Nearly 26 million individuals in the United States have limited English proficiency (LEP), meaning they have a limited ability to read, speak, write, or understand English. This number is likely a significant underestimate as many LEP individuals face stigma and discrimination and may therefore choose to say they are proficient in English out of fear of receiving lower quality care.

The scientific community and policymakers have identified the need to collect and analyze data across differing populations in order to reveal health disparities and address sources of inequities. Despite decades of research demonstrating that patients with LEP experience potentially costly health consequences such as increased difficulty understanding discharge instructions and obtaining prescriptions, longer lengths of stay, and unplanned ED revisits within 72 hours, a person's language preference remains an overlooked demographic in research.

One example of the importance of capturing a person's language preference in research is that it can help researchers to differentiate health outcomes between persons of the same heritage who may have different histories within the same community. For example, in Chicago there is a large Polish community comprising both naturalized citizens and immigrants. Some in the community have LEP, some were born in the United States but grew up speaking Polish (known as heritage speakers), and some may not speak Polish at all. Language is a key factor that could differentiate their health status, health-seeking behaviors, utilization of health services, and other issues that may explain their health outcomes.

The exclusion of individuals with LEP from research is far too common. Rationales such as “It's too much work to include or recruit them” ignore the fact that exclusion of individuals with LEP can hinder uncovering health outcome disparities. If researchers do not have access to data concerning language preference in their analyses, they should identify why so that the structural sources of inequities in data collection can be addressed.

One legitimate reason researchers may not include language as a demographic in their studies is inconsistent language data collection. Patient language preference data are often missing or data-entry systems have limited patient language preference options. Clinicians may input English as an individual's preferred language if the patient can “get by” in English or if there is a family member who can provide ad hoc interpretation. This documentation practice poses major safety risks, in that misinterpreted medical information can lead to worse health outcomes. For these reasons, people with language needs are often undercounted in large datasets based on medical chart data. Although the use of patient language data can be challenging, missing data management, patient self-reporting techniques, and quality improvement checks performed by researchers will translate into more precise measurement of the impact of language preference on patient outcomes.

Another common rationale for not examining study results by language groups is small sample sizes that may have insufficient statistical power to determine differences among groups. However, there are statistical methods available that may compensate for these limitations. For example, researchers should become comfortable reporting findings with overlapping confidence intervals, as these results can still hint at what is happening and inform future studies. Additionally, researchers should refrain from dichotomizing language groups and rather assess outcome differences across groups, which may unveil disparities that were previously unnoticed due to category homogenization.

When considering patient characteristics to control for in our regression models, or making decisions about sample stratification or the presentation of study results, we can consider who is, or perhaps more importantly, who isn't, included. Worse health outcomes for individuals with LEP persist and should not be perpetuated by research methodologies that fail to recognize the importance of language preference. It is critically important that these steps be taken to improve the health outcomes of patients who may not speak the same language as their clinician.