Standard Occupational Classification Codes: Gaps in Federal Data on the Public Health Workforce

Abstract

Objectives. To determine whether US Department of Labor standard occupational classification (SOC) codes can be used for public health workforce research.

Methods. We reviewed past attempts at SOC matching for public health occupations and then used the National Institute for Occupational Safety and Health Industry and Occupation Computerized Coding System (NIOCCS) to match the actual job titles for 26 516 respondents to the 2021 Public Health Workforce Interests and Needs Survey (PH WINS) with SOC codes, grouped by respondents’ choice of job category in PH WINS. We assessed the accuracy of the NIOCCS matches and excluded matches under a cutpoint using the Youden Index. We assessed the percentage of SOC matches with insufficient information and diversity of SOC matches per PH WINS category using the Herfindahl–Hirschman Index.

Results. Several key public health occupations do not have a SOC code, including disease intervention specialist, public health nurse, policy analyst, program manager, grants or contracts specialist, and peer counselor.

Conclusions. Without valid SOC matches and detailed data on local and state government health departments, the US Department of Labor’s data cannot be used for public health workforce enumeration. (Am J Public Health. 2024;114(1):48–56. https://doi.org/10.2105/AJPH.2023.307463)

The local, state, and Tribal government public health workforce is the backbone of the US public health infrastructure, providing services crucial to health, safety, and national biosecurity. The workforce serves as frontline responders to public health emergencies, ensures clean water and air, controls infectious disease outbreaks, prevents chronic disease and injury, and maintains vital statistics such as birth and death certificates.¹ Building and maintaining this workforce is 1 of the 10 essential public health services² and is essential for providing foundational public health services.¹ To determine whether the workforce can provide needed services, we must understand how many individuals are in this workforce, what they do, how much they get paid, and what training they have. Without this basic information, researchers, policymakers, and the public cannot determine whether there are workforce or funding shortages.

An essential aspect of workforce research is an occupational taxonomy that categorizes the unique job functions of different workers. The US Department of Labor Bureau of Labor Statistics (BLS) is charged with collecting, analyzing, and publishing data on the US workforce. It tracks 867 detailed occupations using the standard occupational classification (SOC) system categories.³ Examples of SOCs include such specific positions as denturists, shampooers, and braille transcribers, as well as subspecialists such as nurse anesthetists and advanced practice psychiatric nurses. BLS uses surveys such as the Occupational Requirements Survey, the National Compensation Survey, Employment Projections, and the Occupational Employment and Wage Statistics (OEWS) program to collect data on working conditions, education, training, and experience as well as salaries and benefits.

Another key classification used in workforce research is the North American Industry Classification System (NAICS), used to classify business establishments by industry or sector.⁴ Using SOC codes in conjunction with NAICS codes allows detailed enumeration, categorization, and salary information for workers in a specific industry. Like SOC codes, NAICS codes have broader categories (e.g., category 62, “Health Care and Social Assistance”) with detailed subcategories (“621310 Offices of Chiropractors”), allowing us to know, for example, how many accountants work in chiropractors’ offices.⁵ However, although the NAICS code, “923120 Administration of Public Health Programs,” exists, the OEWS uses only broader NAICS codes, “Local Government, excluding schools and hospitals” and “State Government, excluding schools and hospitals,” perhaps because BLS does not prioritize gathering detailed employment data on government agencies.

The 1997 report “The Public Health Workforce: An Agenda for the 21st Century” (Appendix C)⁶ recommended the development of a standard taxonomy to characterize the public health workforce, specifically one linked to SOC codes. However, it has been noted repeatedly since then^7–13 that SOC codes do not match well with several public health occupations, creating challenges in researching the workforce. “The Public Health Work Force: Enumeration 2000 Report” (Appendix C, part 3) reiterated the need for public health SOCs,¹⁴ but many of the recommended SOC codes were never established. A 2023 report by the President’s Council of Advisors on Science and Technology again recommended new SOCs for public health and recommended that they “include the public health workforce in … the next revision of the SOC manual.”^15(p23)

Public health workforce researchers, in an attempt to better categorize the workforce, established a public health workforce occupational taxonomy, imperfectly connected to SOC codes, in the “Enumeration 2000 Report.”¹⁶ A workforce taxonomy was revised in 2014^17,18 and then adapted for research such as the Public Health Workforce Interests and Needs Survey (PH WINS),¹⁹ but matching with SOC codes was not an explicit aspect of the 2014 taxonomy or PH WINS.

