Improving ICU Risk Predictive Models Through Automation Designed for Resiliency Against Documentation Bias.

OBJECTIVES: Electronic health records enable automated data capture for risk models but may introduce bias. We present the Philips Critical Care Outcome Prediction Model (CCOPM) focused on addressing model features sensitive to data drift to improve benchmarking ICUs on mortality performance. DESIGN: Retrospective, multicenter study of ICU patients randomized in 3:2 fashion into development and validation cohorts. Generalized additive models (GAM) with features designed to mitigate biases introduced from documentation of admission diagnosis, Glasgow Coma Scale (GCS), and extreme vital signs were developed using clinical features representing the first 24 hours of ICU admission. SETTING: eICU Research Institute database derived from ICUs participating in the Philips eICU telecritical care program. PATIENTS: A total of 572,985 adult ICU stays discharged from the hospital between January 1, 2017, and December 31, 2018, were included, yielding 509,586 stays in the final cohort; 305,590 and 203,996 in development and validation cohorts, respectively. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Model discrimination was compared against Acute Physiology and Chronic Health Evaluation (APACHE) IVa/IVb models on the validation cohort using the area under the receiver operating characteristic (AUROC) curve. Calibration assessed by actual/predicted ratios, calibration-in-the-large statistics, and visual analysis. Performance metrics were further stratified by subgroups of admission diagnosis and ICU characteristics. Historic data from two health systems with abrupt changes in Glasgow Coma Scale (GCS) documentation were assessed in the year prior to and after data shift. CCOPM outperformed APACHE IVa/IVb for ICU mortality (AUROC, 0.925 vs 0.88) and hospital mortality (AUROC, 0.90 vs 0.86). Better calibration performance was also attained among subgroups of different admission diagnoses, ICU types, and over unique ICU-years. The CCOPM provided more stable predictions compared with APACHE IVa within an external cohort of greater than 120,000 patients from two health systems with known changes in GCS documentation. CONCLUSIONS: These mortality risk models demonstrated excellent performance compared with APACHE while appearing to mitigate bias introduced through major shifts in GCS documentation at two large health systems. This provides evidence to support using automated capture rather than trained personnel for capture of GCS data used in benchmarking ICUs on mortality performance.

A systematic literature review to identify ethical, legal, and social responsibilities of nonprofit organizations when funding clinical trials in pediatric cancer.

Nonprofit organizations (NPOs) play critical roles as funding sources, research partners, and disseminators of drug developments in pediatric cancer. Yet the literature provides limited guidance about ethical best practices when NPOs make trial funding decisions in this space. We conducted a systematic review of the literature indexed in PubMed and Web of Science to identify the ethical, legal, and social responsibilities of NPOs to four key stakeholder groups in funding pediatric cancer trials: (i) patients/families, (ii) researchers, (iii) industry sponsors, and (iv) donors. We applied the lifecycle framework for patient engagement in drug research and development proposed by Geissler and colleagues to analyze themes related to NPOs' responsibilities across 54 articles that met our inclusion criteria. Emergent themes included transparency surrounding conflicts of interest, the rigor of scientific review, and communication with patients/communities about trial progress. Our research identified critical gaps in best practices for negotiating research partnerships, managing competing research priorities, and pursuing alternative financing models including venture philanthropy. Results from our review informed a set of best practices to guide NPOs in making trial funding decisions that align with stakeholder values and interests.

Associations of self-reported eating disorder behaviors and personality in a college-educated sample.

OBJECTIVE: In order to better understand factors motivating eating disorder (ED) behaviors and better identify persons at-risk for these behaviors, we sought to identify which personality domains and facets were associated with behaviors for weight control. METHODS: ED behavior information was gathered from the University of North Carolina Alumni Heart Study using the question, "have you ever used any of the following to lose weight?" Respondents endorsed any combination of the following: "Vomiting," "Fasting," "Laxatives," "Excessive physical exercise." Personality was measured using the Revised NEO Personality Inventory (NEO-PI-R). One-way ANOVAs were performed comparing personality domains and facets to reported ED behaviors, computed both as separate behaviors and the number of cumulative behaviors. RESULTS: Of 3496 respondents, 9.41% endorsed ever having used at least one ED behavior, with the majority endorsing only a single ED behavior. For both sexes, endorsing greater numbers of ED behaviors was associated with higher scores on Neuroticism and Openness. For women, the strongest associations for behaviors with personality were: excessive exercise with high Impulsiveness; fasting with high Impulsiveness and low Gregariousness; laxative use/purging with high scores on Activity and Feelings. For men, the strongest associations were: excessive exercise with high Impulsiveness; fasting with high Ideas; laxative use/purging with low Modesty. DISCUSSION: Data collected from this sample showed a sex-modulated pattern of association between personality domains and facets with ED behaviors. Our findings support that obtaining personality profiles of individuals exhibiting subclinical eating behaviors will enhance our understanding of who is at risk of developing an ED diagnosis.

Machine learning prediction of the total duration of invasive and non-invasive ventilation During ICU Stay.

Predicting the duration of ventilation in the ICU helps in assessing the risk of ventilator-induced lung injury, ensuring sufficient oxygenation, and optimizing resource allocation. Prior models provided a prediction of total duration without distinguishing between invasive and non-invasive ventilation. This work proposes two independent gradient boosting regression models for predicting the duration of invasive and non-invasive ventilation based on commonly available ICU features. These models are trained on 2.6 million patient stays across 350 US hospitals between 2010 to 2019. The mean absolute error (MAE) for the prediction of duration was 2.08 days for invasive ventilation and 0.36 days for non-invasive ventilation. The total ventilation duration predicted by our model had MAE of 2.38 days, which outperformed the gold standard (APACHE) with MAE of 3.02 days. The feature importance analysis of the trained models showed that, for invasive ventilation, high average heart rate, diagnosis of respiratory infection and admissions from locations other than the operating room were associated with longer ventilation durations. For non-invasive ventilation, higher respiratory rates and having any GCS measurement were associated with longer durations.

Altered neuronal activity relationships between the pedunculopontine nucleus and motor cortex in a rodent model of Parkinson's disease.

The pedunculopontine nucleus (PPN) is a new deep brain stimulation (DBS) target for Parkinson's disease (PD), but little is known about PPN firing pattern alterations in PD. The anesthetized rat is a useful model for investigating the effects of dopamine loss on the transmission of oscillatory cortical activity through basal ganglia structures. After dopamine loss, synchronous oscillatory activity emerges in the subthalamic nucleus and substantia nigra pars reticulata in phase with cortical slow oscillations. To investigate the impact of dopamine cell lesion-induced changes in basal ganglia output on activity in the PPN, this study examines PPN spike timing with reference to motor cortex (MCx) local field potential (LFP) activity in urethane- or ketamine-anesthetized rats. Seven to ten days after unilateral 6-hydroxydopamine lesion of the medial forebrain bundle, spectral power in PPN spike trains and coherence between PPN spiking and PPN LFP activity increased in the approximately 1 Hz range in urethane-anesthetized rats. PPN spike timing also changed from firing predominantly in phase with MCx slow oscillations in the intact urethane-anesthetized rat to firing predominantly antiphase to MCx oscillations in the hemi-parkinsonian rat. These changes were not observed in the ketamine-anesthetized preparation. These observations suggest that dopamine loss alters PPN spike timing by increasing inhibitory oscillatory input to the PPN from basal ganglia output nuclei, a phenomenon that may be relevant to motor dysfunction and PPN DBS efficacy in PD patients.