Jeremy P. Bauer
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Jeremy P. Bauer has a total of 84 co-authors. Here are the top co-authors by number of shared publications:
- Anita Bagley
Shriners Children's Northern California, 2425 Stockton Blvd, Sacramento, CA 95817, the United States of America
5 shared publications - Susan Sienko
Shriners Children's Portland, 3101 SW Sam Jackson Park Rd, Portland, OR 97239, the United States of America
5 shared publications - Ellen Marie Raney
Shriners Hospitals for Children, Portland, OR
3 shared publications - Jon Robert Davids
Department of Orthopaedic Surgery, University of California School of Medicine, Sacramento, CA, USA; Department of Orthopaedic Surgery, Shriners Children's Northern California, Sacramento, CA, USA
3 shared publications Susan Sienko
Shriners Children’s Portland, Portland, OR
3 shared publications- Adam S. Arterbery
The Jackson Laboratory for Genomic Medicine , Farmington, CT
2 shared publications
Latest Publications
- Open Access
- 3 September 2026
- Published by Elsevier in The American Journal of Human Genetics
- Vol. 113 (9) , 1972-1986
A phenotypic paradigm for cerebral palsy genetics
- 3 September 2026
- Published by Elsevier in The American Journal of Human Genetics
- Vol. 113 (9) , 1972-1986
Open AccessA phenotypic paradigm for cerebral palsy genetics
- Open Access
- 1 June 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Pain Severity Classification From Electronic Health Record in Children With Cerebral Palsy: Retrospective Study (Preprint)
BACKGROUND Pain is common and clinically consequential in children with cerebral palsy (CP), but retrospective pain severity assessment from electronic health records (EHRs) is difficult. Pain evidence is often dispersed across long narrative notes, many children have limited ability to self-report, and structured pain scores may be influenced by proxy reporting and encounter context. OBJECTIVE This study introduces the first systematic benchmark for three-class pain severity classification from clinical notes of children with CP and evaluates whether modern Natural Language Processing (NLP) and large language model approaches can recover clinically meaningful severity signals from real-world EHR documentation. METHODS We conducted a retrospective study using 15,969 de-identified clinical notes from 1,467 children with CP treated across the Shriners Children’s Hospital Network over 19 years. Notes were linked to encounter-level numeric pain ratings and grouped into No/Mild (0–3), Moderate (4–6), and Severe (7–10) classes. Using patient-stratified splits, we compared supervised encoder-fusion models, explanation-augmented local language models, frozen instruction-following LLMs, and a late-fusion ensemble. RESULTS Performance improved as models incorporated broader context and complementary decision patterns. The best supervised encoder-fusion model achieved a macro F1 of 0.600, while explanation-augmented local LLMs improved to 0.620 when clinical notes were combined with generated explanations. Few-shot GPT-5 was the strongest standalone model in this held-out test set, achieving a macro F1 of 0.644. The best overall performance was achieved by a late-fusion ensemble of Bio-Mistral-7B, zero-shot GPT-5, and few-shot GPT-5, reaching a macro F1 of 0.66. SHapley Additive exPlanations (SHAP) analysis showed that No/Mild predictions were driven mainly by negated pain language, whereas Moderate and Severe predictions were associated with explicit pain terms and severity cues. CONCLUSIONS Modern NLP and LLM-based approaches can extract clinically meaningful pain severity signals from real-world EHR notes of children with CP. Performance was strongest when models used long-context modeling, explanation-guided supervision, and complementary prediction patterns. Remaining errors reflected the subjective, proxy-mediated, and temporally variable nature of pain documentation. Therefore, these models should support retrospective pain research, cohort identification, and documentation review. To our knowledge, this is the first study to benchmark multi-level pain severity classification from unstructured EHR notes of children with CP.
