Natural Language Processing Support for Evaluating Similarity in Physical Therapy Laws, Rules, and Regulations

July 2023
Researcher: Human Resources Research Organization (HumRRO)

HRRI & HumRRO partnered to evaluate the feasibility of computer-automated methods for detecting similarities and differences among the laws, rules, and regulations (LRRs) governing the practice of physical therapy (PT) across the Federation of State Boards of Physical Therapy’s (FSBPT) member jurisdictions.

Authors

J. Dahlke

Executive Summary

The Healthcare Regulatory Research Institute (HRRI) partnered with the Human Resources Research Organization (HumRRO) to evaluate the feasibility of computer-automated methods for detecting similarities and differences among the laws, rules, and regulations (LRRs) governing the practice of physical therapy (PT) across the Federation of State Boards of Physical Therapy’s (FSBPT) member jurisdictions. Specifically, HRRI asked HumRRO to explore whether natural language processing (NLP) techniques could help to automate comparisons of LRRs across jurisdictions.

NLP is a discipline closely related to computational linguistics with roots in the interdisciplinary use of artificial intelligence and computer science principles to develop computerized models of human language. NLP researchers aim to use such models to understand and analyze text produced by humans, especially in high-volume contexts where systematic manual analyses of the text would be impractical. NLP methods are quite varied, and NLP tools exist to automate a wide variety of tasks, such as language recognition, language parsing, structural/syntactical analysis, representation of text in terms of quantitative dimensions, semantic analysis (of words, sentences, or longer pieces of text), sentiment analysis, and topic modeling, just to highlight a subset of use cases. Such a diverse collection of techniques makes it possible to automate many facets of text-based analyses and streamline tasks that would take human analysts many hours of labor to accomplish without automation. Beyond the diversity of NLP techniques available to support automation, users of these techniques can also select the degree to which a task is automated (e.g., should an NLP tool automate all aspects of an analysis, or only part of a process while leaving room for human intervention?).

HRRI developed an interest in cross-jurisdictional LRR comparisons after learning about NLP-based research HumRRO conducted to identify related occupations for the O*NET program (Dahlke et al., 2022). In that research, the semantic similarity between pieces of text describing O*NET occupations represented two thirds of the information that contributed to a relatedness composite, which was then used to construct a list of related occupations for each O*NET occupation. Theoretically, the types of pairwise comparisons HumRRO used to develop that related occupations framework could be extended to analyses of LRRs across jurisdictions.

HRRI’s desire to explore similarities in LRRs across jurisdictions developed out of FSBPT’s licensure assessment program. In addition to administering the National Physical Therapy Examination (NPTE®), which is required for licensure as a physical therapist or physical therapy assistant in the US (APTA, 2023; FSBPT, 2023c), FSBPT also offers jurisprudence exams and Jurisprudence Assessment Modules (JAMs®) for a subset of jurisdictions. More than half of jurisdictions in the US require licensees to pass a jurisprudence exam, and FSBPT develops, maintains, and administers jurisprudence exams for Arizona, California, the District of Columbia, Florida, and Nebraska (FSBPT, 2023b). Currently, each jurisdiction that requires licensees to pass a jurisprudence exam relies on an exam that is tailored to the PT-related LRRs in that jurisdiction, and the jurisdiction-specific nature of the exams makes scores non-transferrable among jurisdictions. FSBPT also offers JAMs—online assessments that licensees can take to meet requirements in some jurisdictions when renewing their licenses—for Georgia, Hawaii, Kansas, New Hampshire, New Jersey, Ohio, Oregon, and Texas (FSBPT, 2023a).

Identifying areas of overlap and nuance between the LRRs from different jurisdictions could help FSBPT to develop micro-assessments for practitioners interested in changing jurisdictions or practicing in multiple jurisdictions. Such micro-assessments could help practitioners to understand the similarities and differences between the jurisdictions’ LRRs. NLP methods have potential for helping to identify these types of similarities and differences, and the goal of the present research was to evaluate the feasibility of using NLP tools to guide the detection of relevant overlaps and nuances among LRRs from a representative sample of jurisdictions.

We used LRR information provided by HRRI to conduct feasibility analyses in which we used NLP approaches to automate the detection of similar text between jurisdictions. The remainder of this report is dedicated to describing the methods, results, and implications of our feasibility analyses.

Browse Similar Research on HRRI.org

New Technology