SMShivani Mehta

Shivani Mehta

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Shivani Mehta has a total of 37 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Large Language Model–Assisted Semiautomated Data Harmonization Pipeline for the Multiple Chronic Disease Disparities Research Consortium: Framework Development and Comparative Validation Study

    by Hyelee Kim, Shuang Liang, Kathy Lanier, Sarit Helman, Nancy F Cheng, Niloufar Ameli, Jaysón Davidson, Shivani Mehta, Yulin Hswen, Catherine E Oldenburg, et al.

    Background: The increasing availability of machine-readable research data has created a growing need for efficient large-scale data harmonization (DH). Although large language models (LLMs) show promise for reducing the time and labor required for DH, their effective integration into harmonization workflows remains an important methodological challenge. Objective: This study aims to develop and evaluate a semiautomated, human-in-the-loop (HITL) LLM-assisted DH workflow as a proof of concept across 8 projects within a research consortium focused on disparities in multiple chronic conditions. Methods: We developed an LLM-assisted DH workflow that combined preprocessing, semantic mapping, variable response mapping, synthetic data generation, data transformation, and iterative researcher review. Using Generative Pretrained Transformer (GPT-4o) hosted in a secure academic environment, project-specific survey items were mapped to 114 Consortium Common Data Element (CDE) semantic groups. Semantic mapping accuracy was evaluated based on the Consortium’s consensus harmonization results. Complete and incomplete semantic mappings were distinguished according to construct overlap, and incomplete mappings requiring context-dependent human judgment were excluded for mapped item pairs and CDE semantic groups with and without HITL review. Results: Following preprocessing, 884 survey items were included in the DH workflow, of which 812 were mapped to a mean of 79 CDE semantic groups per project. Mean semantic mapping accuracy for item pairs was 95.9% (SD 5.7%) with HITL review and 86.7% (SD 8.6%) without HITL review. At the CDE semantic-group level, the corresponding accuracies were 98.9% (SD 1.7%) and 94.9% (SD 4.0%), respectively. Across projects, omission of valid semantic mappings occurred more frequently than incorrect semantic mappings, particularly for heterogeneous response structures and subjective constructs measured using different instruments. Pipeline components—including concept-based preprocessing, iterative prompt refinement, and targeted HITL review—improved mapping completeness and supported variable harmonization across heterogeneous datasets. Conclusions: This proof-of-concept study demonstrates that the effectiveness of LLM-assisted DH depends more on the design of a structured HITL workflow rather than on the LLM alone. Combining preprocessing, iterative verification, and targeted human oversight improved the efficiency and reliability of DH while preserving researcher judgment for complex mapping decisions.

  • Examining housing insecurity and transportation barriers in pediatric hospital readmissions: insights from structured and unstructured data

    by Shivani Mehta, William Brown, Urmimala Sarkar, Nathan Tran, Yulin Hswen, Matthew S Pantell

    Background: Pediatric hospital readmissions increase healthcare costs and highlight gaps in care. Social determinants of health (SDOH), such as housing and transportation insecurity, significantly impact outcomes but are underexplored in pediatric populations. Objectives: This study evaluates the impact of housing and transportation-related SDOH on pediatric readmissions, comparing structured ICD-10-CM Z-codes alone to a combination of structured and unstructured data extracted via natural language processing (NLP). Materials and Methods: We conducted a retrospective cohort study of pediatric patients (ages 2-17) discharged from UCSF Benioff Children’s Hospital between January 2018 and January 2023. SDOH exposure was identified using structured Z-codes and NLP-extracted data. The primary outcome was hospital readmission within 365 days. Cox proportional hazards models assessed associations between SDOH and readmission risk. Results: Among 8928 patients, only 0.8% were identified as exposed using structured data, compared to 31.7% using combined data. Patients identified through combined data had a higher readmission risk (HR: 2.64, 95% CI: 2.34-2.98) compared to those identified with structured data alone (HR: 1.99, 95% CI: 1.27-3.13). ED utilization was also higher among exposed patients. In the structured-only analysis, exposed patients had a significantly higher hazard of ED readmission (HR: 2.26, 95% CI: 1.65-3.10), whereas the association was slightly attenuated in the combined analysis (HR: 1.49, 95% CI: 1.37-1.62). Conclusion: Leveraging unstructured data enhances SDOH identification and reveals stronger associations with hospital and ED readmissions. A hybrid approach enables improved risk stratification and targeted interventions to address pediatric health disparities.


  • Identifying and Addressing Housing Insecurity in Older Patients: Trends, Referrals, and Inequities in a California Medical System

    by Erin L Ferguson, Shivani Mehta, Silvia Miramontes, Minhyuk Choi, Ye Ji Kim, Tanisha G Hill-Jarrett, Nicolas Cevallos, Yulin Yang, Scott C Zimmerman, Ruijia Chen, et al.

