Research indicates that Latino day laborers (LDLs) perform jobs in landscaping, moving furniture, and demolition after extreme weather events (18–23). Most jobs, however, are reported to be construction related (21–24). Although not specific to LDLs, studies of workers in the construction industry provide insight into the health and safety risks they face. Non-fatal injuries from falls, slips, and trips accounted for about one-third of Occupational Safety and Health Administration recordable incidents in the construction industry (25). Workers who suffered these injuries missed a median of 28 days of work, which illustrates the severity of their injuries (25). An analysis of 8 years of workers compensation data found that ladders were a contributing factor in 38% of non-fatal falls to a lower level (26). Many jobs performed by LDLs involve working at heights (e.g., on ladders, roofs), which is the leading cause of death among construction workers (27). In 2022, nearly 40% of fatal injuries among construction workers were due to falls, slips, and trips, and nearly all involved falls to a lower level (28).

An analysis of construction industry data from the U. S. Bureau of Labor Statistics found significantly higher fatality rates from falls among Latino workers compared to non-Hispanic White workers and among foreign-born Latino workers compared to U. S.-born Latinos (29). In a study on work-related fatalities from traumatic brain injuries (TBIs), 57% were related to falls, and 80% occurred among construction workers. The rate of TBIs was significantly higher for Latino than White workers and more profound among foreign-born Latino workers compared to U. S.-born Latinos (30). An analysis of occupational fatalities in North Carolina over a 25-year period found that the fatality rate for Latino workers was more than twice the rate for White workers. During those years, 58% of the deaths were among construction workers (31).

With respect to non-fatal injuries among construction workers, the available data by race and ethnicity are limited. An analysis of data from the Agency for Healthcare Research and Quality found that Latino construction workers were more likely to report work-related injuries as well as injuries that resulted in lost workdays (32). In one study, work-related injury inequalities among construction workers were examined using data from the National Health Interview Survey. The analysis found that, for the 14-year period ending in 2017, Latino workers reported more severe injuries than did White, non-Hispanic workers (33).

Hazards in the work environment are the most proximal determinants of an injury at work. As day laborers constantly rotate across a wide variety of jobs, they are at risk of an injury due to their exposure to a wide variety of hazards at work. Because many of their jobs are in the construction industry, the hazards include exposure to noise, extreme temperatures, dust (e.g., respirable silica, lead paint), and fumes (e.g., solder, adhesives), and electrocution, ladders, and tools (e.g., nail guns) (34). The research on the extent to which these or other hazards contribute to serious injuries among LDLs is limited. A review of the literature identified three relevant studies, discussed below.

A cross-sectional analysis of hazards reported by 21 day laborers in Chicago, Illinois, found that the most common were roofing without safety equipment, extreme temperatures, and ladders (23). A study of 183 day laborers in Seattle, Washington, found a statistically significant increased risk of exposure to noise, airborne hazards, and work at heights in day laborers who worked in construction jobs as compared to those who did not (22). A study of 217 day laborers in San Francisco, California, found that 26% experienced a work-related injury or health complaint in the previous 12 months (24).

Two reports about day laborers in the U. S. also provide insight into their working conditions. Following Hurricane Sandy in 2012, day laborers from the region were involved in clean-up activities. The most common hazards reported by the workers were unstable structures, contaminated water, hazardous waste, and mold (19). Surveys of LDLs following extreme weather events in New Orleans, Louisiana, and Houston, Texas, showed the prevalence of skin rashes, headaches, and lacerations (18, 20). These adverse health effects suggest exposure to sharp objects, fumes, and dust. Evidence also indicates that exposure to hazards varies by working context and conditions, with more hazardous work being present in disaster recovery jobs.

Currently, there is limited information about the demographic characteristics of workers who experience serious injuries, which limits our ability to understand who is most at risk. A few studies, however, suggest that age, education, and English proficiency may contribute to the risk of an injury at work. Data from the U. S. National Health Interview Survey for construction workers (2004–2017) indicate that being a member of a racial/ethnic minority group was associated with several factors, including being of younger age and having less education. An analysis of injury rates for 2015–2017 found that the highest rates were among workers under age 25 and those between the ages of 45 and 54 years (34). Limited English proficiency also exacerbates the risk of an injury, as it can impair workers’ ability to comprehend task-related instructions and safety information (35). It should be noted, however, that the correlations with demographic factors reported in these studies were weak, indicating the need to evaluate hazards and other characteristics of construction jobs associated with injury risk (33).

