Joanna Shen

Background: Health-related social needs (HRSNs), such as housing instability, food insecurity, and financial strain, are increasingly prevalent among patients. Healthcare organizations must first correctly identify patients with HRSNs to refer them to appropriate services or offer resources to address their HRSNs. Yet, current identification methods are suboptimal, inconsistently applied, and cost prohibitive. Machine learning (ML) predictive modeling applied to existing data sources may be a solution to systematically and effectively identify patients with HRSNs. The performance of ML predictive models using data from electronic health records (EHRs) and other sources has not been compared to other methods of identifying patients needing HRSN services. Methods: A screening questionnaire that included housing instability, food insecurity, transportation barriers, legal issues, and financial strain was administered to adult ED patients at a large safety-net hospital in the mid-Western United States (n = 1,101). We identified those patients likely in need of HRSN-related services within the next 30 days using positive indications from referrals, encounters, scheduling data, orders, or clinical notes. We built an XGBoost classification algorithm using responses from the screening questionnaire to predict HRSN needs (screening questionnaire model). Additionally, we extracted features from the past 12 months of existing EHR, administrative, and health information exchange data for the survey respondents. We built ML predictive models with these EHR data using XGBoost (ML EHR model). Out of concerns of potential bias, we built both the screening question model and the ML EHR model with and without demographic features. Models were assessed on the validation set using sensitivity, specificity, and Area Under the Curve (AUC) values. Models were compared using the Delong test. Results: Almost half (41%) of the patients had a positive indicator for a likely HRSN service need within the next 30 days, as identified through referrals, encounters, scheduling data, orders, or clinical notes. The screening question model had suboptimal performance, with an AUC = 0.580 (95%CI = 0.546, 0.611). Including gender and age resulted in higher performance in the screening question model (AUC = 0.640; 95%CI = 0.609, 0.672). The ML EHR models had higher performance. Without including age and gender, the ML EHR model had an AUC = 0.765 (95%CI = 0.737, 0.792). Adding age and gender did not improve the model (AUC = 0.722; 95%CI = 0.744, 0.800). The screening questionnaire models indicated bias with the highest performance for White non-Hispanic patients. The performance of the ML EHR-based model also differed by race and ethnicity. Conclusion: ML predictive models leveraging several robust EHR data sources outperformed models using screening questions only. Nevertheless, all models indicated biases. Additional work is needed to design predictive models for effectively identifying all patients with HRSNs.

Background: Digital phenotyping, the use of personal digital devices to capture and categorize real-world behavioral and physiological data, holds great potential for complementing traditional clinical assessments. However, missing data remains a critical challenge in this field, especially in longitudinal studies where missingness might obscure clinically relevant insights. Objective: This paper examines the impact of data missingness on digital phenotyping clinical research, proposes a framework for reporting and accounting for data missingness, and explores its implications for clinical inference and decision-making. Methods: We analyzed digital phenotyping data from a study involving 85 patients with chronic musculoskeletal pain, focusing on active (PROMIS-29 survey responses) and passive (accelerometer and GPS measures) data collected via the Beiwe Research Platform. We assessed data completeness and missingness at different timescales (day, hour, and minute levels), examined the relationship between data missingness and accelerometer measures and imputed GPS summary statistics, and studied the stability of regression models across varying levels of data missingness. We further investigated the association between functional status and day-level data missingness in PROMIS-29 subscores. Results: Data completeness showed substantial variability across timescales.Accelerometer-based cadence and imputed GPS-based home time and number of significant locations were generally robust to varying levels of data missingness. However, the stability of regression models was affected at higher thresholds (40% for cadence and 60% for home time). We also identified patterns wherein data missingness was associated with functional status. Conclusion: Data missingness in clinical digital phenotyping studies impacts individual- and group-level analyses. Given these results, we recommend that studies account for and report data at multiple timescales (we recommend day, hour, and minute-level where possible), depending on the clinical goals of data collection. We propose a modified framework for categorizing missingness mechanisms in digital phenotyping, emphasizing the need for clinically relevant reporting and interpretation of missing data. Our framework highlights the importance of integrating clinical with statistical expertise, specifically to ensure that imputing missing data does not obscure but helps capture clinically meaningful changes in functional status.