- 1 June 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessPain Severity Classification From Electronic Health Record in Children With Cerebral Palsy: Retrospective Study (Preprint)
BACKGROUND Pain is common and clinically consequential in children with cerebral palsy (CP), but retrospective pain severity assessment from electronic health records (EHRs) is difficult. Pain evidence is often dispersed across long narrative notes, many children have limited ability to self-report, and structured pain scores may be influenced by proxy reporting and encounter context. OBJECTIVE This study introduces the first systematic benchmark for three-class pain severity classification from clinical notes of children with CP and evaluates whether modern Natural Language Processing (NLP) and large language model approaches can recover clinically meaningful severity signals from real-world EHR documentation. METHODS We conducted a retrospective study using 15,969 de-identified clinical notes from 1,467 children with CP treated across the Shriners Children’s Hospital Network over 19 years. Notes were linked to encounter-level numeric pain ratings and grouped into No/Mild (0–3), Moderate (4–6), and Severe (7–10) classes. Using patient-stratified splits, we compared supervised encoder-fusion models, explanation-augmented local language models, frozen instruction-following LLMs, and a late-fusion ensemble. RESULTS Performance improved as models incorporated broader context and complementary decision patterns. The best supervised encoder-fusion model achieved a macro F1 of 0.600, while explanation-augmented local LLMs improved to 0.620 when clinical notes were combined with generated explanations. Few-shot GPT-5 was the strongest standalone model in this held-out test set, achieving a macro F1 of 0.644. The best overall performance was achieved by a late-fusion ensemble of Bio-Mistral-7B, zero-shot GPT-5, and few-shot GPT-5, reaching a macro F1 of 0.66. SHapley Additive exPlanations (SHAP) analysis showed that No/Mild predictions were driven mainly by negated pain language, whereas Moderate and Severe predictions were associated with explicit pain terms and severity cues. CONCLUSIONS Modern NLP and LLM-based approaches can extract clinically meaningful pain severity signals from real-world EHR notes of children with CP. Performance was strongest when models used long-context modeling, explanation-guided supervision, and complementary prediction patterns. Remaining errors reflected the subjective, proxy-mediated, and temporally variable nature of pain documentation. Therefore, these models should support retrospective pain research, cohort identification, and documentation review. To our knowledge, this is the first study to benchmark multi-level pain severity classification from unstructured EHR notes of children with CP.
- 9 May 2026
- Published by Elsevier in Journal of Biomechanics
- Vol. 204, 113359
Validation, characterization, and utility of markerless motion capture in a large cohort of pediatric patients with complex gait patterns
- 9 May 2026
- Published by Elsevier in Journal of Biomechanics
- Vol. 204, 113359
Validation, characterization, and utility of markerless motion capture in a large cohort of pediatric patients with complex gait patterns
- Open Access
Validation, characterization, and utility of markerless motion capture in a large cohort of pediatric patients with complex gait patterns
Markerless motion capture has emerged as a potential substitute for traditional marker-based systems, offering scalable, non-invasive acquisition of human movement. Despite increasing adoption in research and sports applications, its clinical utility for children with complex gait patterns remains an open question. To address this gap, simultaneous marker-based and markerless data were collected in 202 pediatric children (12.1 ± 3.9 years). Marker-based kinematics were processed using the Shriners Children’s Gait Model (SCGM), while markerless outputs were computed using Theia3D with identical Cardan sequences. Agreement between systems was evaluated using statistical parametric mapping (SPM), root-mean-square error (RMSE), and a gait pattern classification based on the plantarflexor–knee extension index. Markerless output systematically underestimated pelvic tilt, hip rotation, and knee rotation and demonstrated reduced between-subject variance in the transverse plane. SPM revealed widespread waveform differences, although most were of negligible effect, especially in the sagittal plane. Mean sagittal-plane RMSEs were < 5° for the knee and ankle and < 8° for the pelvis and hip. Coronal-plane deviations were < 7°, whereas transverse-plane errors exceeded 10°. RMSE increased significantly with body mass index and use of a walker (p < 0.001). Agreement in sagittal-plane gait classification was moderate between systems (κ = 0.60; 67% overall concordance). These results indicate that markerless motion capture is suitable for analyses emphasizing sagittal deviations but remains limited for applications requiring precise axial or frontal-plane estimation. Future work should address algorithmic underestimation of transverse motion and evaluate markerless performance across increasing severity of gait deviation.