    Background and Objectives: Housing insecurity is rising among older adults, especially for racially minoritized people. Few studies have evaluated whether healthcare institutions are meeting that challenge. Using data from a large California medical system, we examined how often older patients are (A) identified as housing insecure and then (B) referred to social services, as well as inequities in those rates. Research Design and Methods: We analyzed electronic health records (2013–2022) for 119,127 older adults (55+) receiving primary or emergency care. We used a natural language processing model to identify housing insecurity and social services referrals/connections from unstructured notes, with referrals also captured via structured data. Trends in identification were compared to eviction and homelessness trends in the general population. Racial inequities in referrals were evaluated using logistic regression. Results: 0.6% of encounters (n = 6,253) screened positive for housing insecurity. Positive screening trends were nonlinear, with initial increases followed by declines to baseline (roughly tracking regional eviction trends). Only 7% of patients identified as housing insecure were referred to social services, and connections were more likely in primary than emergency care (odds ratio [OR] = 2.04, 95% confidence interval [95% CI]: 1.41–2.96). Asian patients had lower odds of referral than non-Hispanic White patients (OR = 0.51, 95% CI: 0.28–0.95). Discussion and Implications: We identified urgent unmet needs for housing intervention among older patients. Healthcare systems must do more to ensure older patients, especially racially minoritized older adults, are screened for housing insecurity and connected to housing services.


  • Open Access

    Social Determinants of Health Documentation in Structured and Unstructured Clinical Data of Patients With Diabetes: Comparative Analysis

    by Shivani Mehta, Courtney R Lyles, Anna D Rubinsky, Kathryn E Kemper, Judith Auerbach, Urmimala Sarkar, Laura Gottlieb, William Brown Iii

    Background: Electronic health records (EHRs) have yet to fully capture social determinants of health (SDOH) due to challenges such as nonexistent or inconsistent data capture tools across clinics, lack of time, and the burden of extra steps for the clinician. However, patient clinical notes (unstructured data) may be a better source of patient-related SDOH information. Objective: It is unclear how accurately EHR data reflect patients’ lived experience of SDOH. The manual process of retrieving SDOH information from clinical notes is time-consuming and not feasible. We leveraged two high-throughput tools to identify SDOH mappings to structured and unstructured patient data: PatientExploreR and Electronic Medical Record Search Engine (EMERSE). Methods: We included adult patients (≥18 years of age) receiving primary care for their diabetes at the University of California, San Francisco (UCSF), from January 1, 2018, to December 31, 2019. We used expert raters to develop a corpus using SDOH in the compendium as a knowledge base as targets for the natural language processing (NLP) text string mapping to find string stems, roots, and syntactic similarities in the clinical notes of patients with diabetes. We applied advanced built-in EMERSE NLP query parsers implemented with JavaCC. Results: We included 4283 adult patients receiving primary care for diabetes at UCSF. Our study revealed that SDOH may be more significant in the lives of patients with diabetes than is evident from structured data recorded on EHRs. With the application of EMERSE NLP rules, we uncovered additional information from patient clinical notes on problems related to social connectionsisolation, employment, financial insecurity, housing insecurity, food insecurity, education, and stress. Conclusions: We discovered more patient information related to SDOH in unstructured data than in structured data. The application of this technique and further investment in similar user-friendly tools and infrastructure to extract SDOH information from unstructured data may help to identify the range of social conditions that influence patients’ disease experiences and inform clinical decision-making.


  • The association between prescription drug monitoring programs and controlled substance prescribing: a cross-sectional study using data from 2019 National Electronic Health Records Survey

    by Shivani Mehta, William Brown, Erin Ferguson, James Najera, Matthew S Pantell

    Objective: The use of controlled medications such as opioids, stimulants, anabolic steroids, depressants, and hallucinogens has led to an increase in addiction, overdose, and death. Given the high attributes of abuse and dependency, prescription drug monitoring programs (PDMPs) were introduced in the United States as a state-level intervention.Materials and methods: Using cross-sectional data from the 2019 National Electronic Health Records Survey, we assessed the association between PDMP usage and reduced or eliminated controlled substance prescribing as well as the association between PDMP usage and changing a controlled substance prescription to a nonopioid pharmacologic therapy or nonpharmacologic therapy. We applied survey weights to produce physician-level estimates from the survey sample.Results: Adjusting for physician age, sex, type of medical degree, specialty, and ease of PDMP, we found that physicians who reported “often” PDMP usage had 2.34 times the odds of reducing or eliminating controlled substance prescriptions compared to physicians who reported never using the PDMP (95% confidence interval [CI] 1.12–4.90). Adjusting for physician age, sex, type of doctor, and specialty, we found that physicians who reported “often” use of the PDMP had 3.65 times the odd of changing controlled substance prescriptions to a nonopioid pharmacologic therapy or nonpharmacologic therapy (95% CI: 1.61–8.26).Discussion: These results support the continued use, investment, and expansion of PDMPs as an effective intervention for reducing controlled substance prescription and changing to nonopioid/pharmacologic therapy.Conclusion: Overall, frequent usage of PDMPs was significantly associated with reducing, eliminating, or changing controlled substance prescription patterns.