Evaluating the association of state regulation of community health workers on adoption of standard roles, skills, and qualities by employers in select states: a mixed methods study

This study of CHW employment trends and state policy changes was conducted in three phases that were comprised of qualitative interviews, analysis of job ads using Natural Language Processing (NLP), and job ad content analysis. The first phase included semi-structured interviews with nine CHW subject matter experts. The experts included individuals and representatives from a state-level CHW program, CHW certifying organization, state CHW association, Medicaid managed care CHW program director, and chair of a state level CHW council, to name a few. These experts provided valuable insights that helped shape the design of this study.

The second phase, NLP, included extracting jobs ads derived from a particular performance period and analyzing the text of job ad content. Job ad data were obtained from a labor market research software company named Chmura. The company provides proprietary labor market software, JobsEQ that collects job ad data daily using Real-Time Intelligence. A sample of job ads posted by employers in Alabama, Maryland, Rhode Island, Tennessee, Texas, Virginia, and Wyoming from 2017 to 2020 were retrieved from JobsEQ using Standard Occupation Code (SOC) 21–1094 Community Health Workers and the following titles: community health worker, peer health educator, peer specialist, peer support specialist, doula, and promotor de salud. These commonly used titles for CHWs were identified through literature review and the SME interviews. The query resulted the job title, employer, and job ad URL for each job ad.

The job ad URLs were used to generate a unique identifier for each observation and retrieve the complete job description for each ad. The observations were deduplicated using the unique identifier. Employers were categorized by type as (1) hospitals/health systems, (2) other non-hospital healthcare, (3) health department, (4) health plans, (5) community-based organizations, (6) other, and (0) unknown. The ads categorized with employer type unknown or other and those that did not include one of the CHW titles in the job title or job description were removed to reduce false positive results for CHW related job ads. The final sample included 4804 deduplicated job ads. Next, job ad text was parsed and tokenized into three-word patterns (trigrams). Then the list of roles, skills, and qualities for community health workers identified by the Community Health Worker Consensus Project [41] were lemmatized, a text pre-processing technique in which words are reduced to their root. A binary variable was created for each lemmatized key word and assigned 0 if the ad did not include the word(s) or 1 if it included the word(s). A composite variable for roles, skills, and qualities was generated from the sum of values for individual key words under each category.

In the third phase of this study, a series of hypotheses were tested using one-way analysis of variance (ANOVA), chi-square analysis (CHI2), and multivariate ANOVA (MANOVA) to examine the association between state CHW regulation (policy) and CHW roles, skills, and qualities. The first ANOVA tested differences in composite scores for skills, qualities, and roles between state policy types (i.e., no policy, new policy, mature policy). A Bonferroni post-hoc test was conducted to identify specific differences between types of policy. The second analysis used chi-square to test the association between policy type and individual key words representing CHW skills, qualities, and roles (e.g., assessment, care, and advocate). The final analysis used MANOVA to address possible threats to validity, because the data were not normally distributed and the sphericity assumption is often violated.

A total of 4804 job ads were included in this analysis. Twenty three percent of the ads were from states with no state certification program, 63% were derived from states with certification programs less than 5 years (new policy states), and approximately 14% were from a state with long standing CHW certification program (i.e., mature policy states). The majority of the job ads (63.24%) were for ‘community health workers’. An additional 1766 ads (36.76%) were identified for jobs entitled doula, peer health education, peer specialist, and peer support specialist.

The largest number of job ads were posted by community-based organizations (30.47%), followed by hospital/health systems (29.87%), and non-hospital healthcare organizations (18.61%). Table 2 displays the distribution of ads by state policy type, employer type, and occupational title. Also included in Table 2 is the frequency count of job ads identified with the standardized skills, qualities, and roles recommended for state certification programs. Across all job ads, professional (88%), health disparity (38.6%), and relationship building (35.9%) were the most commonly identified skills specified in job ads. Motivate (68.4%), self-direct (38.2%), and care (17.9%) were the most common qualities included. System navigation (35.5%), social support (16.7%), and coach (15.3%) were the top roles identified in ads.

The average number of jobs ads by identified roles (F(2, 4801) = 27.97, p = 0.000), skills (F(2, 4801) = 38.17, p = 0.000), and qualities (F(2, 4801) = 2.23, p = 0.006) varied significantly based on state policy type (Table 3). The Bonferroni post hoc test indicates that the mean job ads that include roles and skills are significantly different between all state policy types (p < 0.05).

