Shant Ayanian
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Shant Ayanian has a total of 158 co-authors. Here are the top co-authors by number of shared publications:
- Alexander J. Ryu
Mayo Clinic
5 shared publications - Elena Myasoedova
Division of Rheumatology, Mayo Clinic College of Medicine and Science, Rochester, Minnesota, USA
3 shared publications - Juan A. Reyes
Department of Medicine. George Washington University Hospital, Washington, DC USA
3 shared publications - Sagar B. Dugani
Mayo Clinic
3 shared publications Bhargav Kanakiya
Cerebras Systems , Inc., 1237 E. Arques Ave, Sunnyvale 94085, CA
2 shared publicationsCarl Molnar
Mayo Clinic , 200 1st St SW, Rochester 55905, MN
2 shared publications
Latest Publications
- Open Access
- 31 July 2026
- Published by Elsevier in SSRN Electronic Journal
Validated Transformation of Inpatient EHR Data into a Deployment-Ready Analytic Infrastructure
ImportanceHospital electronic health records (EHRs) contain extensive clinical data used for predictive analytics and quality improvement. However, raw EHR data are heterogeneous, temporally complex, and prone to inconsistencies that limit reproducibility and impede reliable translation into real-world clinical practice.ObjectiveTo develop a reproducible framework for transforming inpatient EHR data into a validated, analysis-ready dataset aligned with operational EHR semantics.Materials and MethodsWe extracted EHR data for adult inpatient admissions with research authorization across 19 hospitals within a US academic health system spanning four states from July 2017 through May 2024. Data domains included demographics, laboratory results, vital signs, medications, comorbidities, and utilization measures. Processing involved normalization of semi-structured values, categorical harmonization, handling of missingness, and two-layer outlier management using modified Z-scores and physiologic plausibility thresholds. Temporal alignment was validated, including large-scale manual review of 144,283 data fields. Feature engineering generated admission-level summaries, laboratory trends, time-normalized care-intensity measures, medication exposures across care phases, and age-adjusted Elixhauser Comorbidity Index scores using a two-year ICD-10 lookback window.ResultsThe final dataset included 889,870 admissions among 457,543 patients, with 277 curated features per admission. Manual validation demonstrated 99.36% field-level accuracy and revealed discrepancies in timestamp semantics, which were corrected through refinement of extraction logic.DiscussionBy embedding traceability, empirical validation, and standardized transformation rules, this framework supports reproducibility and reduces training-serving mismatch between model development and deployment.ConclusionValidated extraction and semantically consistent transformation of EHR data are essential for enabling reliable and reproducible use of EHR data in clinical research and real-world clinical practice.
- 31 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessValidated Transformation of Inpatient EHR Data into a Deployment-Ready Analytic Infrastructure
ImportanceHospital electronic health records (EHRs) contain extensive clinical data used for predictive analytics and quality improvement. However, raw EHR data are heterogeneous, temporally complex, and prone to inconsistencies that limit reproducibility and impede reliable translation into real-world clinical practice.ObjectiveTo develop a reproducible framework for transforming inpatient EHR data into a validated, analysis-ready dataset aligned with operational EHR semantics.Materials and MethodsWe extracted EHR data for adult inpatient admissions with research authorization across 19 hospitals within a US academic health system spanning four states from July 2017 through May 2024. Data domains included demographics, laboratory results, vital signs, medications, comorbidities, and utilization measures. Processing involved normalization of semi-structured values, categorical harmonization, handling of missingness, and two-layer outlier management using modified Z-scores and physiologic plausibility thresholds. Temporal alignment was validated, including large-scale manual review of 144,283 data fields. Feature engineering generated admission-level summaries, laboratory trends, time-normalized care-intensity measures, medication exposures across care phases, and age-adjusted Elixhauser Comorbidity Index scores using a two-year ICD-10 lookback window.ResultsThe final dataset included 889,870 admissions among 457,543 patients, with 277 curated features per admission. Manual validation demonstrated 99.36% field-level accuracy and revealed discrepancies in timestamp semantics, which were corrected through refinement of extraction logic.DiscussionBy embedding traceability, empirical validation, and standardized transformation rules, this framework supports reproducibility and reduces training-serving mismatch between model development and deployment.ConclusionValidated extraction and semantically consistent transformation of EHR data are essential for enabling reliable and reproducible use of EHR data in clinical research and real-world clinical practice.
