1.1. Background

Migrant and seasonal farmworkers (MSFW) face enormous structural barriers to health, including poverty, food insecurity, poor working conditions, high occupational hazard, and limited access to healthcare [1]. Although data are lacking, those available suggest that MSFW experience inequities in multiple health outcomes [2, 3, 4]. At work they face challenging environmental conditions including heat, sun, pesticide exposure, and dust and particle inhalation [1]. Additionally, most MSFW are uninsured and must overcome language barriers, long work hours, and lack of transportation to attend health appointments, with care often only available at Federally Qualified Health Centers (FQHC) [5]. Consequently, the majority of MSFW do not have primary care physicians [6].

The COVID-19 pandemic has only deepened pre-existing health inequities. Data demonstrate disproportionate rates of COVID-19 morbidity and mortality among Black, Latinx and immigrant populations, who make up the majority of MSFW in the U.S [7, 8, 9]. MSFW are at especially high risk for COVID-19 exposure and subsequent infection due to high density housing and ‘essential worker’ status, which has exempted them from the protection of working from home enjoyed by other industries [10, 11]. Typical working conditions are not conducive to distancing, as many MSFW work side-by-side and share transportation to and from work [12]. Once infected, spread within MSFW communities is facilitated by lack of testing and contact tracing, and by housing conditions which preclude distancing or effective quarantine [13]. Finally, those infected are at higher risk of severe COVID-19 disease due to underlying health conditions that predispose them to morbidity and mortality.

Florida houses one of the largest populations of MSFW in the country, and though the plight of MSFW during the COVID-19 pandemic has gained media attention, there is minimal data around the impact of the pandemic on MSFW communities. Specifically, though it is known that MSFW are a high-risk group, there is no data on the true death rates from COVID-19 for MSFW and there is little information regarding specific workplace conditions, sources of COVID-19 information, ability to follow guidelines, nor their experiences with testing and contact tracing. All of this information is key for improving public health programming.

1.2. Study Site: Immokalee, Florida

Immokalee is a rural community located in southwest Florida’s Collier County, the center of the nation’s tomato growing industry. The total population fluctuates with the agricultural seasons and is estimated at about 25,000, with 37.4% living below the poverty line [14]. Census data suggests 22% of the population works in agriculture and related fields, though other sources estimate as many as 15,000–20,000 MSFWs live in Immokalee for some parts of the year [14, 15, 16]. These discrepancies likely exist as MSFWs often live in temporary or mobile housing, which are underrepresented in census data [1]. Residents are primarily from Mexico, Central America, and Haiti; 68% of the population speak Spanish at home and 13% speak Haitian Creole, along with some bilingual or monolingual speakers of indigenous languages [6, 17]. Previous studies in Immokalee have noted that residents face increased everyday stressors and workplace risks that can negatively impact health [18, 19].

Occupational, economic, and linguistic factors confer substantial vulnerability to COVID-19 infection in Immokalee; despite its small size, Immokalee was the Florida zip code with the highest number of COVID-19 cases in the state in June 2020 [20]. Positivity rates at the time were as high as 36% compared to 5.6% in the state [21]. After observing the disease burden and difficulties in accessing COVID-19 testing and contract tracing among MSFWs, the Coalition of Immokalee Workers (CIW)—a local human rights organization—facilitated partnerships between the Collier County Department of Health (DOH), Partners in Health (PIH), Doctors Without Borders (Medicins Sans Frontieres (MSF)), and a local FQHC (Collier Health Services, Inc., d/b/a Healthcare Network (HCN)). The DOH led all contract tracing efforts in Immokalee, and offered nasal PCR testing by appointment until November 2020, with results given via an English-language online portal which asked for a social security number. Starting in November 2020, HCN began offering walk-up rapid testing in Immokalee, with results given within an hour by Spanish- and Haitian Creole-speaking staff. Both HCN and DOH created community health worker programs for COVID-19 outreach, but all positive results from HCN’s testing were forwarded to the DOH’s contact tracing system (see timeline in Figure 1 below).

In this study, we first aimed to estimate the impact of COVID-19 on Immokalee, FL by calculating excess mortality from publicly available data. Then, we conducted a household survey to collect sociodemographic information and assess community experiences with workplace conditions, access to testing, sources of information, and the DOH contact tracing program. Given the linguistic diversity in Immokalee and MSFW populations in general, and limited English proficiency as a known barrier to care, we evaluated for differences in experience by language.

