Abstract
This study identifies the factors that predict the sanctioning outcomes of occupational therapy practitioners.
Since the advent of the 10th amendment in 1791 and the development of the first health occupations state licensure boards (SLBs), one primary function of SLBs is the discipline of ethical misconduct by licensed health care practitioners (Landess, 2019). To this day, the public interacts with SLBs by being able to lodge complaints against practitioners who have behaved contrary to their professional ethical standards. The public, professionals, and institutions can file complaints; SLBs either choose to further investigate the actions of the offending practitioner, known as the respondent, to assess an appropriate sanctioning outcome, close the case, or refer the case to a more appropriate regulatory body. In cases that rise beyond nominal levels of severity, sanctioning decisions of SLBs are communicated publicly in the form of final consent orders—an outline of investigation findings and the final case verdict—or in databases. Many SLBs incorporate the 2020 Occupational Therapy Code of Ethics (the Code) into regulations to help guide and define decision-making parameters (American Occupational Therapy Association [AOTA], 2020a; Maine Board of Occupational Therapy Practice, 2023; South Carolina Board of Occupational Therapy, 2024).
SLB regulation is directly relevant to occupational therapy practice because all occupational therapy practitioners require SLB licensure to practice in the United States. Practice acts, which are minimum competence requirements that occupational therapy practitioners must follow to maintain licensure, include practice expectations in the form of statutes that are often informed by guidelines established by national associations and credentialing bodies, which include AOTA and the National Board for Certification for Occupational Therapy (NBCOT; Neville & Willmarth, 2019). Failure to abide by these expectations are generally grounds for sanctioning, which remains poorly investigated throughout the occupational therapy literature. Thus, investigation into occupational therapy sanctioning has hardly occurred at all, with the exception of studies exploring the effect of jurisprudence examinations on sanctioning frequency against occupational therapists and certified occupational therapy assistants for Alabama and Tennessee SLBs (McDonald, 2013) as well as reasons for licensure revocation of occupational therapy licenses from 2005 to 2015 (Drummond et al., 2019).
Researchers exploring predictive factors regarding discipline have exclusively focused on physician sanctioning. Recent reports on physician discipline indicate that more than 4,000 physicians are disciplined annually (Federation of State Medical Boards, 2018); most physicians are often sanctioned for negligence, practice under the influence, inappropriate prescribing, sexual misconduct, and fraud (Cardarelli & Licciardone, 2006; Clay & Conatser, 2003). Factors associated with physician sanctioning include maleness, increasing age, lack of board certification, international education, Black practitioners, poor medical school performance, and higher years of practice (Khaliq et al., 2005; Kohatsu et al., 2004; Papadakis et al., 2005).
There has yet to be a study that models the disciplinary tendencies of occupational therapy SLBs. The purpose of this study was to predict sanctioning outcomes from all publicly available final consent orders and database items provided by occupational therapy SLBs. We hypothesized that the complaint reason would have the highest influence on sanction severity, with other respondent and complaint attribute variables modifying sanction outcomes odds.
Method
This retrospective study was used to predict sanctioning outcomes from all publicly available final consent order documents and database items provided by occupational therapy SLBs, ranging from July 1985 to May 2023. Institutional review board approval was obtained before data collection. Final consent orders and database items were reviewed for the following respondent and complaint attribute variables: gender, geographic region, professional designation (occupational therapist or certified occupational therapy assistant), practice setting, NBCOT certification, previous disciplinary history, complaint source, complaint reason, number of complaints in a case, months practicing in state at time of sanction, investigation length in months, sanction outcome, total time practicing on expired license, monetary fine amount, probation terms in months, suspension terms in months, presence of conflict resolution conference (CRC), and CRC attendance rate. All available data were used for analysis despite missing data because of state-to-state variabilities in reporting. States were excluded from analysis if they did not provide public sanctioning outcome data to eliminate the need for correspondence with SLBs regarding potentially sensitive or identifying information irrelevant to the study. Cases regarding receipt of a temporary license, or having a license returned to an active status from inactive or sanctioned statuses, were also excluded. A total of 2,400 cases were analyzed across 47 states and Washington, DC.