PH WINS is the largest national source of data on individual state and local governmental public health workers.^20,21 Respondents select their job function from a list of 70 job categories. Matching these categories with SOC codes could improve workforce research, allow salary benchmarking, and contribute to workforce enumeration. We sought to determine which public health job titles clearly match to a SOC code and where there are gaps between SOC codes and public health occupations.

METHODS

Our fundamental approach to answering our project’s research questions was to use a federally supported machine learning system that autocodes industry and occupation data into SOC code standards. This allowed us to (1) compare administrative and civil service data on job titles from PH WINS fielding, which was not previously available to researchers; and (2) examine where SOC codes do not capture the occupations of public health workers.

We conducted a literature review on past schema-matching efforts. The National Institute for Occupational Safety and Health’s Industry and Occupation Computerized Coding System (NIOCCS)²² matched the job titles with SOC codes, giving a probability of the match being correct, with some titles coded as “insufficient information.” We manually assessed the validity of the NIOCCS matches for the 200 most common civil service job titles, coded the matches as correct or incorrect using a receiver operating characteristic (ROC), and determined a cutpoint using the Youden Index, a statistic that conveys the performance of dichotomous diagnostic tests. We then excluded SOC matches with a predicted SOC probability below the threshold. Finally, we analyzed each PH WINS job category to assess which had the largest proportion of “insufficient information” instead of clear matches, which had lower SOC-matching probability and had the most diversity in SOC matches using the Herfindahl–Hirschman Index (HHI)—a metric that expresses how evenly dispersed a group is throughout a selection of states—to determine which PH WINS categories had the least clear SOC matches.

Past Schema-Matching Efforts

A 2021 study²³ attempted to match PH WINS titles with SOC codes. First, the study’s authors used the SOC codes from the Enumeration 2000 public health workforce taxonomy¹⁴ and connected them to the 2014 taxonomy¹⁷; second, they used machine-learning software developed by the National Institutes of Health to match the job descriptions in the 2014 taxonomy with SOC codes²⁴; and, finally, they validated the matches with a coauthor of the 2014 taxonomy. Additionally, the authors analyzed a data set of 38 533 full job descriptions from the Burning Glass database, in both government and nongovernment organizations, for which employers sought public health degree candidates. The Burning Glass system used machine learning to match these job descriptions with SOC codes, allowing an analysis of which organizations compete to hire public health graduates, as well as which occupations in the public health workforce require public health degrees. We used the taxonomy matching from the 2021 as a framework to begin the SOC to PH WINS matching process in our study.

Code Matching With NIOCCS

We accessed administrative data used to field 2021 PH WINS, specifically the survey recipients’ actual job titles (referred to in this article as “civil service job titles”), and linked these job titles with respondents’ choice of job category from the list of 70 categories in the PH WINS instrument. There were 108 044 records from the PH WINS 2021 distribution. Of these, 65 115 were nonresponders to PH WINS, so we excluded them; 16 413 respondents to PH WINS did not have a job title (or the title was a string of numbers with no text), so we excluded them as well. We analyzed the remaining 26 516 PH WINS respondents who also had a civil service job title. We then cleaned the job title data so that it could be coded by NIOCCS, which matches job titles with the 2018 (latest) SOC codes using machine learning.²² Data cleaning examples are provided in Appendix A (available as a supplement to the online version of this article at http://www.ajph.org).