The previous research indicates that LDLs are at a high risk for a serious injury at work. While research on reported injury among these workers is limited, the evidence on specific hazards and worker characteristics associated with this risk is even more limited. To assist workers in reducing their risk for injury at work, it is critical to understand how work and worker characteristics influence this risk. Furthermore, it is important to understand what hazardous chemical exposures are recurrent across different time points and which ones are associated with injury when data is aggregated regardless of the time period.

The purpose of this study is to fill this knowledge gap by exploring the association between worker (demographics) and work (hazardous chemical exposures and conditions) characteristics and injuries reported by LDLs. Our guiding research questions are: (1) is there variability in the worker and work characteristics associated with LDLs reported serious injury at specific time points? (2) Is there a set of common worker and work characteristics associated with reported serious injury regardless of time point? Our approach is unique in that it uses a rigorously tested field methodology and, for the first time, provides a systematic analysis of reported serious injuries and hazardous conditions in this vulnerable population.

We defined a “corner” as a public location (e.g., park, home improvement store, parking lot or street corner) where LDLs routinely gather and wait to be hired by contractors and homeowners. Candidate corners for each survey were observed, and their group size and location were entered into a database later used to randomly select actual interview locations. Despite variations in the main objectives of each survey, they all collected injury-related data among LDLs using the same corner selection, participant recruitment, survey measures, and administration methodology. Participants were included in surveys if they self-identified as Hispanic or Latino, they were Spanish speakers, had been looking for work at the corner for at least 3 months, and were over 18 years old. Spanish-speaking interviewers were trained extensively in survey administration at the corners, and their coding of participant responses were monitored for data quality and accuracy during the data collection period.

To ensure comparability between surveys, we assessed consistency of question wording and response options. Of the seven available datasets, one survey was excluded because its time frame to assess exposures differed from the other surveys. A second dataset was excluded because it did not include questions assessing exposure to hazardous chemicals, and a third one was excluded because it did not include questions related to hazardous working conditions (Figure 1). In the end, we selected data from four surveys representing different time points before (2014, 2019) and during (2020, 2021) the COVID pandemic. While secular trends in the local area indicate that general working conditions may have worsened during the pandemic, our current data analysis sheds light on this assumption.

After ascertaining the comparability of data sets, we proceeded to integrate data from four selected surveys. Any records with missing values were excluded (Figure 1). Redundant records were removed and date of birth and place of origin were used to identify and eliminate duplicates. Out of 801 records, 61 (7.6%) were duplicates. We kept only the most recent record among the identified duplicates. Data was aggregated from four surveys that were conducted in 2014 (n = 296), 2019 (n = 136), 2020 (n = 44), and 2021 (n = 264). This resulting dataset contained records of 740 participants, with five demographic variables and 18 variables on hazardous chemical exposures and working conditions as described below.

The main outcome of interest was reported serious injury experienced in the previous year, defined as one for which the LDL missed at least 1 day of work, went to work when he was injured but thought he should have stayed home, or needed to receive medical attention from a doctor or clinic. This broad definition was chosen to facilitate comprehension and relevance among LDLs who may not be familiar with formal injury classifications. Importantly, this definition differs from OSHA’s definition of a severe injury, which includes defined outcomes such as amputation, in-patient hospitalization, or loss of an eye (39). To assess injury events, interviewers first read the definition of serious injury and then read the following question: “In the past year, have you had a serious injury related to your job as a day laborer?”

To answer the first research question regarding the variability in worker and work characteristics associated with serious injury, we tested whether hazardous chemical exposures and working conditions predicted the odds of reported serious injury at three different time points. We then tested for the commonality of worker and work characteristics associated with reported serious injury across time in the integrated data set. We adopted the same statistical analysis approach to answer both research questions. First, descriptive statistics were calculated for all demographics, hazardous chemical exposures, and hazardous working conditions. Frequencies and percentages are reported for categorical variables and mean and standard deviations for continuous variables. Statistical differences in exposures and working conditions by survey year were examined using Pearson’s chi-square or Fisher’s exact test, depending on the expected frequency in each contingency table cell. Differences by injury status were tested using similar statistics when we used the integrated dataset.