- Open Access
- 8 June 2026
- Published by Springer Nature in Journal of General Internal Medicine
- Vol. 41 (15) , 4388-4397
Trends and Sex-Specific Rates of Obesity and Severe Obesity in US Hospitalized Adults, 2012–2021
Background: Rising national obesity rates may increase clinical and operational challenges for hospitals. However, trends and distributions in obesity and severe obesity among hospitalized adults are incompletely described. Objective: To examine trends in hospital-coded obesity and severe obesity rates among adults, by year, sex, age-groups, race/ethnicity, and US Census divisions. Design: Cross-sectional analyses of patients with hospital-coded obesity and severe obesity, using US National Inpatient Sample data, from 2012 through 2021. Rates were stratified by demographic group, and trends were calculated using logistic regression models, and prevalence ratios (PR) were used to compare male-to-female rates. Participants: Hospitalized adults with obesity (body mass index [BMI] ≥ 30 kg/m 2 ) or severe obesity (BMI ≥ 40 kg/m 2 ) codes excluding pregnancy-related hospitalizations. Main Measure(s): Hospital-coded obesity and severe obesity rate per year, annual average change, absolute percent change, and PR of male-to-female obesity and severe obesity. Key Results: From 2012 through 2021, among 51,669,013 hospitalizations, 16.9% ( n = 8,733,196) had hospital-coded obesity and 9.0% ( n = 4,638,219) had severe obesity. Obesity rates increased from 13.0% to 21.4% (0.88% per year [95% CI, 0.81–0.95%], P < 0.001) while severe obesity rates increased from 6.6% to 11.2% (0.48% per year [95% CI, 0.43–0.53%], P < 0.001). Compared to other female age-groups and all male age-groups, females aged 30–49 years had the largest annual increase in obesity (1.25% [95% CI, 1.09–1.41%], P < 0.01) and severe obesity (0.95% [95% CI, 0.84–1.05%], P < 0.01). When compared across age and racial/ethnic groups, the largest sex differences in obesity rates were observed among adults aged 18–29 years, particularly of White (weighted average PR, 0.49 [95% CI, 0.48–0.49]) and Black (weighted average PR, 0.43 [95% CI, 0.43–0.44]) race. Conclusions: From 2012 through 2021, US hospital-coded obesity and severe obesity rates increased and revealed sex differences, underscoring the need for hospitals to adapt care delivery and operations.
