Labor Market Intelligence Using Natural Language Processing for the U.S. Direct Care Workforce: Predicting Skill Gaps, Recruitment Needs, and Emerging Healthcare Workforce Demand

Abstract

Background: The U.S. direct care workforce—home health aides, personal care aides, and nursing assistants—is the largest and fastest-growing occupational group in the healthcare economy and is chronically undersupplied and understudied. With demographic ageing and the shift in care delivery from institutional to home and community-based care, there is a critical need for accurate and near real-time labour market intelligence to predict skill shortages, recruitment challenges, and new opportunities for care.

Objective: This study introduces a Natural Language Processing (NLP)-enhanced labor-market intelligence framework and applies it in an empirical application to a national survey of 350 direct care workers (DCWs) in 20 states in the U.S. to (i) quantify the skill-gap indices across the 10 competency domains; (ii) create a composite Recruitment-Need Score; and (iii) project the workforce demand through 2033.

Methods: A structured cross-sectional questionnaire was used to collect demographic information, wage-hour information, 10 skill competencies on a Likert scale, job-experience information, and 2 open-ended questions. Descriptive statistics, Pearson correlations and stratified subgroup analysis were used for quantitative data. The NLP pipeline of tokenization, lemmatization and TF-IDF was used on open-ended answers to extract latent themes related to skills and recruitment, which was then complemented with the contextual embedding techniques.

Results: The top three skill gaps were Telehealth (2.67), Technology/EHR (2.32), and Dementia Care (2.07). NLP extraction validated these themes, with the terms “training,” “telehealth,” “technology,” “dementia,” and “documentation” being the most prominent in the corpus. Turnover intent was high (mean 3.45/5) and negatively (but only weakly) associated with wage (r = 0.11 in the observed direction), suggesting that non-monetary factors play a significant role in the decision to leave. Personal Care Aides (90.5) and CNAs (90.3) had the highest Composite Recruitment-Need Scores. The total demand for the region is estimated to be about 11.7 million DCWs by 2033.

Conclusions: The NLP-augmented workforce intelligence provides a scalable, granular and policy-relevant tool for predicting skill shortages in the direct care sector. Federal and State workforce plans should prioritize investment in telehealth and EHR competency training, wage stabilization, and pathway credentialing.