We adopted a multivariable logistic regression approach to assess the extent to which worker (demographics) and work (exposure to hazardous chemicals and conditions) characteristics were associated with the odds of reporting a serious injury. Due to the limited sample size and statistical power of the 2020 survey, we excluded it from data analysis. Initially, we built univariable logistic regression models, using all 18 exposures and working conditions to select the variables and covariates significantly associated with a reported injury. Next, we used multivariable logistic regression with all the significant variables obtained from the initial univariate models. To refine the model, we applied purposeful variable selection. This method is an iterative process whereby variables are removed one at a time, starting with the variable that has the highest p-value. Variables were removed if their p-value exceeded 0.05 and if their removal did not cause a change of more than 15% in the coefficients of the remaining significant variables (43). We applied forward and backward stepwise selection to assess the final model’s robustness and the sensitivity of the model to the variable selection process (44). We computed the variance inflation factor to detect potential collinearity among the variables in our model and found it to be adequate. We deemed p-values below 0.05 as statistically significant. All analyses were executed in R 4.3.0.

The proportion of LDLs in Houston, Texas, who reported experiencing a serious injury in the previous year is substantially higher than the 15% reported in a survey conducted in 2006 in Seattle, Washington (22), and the 19% that was reported in a national survey of day laborers conducted in 2004 (12). We observed variations by survey year in the proportion of study participants who reported a serious injury, ranging from 19.6% in 2014 to 39% in 2019. Furthermore, analyzing results over time in the integrated data set indicates that, overall, the proportion of workers who reported an injury is 27%. This is similar to the proportion (26%) reported by LDLs in San Francisco, California (24); but our definition of serious injury differed slightly. We included incidents in which the participant reported missing at least 1 day of work due to injury, went to work when injured but thought he should have stayed home, or had to receive medical attention from a doctor or clinic. By contrast, Burgel et al. (24) limited their definition of serious injury to one that leads to “missing work” but added “health complaint” as part of this measure. Comparability across studies would necessitate the adoption of a more standard definition and measurement of reported serious injury, as previously proposed (36).

We assessed the extent to which demographic characteristics of LDL would predict reported serious injury. There were differences between the samples recruited in 2014, 2019, 2020 and 2021. The results indicate that there is significant variability in the demographic characteristics of LDLs recruited for the four surveys and trends that need to be explored further. For example, LDLs from Mexico have been the most prevalent over time, while the proportions of those from Cuba or El Salvador are the most variable. These changes in population composition confirm our own observations of how workers at the day labor corners have changed over time. However, none of the demographic characteristics, except for education in 2021, were significantly associated with a reported serious injury. Results were similar for the aggregated dataset, indicating that worker demographic characteristics do not increase the odds of reported serious injury. These results reinforce the necessity of addressing the characteristics of the work environment for injury risk reduction rather than the personal characteristics of the LDLs.

Our study has several limitations. First, we were able to explore the association between exposures and reported injury but were unable to determine the association with specific types of injuries. The nature of the injury related to each significant exposure remains a question to be addressed in a future study. Second, although we explored hazards relevant to LDL experiences, our scales were adapted from other studies, not developed locally and thus, there might be other important hazards that were not included in our surveys. Third, relying on cross-sectional surveys allows us to ascertain only the associations between exposures and serious injuries at discrete time points as well as comprehensively, but the nature of our data does not allow us to establish true temporality or the causal connections between exposures and serious injuries. Finally, we acknowledge that some changes in the significance level of exposures can be due to variation in sample sizes or simultaneous exposures in the multivariable model, and wide confidence intervals are mainly results of small sample size and high variance for estimations.

Our study has several strengths. First, by integrating data from four surveys, we had responses from 740 LDLs recruited, using the same random corner selection procedures at four points over 8 years. The large sample size provided stability to the reported results and increased the possible comparability to other groups of workers who experience similar working conditions. Second, we were able to analyze the relationship between reported serious injuries and a wide variety of potential exposures and working conditions relevant to the work experience of LDLs.