Search

Hospital length-of-stay is a central operational metric for inpatient capacity planning, discharge coordination, and resource allocation. Accurate prediction remains difficult because patient trajectories are heterogeneous, nonlinear, and shaped by evolving clinical events during admission. Traditional statistical models often have limited flexibility for high-dimensional and sequential electronic health record data. Across the literature, there is no settled consensus regarding optimal model architecture, feature representation, validation design, or clinical implementation strategy. This systematic review synthesizes machine learning approaches for hospital length-of-stay prediction published from 2017 to 2022. It focuses on EHR feature types, model architectures, validation methods, interpretability strategies, and reported operational outcomes. A structured review of peer-reviewed literature was conducted using targeted search strings related to machine learning, deep learning, electronic health records, discharge prediction, and hospital length-of-stay. The review included studies across emergency, inpatient, surgical, pediatric, cardiovascular, and intensive care settings. The literature suggests that gradient boosting, random forest, ensemble learning, and recurrent neural networks are common approaches for LOS prediction. However, external validation remains uncommon, prediction horizons vary widely, and operational implementation outcomes are reported less consistently than model development results. Future research should prioritize external validation, prospective implementation studies, standardized outcome definitions, and transparent reporting of workflow barriers. Shared benchmarking datasets and multi-center validation consortia would strengthen comparability across LOS prediction studies.

Failure to follow up after abnormal screening results is a persistent ambulatory safety problem. Because the diagnostic process often spans patient notification, scheduling, primary care review, and reminder outreach, delays may emerge gradually before they become visible in registry reports. Existing care gap reports commonly classify results as closed or open after a defined time window. This retrospective framing limits the ability to detect patients who are currently moving toward delayed diagnostic resolution. The objective of this article is to describe a predictive model that estimates the probability of delayed diagnostic follow-up after an abnormal screening result. The proposed model integrates patient engagement signals, scheduling activity, clinician workload, result severity, and reminder history. The model would use a gradient-boosted classification framework trained on historical abnormal results and longitudinal care-process features. Inputs would be extracted from patient portals, scheduling systems, primary care workload records, laboratory and radiology result metadata, and reminder logs. Conceptually, the model would generate a daily updated risk score for each unreconciled abnormal result. It would also identify the dominant drivers of risk, such as unread portal messages, repeated appointment cancellations, limited primary care availability, high-severity findings, or escalating reminder activity. A predictive approach could shift delayed follow-up management from retrospective audit to proactive prioritization. By identifying patients at greatest risk before the diagnostic window closes, care teams could better allocate outreach and navigation resources.