We completed data analysis using JASP (Version 0.18.1; https://jasp-stats.org/). Descriptive and frequency analyses were used to examine variable central tendency and incidence. Chi-square multinomial testing was used to compare attribute variable data counts with current workforce estimates from the AOTA workforce survey (AOTA, 2020b) regarding gender, professional designation, and geographic region; the NBCOT certification databook (NBCOT, 2023) was used for NBCOT certification counts. Multinomial testing was also used to identify significant overrepresentations in practice setting, sanction outcome, and complaint reason.
Supervised gradient boosting machine learning was used to extract linkages and structured relationships throughout the data set to create predictive models (James et al., 2013). Models were trained on each respondent and complaint attribute variable to determine which variables had the most relative influence on sanction outcomes. Training parameters are reported in Table 1; validation accuracy and test accuracy were chosen as assessments of model predictive performance, which summarize both model true and false positives and negatives. We further analyzed outputs using logistic regression and contingency table modeling to generate odds ratios as an index of predictive effect size.
Machine Learning Training Parameters
The α for this study was set at .05. Numerous protective measures were used to minimize false positivity risk: First, concurrent Bayesian hypothesis testing was used. Bayesian versions of all multinomial testing, logistic regressions, and contingency tables were performed with a Cauchy prior of .707 to generate Bayes factors (BF10), which are ratio outputs that represent the strength of evidence for accepting either alternative or null hypotheses (Andraszewicz et al., 2014). Outputs range from 0 to 1 for evidence in support of the null hypothesis, 1 to 3 for weak evidence in support of the alternative hypothesis, 3 to 10 for moderate evidence, 10 to 30 for strong evidence, 30 to 100 for very strong evidence, and ≥100 for decisive evidence with scores capable of reaching infinity for outcomes that are “overwhelmingly informative” (van Ravenzwaaij et al., 2019). Bayesian outputs demonstrate better accuracy than p values for large data sets, and supplementation of p values with Bayesian outputs is recommended for reducing false positivity risk (Garcia & Puga, 2017). The minimum threshold of acceptable standardized Pearson residuals was set to 2.00 for contingency table outputs; continuity corrections and Fisher’s exact test were used when necessary. Keppel’s modification of the Bonferroni correction was used to adjust for α inflation for post hoc testing.
Results
A total of 2,400 cases were analyzed across 47 states and Washington, DC. Colorado, Alabama, and Arkansas were excluded because of not having public sanctioning data.
Respondent Attribute Variables
Although the majority of cases had female respondents (n = 1,948; 81.17%), male respondents (n = 452; 18.83%) were overrepresented in the data at over twice the rate of the expected national gender distribution of practitioners, χ2(1) = 283.68, p < .001, BF10 = 3.22216. Similarly, although there were more occupational therapist respondents (n = 1,407; 58.63%), certified occupational therapy assistants (n = 993; 41.38%) were overrepresented at a rate 2.50 times greater than the expected national professional designation distribution, χ2(1) = 1,011.51, p < .001, BF10 = 3.44169. Although more respondents were NBCOT certified (n = 1,209; 50.38%), noncertified respondents (n = 1,191; 49.63%) were overrepresented at roughly 5.50 times the expected rate, χ2(1) = 10,061.76, p < .001, BF10 = ∞. Southern (n = 910; 37.92%) and midwestern (n = 889; 37.04%) states had higher respondent counts than northeastern (n = 377; 15.71%) and western (n = 224; 9.33%) states when assessed against practitioner distributions per geographic region, χ2(3) = 344.92, p < .001, BF10 = 8.4777.
Only 26.12% (n = 628) of cases reported practice setting. Respondents in subacute rehabilitation (n = 293; 46.66%) and home health (n = 118; 18.79%) were overrepresented compared with school based (n = 54; 8.60%), acute care (n = 47; 7.48%), orthopedic and neurologic outpatient (n = 45; 7.17%), outpatient pediatrics (n = 36; 5.73%), early intervention (n = 19; 3.03%), inpatient acute rehabilitation (n = 8; 1.27%), psychosocial rehabilitation (n = 6; 0.96%), and academia (n = 2; 0.32%), χ2(9) = 1,102.64, p < .001, BF10 = 1.94167. No prior disciplinary history was found for 91.4% of respondents.
