Crowdsourcing with Enhanced Data Quality Assurance: An Efficient Approach to Mitigate Resource Scarcity Challenges in Training Large Language Models for Healthcare.

Large Language Models (LLMs) have demonstrated immense potential in artificial intelligence across various domains, including healthcare. However, their efficacy is hindered by the need for high-quality labeled data, which is often expensive and time-consuming to create, particularly in low-resource domains like healthcare. To address these challenges, we propose a crowdsourcing (CS) framework enriched with quality control measures at the pre-, real-time-, and post-data gathering stages. Our study evaluated the effectiveness of enhancing data quality through its impact on LLMs (Bio-BERT) for predicting autism-related symptoms. The results show that real-time quality control improves data quality by 19% compared to pre-quality control. Fine-tuning Bio-BERT using crowdsourced data generally increased recall compared to the Bio-BERT baseline but lowered precision. Our findings highlighted the potential of crowdsourcing and quality control in resource-constrained environments and offered insights into optimizing healthcare LLMs for informed decision-making and improved patient care.

Transparent deep learning to identify autism spectrum disorders (ASD) in EHR using clinical notes.

OBJECTIVE: Machine learning (ML) is increasingly employed to diagnose medical conditions, with algorithms trained to assign a single label using a black-box approach. We created an ML approach using deep learning that generates outcomes that are transparent and in line with clinical, diagnostic rules. We demonstrate our approach for autism spectrum disorders (ASD), a neurodevelopmental condition with increasing prevalence. METHODS: We use unstructured data from the Centers for Disease Control and Prevention (CDC) surveillance records labeled by a CDC-trained clinician with ASD A1-3 and B1-4 criterion labels per sentence and with ASD cases labels per record using Diagnostic and Statistical Manual of Mental Disorders (DSM5) rules. One rule-based and three deep ML algorithms and six ensembles were compared and evaluated using a test set with 6773 sentences (N = 35 cases) set aside in advance. Criterion and case labeling were evaluated for each ML algorithm and ensemble. Case labeling outcomes were compared also with seven traditional tests. RESULTS: Performance for criterion labeling was highest for the hybrid BiLSTM ML model. The best case labeling was achieved by an ensemble of two BiLSTM ML models using a majority vote. It achieved 100% precision (or PPV), 83% recall (or sensitivity), 100% specificity, 91% accuracy, and 0.91 F-measure. A comparison with existing diagnostic tests shows that our best ensemble was more accurate overall. CONCLUSIONS: Transparent ML is achievable even with small datasets. By focusing on intermediate steps, deep ML can provide transparent decisions. By leveraging data redundancies, ML errors at the intermediate level have a low impact on final outcomes.

Intentional interprofessional leadership in maternal and child health.

PURPOSE: This paper aims to describe an interprofessional leadership training program curriculum implemented by a new maternal and child health leadership training program, its collaboration with a well-established leadership consortium, the measures taken to evaluate this training and implications for other leadership programs. DESIGN/METHODOLOGY/APPROACH: The intentional leadership program weaves together the complementary core threads to create strong sets of skills in the areas of personal leadership, leading and influencing others and creating effective interprofessional partnerships with others around women and children's health. FINDINGS: The strong emphasis on the incorporation of leadership competencies coupled with evidence-based leadership training strengthens students' clinical skills, enhances workforce development and increases interdisciplinary health care practices. RESEARCH LIMITATIONS/IMPLICATIONS: The findings presented in this paper are limited to self-reported changes in understanding components of leadership skills for self, others and the wider community and attitudes and beliefs related to interdisciplinary training and interprofessional team decision-making. SOCIAL IMPLICATIONS: The in-depth focus on one's self, teams and on the wider community enhances each individual's grasp of how people and organizations approach women and children's health challenges and strengthens their ability to negotiate among the diverse disciplines and cultures. ORIGINALITY/VALUE: This paper details the intentional incorporation of leadership skill development throughout an academic program and brings to focus the importance of thoughtful leadership development to prepare participants to anticipate, manage and take advantage of changes in knowledge and health care delivery systems.

Incidence of delirium in children with cancer.

