Relationships among language proficiency, communication patterns, and safety performance in small work crews in the United States

The construction industry employs approximately 6% of all U.S. workers but accounts for over 16% of all occupational fatalities. Recent statistics indicate that over 40% of all U.S. construction laborers cannot speak English proficiently. To ensure strong safety performance, it is vital to include every individual in a construction crew in safety-related communications, regardless of language proficiency. Considering that most safety communication is delivered in English, it is not surprising that Spanish-speaking construction workers are fatally injured at a disproportionate rate. To conduct the first exploration of the characteristics of strong, multilingual safety networks in the Denver Metropolitan region of the United States, a multilingual research team conducted interviews with the members of 14 construction crews. Demographic attribute data for each individual (e.g., language proficiency, years of experience, position in the company) and network data were collected to analyze the safety communication network for each crew. The units of analysis included the individual actors in the network and the networks as a whole. The exploratory results contribute to the body of knowledge by revealing that unilingual work crews have safety performance that is 51% better than multilingual work crews (p=0.10), bilingual workers play a more central role than unilingual workers when more than one language is spoken (p<0.001), workers less than 35 years of age have a higher degree of centrality than do workers who are older than 35 years old (p=0.11), and managers play an important role in the exchange and diffusion of safety knowledge regardless of language proficiency (p<0.001). Most importantly, social-network analysis metrics show that these language-boundary spanners often form the core of a network that connects disparate groups of individuals. By contrast, crews with relatively weak safety performance tend to have clear and disparate subnetworks distinguished by language and high turnover rates. Such characteristics are of concern because individual actors are not able to effectively warn one another of uncontrolled hazardous exposures or work in transition.