ROC Curve and Cutoff Value

Although NIOCCS provides a match probability, we wanted to establish an independent threshold for confidence in the system’s ability to assign occupation codes. To find the most valid SOC code matches for each PH WINS job category, we reviewed the SOC matches of the NIOCCS for the 200 most common civil service titles (representing 13 257 of 26 497 respondents).

The lead researcher (H. K.) identified whether NIOCCS had matched the job titles to a correct SOC code and made observations of patterns in the code matching (Appendix B, available as a supplement to the online version of this article at http://www.ajph.org). When it was unclear whether NIOCCS had matched to the correct SOC code, we looked up the civil service title using O*NetOnline, which draws from SOC codes but also contains multiple alternative job titles per occupation.²⁵ In some cases, we also searched for the civil service job titles in job boards, including Indeed.com, to review currently posted job descriptions.

We looked at performance using both a ROC curve²⁶ and a Youden Index.²⁷

ROC classes:

$$\text{y}=1 \text{if} \text{NIOCCS} \text{probability}\ge 0.5$$ (1)

$$\text{y}=0 \text{if} \text{NIOCCS} \text{probability}<0.5$$ (2)

The ROC curve is a probability curve that plots the true positive rate versus false positive rate at various cutpoint values.²⁸

The ideal cutpoint for the Youden Index is the one that maximizes the Youden function, which represents the difference between the true positive rate and the false positive rate among all plausible cutpoint values. The formula for the index is shown as

$$J_{\max }={\max }_t\{\text{sensitivity}(t)+\text{specificity}(t)-1\},$$ (3)

where t denotes the classification cutpoint for which J is maximal, with values for J ranging between 0 and 1.²⁸

We used the cutpoint function in R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria) to fit an ROC curve on the NIOCCS coder– predicted SOC probabilities and our determined actual value (1 = predicted correctly, 0 = predicted improperly). The coder had a fair predictive ability with an area under the curve of 0.8072. The optimal cutpoint using the Youden Index was 0.4811. We used the Youden Index to determine the cutpoint, as it conceptually allowed us to balance sensitivity and specificity. We excluded all rows with a NIOCCS-predicted SOC probability less than this value in the subsequent analysis (Appendix C, available as a supplement to the online version of this article at http://www.ajph.org).

Analysis

We analyzed the SOC codes for respondents, grouped by the PH WINS job category they selected, to determine which titles had the largest proportion of insufficient information instead of a SOC code match, the lowest SOC probability, and an examined distribution of SOC codes per PH WINS category using the HHI. We also examined which, if any, had insufficient SOC matches. We conducted these analyses using SAS on Demand for Academics, 2023 (SAS Institute, Cary, NC).

Herfindahl–Hirschman Index

We used the HHI²⁹ in this analysis to assess the diversity of SOC codes per PH WINS category. In Equation 4, s_i is the proportion of SOC title guesses for each PH WINS title i. A higher HHI implies that the SOC occupation “guesses” are less diverse, and a lower HHI implies that the SOC occupation guesses are more diverse.

$$H={\displaystyle \sum _{i=1}^N{s}_i^2}$$ (4)

For each PH WINS title, we calculated s_i by dividing the count of each SOC title guess by the sum of all SOC titles guesses; we then squared and summed up s_i to get the final HHI.

RESULTS

Our analysis allowed us to identify which occupations within health departments were well matched to SOCs, and which were poorly matched.

Occupations With No or Poorly Matched SOCs

Business support roles and those with clinical titles were well matched by NIOCCS to an accurate SOC code. Epidemiologist matched well, as did health educator, nutritionist or dietitian, pharmacist, dentist, attorney, and economist.

