Using a Multilingual AI Care Agent to Reduce Disparities in Colorectal Cancer Screening: Higher FIT Test Adoption Among Spanish-Speaking Patients
Bibliographic record
Abstract
Abstract Background Colorectal cancer (CRC) screening rates remain disproportionately low among Hispanic and Latino populations compared to non-Hispanic whites. While artificial intelligence (AI) shows promise in healthcare delivery, concerns exist that AI-based interventions may disadvantage non-English-speaking populations due to biases in development and deployment. Objective To evaluate the effectiveness of a bilingual AI care agent in engaging Spanish-speaking patients for CRC screening compared to English-speaking patients. Methods This retrospective analysis examined an AI-powered outreach initiative at WellSpan Health in Pennsylvania and Maryland during September 2024. The study included 1,878 patients (517 Spanish-speaking, 1,361 English-speaking) eligible for CRC screening who lacked active online health profiles. A bilingual AI conversational agent conducted personalized telephone calls in the patient’s preferred language to provide education about CRC screening and facilitate fecal immunochemical test (FIT) kit requests. Primary outcome was FIT test opt-in rate, with secondary outcomes including connect rates and call duration. Statistical analysis included descriptive statistics, bivariate comparisons, and multivariate logistic regression. Results Spanish-speaking patients demonstrated significantly higher engagement across all measures compared to English-speaking patients: FIT test opt-in rates (18.2% vs. 7.1%, p<0.001), connect rates (69.6% vs. 53.0%, p<0.001), and call duration (6.05 vs. 4.03 minutes, p<0.001). Demographically, Spanish-speaking patients were younger (mean age 57 vs. 61 years, p<0.001) and more likely to be female (49.1% vs. 38.4%, p<0.001). In multivariate analysis, Spanish language preference remained an independent predictor of FIT test opt-in (adjusted OR 2.012, 95% CI 1.340-3.019, p<0.001) after controlling for demographic factors and call duration. Conclusions AI-powered outreach achieved significantly higher engagement among Spanish-speaking patients, challenging the assumption that technological interventions inherently disadvantage non-English speaking populations. The 2.6-fold higher FIT test opt-in rate among Spanish-speaking patients represents a notable departure from historical patterns of healthcare disparities. These findings suggest that language-concordant AI interactions may help address longstanding disparities in preventive care access. Study limitations include its single healthcare system setting, short duration, and lack of follow-up data on completed screenings. Future research should assess long-term adherence and whether higher engagement translates to improved clinical outcomes.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.006 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.005 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".