Effects of a Health Education Intervention for COVID-19 Prevention in Latinx Communities: A Cluster-Randomized Controlled Trial

Abstract

To promote COVID-19 preventive attitudes and behaviors among Latinx individuals, researchers and community partners implemented a culturally tailored health education intervention across 12 Oregon counties from February 2021 through April 2022. We did not identify any significant intervention effects on preventive attitudes and behaviors but did observe significant decreases in psychological distress. Although Latinx individuals’ preventive attitudes and behaviors were not associated with the health education intervention, findings suggest the intervention has value in promoting their well-being (ClinicalTrials.gov Identifier: NCT04793464). (Am J Public Health. 2022;112(S9):S923–S927. https://doi.org/10.2105/AJPH.2022.307129)

Latinx communities are more likely to face COVID-19 illness and death than are their non-Latinx White counterparts in the United States.¹ In Oregon, Latinx residents made up 14% of the state’s population but comprised 31.7% of COVID-19 cases in May 2020.² To address these health disparities, there was a need to increase access to preventive services among Latinx communities, such as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) testing and culturally tailored, evidence-informed COVID-19 health education in Latinx community networks.^3,4

INTERVENTION AND IMPLEMENTATION

We implemented the Promotores de Salud intervention across Oregon,⁵ which included two culturally and trauma-informed components with a focus on the Latinx community: (1) outreach, and (2) COVID-19 health education. The randomized controlled trial evaluation of the outreach component showed Promotores de Salud was effective at engaging more Latinx individuals in SARS-CoV-2 testing, with nearly four times as many Latinx individuals tested per event in the intervention condition than in outreach as usual (OAU).⁶

We evaluated the COVID-19 health education intervention component. The primary hypothesized outcome was greater endorsement of COVID-19 preventive attitudes and behaviors among those who received health education than among those who did not. For the secondary outcome, psychological well-being, we expected less psychological distress in attendees at intervention relative to OAU sites, given the trauma-informed and cultural tailoring of the health education intervention. We also examined whether intervention effects differed among Latinx versus non-Latinx individuals and as a function of Spanish language use at home.

PLACE, TIME, AND PERSONS

The Promotores de Salud intervention was implemented between February 2021 and April 2022 (data collection began March 2021) in 12 Oregon counties. Community members who were aged 15 years or older, had proficiency in English or Spanish, and visited one of the project’s 43 SARS-CoV-2 testing sites were eligible to complete baseline and follow-up surveys.

PURPOSE

The health education promoted COVID-19 preventive behaviors, knowledge, and receptive attitudes among Latinx individuals. The collaborative intervention development details appear elsewhere.⁵ Promotores, akin to community health workers,⁷ delivered the health education. Promotores (n = 21) were bicultural, bilingual (English and Spanish), and hired by Latinx-serving community-based organizations.

When community members arrived at a test site, promotores provided a five-minute overview of why and how to practice social distancing, mask wearing, hand washing, repeated testing, and (after April 2021) vaccination. The health education and print materials were available in Spanish and English. Messages focused on engaging in preventive behaviors to protect family and community, reflecting Latinx cultural values related to collective responsibility.⁸ Promotores were trained in trauma-informed practices and motivational interviewing to address community members’ questions and concerns, and they supported access to resources.⁸ OAU sites served as controls and did not have promotores.

EVALUATION AND ADVERSE EFFECTS

We collected data before respondents received the health education via an onsite survey offered in Spanish and English and then again approximately one to two months after respondents received the health education via an online follow-up survey (median [Quartile 1 (Q1), Quartile 3 (Q3)] = 47 [36, 188] days later). Measures assessing mitigating attitudes and behaviors were from Rapid Acceleration of Diagnostics-Underserved Populations (RADx-UP) common data elements and the PhenX Toolkit.^9,10 We evaluated six scores:

• 1.a six-item Likert mean scale of safety attitudes about preventive behaviors,
• 2.a five-item Likert mean scale of exposure attitudes about risky behaviors,
• 3.a vaccine attitude item,
• 4.a six-item sum index of risky health behaviors,
• 5.a 12-item sum score of preventive behaviors, and
• 6.a two-item scale score of psychological distress (PhenX Broadband depression and anxiety; see Table A for details [available as a supplement to the online version of this article at http://www.ajph.org]).