However, certain critical public health occupations had insufficient information as the most frequent SOC response or had SOC matches that are clearly incorrect, indicating that they have no SOC code. Others matched a SOC code, but the description of the SOC occupation did not accurately describe the job tasks for public health professionals in that occupation. In particular, we found the following:

• • Disease intervention specialist had insufficient information as the most frequent match result. Titles including “disease intervention specialist” and “communicable disease investigator” matched “epidemiologist.” “Contact tracer” matched “community health worker.” “Case investigator” matched “social workers, all other.” “Disease investigator” or “public health investigator” matched “private investigator” (although these matches had low certainty so we excluded them after ROC analysis). We excluded 28% of PH WINS responders who selected disease intervention specialist as their classification under the ROC cutpoint, and it had a low HHI of 0.146. This lack of SOC code is consistent with the recent 2021 taxonomy matching article, which found “private investigators” as a common SOC code for contact tracers or disease intervention specialist.²³

• • Policy analyst matched operations research analysts, software developers, “managers, all other,” or management analysts, whereas the correct match using O*NetOnline is political scientist. We excluded 26.3% of PH WINS respondents in this classification, and the role had a low HHI of 0.13.

• • Public health agency director (HHI 0.29), department/bureau director (HHI 0.22), deputy director (HHI 0.21), health officer (0.14), program director (0.2), and program manager (0.25) all match the very broad SOC title “managers, all other” or “general and operations managers,” with some matching “medical and health services managers” or “social and community services managers.”

• • Public health nurse and other nurse titles matched well with registered nurse and had high HHIs and low percentages of insufficient data, but the job duties of registered nurses are different from those of public health nurses. Public health nurses have their own certification in certain states, their own professional association, their own competency instrument, and unique job functions.³⁰

• • Grants or contracts specialist matched “managers, all other” or “office and administrative support workers, all other.” No other SOC clearly aligns with the role of a grants or contracts specialist in a government agency. The title had a very low HHI of 0.075, and we excluded 31% of titles under the ROC curve.

• • Peer counselor matched “counselors, all other,” an incorrect match because counselors require credentialling and peer counselors are typically paraprofessionals.

• • Program evaluator had insufficient information as the highest match followed by “managers, all other” and had a low HHI of 0.086.

• • Public health informatics specialist had an HHI of 0.09 and insufficient information as the top match, and information systems manager had an HHI of 0.09.

• • Vital records staff had an HHI of 0.07 and insufficient information as the top match.

• • Inspector/compliance roles: licensure/regulation/enforcement worker, quality improvement, sanitarian or inspector, and nurse appear to have some of the same civil service job titles relating to compliance. Licensure/regulation/enforcement matched registered nurses, perhaps because health facilities inspectors are often nurses.

• • Quality improvement worker had a low HHI of 0.085 and did not match a SOC clearly.

Because the PH WINS study did not have a goal of matching with specific SOC codes, certain job categories in PH WINS were too vague to match with particular SOC codes. Examples include “other business support services,” “other (please specify),” “other program staff.” Similarly, certain PH WINS job classifications did not match SOCs clearly because they encompassed diverse occupations, such as “behavioral health professional” (HHI of 0.1, matching social workers, medical and health services managers, substance abuse counselors etc.); “data or research analyst” (HHI of 0.07, with insufficient information as the highest match). “Implementation specialist” (47% insufficient information), and “population specialist” did not have consistent civil service job titles, and the civil service titles rarely included these terms.

We created a table (Appendix D, available as a supplement to the online version of this article at http://www.ajph.org) illustrating the most common civil service titles, and SOC code matches, by PH WINS title. To our knowledge this cross-walking between the PH WINS titles (and the previous public health workforce taxonomy) is the first evidence-based taxonomy mapping between public health titles and SOC codes since 1997. When the SOC matches are inadequate, this is noted in Table 1.

TABLE 1—

PH WINS Job Categories With Percentages Excluded, Percentages of Insufficient Information, and HHI: United States, 2021