There are limited data on the incidence of delirium in children with cancer. We performed a retrospective chart review of all pediatric oncology admissions over a 1 year period to determine the incidence of delirium in this population. We identified seven patients with delirium (10% incidence). Delirium is associated with significant morbidity and mortality, and is likely under-recognized in this population. Improved diagnosis and treatment of delirium may improve outcomes in children with cancer.

Rehabilitation of children with traumatic brain injury: descriptive analysis of a nationwide sample using the WeeFIM.

OBJECTIVE: To describe functional capability at admission and discharge of children with traumatic brain injury (TBI) in rehabilitation settings. DESIGN: Descriptive analysis. SETTING: Inpatient pediatric rehabilitation hospitals in the United States. PARTICIPANTS: Children (N=3815) in 56 pediatric inpatient rehabilitation facilities who were discharged during 1999 to 2001. INTERVENTIONS: Not applicable. MAIN OUTCOME MEASURES: Admission and discharge WeeFIM scores. RESULTS: Admission and discharge WeeFIM scores correlated positively with age at admission, time from injury to rehabilitation admission, and length of stay (LOS). Higher admission WeeFIM scores correlated with shorter LOS, shorter time from injury to admission to rehabilitation, and higher discharge WeeFIM scores. CONCLUSIONS: Children with TBI demonstrated significant improvement in functional measures during rehabilitation. Discharge function and LOS correlated with admission severity, with children who had higher functional status and shorter time between injury and rehabilitation care having higher discharge function and shorter LOS.

Effect of shortened length of stay on functional and educational outcome after pediatric rehabilitation.

OBJECTIVE: The purpose of this study was to assess changes in the length of stay and its effect on effectiveness and return to school in an inpatient pediatric rehabilitation unit during a 5-yr period from fiscal year 1997 through 2001. DESIGN: We reviewed prospectively collected data for a cohort of 321 children during fiscal years 1997-2001. RESULTS: Length of stay was significantly shortened, with mean lengths of stay of 58.9, 43.5, 30.7, 40.9, and 24.0 days in years 1997 through 2001, respectively. Change in length of stay remained significantly decreased after adjusting for age, sex, admission diagnosis, admission severity, and type of health insurance. There was no difference in mean change in effectiveness measured by change in admission and discharge WeeFIM ratings. There were significant differences across years in the educational placement of children at discharge, with a declining trend in the proportion of children discharged to classroom-based educational services. CONCLUSION: There was a reduction in inpatient length of stay during a 5-yr period for children in this pediatric rehabilitation setting. During this time, there was no change in the effectiveness of rehabilitation as measured by functional outcome. However, using return to a classroom setting as a marker of reintegration into routine activities, fewer children returned to a similar level of community participation.

Brain imaging as a predictor of early functional outcome following traumatic brain injury in children, adolescents, and young adults.

OBJECTIVES: A depth of lesion (DOL) model using brain imaging has been proposed to aid in medical decision-making and planning for rehabilitation resource needs. The purpose of this study was to determine the early prognostic value of a DOL classification system for children and young adults following severe traumatic brain injury. METHODS AND OUTCOME MEASURES: CT/MRI brain imaging studies on 92 patients, aged 3 to 21, admitted to the Kluge Children's Rehabilitation Center, University of Virginia, were evaluated to determine DOL. Images were classified according to 5 DOL levels (cortical to brainstem). Functional outcomes in mobility, self-care, and cognition, as rated on the WeeFIM instrument, were compared by DOL levels. RESULTS: Admission WeeFIM scores were significantly different for the DOL levels with the highest score for frontal and/or temporal lesions and the lowest for lesions including the brainstem or cerebellum (P<.001). However, the deeper the lesion, the greater the functional gains (P=.05), resulting in discharge WeeFIM scores that were not significantly different across DOL levels. Patients with deeper lesions tended to have longer lengths of stay in rehabilitation but were able to "catch up" with patients who had more superficial lesions. CONCLUSIONS: While relatively simple and convenient, the DOL classification system is limited in its usefulness as an early prognostic tool. It may not be possible to predict outcome in the early acute phase in the intensive care unit on the basis of standard brain imaging alone. Patients with deeper lesions may enter rehabilitation at a more impaired level but can make remarkable progress, though it may take longer than for less severely injured individuals.