The percentage of job ads with the roles advocate, care coordination, coach, direct service, health education, mediation, social support, and system navigation are significantly different by state policy type (p ≤ 0.05). New and mature policy states are more likely to include these roles in job ads than states without certification programs (Table 4). The percentage of job ads with skills assessment, capacity building, communication, community, evaluation, outreach, professional, public health, social determinant, and social service system are also significantly different by state policy type (p ≤ 0.05).

A higher percentage of job ads in states with new or mature CHW certification programs included these skills than states without a program, with the exception of assessment, where a higher percentage of ads are found in states without a certification program) (Table 5). The percentage of job ads that included the CHW qualities care, honest, motivate, patient, reliable, and self-direct were significantly different (p ≤ 0.05). States with no CHW certification programs had a higher percentage of job ads that included care, honest, reliable, and self-direct. Motivate was found more often in new and mature policy states (Table 6).

In the analysis to determine if C3 defined roles included in job ads varied by employer type, we observed statically significant differences between organizations and policy settings. The mean number of job ads that included roles were highest in job ads posted by non-hospital healthcare employers in states without state CHW policies (F(4, 1107) = 60.33, p = 0.000) and by hospital/health system employers in states with new policies (F(4, 13,021) = 18.49, p = 0.000) or mature CHW policies (F(4, 661) = 6.80, p = 0.000).

The mean number of job ads that included C3 defined skills were greater in community-based organizations in states without a CHW policy (F(4, 1107) = 40.32, p = 0.000), by hospital/health system employers in new policy states (F(4, 3021) = 11.24, p = 0.000), and by health plans in mature policy states (F(4, 661) = 5.65, p = 0.000). The differences in adoption or inclusion of qualities were not significantly different by employer type in states without CHW policies and in the mature policy state. The mean number of ads that included qualities were higher for hospitals/health systems in new policy states (F(4, 3,021) = 115.06, p = 0.000). The Bonferroni post-hoc tests indicated the means for roles, skills, and qualities were not significantly different between some employer types and the results varied based on state policy type. Further exploration of differences by type of employer is needed. Table 7 displays the results of the analysis of variance between by type of employer and state policy type.

MANOVA was utilized as an alternative test of validity, to address the limitations of ANOVA for susceptibility to violations of the assumption of sphericity. Using Wilk’s lamba analysis, we reject the null hypothesis that state CHW policy type (F(2, 4801) = 26.21), p = 0.000) and employer type (F(4, 4799) = 69.08, p = 0.000) have no effect on roles, skills, and qualities identified in job ads.

This research represents an important contribution to understanding the diffusion and adoption of occupational standards by employers. This study found state CHW policies and types of CHW employers were associated with variation in adoption of nationally defined occupational roles, skills, and qualities. The mean number of jobs that included the roles and skills were significantly higher in mature policy states. Health plans in such states may have greater standardization in how CHWs are employed and, therefore, are more likely to have job ads that incorporated the specific descriptive terms utilized in job ads. Among organizations that employ CHWs in a greater variety of roles, less standardization in the roles, skills, and qualities was evident in job ads. Although the findings were not statistically significantly, hospitals/health systems job ads were associated with a higher number of the C3 qualities. Given that CHWs and CHW programs are being leveraged by health systems for these qualities and their ability to connect with community members outside clinical healthcare settings, this finding is not surprising. Adoption of a uniform framework for regulation that specifies CHW roles, skills, and qualities needed to function across various states, organizations, and practice types may improve recognition of the CHW workforce, reduce role confusion, and ensure that the unique skillset of CHWs is utilized consistently by employers, policy makers and the public.

We acknowledge some limitations of the study. The results from this study may not be generalizable, because the study sample was derived from purposive sampling of job ads from specific states. In addition, there are important differences in how states regulate CHWs. These differences may affect employer behavior and influence the adoption of occupational standards set by state and national CHW associations. Regardless, future studies on how differences in state-level regulation may affect the professionalization of CHW occupations and influence the adoption of skills, roles, and qualities utilized by employers are needed.

Unmet social needs are seen as contributing to poor health as much as lack of access to healthcare [42]. CHWs are an important service delivery resource for addressing the health of underserved communities and disparities in health, especially disparities that stem from unmet social needs. Adoption of a uniform framework for regulation that specifies CHW roles, skills, and qualities needed to function across various states and employer types may improve recognition of the workforce, reduce role confusion, and ensure that the unique skillset of CHWs are available to those with the greatest needs.