PH WINS Job Category (2021)	Total Respondents in PH WINS 2021 Data Subseta	% Excluded After Youden Removedb	% Insufficient Information SOC Matchc	HHId	NIOCCS-Reported SOC Match Probability,e Median (IQR)	Total Unique SOC Matchesf
Animal control worker	18	55.6	25.0	0.25	0.90 (0.72–0.99)	5
Attorney or legal counsel	110	12.7	2.1	0.43	0.61 (0.94–0.99)	13
Behavioral health professional	230	29.6	8.6	0.10	0.76 (0.96–0.97)	32
Business support–accountant/fiscal	822	14.6	8.0	0.14	0.74 (0.96–1.00)	36
Business support services–administrator	385	18.7	12.5	0.09	0.68 (0.87–0.97)	41
Business support services–coordinator	232	19.0	11.7	0.13	0.73 (0.88–0.96)	33
Clerical personnel–administrative assistant	1650	11.8	7.4	0.17	0.67 (0.82–0.96)	62
Clerical personnel–secretary	403	12.9	7.4	0.18	0.74 (0.97–1.00)	37
Community health worker	574	17.9	8.1	0.14	0.64 (0.80–0.97)	52
Custodian	22	0.0	13.7	0.33	0.83 (0.90–1.00)	8
Customer service/support professional	597	13.9	10.5	0.18	0.83 (0.98–1.00)	63
Data or research analyst	848	21.5	14.9	0.07	0.71 (0.81–0.95)	59
Department/bureau director	414	13.8	6.2	0.22	0.64 (0.80–0.93)	31
Deputy director	170	13.5	11.6	0.21	0.72 (0.80–0.94)	16
Disability claims/benefits examiner or adjudicator	25	20.0	0.0	0.17	0.55 (0.97–1.00)	10
Disease intervention specialist/ contact tracer	806	28.0	25.5	0.15	0.60 (0.83–0.97)	38
Economist	5	20.0	0.0	0.63	0.94 (0.94–0.97)	2
Emergency medical services worker	18	16.7	40.0	0.26	0.60 (0.83–0.95)	6
Emergency medical technician/ advanced emergency medical technician/paramedic	22	9.1	35.0	0.18	0.77 (0.95–0.99)	10
Emergency preparedness/management worker	410	22.4	13.5	0.09	0.76 (0.88–0.97)	40
Engineer	119	20.2	11.6	0.15	0.77 (0.96–0.98)	21
Environmental health worker	1056	12.6	1137.6	0.21	0.70 (0.78–0.88)	48
Epidemiologist	1350	9.1	9.5	0.44	0.86 (1.00–1.00)	38
Grants or contracts specialist	382	31.2	12.2	0.08	0.73 (0.85–0.96)	33
Health educator	911	17.1	12.2	0.27	0.73 (0.95–0.97)	48
Health navigator	55	12.7	10.4	0.09	0.64 (0.94–0.99)	17
Health officer	158	26.6	9.5	0.14	0.69 (0.78–0.91)	26
Human resources personnel	235	11.1	6.7	0.10	0.80 (0.91–0.98)	29
Implementation specialist	56	35.7	47.2	0.27	0.61 (0.82–0.94)	11
Information systems manager/ information technology specialist	468	14.1	12.2	0.09	0.75 (0.94–0.99)	42
Laboratory aide or assistant	52	36.5	39.4	0.20	0.65 (0.96–0.98)	13
Laboratory quality control worker	31	3.2	13.3	0.12	0.77 (0.93–0.99)	14
Laboratory scientist/medical technologist	712	12.6	10.0	0.15	0.64 (0.85–0.99)	22
Laboratory technician	188	18.6	9.2	0.18	0.56 (0.68–0.97)	17
Licensed practical or vocational nurse	91	4.4	6.9	0.59	0.99 (1.00–1.00)	8
Licensure/regulation/enforcement worker	629	29.3	15.5	0.08	0.65 (0.86–0.97)	54
Medical examiner	42	4.8	0.0	0.14	0.61 (0.94–1.00)	13
Medical/vital records staff	194	23.7	16.5	0.07	0.67 (0.89–0.98)	35
Nurse practitioner	111	9.0	4.0	0.49	0.93 (0.99–0.99)	10
Nursing and home health aide	120	5.0	3.5	0.51	0.92 (0.99–1.00)	13
Nutritionist or dietitian	633	9.5	5.2	0.41	0.75 (0.98–1.00)	24
Other (please specify)	573	24.4	15.5	0.05	0.71 (0.88–0.97)	86
Other business support services	333	17.7	17.9	0.06	0.72 (0.92–0.98)	59
Other facilities or operations worker	138	19.6	9.0	0.05	0.68 (0.84–0.96)	40
Other health professional/clinical support staff	506	20.8	16.2	0.05	0.68 (0.93–0.99)	64
Other Nurse–clinical services	137	5.8	2.3	0.44	0.86 (0.99–0.99)	14
Other oral health professional	84	14.3	15.3	0.20	0.85 (0.99–1.00)	14
Other program staff	1296	31.8	20.4	0.09	0.66 (0.82–0.95)	69
Peer counselor	85	22.4	4.5	0.40	0.94 (0.97–0.97)	14
Pharmacist	47	19.2	7.9	0.53	0.97 (1.00–1.00)	6
Physical/occupational/rehabilitation therapist	147	15.7	34.7	0.25	0.89 (0.97–0.98)	11
Physician assistant	9	11.1	12.5	0.31	0.78 (0.81–0.99)	5
Policy analyst	224	26.3	7.3	0.13	0.70 (0.90–0.97)	25
Population health specialist	99	44.4	20.0	0.14	0.57 (0.80–0.91)	17
Program director	589	16.5	6.9	0.20	0.70 (0.80–0.94)	40
Program evaluator	144	23.6	18.2	0.09	0.71 (0.83–0.94)	29
Public health agency director	90	8.9	3.7	0.29	0.72 (0.80–0.92)	11
Public health dentist	19	10.5	5.9	0.46	0.72 (0.85–0.85)	5
Public health informatics specialist	189	23.3	23.5	0.09	0.68 (0.91–0.97)	37
Public health manager or program manager	1611	18.7	8.4	0.25	0.67 (0.80–0.94)	65
Public health veterinarian	18	11.1	0.0	0.38	0.79 (0.80–0.82)	5
Public health/preventive medicine physician	100	18.0	14.6	0.13	0.58 (0.74–0.95)	11
Public information specialist	217	26.8	16.9	0.12	0.65 (0.87–0.95)	40
Quality improvement worker	179	23.5	13.1	0.08	0.67 (0.82–0.96)	30
Registered nurse–public health or community health nurse	2086	4.8	5.9	0.70	0.97 (0.99–1.00)	31
Registered nurse–unspecified	519	9.6	7.3	0.52	0.82 (0.98–1.00)	18
Sanitarian or inspector	736	10.3	7.0	0.43	0.77 (0.87–0.97)	37
Social worker/social services professional	671	18.6	19.4	0.18	0.78 (0.91–0.95)	34
Statistician	44	13.6	5.3	0.19	0.60 (0.78–0.97)	11
Student, professional or scientific	51	23.5	53.9	0.31	0.77 (0.85–0.92)	14