- 8 June 2026
- Published by Springer Nature in Journal of General Internal Medicine
- Vol. 41 (15) , 4388-4397
Open AccessTrends and Sex-Specific Rates of Obesity and Severe Obesity in US Hospitalized Adults, 2012–2021
Background: Rising national obesity rates may increase clinical and operational challenges for hospitals. However, trends and distributions in obesity and severe obesity among hospitalized adults are incompletely described. Objective: To examine trends in hospital-coded obesity and severe obesity rates among adults, by year, sex, age-groups, race/ethnicity, and US Census divisions. Design: Cross-sectional analyses of patients with hospital-coded obesity and severe obesity, using US National Inpatient Sample data, from 2012 through 2021. Rates were stratified by demographic group, and trends were calculated using logistic regression models, and prevalence ratios (PR) were used to compare male-to-female rates. Participants: Hospitalized adults with obesity (body mass index [BMI] ≥ 30 kg/m 2 ) or severe obesity (BMI ≥ 40 kg/m 2 ) codes excluding pregnancy-related hospitalizations. Main Measure(s): Hospital-coded obesity and severe obesity rate per year, annual average change, absolute percent change, and PR of male-to-female obesity and severe obesity. Key Results: From 2012 through 2021, among 51,669,013 hospitalizations, 16.9% ( n = 8,733,196) had hospital-coded obesity and 9.0% ( n = 4,638,219) had severe obesity. Obesity rates increased from 13.0% to 21.4% (0.88% per year [95% CI, 0.81–0.95%], P < 0.001) while severe obesity rates increased from 6.6% to 11.2% (0.48% per year [95% CI, 0.43–0.53%], P < 0.001). Compared to other female age-groups and all male age-groups, females aged 30–49 years had the largest annual increase in obesity (1.25% [95% CI, 1.09–1.41%], P < 0.01) and severe obesity (0.95% [95% CI, 0.84–1.05%], P < 0.01). When compared across age and racial/ethnic groups, the largest sex differences in obesity rates were observed among adults aged 18–29 years, particularly of White (weighted average PR, 0.49 [95% CI, 0.48–0.49]) and Black (weighted average PR, 0.43 [95% CI, 0.43–0.44]) race. Conclusions: From 2012 through 2021, US hospital-coded obesity and severe obesity rates increased and revealed sex differences, underscoring the need for hospitals to adapt care delivery and operations.
- Open Access
- 12 May 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Artificial intelligence for predicting hospital admissions from the emergency department: a prospective, quasi-experimental study
The use of certain artificial intelligence (AI) tools may improve hospital operational efficiency, in particular in overcrowded emergency departments (ED). Here, we conduct a prospective, quasi experimental study evaluating an AI model predicting hospital admission risk, alternately displaying and hiding its outputs to clinicians in 2 week blocks over 11 months in 2023. Among 54,394 eligible ED visits, the AI tool does not change the number of ED patients discharged per day but reduces median ED length of stay by 12 min without increasing 72 h bounceback visits. Model performance remained stable (AUC 0.80-0.82), and hospitalist clinicians reported greater perceived usefulness than ED clinicians. These findings show that integrating a low burden AI prediction tool into ED workflows can improve operational efficiency. The study was registered on Clinicaltrials.gov (NCT05683899).
- 12 May 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessArtificial intelligence for predicting hospital admissions from the emergency department: a prospective, quasi-experimental study
The use of certain artificial intelligence (AI) tools may improve hospital operational efficiency, in particular in overcrowded emergency departments (ED). Here, we conduct a prospective, quasi experimental study evaluating an AI model predicting hospital admission risk, alternately displaying and hiding its outputs to clinicians in 2 week blocks over 11 months in 2023. Among 54,394 eligible ED visits, the AI tool does not change the number of ED patients discharged per day but reduces median ED length of stay by 12 min without increasing 72 h bounceback visits. Model performance remained stable (AUC 0.80-0.82), and hospitalist clinicians reported greater perceived usefulness than ED clinicians. These findings show that integrating a low burden AI prediction tool into ED workflows can improve operational efficiency. The study was registered on Clinicaltrials.gov (NCT05683899).
- 9 May 2026
- Published by Taylor & Francis in Postgraduate Medicine
- Vol. 138 (4) , 443-449
Characteristics and outcomes among patients with stercoral colitis – a retrospective cohort study
Stercoral colitis is a rare, but potentially serious, complication of chronic constipation. Limited literature exists, pertaining to stercoral colitis-related management and outcomes. We aim to study characteristics and outcomes of stercoral colitis patients presenting to emergency department. We compiled a cohort of patients who presented to one of 21 EDs between 1 January 2024 and 31 January 2025 and had a radiological finding of stercoral colitis on CT imaging. We manually extracted details regarding their ED presentation and disposition, hospital course, and any acute adverse outcomes defined as inpatient endoscopy, inpatient surgery, and death within 90 days of the ED visit related to stercoral colitis. Additional chart review was performed in patients who experienced adverse outcomes to study the clinical circumstances. Three hundred and eighty-one patient encounters (358, unique patients) were identified with stercoral colitis. Only seven patients (2.0%) experienced acute adverse outcomes, though four of these were not attributed to stercoral colitis. Two hundred and seventy-two patient encounters (71.4%) were admitted to the hospital, among which 114 encounters (29.9%) had primary admitting diagnoses related to stercoral colitis. Of the 109 (28.6%) encounters discharged from ED, 2 patients returned to ED within 72 hours. Management of hospitalized stercoral colitis patients mostly included disimpaction and bowel regimen medications. Stercoral colitis is an uncommon diagnosis, and adverse outcomes are rare. The decision to hospitalize should be based on the overall clinical picture, considering need for hospital-based interventions, or assistance with bowel regimen. The Age-Friendly Health System framework can facilitate an optimal management approach.
