Evaluating the impact of language on health care utilization outcomes in paitents with diffuse large B-cell lymphoma at a large safety-net hospital

Abstract

Abstract Introduction: In the United States 25 million people have limited English proficiency. Individuals with limited English proficiency (LEP) are at increased risk for poor cancer health outcomes and higher health care utilization. While the impact of language on healthcare utilization in solid tumors is better studied, little is known about the impact of language on healthcare utilization in patients with diffuse large B-cell lymphoma (DLBCL). Between 2020-2025 there has been a projected 11% increase in diagnoses of DLBCL. Understanding healthcare utilization patterns is necessary to identify and address barriers to optimal care. Methods Our primary independent variable for this analysis was primary spoken language recorded in the electronic health record: English, Spanish, and other (Haitian Creole, Cape Verdean Creole, and Vietnamese, Italian, Albanian, Amharic, Arabic, Bosnia, Brazilian Portuguese, Burmese, Cambodian, French, Greek, Moroccan Arabic, Punjabi, and Turkish by descending prevalence). We conducted a retrospective cohort study of adult patients diagnosed with DLBCL at Boston Medical Center between October 2015 to December 2023. Inclusion criteria included age > 18 years and diagnosis of DLBCL. Healthcare utilization was measured using the number of all-cause hospital admissions within the first 6 months after diagnosis and average admission length of stay (LOS). Variables included age, primary language, insurance, Charles Comorbidity Index (CCI), marital status, social vulnerability index (SVI), disease stage, and disease response at time of restaging. Results Among 136 eligible DLBCL patients, 76 (55.8%) were English speakers, 33 (24.26%) were Spanish speakers, and 27 (19.85%) spoke a non-English, non-Spanish language. Median age at diagnosis was 57.6, 61.6, and 70.0 years (p = 0.0018) for English, Spanish, and other languages, respectively. The proportion of married patients was 32.9%, 42.4%, and 48.2% (p=0.0319) for English, Spanish, and other languages, respectively. There were no statistically significant differences between these three language groups for the following variables: 1) disease characteristics, CCI ( 4.18, 4.03, and 4.59, p = 0.6710), Stage IV at time of diagnosis (53.9%, 40.6%, and 40.7% p = 0.3268), 2) health care utilization: hospital admissions in first 6 months (2.03, 1.67, and 1.67 p = 0.6050) and average hospital admission LOS in days (6.23, 5.72, and 6.38, p = 0.8841) or 3) social determinants: SVI (0.706, 0.859, and 0.7210) and distance to hospital in miles (16.0, 16.9, and 14.6),respectively. We further examined the relationship between language category with dichotomized stage at diagnosis (Stage IV vs. Not) and dichotomized completeness of response (Complete vs. Not) in cross tabulations with chi-square testing. As in our comparisons of mean LOS and hospitalizations in the first 6 months after diagnosis by language category, we found no statistically significant differences though the proportions of the outcomes were descriptively different. Spanish speakers had a greater proportion of patients with complete response (81%) versus English speakers (64%) and those who spoke other languages (65%), p=0.20. Fifty-five percent of English speakers were in stage IV at diagnosis compared to Spanish speakers (41%) and those who spoke other languages (44%), p=0.32. In multivariable analyses adjusting for age at diagnosis, Charlson Index, SVI, and distance to the hospital, we likewise did not find any significant differences in outcomes between the language categories. Conclusion In our cohort, key differences were a 13-year age gap between English and “other language” speakers. It is encouraging that despite demographic differences, we found no statistically significant differences in disease severity (stage IV rates, CCI), healthcare utilization (all-cause hospital admissions in first 6 months and average admission LOS in days), SVI, or treatment outcomes (complete response rates at first restaging). Next steps, include (1) pooling data for multisite analysis of system, disease, and patient characteristics (2) LEP subgroup outcome analysis and (3) assessment of overall survival and quality of life. These steps will aid in safety-net system and LEP sub-group characteristics contributing significant differences in age at time of diagnosis with preserved health care utilization to inform age-appropriate recommendations for individuals within the LEP population.