We propose a novel recursive partitioning method for identifying subgroups of subjects with enhanced treatment effects based on a differential effect search algorithm. The idea is to build a collection of subgroups by recursively partitioning a database into two subgroups at each parent group, such that the treatment effect within one of the two subgroups is maximized compared with the other subgroup. The process of data splitting continues until a predefined stopping condition has been satisfied. The method is similar to 'interaction tree' approaches that allow incorporation of a treatment-by-split interaction in the splitting criterion. However, unlike other tree-based methods, this method searches only within specific regions of the covariate space and generates multiple subgroups of potential interest. We develop this method and provide guidance on key topics of interest that include generating multiple promising subgroups using different splitting criteria, choosing optimal values of complexity parameters via cross-validation, and addressing Type I error rate inflation inherent in data mining applications using a resampling-based method. We evaluate the operating characteristics of the procedure using a simulation study and illustrate the method with a clinical trial example.
✓ The authors have analyzed their experience with intracranial pressure (ICP) monitoring in 207 patients over a 4-year period. Patients with either high-density or low-density lesions on computerized tomography (CT) at admission had a high incidence (53% to 63%) of intracranial hypertension (ICP persistently over 20 mm Hg). In contrast, patients with normal CT scans at admission had a relatively low incidence of ICP elevation (13%). Among these patients, three features were found to be strongly associated with the development of intracranial hypertension: 1) age over 40 years; 2) systolic blood pressure under 90 mm Hg; and 3) motor posturing — unilateral or bilateral. When two or more of these features were noted at admission, the incidence of intracranial hypertension was 60%, as compared to 4% when only one, or none, of these features were present. Thus, the patients at high risk for developing intracranial hypertension after severe head injury are those with abnormal CT scans at admission, and those with normal CT scans who demonstrate two or more of the above-mentioned adverse features. Based on these criteria, only 16% of this series of patients with normal CT scans would have qualified for monitoring. In addition to the three clinical features noted above, multimodality evoked potential (MEP) studies were also found to be strong predictors of ICP elevation in the normal CT scan group, with a 75% incidence of intracranial hypertension in patients with disseminated deficits. There was no statistically significant correlation between the Glasgow Coma Scale score, eye movements, pupillary reaction, hypoxia, or anemia at admission and subsequent ICP elevation in the group with normal CT scans. In this series, an intraventricular catheter was used as the sole monitoring device in 91% of the cases. In the remaining 9%, subarachnoid screws were employed, either alone, or upon failure of the ventriculostomy. While no mortality was directly ascribed to the monitoring process, there was a 7.7% complication rate (infection 6.3% + hemorrhage 1.4%). Eighty-five percent of the infections occurred in patients who had been monitored for 5 days or more, while no infections were noted in those monitored for less than 3 days. Used judiciously, this technique can be valuable in the monitoring and treatment of the brain-injured patient.
An analysis of clinical signs, singly or in combination, multimodality evoked potentials (MEP's), computerized tomography scans, and intracranial pressure (ICP) data was undertaken prospectively in 133 severely head-injured patients to ascertain the accuracy, reliability, and relative value of these indicants individually, or in various combinations, in predicting one of two categories of outcome. Erroneous predictions, either falsely optimistic (FO) or falsely pessimistic (FP), were analyzed to gain pathophysiological insights into the disease process. Falsely optimistic predictions occurred because of unpredictable complications, whereas FP predictions were due to intrinsic weakness of the indicants as prognosticators. A combination of clinical data, including age, Glasgow Coma Scale (GCS) score, pupillary response, presence of surgical mass lesions, extraocular motility, and motor posturing predicted outcome with 82% accuracy, 43% with over 90% confidence. Nine percent of predictions were FO and 9% FP. The GCS score alone was accurate in 80% of predictions, but at a lower level of confidence (25% at the over-90% level), with 7% FO and 13% FP. Computerized tomography and ICP data in isolation proved to be poor prognostic indicants. When combined individually with clinical data, however, they increased the number of predictions made with over 90% confidence to 52% and 55%, respectively. Data from MEP's represented the most accurate single prognostic indicant, with 91% correct predictions, 25% at the over-90% confidence level. There were no FP errors associated with this indicant. Supplementation of the clinical examination with MEP data yielded optimal prognostic power, an 89% accuracy rate, with 64% over the 90% confidence level and only 4% FP errors. The clinical examination remains the strongest basis for prognosticating outcome in severe head injury, but additional studies enhance the reliability of such predictions.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.