- 9 May 2026
- Published by Taylor & Francis in Postgraduate Medicine
- Vol. 138 (4) , 443-449
Characteristics and outcomes among patients with stercoral colitis – a retrospective cohort study
Stercoral colitis is a rare, but potentially serious, complication of chronic constipation. Limited literature exists, pertaining to stercoral colitis-related management and outcomes. We aim to study characteristics and outcomes of stercoral colitis patients presenting to emergency department. We compiled a cohort of patients who presented to one of 21 EDs between 1 January 2024 and 31 January 2025 and had a radiological finding of stercoral colitis on CT imaging. We manually extracted details regarding their ED presentation and disposition, hospital course, and any acute adverse outcomes defined as inpatient endoscopy, inpatient surgery, and death within 90 days of the ED visit related to stercoral colitis. Additional chart review was performed in patients who experienced adverse outcomes to study the clinical circumstances. Three hundred and eighty-one patient encounters (358, unique patients) were identified with stercoral colitis. Only seven patients (2.0%) experienced acute adverse outcomes, though four of these were not attributed to stercoral colitis. Two hundred and seventy-two patient encounters (71.4%) were admitted to the hospital, among which 114 encounters (29.9%) had primary admitting diagnoses related to stercoral colitis. Of the 109 (28.6%) encounters discharged from ED, 2 patients returned to ED within 72 hours. Management of hospitalized stercoral colitis patients mostly included disimpaction and bowel regimen medications. Stercoral colitis is an uncommon diagnosis, and adverse outcomes are rare. The decision to hospitalize should be based on the overall clinical picture, considering need for hospital-based interventions, or assistance with bowel regimen. The Age-Friendly Health System framework can facilitate an optimal management approach.
- Open Access
Interpretable variant effect prediction from genomic foundation model embeddings
Scientific foundation models learn high-dimensional representations from diverse data modalities, yet what they encode and how to extract that knowledge remain open questions. Here we show that probing the internal representations of Evo 2, a 7-billion-parameter genomic foundation model, enables accurate and interpretable genetic variant effect prediction. We introduce a covariance-based probe that captures second-order structure from Evo 2 sequence embeddings to predict variant pathogenicity across variant types and functional consequences, matching or exceeding specialized predictors within their domains. To ground these predictions in known biological mechanisms, we train a complementary panel of probes on existing annotations to detect which genomic properties are disrupted by a variant. This categorized evidence is then integrated with each variant's genomic context through a language model to generate variant-specific mechanistic hypotheses. Our pathogenicity predictions correlate with experimental measures of variant function, clinical penetrance, and biobank disease associations while the mechanistic hypotheses are consistent with expert reviews, known mechanism classes, and downstream molecular readouts. We release pathogenicity scores, disruption profiles, and contextualized interpretations for 4.2 million variants from the ClinVar database as an open resource through the Evo Variant Effect Explorer (EVEE). More broadly, this structured probing approach offers a general framework for interrogating foundation models across scientific disciplines and grounding their outputs in existing domain concepts.
