Multilevel determinants of hepatocellular carcinoma surveillance in eligible Hispanic and Asian American patients: Language concordance and ethnic enclaves

Low hepatocellular carcinoma (HCC) surveillance uptake among high-risk Hispanic and Asian American populations may be influenced by provider communication and/or neighborhood context. We investigated associations between patient-provider language concordance, neighborhood ethnic enclave and HCC surveillance in a large, diverse California health system. We analyzed electronic health record data from Sutter Health (2000-2016) appended to census tract-level ethnic enclave indices (% recent immigrants, Spanish/Asian language-speaking households and limited English proficiency, % Hispanic/Asian residents). Eligible adults had an assigned primary care provider (PCP), ≥1 encounter, valid address, no liver cancer history, surveillance indication (cirrhosis or chronic hepatitis B/C), ≥18 months follow-up, and self-identified as Hispanic (HA) or Asian American (AA). Surveillance was ultrasound every 6 months and/or contrast-enhanced CT/MRI (with/without alpha-fetoprotein). We defined uptake as ≥1 test and adherence as percent time up-to-date (PTUDS) among those surveilled. We assessed language concordance (PCP spoke patient's non-English language) and ethnic enclave (quintiles; Q5=highest). Hierarchical logistic/linear regression with facility clustering estimated ORs (uptake) and betas (PTUDS), adjusting for demographics, follow-up time, utilization. N=11,311 patients were included (3,444 HA, 7,867 AA); 19% of HAs and 28% of AAs were non-English speakers; 38% of HAs and 55% of AAs lived in Q5 enclaves. Overall, 36% of HAs and 43% of AAs were surveilled (median PTUDS: 0.16 HA, 0.15 AA). Among AAs, compared to English speakers, language-concordant non-English speakers had higher uptake (OR=1.35, 95%CI 1.12-1.62) and adherence (beta=0.024, 95%CI 0.009-0.039), while non-concordant non-English speakers had similar uptake (OR=1.01, 95%CI 0.88-1.16) but higher adherence (beta=0.020, 95%CI 0.008-0.032). Among HAs, language-concordant non-English speakers showed similar uptake/adherence to English speakers, but non-concordant non-English speakers had higher uptake (OR=1.43, 95%CI 1.16-1.76). Ethnic enclave showed no AA association (p-trend=0.46) but HA uptake was inversely associated with enclave (OR=0.79 comparing Q1 vs. Q5, 95%CI 0.59-1.07, p-trend=0.01). Adherence was not associated with enclave. Patient-provider language concordance and neighborhood ethnic enclave influenced HCC surveillance differently by race/ethnicity. AA surveillance was associated with language-concordant providers but not enclave; HA surveillance was associated with enclave but not provider language. These differential patterns suggest tailored intervention targets: multilingual provider recruitment for Asian languages, and community-based outreach leveraging existing networks in Hispanic neighborhoods.