffs had worse sensitivity (94.1%–96.6%) but improved specificity (67.4%–81.3%). The CHIIDA model lacked the most conservative cutoff and otherwise showed the same or slightly worse performance compared to the other two cutoffs. Conclusions The KIIDS-TBI model has high sensitivity and moderate specificity for risk stratifying children with mTBI and intracranial injuries. Use of this CDS tool may help improve the safe, resource-efficient management of this important patient population.
  • Thumbnail Image
    Item
    Measures of Intracranial Injury Size Do Not Improve Clinical Decision Making for Children With Mild Traumatic Brain Injuries and Intracranial Injuries
    (Wolters Kluwer, 2022) Greenberg, Jacob K.; Olsen, Margaret A.; Johnson, Gabrielle W.; Ahluwalia, Ranbir; Hill, Madelyn; Hale, Andrew T.; Belal, Ahmed; Baygani, Shawyon; Foraker, Randi E.; Carpenter, Christopher R.; Ackerman, Laurie L.; Noje, Corina; Jackson, Eric M.; Burns, Erin; Sayama, Christina M.; Selden, Nathan R.; Vachhrajani, Shobhan; Shannon, Chevis N.; Kuppermann, Nathan; Limbrick, David D., Jr.; Neurological Surgery, School of Medicine
    Background: When evaluating children with mild traumatic brain injuries (mTBIs) and intracranial injuries (ICIs), neurosurgeons intuitively consider injury size. However, the extent to which such measures (eg, hematoma size) improve risk prediction compared with the kids intracranial injury decision support tool for traumatic brain injury (KIIDS-TBI) model, which only includes the presence/absence of imaging findings, remains unknown. Objective: To determine the extent to which measures of injury size improve risk prediction for children with mild traumatic brain injuries and ICIs. Methods: We included children ≤18 years who presented to 1 of the 5 centers within 24 hours of TBI, had Glasgow Coma Scale scores of 13 to 15, and had ICI on neuroimaging. The data set was split into training (n = 1126) and testing (n = 374) cohorts. We used generalized linear modeling (GLM) and recursive partitioning (RP) to predict the composite of neurosurgery, intubation >24 hours, or death because of TBI. Each model's sensitivity/specificity was compared with the validated KIIDS-TBI model across 3 decision-making risk cutoffs (<1%, <3%, and <5% predicted risk). Results: The GLM and RP models included similar imaging variables (eg, epidural hematoma size) while the GLM model incorporated additional clinical predictors (eg, Glasgow Coma Scale score). The GLM (76%-90%) and RP (79%-87%) models showed similar specificity across all risk cutoffs, but the GLM model had higher sensitivity (89%-96% for GLM; 89% for RP). By comparison, the KIIDS-TBI model had slightly higher sensitivity (93%-100%) but lower specificity (27%-82%). Conclusion: Although measures of ICI size have clear intuitive value, the tradeoff between higher specificity and lower sensitivity does not support the addition of such information to the KIIDS-TBI model.
  • Thumbnail Image
    Item
    Predicting Incident Heart Failure in Women With Machine Learning: The Women's Health Initiative Cohort
    (Elsevier, 2021) Tison, Geoffrey H.; Avram, Robert; Nah, Gregory; Klein, Liviu; Howard, Barbara V.; Allison, Matthew A.; Casanova, Ramon; Blair, Rachael H.; Breathett, Khadijah; Foraker, Randi E.; Olgin, Jeffrey E.; Parikh, Nisha I.; Medicine, School of Medicine
    Background: Heart failure (HF) is a leading cause of cardiac morbidity among women, whose risk factors differ from those in men. We used machine-learning approaches to develop risk- prediction models for incident HF in a cohort of postmenopausal women from the Women's Health Initiative (WHI). Methods: We used 2 machine-learning methods-Least Absolute Shrinkage and Selection Operator (LASSO) and Classification and Regression Trees (CART)-to perform variable selection on 1227 baseline WHI variables for the primary outcome of incident HF. These variables were then used to construct separate Cox proportional hazard models, and we compared these results, using receiver-operating characteristic (ROC) curve analysis, against a comparator model built using variables from the Atherosclerosis Risk in Communities (ARIC) HF prediction model. We analyzed 43,709 women who had 2222 incident HF events; median follow-up was 14.3 years. Results: LASSO selected 10 predictors, and CART selected 11 predictors. The highest correlation between selected variables was 0.46. In addition to selecting well-established predictors such as age, myocardial infarction, and smoking, novel predictors included physical function, number of pregnancies, number of previous live births and age at menopause. In ROC analysis, the CART-derived model had the highest C-statistic of 0.83 (95% confidence interval [CI], 0.81-0.85), followed by LASSO 0.82 (95% CI, 0.81-0.84) and ARIC 0.73 (95% CI, 0.70-0.76). Conclusions: Machine-learning approaches can be used to develop HF risk-prediction models that can have better discrimination compared with an established HF risk model and may provide a basis for investigating novel HF predictors.