2.1. Descriptive Analysis

We collated data from the Florida Department of Health and Medical Examiners Offices for Collier, Lee, Hendry, Glades and Orange Counties. First, we tabulated the total deaths from COVID-19 for residents of Collier County. Data for the Immokalee zip code alone were not available. We then compared deaths for Collier County during May-August 2015–2019 with deaths from May-August 2020. Excess mortality was calculated in accordance with CDC guidelines, where death data were grouped weekly to account for temporal effects [22]. Excess mortality was calculated for each week and then summed together to find excess mortality for the period beginning 4/27/20 and ending 8/16/20. Data were disaggregated by age (under 60 or ≥ 60 years) and sex.

2.2. Questionnaire

The questionnaire was created using demographic and social screening tools from the National Agricultural Workers Survey and various farmworker healthcare organizations [7, 8, 9]. Spanish- and Haitian Creole-speaking CIW staff reviewed the tool for understandability by the local population. The final questionnaire assessed the following thematic areas [dataset 23]:

• Demographic and socioeconomic information
• Sources of news and information on COVID-19
• Ability to follow COVID-19 precautions
• Experiences with contact tracing

2.3. Sample Size

Our target population was adults living in Immokalee during the months of March-November 2020. We estimated a population of 12,000 adults in Immokalee, a 95% confidence interval, a precision of 5%, and a prevalence of 15–20% of the population meeting requirements to be contacted by contact tracers (given an estimated local COVID-19 prevalence of 8%) to calculate a prevalence sample size. From our initial prevalence sample size of 193–241 participants, we accounted for intercorrelation that occurs when surveying multiple people per household. We used an intra-cluster correlation (ICC) of 0.1–0.33 and an estimated household size (m) of ~4 to calculate a design effect of 1.3–2, giving us a final sample size goal of 300–350 participants.

To choose addresses, we first obtained a publicly available address list from the Collier County Property Appraiser’s office. Then, given that many MSFW live in temporary or mobile housing [1] that is likely underrepresented in census data, we examined the public address list and added apartment complexes and clusters of mobile homes that were missing. These additional 669 addresses were added in consultation with local organizations, like CIW, who were familiar with the community. We then extracted 350 addresses from this compiled list using a random number generator. Study staff visited each address, and after exhausting that first address list, extracted 100 further addresses at a time to reach a sample size of 300–350 participants.

2.4. Study Procedures

From January 18-March 11, 2021, study staff (FM and BN) visited addresses from the random address list on weekends and evenings, when adults were most likely to be home from work. They wore appropriate personal protective equipment including a mask and eye protection at all times and conducting interviews outside to maintain social distance. They conducted multiple surveys per household as household members may have had different experiences, and it is common for multiple families in Immokalee to cohabitate.

Study staff started each visit with an initial screening to confirm each participant 1) was 18 years or older, 2) lived in Immokalee, and 3) lived in Immokalee for at least 2 weeks between March-November 2020. If inclusion criteria were met, study staff obtained verbal informed consent in the participants’ preferred language. FM is bilingual in Haitian Creole/English and BN is bilingual in Spanish/English. Verbal consent was obtained due to the population’s literacy level and to maintain participant anonymity. If participants preferred to participate at a different time or no adults were home, one follow-up visit was done at the same address. All households visited received a paper pamphlet detailing available health resources in their preferred language.

Questionnaires were administered on a secure tablet by the study staff. Participants’ responses were recorded by study staff directly into REDCap at the time of survey administration [24, 25]. The data were checked weekly by the co-investigators to ensure internal validity. The study was reviewed and exempted by the Institutional Review Boards at Mass General Brigham (Protocol 2020P003045) and the Harvard School of Public Health (IRB20-1755). All data was de-identified and published online [dataset 23].

2.5. Data Analyses

All analyses were conducted utilizing R (Version 4.0.2, The R Foundation, 2021) and RStudio (Version 1.3.959, RStudio Team, 2021). Chi-square and Fisher’s exact tests were used to compare variables as appropriate. Of note, data from a question on sick leave was not analyzed due to participant misinterpretation of the question.

3.1. Descriptive Analysis

As shown in Table 1, analysis of Collier County mortality data from April 27^th through August 16^th revealed an average excess mortality of 108% (167 excess deaths). When data were disaggregated by sex alone, age alone, and both sex and age, excess mortality findings were largely consistent (107%, 115%, and 115%, respectively).

3.2. Study Population

Of 550 households randomized, 131 (23.8%) did not answer the door, 140 (25.5%) were not interested in participating, and 279 (50.7%) agreed to participate. From these 279 households, 318 individuals consented to participate and were surveyed (Figure 2).