Two registered items about transmission knowledge did not exhibit adequate variance, so we omitted them.

At baseline, we collected 1443 surveys from 287 testing events (198 intervention; 87 OAU; 2 missing; median [Q1, Q3] = 3 [1, 6] surveys per event). Of these, 787 (54.5%) respondents were lost to follow-up and 44 (3.0%) had incomplete baseline data. The analysis sample included the remaining 608 respondents with follow-up data (390 intervention; 218 OAU; Table 1 provides descriptors and Figure A [available as a supplement to the online version of this article at http://www.ajph.org] depicts participant flow). Community members were permitted to participate in the survey no more than once at baseline and once at follow-up. The study was a clustered randomized trial. Randomization occurred at the event level (rather than at the individual level) and included collaboration with community partners and county health agencies to identify up to six optimized sites, which we then randomized within county to minimize threats to internal validity.

TABLE 1—

Description of SARS-CoV-2 Testing Participants in Promotores de Salud Intervention: Oregon, March 2021–April 2022

Variable	Overall (n = 680), Count (%) or Median (Q1, Q3)	Intervention (n = 390), Count (%) or Median (Q1, Q3)	OAU (n = 218), Count (%) or Median (Q1, Q3)	SMDa
Female	403 (66)	259 (66)	144 (66)	0.01
Age, y	40 (29, 50)	40 (30, 51)	39 (28, 49)	0.15
Race/ethnicityb
Latinx	390 (64)	248 (64)	142 (65)	0.03
American Indian/Alaska Native	90 (15)	49 (13)	41 (19)	0.17
Asian	18 (3.0)	9 (2.3)	9 (4.1)	0.10
Black/African American	13 (2.1)	7 (1.8)	6 (2.8)	0.06
Middle Eastern/North African	5 (0.8)	5 (1.3)	0 (0.0)	0.16
Native Hawaiian/other Pacific Islander	1 (0.2)	0 (0.0)	1 (0.5)	0.10
White	233 (38)	151 (39)	82 (38)	0.02
Spanish spoken at home	364 (60)	233 (60)	131 (60)	0.01
Employed	401 (66)	254 (66)	147 (67)	0.04
Essential worker	269 (44)	171 (44)	98 (45)	0.02
Educationc				0.15
< high school diploma	147 (24)	97 (25)	50 (23)
High school diploma or GED	126 (21)	72 (18)	54 (25)
≥ some college	335 (55)	221 (57)	114 (52)

Note. GED = general equivalency diploma; OAU = outreach as usual; Q1 = Quartile 1; Q3 = Quartile 3; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SMD = standardized mean difference.

^aThe SMD is a data type agnostic comparison between the intervention and OAU groups. Values given are mean differences in units of SDs. Education is presented as a single nominal variable and thus one SMD is presented.

^bRace/ethnicity were collected as a single question in which participants could endorse any number of options. All participants endorsed at least one option.

^cEducation was collected as an eight-point scale with choices ranging from “I have never gone to school” to “other advanced degree.”

Participants attended events that were randomly assigned as either intervention or control. Control sites received the intervention after a waitlist period.⁶ The total analysis sample was from 197 SARS-CoV-2 testing events at which 2405 COVID-19 tests were performed. Most respondents self-identified as Latinx (64.1%), female (66.3%), and non-White (61.7%) and spoke Spanish at home (59.9%). The median (Q1, Q3) age was 40 (29, 50) years.

Educational attainment varied (44.9% completed high school or less, and 55.1% completed some college or more). Almost half of the respondents were essential workers (44.2%). Loss to follow-up did not differ significantly by group (odds ratio [OR] = 0.80; 95% confidence interval [CI] = 0.56, 1.04; P = .07), although those in the intervention group had a lower rate of study dropout. Respondents who completed follow-up were more likely to identify as female, White, and not Latinx and were more likely to have private health insurance, use only English at home, and have US-born parents compared with those not retained.