Note. HHI = Herfindahl-Hirschman Index; IQR = interquartile range; NIOCCS = NIOSH’s Industry and Occupation Computerized Coding System; NIOSH = National Institute for Occupational Safety and Health; PH WINS = Public Health Workforce Interests and Needs Survey; SOC = standard occupational classification.

^a Total number of PH WINS respondents who selected this job category and who also had a civil service job title.

^b % responses after excluding inaccurate matches via receiver operating characteristic matching.

^c % of civil service job titles that the NIOCCS could not match to any SOC code.

^d A lower number indicates more diverse SOC matches.

^e The NIOCCS-reported probability of a correct match between titles in this PH WINS category and a SOC category. This includes median, first quartile, and third quartile.

^f The total number of unique SOC codes that matched to the job titles in each PH WINS category.

Insufficient Information

To show which PH WINS job categories had a higher percentage of insufficient information SOC titles (i.e., could not match) as well as the diversity of SOC matches per category, we created Table 1. This table contains the PH WINS job category, total number respondents per job category in the PH WINS 2021 data subset used in this research, the percentage of responses excluded below the cutpoint, the percentage of titles resulting in an insufficient information SOC match, and the HHI. It also lists the median and interquartile range for NIOCCS-reported SOC match probability and the total number of unique SOC matches per PH WINS category.