Baseline participant and household demographics are depicted in Table 2. All race/ethnicity/language data is per participants’ self-identification. Spanish (42.1%) was the most frequent preferred language, with English (37.3%) and Haitian Creole (18.7%) accounting for most other participants. Nearly one-fifth of participants reported food insecurity during the past month (19.2%). The mean household size was 3.95 persons with homes averaging 0.856 bedrooms per person and 0.483 bathrooms per person. Twenty-five percent of those surveyed were farm or packinghouse workers, 16.7% were unemployed, and many in the other category worked in service occupations including childcare, cleaning, maintenance, and sales.

3.3. Essential Worker Status and Workplace Policies

Table 3 details participants’ occupational status and policies. Most employed individuals in all language groups were categorized as essential workers by the state of Florida based on their occupation [26]. Twenty-six percent of English speakers stated they were offered the option to work from home during the pandemic. By contrast, only 3% of both Spanish and Haitian Creole speakers reported being given the option to work from home (p < 0.001 English vs. Spanish; p = 0.0023 English vs. Haitian Creole).

3.4. COVID-19 Testing Experiences

Table 4 shows participants’ testing experiences, compared by preferred language. Of English, Spanish, and Haitian Creole speakers, 38.1%, 48.1%, and 57.6% respectively reported that they had never been tested for COVID-19. Significantly fewer Haitian Creole than English speakers reported being tested (p = 0.014). English and Spanish speakers reported being tested at similar sites, primarily by the DOH, while Haitian Creole speakers were more likely to report being tested at the HCN Immokalee Clinic or HCN-affiliated mobile testing sites (p < .0001). English speakers were more likely to report being tested at sites outside of Immokalee, namely Fort Myers and Naples, compared to Spanish and Haitian Creole speakers.

There were no differences in the length of time that lapsed before results were available, though 31.5% of English and 42.6% of Spanish speakers reported waiting one week or more before receiving their results. More positive COVID-19 test results were reported for English-speaking participants (27.4%) than Haitian Creole-speaking participants (0%), though no significant difference was reported between English and Spanish speaking (26.1%) participants.

When asked whether participants would utilize testing resources if exposed to COVID-19, a large share of Spanish (89.5%), Haitian Creole (91.4%) and English speakers (87.3%) reported they would be willing to be tested. Most English (53.4%), Spanish (65.4%), and Haitian Creole speakers (66.7%) reported that they would be willing to isolate themselves in a temporary shelter if necessary and they were unable to do so inside their own home.

3.5. Quality of Contact Tracing

Per the DOH guidelines at the time of the study, those who tested positive for COVID-19 should have been called by the DOH to trace close contacts, inquire about their ability to quarantine, and provide information about local resources supporting quarantine. Calls were expected to be completed in the patient’s preferred language. Table 5 reflects participants’ experiences with the contact tracing process. No Haitian Creole speakers from our sample reported having tested positive. Only 35% of English speakers and 33% of Spanish speakers who tested positive were asked for names and phone numbers of individuals with whom they had been in close contact. Seventy percent of English speakers were asked about their ability to safely quarantine, compared to only 39% of Spanish speakers (p = 0.041). Forty-five percent of English speakers reported being provided information on local resources helping with quarantine, compared to 28% of Spanish speakers. Only 26% of English speakers reported being asked about their language preference compared to 82% of Spanish speakers, suggesting that most calls started in English and switched to Spanish if the recipient stated they did not understand English.

Individuals that were identified as close contacts of someone that had tested positive for COVID-19 were also supposed to be called about their exposure and need to quarantine. Thirty-three percent of English speakers and 42% of Spanish speakers reported being called by the DOH about their positive close contact.

4.1 Limitations

First, though we added additional mobile homes and apartments to create a more representative sample, most addresses came from publicly available housing data, which likely missed some MSFW living in mobile and temporary housing. While our sample included many essential workers, only 20% were MSFW, which was less than anticipated. Also, the survey response rate was just over 50%, raising the possibility of non-response bias. Our suspicion is that non-respondents were more likely to view outreach efforts as intrusive during a pandemic and/or have concerns about migration status; we hypothesize that non-respondents were probably less likely to be reached by public health programs. Additionally, respondents who were migratory from June-October for farm work may have been less likely to receive services. Finally, three participants whose primary language was Mam (a Central American Indigenous language) completed the questionnaire in Spanish, potentially affecting their answers. Future surveys should include all languages in Immokalee. There may have been participants who would have preferred answering in Mam or another Indigenous language if offered this option, as there are known bilingual speakers in Immokalee [17].