For generalizability to the state of Oregon, we calculated propensity scores and inverse probability weights using social disparity indices from the RADx-UP common data elements¹⁰: the child opportunity index, the social vulnerability index, area health and research quality, pandemic vulnerability, and the pandemic vulnerability–vaccine model. Multilevel models adjusted for the nonindependence of participants clustered in randomized testing events by appropriately estimating SEs. Specifically, negative binomial generalized linear mixed models were specified for count data with fewer than seven categories and linear mixed models for continuous scale scores. We adjusted models for sex, race, ethnicity, age, education, essential worker status, time to follow-up, and the National Institutes of Health pandemic vulnerability index.¹¹

Results supported neither the hypothesized main effects nor the moderators (Latinx ethnicity and Spanish language use at home) of the intervention on COVID-19 preventive attitudes and behaviors (Table 2). The intervention was, however, associated with pre–post reductions in psychological distress (b = −0.14; 95% CI = −0.29, −0.02; P < .05), yielding a small effect size (d = 0.15; see Figure B [available as a supplement to the online version of this article at http://www.ajph.org]). Relative to the control condition, the promotores group was −0.15 lower in psychological distress after the intervention, controlling for baseline. Overall, at follow-up Latinx respondents in both conditions exhibited greater safety (b = 0.2; 95% CI = 0.01, 0.39; P < .05) and exposure attitudes (b = 0.33; 95% CI = 0.09, 0.57; P < .05). That is, compared with non-Latinx participants, Latinx respondents’ attitudes about safety and exposure were b = 0.2 and 0.33 higher after the intervention, controlling for baseline.

TABLE 2—

Sample Weighted Tests of Effectiveness Hypotheses for Health Behaviors, Attitudes, Psychological Distress, and Vaccine Acceptance Among SARS-CoV-2 Testing Participants: Promotores de Salud Intervention, Oregon, March 2021−April 2022

	Safety Attitudes (n = 509),a b (95% CI)	Exposure Attitudes (n = 539),a b (95% CI)	Psychological Distress (n = 591),a b (95% CI)	Preventive Behaviors (n = 504),a b (95% CI)	Risky Behaviors (n = 608),b IRR (95% CI)	Vaccine Attitude (n = 377),b IRR (95% CI)
Intercept	3.63 (3.34, 3.93)	3.19 (2.84, 3.53)	1.60 (1.31, 1.89)	3.60 (2.60, 4.70)	0.66 (0.438, 1.001)	1.02 (0.66, 1.58)
Intervention (ITT)	0.03 (−0.09, 0.14)	0.05 (−0.08, 0.19)	−0.14 (−0.26, −0.02)	−0.05 (−0.44, 0.34)	0.97 (0.82, 1.15)	1.04 (0.87, 1.25)
BL score	0.40 (0.32, 0.48)	0.28 (0.21, 0.36)	0.58 (0.51, 0.64)	0.33 (0.26, 0.41)	1.42 (1.34, 1.51)	1.36 (1.25, 1.47)
Time since BL	−0.003 (−0.007, 0.001)	0.001 (−0.004, 0.006)	−0.002 (−0.006, 0.002)	−0.03 (−0.04, −0.02)	1.008 (1.003, 1.014)	1.002 (0.996, 1.009)
Latinx	0.197 (0.002, 0.393)	0.33 (0.09, 0.57)	−0.05 (−0.25, 0.16)	0.73 (0.05, 1.40)	0.74 (0.57, 0.96)	0.94 (0.70, 1.25)
Female	0.16 (0.04, 0.28)	0.04 (−0.10, 0.19)	0.13 (0.01, 0.25)	0.09 (−0.31, 0.50)	0.92 (0.79, 1.09)	0.97 (0.81, 1.17)
White	0.09 (−0.10, 0.28)	0.07 (−0.16, 0.31)	0.11 (−0.09, 0.30)	−0.30 (−0.90, 0.40)	0.89 (0.70, 1.15)	0.97 (0.73, 1.29)
Age	0.002 (−0.002, 0.006)	−0.001 (−0.006, 0.003)	−0.005 (−0.010, −0.001)	0.01 (−0.01, 0.02)	0.997 (0.992, 1.003)	1.0003 (0.99, 1.007)
Education	0.03 (−0.01, 0.07)	0.04 (−0.01, 0.08)	0.03 (−0.01, 0.07)	−0.06 (−0.19, 0.06)	1.06 (1.002, 1.120)	0.97 (0.92, 1.03)
Essential worker	0.05 (−0.06, 0.16)	−0.132 (−0.268, 0.004)	0.11 (−0.01, 0.22)	−0.17 (−0.55, 0.21)	1.12 (0.96, 1.31)	1.02 (0.85, 1.21)
County PVI	−0.50 (−2.30, 1.20)	−0.60 (−2.60, 1.40)	2.10 (0.20, 4.00)	−5.40 (−11.10, 0.20)	7.40 (0.40, 141.70)	0.6 (0.2, 2.1)
Latinxc	0.28 (0.04, 0.53)	0.21 (−0.09, 0.52)	−0.02 (−0.27, 0.23)	0.70 (−0.20, 1.50)	0.77 (0.55, 1.07)	0.89 (0.63, 1.27)
Latinx × ITTc	−0.15 (−0.40, 0.11)	0.20 (−0.11, 0.50)	−0.04 (−0.29, 0.21)	0.10 (−0.80, 1.00)	0.95 (0.68, 1.34)	1.09 (0.74, 1.59)
Spanishd	−0.01 (−0.28, 0.26)	0.13 (−0.20, 0.45)	−0.31 (−0.59, −0.03)	0.70 (−0.20, 1.70)	1.18 (0.79, 1.76)	1.12 (0.73, 1.72)
Spanish × ITTd	−0.06 (−0.31, 0.18)	0.13 (−0.16, 0.43)	0.05 (−0.19, 0.29)	0.20 (−0.60, 1.00)	0.83 (0.59, 1.16)	1.13 (0.77, 1.64)