Open AccessInterpretable variant effect prediction from genomic foundation model embeddings
Scientific foundation models learn high-dimensional representations from diverse data modalities, yet what they encode and how to extract that knowledge remain open questions. Here we show that probing the internal representations of Evo 2, a 7-billion-parameter genomic foundation model, enables accurate and interpretable genetic variant effect prediction. We introduce a covariance-based probe that captures second-order structure from Evo 2 sequence embeddings to predict variant pathogenicity across variant types and functional consequences, matching or exceeding specialized predictors within their domains. To ground these predictions in known biological mechanisms, we train a complementary panel of probes on existing annotations to detect which genomic properties are disrupted by a variant. This categorized evidence is then integrated with each variant's genomic context through a language model to generate variant-specific mechanistic hypotheses. Our pathogenicity predictions correlate with experimental measures of variant function, clinical penetrance, and biobank disease associations while the mechanistic hypotheses are consistent with expert reviews, known mechanism classes, and downstream molecular readouts. We release pathogenicity scores, disruption profiles, and contextualized interpretations for 4.2 million variants from the ClinVar database as an open resource through the Evo Variant Effect Explorer (EVEE). More broadly, this structured probing approach offers a general framework for interrogating foundation models across scientific disciplines and grounding their outputs in existing domain concepts.
- Open Access
- 27 March 2026
- Published by Elsevier in American Journal of Medicine Open
- Vol. 15
Machine Learning Models to Evaluate County-Level Incidence of Diagnosed Diabetes and Sociodemographic Factors
- 27 March 2026
- Published by Elsevier in American Journal of Medicine Open
- Vol. 15
Open AccessMachine Learning Models to Evaluate County-Level Incidence of Diagnosed Diabetes and Sociodemographic Factors
- Open Access
- 1 November 2025
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 26 (6)
Introducing a foundational sequence transformer for range adaptive nucleotide decoding (STRAND)
The advent of high-throughput sequencing has led to an exponential increase in genomic data, highlighting the need for efficient and accurate models to analyze and interpret this information. In this study, we introduce a novel, exomic foundational model that leverages a combination of the human reference genome and multispecies data to improve variant detection and interpretation. Our model utilizes a short-range transformer architecture and is trained on a large dataset of human exomic sequences derived from the Tapestry study. Through a series of ablation studies and scaling experiments, we demonstrate the effectiveness of our model in predicting next token accuracy and identifying clinically pathogenic variants. We also show that our model outperforms existing models in a range of downstream tasks, including variant effect prediction and disease state identification. In fact, our largest sequence transformer for range adaptive nucleotide decoding variant (1B parameters) surpassed previous benchmarks, demonstrating a mean accuracy of 0.880 [an 8.2% improvement over the original nucleotide transformer (NT) and a 7% improvement over NT-v2]. Furthermore, we construct a unique exomic ClinVar dataset to evaluate the model’s performance on pathogenicity and disease states. Our results highlight the potential of this model to improve our understanding of the human exome and its role in disease. The model and its applications have significant implications for genomics-based diagnosis and personalized medicine, including tailored therapeutic development.
- 1 November 2025
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 26 (6)
Open AccessIntroducing a foundational sequence transformer for range adaptive nucleotide decoding (STRAND)
The advent of high-throughput sequencing has led to an exponential increase in genomic data, highlighting the need for efficient and accurate models to analyze and interpret this information. In this study, we introduce a novel, exomic foundational model that leverages a combination of the human reference genome and multispecies data to improve variant detection and interpretation. Our model utilizes a short-range transformer architecture and is trained on a large dataset of human exomic sequences derived from the Tapestry study. Through a series of ablation studies and scaling experiments, we demonstrate the effectiveness of our model in predicting next token accuracy and identifying clinically pathogenic variants. We also show that our model outperforms existing models in a range of downstream tasks, including variant effect prediction and disease state identification. In fact, our largest sequence transformer for range adaptive nucleotide decoding variant (1B parameters) surpassed previous benchmarks, demonstrating a mean accuracy of 0.880 [an 8.2% improvement over the original nucleotide transformer (NT) and a 7% improvement over NT-v2]. Furthermore, we construct a unique exomic ClinVar dataset to evaluate the model’s performance on pathogenicity and disease states. Our results highlight the potential of this model to improve our understanding of the human exome and its role in disease. The model and its applications have significant implications for genomics-based diagnosis and personalized medicine, including tailored therapeutic development.
