Association of Primary Language with Outcomes After Operations Typically Performed to Treat Cancer: Analysis of a Statewide Database.

Abstract

Few studies have evaluated the effect of primary language on surgical outcomes, and no studies have addressed operations typically performed for cancer diagnoses. This study aimed to determine the effect of primary languages other than English on outcomes after surgical oncology operations. This study retrospectively analyzed adults undergoing operations typically performed to treat cancer using the NJ Healthcare Cost and Utilization Project State Inpatient Database during the interval of 2009-2014. Language was grouped according to English-, Spanish-, and non-English/non-Spanish (NENS)-speaking groups. The study evaluated in-hospital mortality, 7-day readmission, and hospital length of stay (LOS). Logistic and negative binomial regression methods were applied, and generalized linear mixed models were used to account for nesting within a hospital. This study analyzed 37,531 cases. Non-English speakers were of lower economic status, more likely to be admitted on the weekend, and more likely to undergo higher-risk operations. The likelihood of death in the risk-adjusted multi-level models did not differ between Spanish speakers (odds ratio [OR], 0.67; 95% confidence interval [CI], 0.41-1.10) and NENS speakers (OR, 1.16; 95% CI, 0.77-1.75). Readmission rates exhibited high inter-hospital variability (intra-class correlation, 53%). The odds of readmission among Spanish speakers in the non-hierarchical model was increased (OR, 1.50; 95% CI, 1.11-2.02), but this was ameliorated in the multilevel modeling that accounted for variability between hospitals (OR, 1.29; 95% CI, 0.93-1.80). No changes in LOS were observed. No independent association was observed between primary language and outcomes after operations typically performed to treat cancer in the study population. The higher proportion of weekend admissions may suggest more acute or advanced presentations for non-English speakers. Long-term outcomes may be necessary to discern an impact.