Note. BL = preintervention baseline; IRR = incident rate ratio; ITT = intent to treat; PVI = pandemic vulnerability index; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2. Multilevel models are generalized linear negative binomial for count and categorical data and multilevel linear models for scale score data. Multilevel models adjust for the nonindependence of participants clustered in randomized testing events by appropriately estimating SEs.

^aOutcome modeled using a normal distribution.

^bOutcome modeled using a negative binomial distribution.

^cModerator analysis conducted separately from main effects.

^dModerator analysis conducted separately from main effects; Latinx removed from model because of multicollinearity with Spanish language.

Latinx respondents also reported less engagement in risky behaviors (incident rate ratio = 0.74; 95% CI = 0.57, 0.96; P < .05) and more engagement in preventive behaviors (b = 0.73; 95% CI = 0.05, 1.40; P < .05) than did non-Latinx respondents. Latinx participants had b = 0.73 higher engagement than did non-Latinx respondents in COVID-19 prevention behaviors, controlling for baseline. Limitations of the study were related to the urgency of implementation despite the nascent literature base at the start of the COVID-19 pandemic. Stronger outcome measures are needed, as are more effective follow-up procedures to mitigate demographic shifts from baseline. There were no adverse or other unintended consequences of the intervention.

SUSTAINABILITY

Several factors indicate favorable sustainability, including the flexibility of the health education in responding to evolving COVID-19 prevention understanding, the collaborative intervention development and implementation with Latinx community-based organizations, and the state public health department’s adoption of project test sites. A formal evaluation of sustainability is under way.

PUBLIC HEALTH SIGNIFICANCE

Psychological well-being is associated with reduced risk of COVID-19 hospitalization.¹² Thus, the finding that our promotores-delivered intervention decreases psychological distress is notable. Promotores may lessen psychological distress by improving access to reliable information, promoting confidence in navigating challenges, and affirming engagement in preventive behaviors. However, given the relatively small effect size for a secondary outcome, the intervention must be developed further and coupled with other approaches to minimize barriers and maximize health efficacy. For the Latinx community, specific improvements in COVID-19 preventive attitudes and behaviors were not associated with the intervention. Findings across studies suggest that promotores are ideal for improving access to SARS-CoV-2 testing and outreach provided in Spanish by bicultural staff.⁶