Complaint Attribute Variables
A total of 3,752 sanctions were imposed on respondents, with an average of 1.56 sanctions per case. Monetary fines (n = 1,299; 34.62%), reprimands (n = 721; 19.22%), probations (n = 691; 18.42%), and suspensions (n = 593; 15.80%) were overrepresented compared with revocation (n = 244; 6.50%), letter of surrender (n = 144; 3.84%), and summary suspension sanctions (n = 60; 1.60%), χ2(6) = 2,069.35, p < .001, BF10 = ∞.
Contextual sanctioning information demonstrated considerable variability across states. The average monetary fine was $539.57 (SD = $1,195.47, Mdn = $215.00); the largest fine was $13,972.00. The average probation term was 22.41 mo (n = 379; SD = 13.49, Mdn = 24.00), whereas the average suspension term was 7.72 mo (n = 498; SD = 13.00, Mdn = 1.50). The average respondent length of practice at sanctioning was 118.39 mo (n = 2,032; SD = 92.53, Mdn = 96.00); the shortest length was 0 mo, whereas the longest length was 517 mo. The average length of an SLB complaint investigation was 4.77 mo (n = 1,277; SD = 7.10, Mdn = 2.00); the shortest investigation was 0 mo, whereas the longest investigation was 67 mo.
A total of 2,320 instances of complaint source were reported, often with more than one source in a case. SLB (n = 1,230; 53.02%) and employer-initiated (n = 537; 23.15%) complaints were overrepresented compared with judiciary branch (n = 230; 9.91%), consumer and patient (n = 195; 8.41%), self-report (n = 100; 4.31%), and coworker- and employee-initiated (n = 28; 1.21%) complaints, χ2(5) = 2,653.25, p < .001, BF10 = ∞.
There were 2,705 recorded reasons for complaint, with an average of 1.14 reasons per case. Continuing education unit (CEU) audit failure (n = 557; 20.59%), practice on an expired license (n = 397; 14.68%), fraudulent billing (n = 361; 13.35%), falsification of records (n = 342; 12.64%), criminal conviction (n = 215; 7.95%), and failure to comply with SLB orders (n = 139; 5.14%) were overrepresented compared with other complaints, including submission of false statement to an SLB (n = 111; 4.10%), practice without initial licensure (n = 95; 3.51%), incompetent practice (n = 73; 2.70%), civil charge (n = 71; 2.62%), patient negligence (n = 49; 1.81%), failure to submit timely documentation (n = 42; 1.55%), patient abandonment (n = 37; 1.37%), practice of physical agent modalities without certification (n = 30; 1.11%), failure to supervise (n = 21; 0.78%), sexual misconduct (n = 21; 0.78%), operating outside of scope of practice (n = 20; 0.74%), aiding and abetting (n = 17; 0.63%), theft (n = 15; 0.55%), exploitation of a vulnerable adult (n = 9; 0.33%), and Health Insurance Portability and Accountability Act of 1996 (HIPAA) violation (n = 7; 0.26%), χ2(22) = 4,310.73, p < .001, BF10 = ∞. Respondents sanctioned for practicing on an expired license practiced an average of 177.77 days beyond their renewal deadline (n = 323; SD = 341.42, Mdn = 64.00); the shortest length was 1 day, whereas the longest length was 3,313 days. Of 421 documented CRCs, the total attendance rate was 78.85%.
Predictive Factors for Sanctioning Outcomes
Refer to Table 2 for predictive factor outcomes by sanction and Table 3 for machine learning model accuracies.
Predictive Factors for Sanction Outcome
Note. BF10 = Bayes factors; CEU = continuing education unit; CI = confidence interval; COTA = certified occupational therapy assistant; CRC = conflict resolution conference; OR = odds ratio; PAMs = physical agent modalities; SLB = state licensure board.
Gradient Boosting Model Accuracies
Discussion
The purpose of this study was to model factors that predict sanctioning outcomes. The most influential factors in predicting sanctioning outcome, in order of influence, were complaint reason, practice setting, and complaint source. Geographic region, number of complaints in a given case, and length of investigation in months were secondary in their overall influence.