The positions with the largest percentage of insufficient information rather than an estimated SOC code include student, implementation specialist, emergency medical services worker, laboratory aide or assistant, emergency medical technician/advanced emergency medical technician/paramedic, physical/occupational/rehabilitation therapist, and disease intervention specialist/contact tracer.

DISCUSSION

BLS maintains SOC coding guidelines, which state, “Each occupation is assigned to only one occupational category at the most detailed level of the classification” and “Occupations are classified based on work performed and, in some cases, on the skills, education and/or training needed to perform the work.”^3(p24) From our analyses, it is clear that although certain public health occupations have a SOC match, many important public health occupations do not.

Several public health occupations, such as disease intervention specialist and public health nurse, have job task analyses,^30,31 professional associations, and even certifications, attributes that denote a clearly defined occupation. Many of these roles, such as public health nurse, “public health program specialist,” and “infection control/disease investigator” were requested as new SOCs in the Enumeration 2000 study.^6,14 Certain roles without clear SOCs, such as policy analyst and grant manager, are common across government agencies and nonprofit organizations; creating a clear SOC for them could improve workforce research beyond health departments. In the 2018 update to the SOCs, several new occupations or revisions for occupations in the public health workforce were requested,³² but most were denied. SOC codes are typically updated every 10 years. Public health researchers, collaborating with relevant professional associations, must take the next opportunity to document how their occupations differ from existing SOC occupations and present a case for new codes to the interagency SOC Policy Committee when it convenes.

An important, related challenge to public health workforce research is the lack of specificity in NAICS codes. The OEWS survey provides occupational and salary information by NAICS industry for many detailed industries. Even though an “administration of public health programs” NAICS code exists,³³ it is not used in OEWS; only “local government, excluding schools and hospitals” and “state government, excluding schools and hospitals” are used.

This is especially problematic for public health leadership and program management roles, most of which match “managers, all other,” a SOC too vague to capture the specialist nature of these roles. Without the use of the public health NAICS code, OEWS cannot currently be used for enumeration of the public health workforce, especially for vague SOC occupations (e.g., “managers, all other”) or occupations that could exist in many local or state government departments, such as social workers or nutritionists.

Limitations

The PH WINS survey asked respondents to select job categories, but the survey instrument did not include descriptions of each category, leaving the survey respondents to self-identify without detailed guidance. There is also always a chance of survey responder error.

Some mismatches are attributable to generic civil service job titles used in government, such as “research scientist,” which mismatched with the SOC for “physical scientist,” and “health services consultant,” which mismatched with “health technologists and technicians, all other.” Some mismatches reflect the challenge of cleaning the job titles and removing abbreviations. Certain job titles have the name of the program in the title (e.g., tuberculosis epidemiologist or maternal/child health program manager), which may make NIOCCS matching more challenging. We had access only to job titles, not full job descriptions; full descriptions might have matched more accurately.

Occupations with fewer responses may have had skewed high HHI’s because of the lower number of responses.

Public Health Implications

Because of a lack of accurate, public health–specific SOC codes and lack of the use of NAICS codes specific to health departments, the US Department of Labor does not provide data that can be used to enumerate the entire public health workforce. As a result, the crucial task of public health workforce enumeration has been left primarily to professional associations or philanthropically funded efforts, and a full workforce enumeration has not been conducted since 2014.³⁴ If BLS used appropriately detailed NAICS codes for OEWS, the federal government would enumerate the public health workforce at least every 3 years.

Future research could use the SOC codes we identified within the constraints of the excessive breadth of NAICS codes to contrast salaries of public health occupations in government versus competitor organizations or contrast education or experience levels between workers in government compared with other sectors.

Without clear SOC codes, the public health workforce—the backbone of our nation’s biodefense—will lack data on workforce shortages and recruitment challenges, and our nation—and our nation’s public health systems—will be less healthy as a result.