- Open Access
- 26 March 2025
- Published by Springer Nature in Research Square
Introducing STRAND: A Foundational Sequence Transformer for Range Adaptive Nucleotide Decoding
The advent of high-throughput sequencing has led to an exponential increase in genomic data, highlighting the need for efficient and accurate models to analyze and interpret this information. Here, we introduce a novel, exomic foundational model that leverages a combination of human reference genome and multispecies data to improve variant detection and interpretation. Our model utilizes a short- range transformer architecture and is trained on a large dataset of human exomic sequences derived from the Tapestry study. Through a series of ablation studies and scaling experiments, we demonstrate the effectiveness of our model in pre- dicting next token accuracy and identifying clinically pathogenic variants. We also show that our model outperforms existing models in a range of downstream tasks, including variant effect prediction and disease state identification. In fact, our largest STRAND variant (1B parameters) surpassed previous benchmarks, demonstrating a mean accuarcy of 0.880 (8.2% improvement over the original NT and 7% improvement over NT-v2). Furthermore, we construct a unique exomic ClinVar dataset to evaluate the model’s performance on pathogenicity and disease states. Our results highlight the potential of this model to improve our understanding of the human exome and its role in disease. The model and its applications have significant implications for genomic based diagnosis and personalized medicine including tailored therapeutic development.
- 26 March 2025
- Published by Springer Nature in Research Square
Open AccessIntroducing STRAND: A Foundational Sequence Transformer for Range Adaptive Nucleotide Decoding
The advent of high-throughput sequencing has led to an exponential increase in genomic data, highlighting the need for efficient and accurate models to analyze and interpret this information. Here, we introduce a novel, exomic foundational model that leverages a combination of human reference genome and multispecies data to improve variant detection and interpretation. Our model utilizes a short- range transformer architecture and is trained on a large dataset of human exomic sequences derived from the Tapestry study. Through a series of ablation studies and scaling experiments, we demonstrate the effectiveness of our model in pre- dicting next token accuracy and identifying clinically pathogenic variants. We also show that our model outperforms existing models in a range of downstream tasks, including variant effect prediction and disease state identification. In fact, our largest STRAND variant (1B parameters) surpassed previous benchmarks, demonstrating a mean accuarcy of 0.880 (8.2% improvement over the original NT and 7% improvement over NT-v2). Furthermore, we construct a unique exomic ClinVar dataset to evaluate the model’s performance on pathogenicity and disease states. Our results highlight the potential of this model to improve our understanding of the human exome and its role in disease. The model and its applications have significant implications for genomic based diagnosis and personalized medicine including tailored therapeutic development.
- Open Access
- 26 April 2022
- Published by Elsevier in Mayo Clinic Proceedings: Innovations, Quality & Outcomes
- Vol. 6 (3) , 193-199
Assessing the Generalizability of a Clinical Machine Learning Model Across Multiple Emergency Departments
- 26 April 2022
- Published by Elsevier in Mayo Clinic Proceedings: Innovations, Quality & Outcomes
- Vol. 6 (3) , 193-199
Open AccessAssessing the Generalizability of a Clinical Machine Learning Model Across Multiple Emergency Departments
- 25 June 2021
- Published by Elsevier in Journal of the National Medical Association
- Vol. 113 (6) , 626-635
Better clinical outcomes in hospitalized COVID-19 minority patients with accompanying gastrointestinal symptoms
- 25 June 2021
- Published by Elsevier in Journal of the National Medical Association
- Vol. 113 (6) , 626-635
Better clinical outcomes in hospitalized COVID-19 minority patients with accompanying gastrointestinal symptoms