The influence of complaint reason suggests that sanctioning may be well predicted by offense severity. Complaint reasons associated with higher odds of severe sanctioning outcomes included failure to comply with SLB order, criminal conviction, sexual misconduct, practice under the influence, and incompetent practice, some of which have been previously identified as risk factors for severe sanctioning (Cardarelli & Licciardone, 2006; Clay & Conatser, 2003; Drummond et al., 2019). Thematically, these reasons for complaint suggest that immediate and direct harm to the public are met with appropriately severe SLB responses. Contrast this with offenses associated with less severe sanctioning outcomes—including CEU audit failure, failure to submit timely documentation, practice on expired license, and practice without initial licensure— and the difference in scope regarding public threat is evident.
The influence of complaint source suggests that SLBs may also interpret case severity by who reported the offense. The most severe sanctioning outcomes were associated with client or legal sources, including employer complaints, patient and consumer complaints, and judiciary branch complaints; less severe outcomes were associated with SLB-sourced complaints. There may also be underlying relationships between complaint reason and complaint sources; judiciary branch complaints are likely associated with criminal convictions that would justify severe sanctions. However, SLB-initiated complaints, which are likely initiated because of a respondent’s inability to follow practice act procedure during CEU audits or licensure renewal, may be perceived as less directly harmful to the public faith in occupational therapy, warranting less severe sanctioning.
Practice setting also appeared influential. Acute care, the setting with the lowest productivity demands (∼75%)—productivity being the expected percentage of time worked that produces billable units—was associated with less severe sanctioning outcomes, whereas subacute rehabilitation, the setting with the highest demands (∼90%), was associated with severe sanctioning outcomes (Tammany et al., 2019). This finding supports Tammany et al. (2019), who asserted that higher productivity expectations are associated with unethical practice, particularly for falsification of records and fraudulent billing, which increased risk of suspension outcomes by nearly triple and quadruple, respectively, in this study. The only practice setting associated with revocation was orthopedic and neurologic outpatient, which may be because of the complex Medicare and Medicaid fraud seen in this setting (Federal Bureau of Investigation, 2014; Office of Public Affairs, 2016).
Although modestly influential, there may be differences in sanctioning patterns between geographic regions similar to those noted in other sources (Harris & Byhoff, 2017). Western SLBs may have more stringent disciplinary practices than midwestern and northeastern SLBs given their associations with probation and revocation outcomes. Further exploration of SLB minimum sanctioning guidelines, alongside individual state differences regarding ethical behavior and applications of the Code, may be necessary to understand regional differences in sanctioning.
Maleness was predictive of revocation, and certified occupational therapy assistant professional designation was predictive of suspension; the former is congruent with existing literature (Khaliq et al., 2005; Kohatsu et al., 2004). Although associations between these attributes and severe sanctioning should be approached with caution given their BF10 outcomes, Bayesian evidence regarding their overrepresentation in disciplinary cases compared with national practitioner distributions was decisively strong. The overrepresentation of respondents without NBCOT certification also may suggest protective elements of certification in terms of following SLB compliance procedures and avoiding behaviors implicated in sanctioning.
Limitations
The largest limitation of this study was the inconsistency of SLB data reporting in databases and consent orders because missing values likely influenced modeling accuracy. Second, although this study technically included the entire population of respondents in occupational therapy, SLB disciplinary information is updated in real time; thus, this study likely did not capture every case at the time of publication. Lastly, given the amount of data collected across 48 sources, coding and reporting errors may exist.
Implications for Occupational Therapy Practice
These findings have the following implications for occupational therapy practice: ▪ An increased awareness of SLB sanctioning patterns among complaint reason, practice setting, complaint source, and geographic regional factors may shape future behaviors of practitioners, particularly overrepresented groups, alongside regulators and academicians, because these behaviors relate to SLB sanctioning patterns, outcomes, and education for occupational therapists and certified occupational therapy assistants. ▪ SLBs and academicians should consider their processes on how they educate stakeholders on the value of comprehending state practice acts, following SLB compliance procedures, and the apparent value of NBCOT certification in matters of compliance.
Conclusion
In this study, we modeled factors that predict sanctioning outcomes onto respondents by SLBs, with multiple complaint and respondent attributes being associated with all levels of sanction outcomes. Further research of new disciplinary cases as they are recorded and practical applications of this knowledge in contemporary practice are warranted.
Footnotes
Acknowledgments
We thank the AJOT reviewers and editors of this article for their valuable feedback and time—it does not go unappreciated. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.