Open AccessValidation, characterization, and utility of markerless motion capture in a large cohort of pediatric patients with complex gait patterns
Markerless motion capture has emerged as a potential substitute for traditional marker-based systems, offering scalable, non-invasive acquisition of human movement. Despite increasing adoption in research and sports applications, its clinical utility for children with complex gait patterns remains an open question. To address this gap, simultaneous marker-based and markerless data were collected in 202 pediatric children (12.1 ± 3.9 years). Marker-based kinematics were processed using the Shriners Children’s Gait Model (SCGM), while markerless outputs were computed using Theia3D with identical Cardan sequences. Agreement between systems was evaluated using statistical parametric mapping (SPM), root-mean-square error (RMSE), and a gait pattern classification based on the plantarflexor–knee extension index. Markerless output systematically underestimated pelvic tilt, hip rotation, and knee rotation and demonstrated reduced between-subject variance in the transverse plane. SPM revealed widespread waveform differences, although most were of negligible effect, especially in the sagittal plane. Mean sagittal-plane RMSEs were < 5° for the knee and ankle and < 8° for the pelvis and hip. Coronal-plane deviations were < 7°, whereas transverse-plane errors exceeded 10°. RMSE increased significantly with body mass index and use of a walker (p < 0.001). Agreement in sagittal-plane gait classification was moderate between systems (κ = 0.60; 67% overall concordance). These results indicate that markerless motion capture is suitable for analyses emphasizing sagittal deviations but remains limited for applications requiring precise axial or frontal-plane estimation. Future work should address algorithmic underestimation of transverse motion and evaluate markerless performance across increasing severity of gait deviation.
- Open Access
- 19 February 2026
- Published by Elsevier in SSRN Electronic Journal
Automated Gait Classification: Comparison of Automated Algorithms to Expert Classification
Background: Accurate classification of gait patterns in children with cerebral palsy is critical for guiding treatment but typically requires expert interpretation of gait data. Automated methods have potential to improve consistency and scalability; however, clinically meaningful automated tools are limited. This work aimed to determine if automated algorithms based on established gait classification frameworks can reproduce expert clinical classification in children with cerebral palsy.Methods: An automated MATLAB algorithm was developed to assign gait classifications using two established systems. A sample of children with cerebral palsy who met criteria for crouch gait underwent automated classification. Three expert gait analysts independently classified all trials using standardized definitions corresponding to each system. Fleiss’s κ quantified inter-rater reliability among reviewers, while Cohen’s κ, weighted κ, percent agreement, and macro F1 scores quantified agreement between each reviewer and the automated classifications.Findings: Inter-rater reliability among reviewers was substantial for Rodda & Graham (κ=0.753), with high agreement between reviewers. Inter-rater reliability among reviewers was moderate for Rozumalski & Schwartz (κ=0.456) and agreement between raters was lower and more variable, with some classifications demonstrating full disagreement among reviewers.Interpretation: Automated gait classification based on quantitatively defined gait criteria can achieve expert-level performance for systems with clear biomechanical boundaries. More complex cluster–based systems yield lower agreement, reflecting inherent ambiguity in cluster overlap. These findings support use of automated tools as reliable, objective ground truth for large-scale analyses and for training future markerless or video-based gait assessment algorithms aimed at expanding gait evaluation beyond specialized motion laboratories.
- 19 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessAutomated Gait Classification: Comparison of Automated Algorithms to Expert Classification
Background: Accurate classification of gait patterns in children with cerebral palsy is critical for guiding treatment but typically requires expert interpretation of gait data. Automated methods have potential to improve consistency and scalability; however, clinically meaningful automated tools are limited. This work aimed to determine if automated algorithms based on established gait classification frameworks can reproduce expert clinical classification in children with cerebral palsy.Methods: An automated MATLAB algorithm was developed to assign gait classifications using two established systems. A sample of children with cerebral palsy who met criteria for crouch gait underwent automated classification. Three expert gait analysts independently classified all trials using standardized definitions corresponding to each system. Fleiss’s κ quantified inter-rater reliability among reviewers, while Cohen’s κ, weighted κ, percent agreement, and macro F1 scores quantified agreement between each reviewer and the automated classifications.Findings: Inter-rater reliability among reviewers was substantial for Rodda & Graham (κ=0.753), with high agreement between reviewers. Inter-rater reliability among reviewers was moderate for Rozumalski & Schwartz (κ=0.456) and agreement between raters was lower and more variable, with some classifications demonstrating full disagreement among reviewers.Interpretation: Automated gait classification based on quantitatively defined gait criteria can achieve expert-level performance for systems with clear biomechanical boundaries. More complex cluster–based systems yield lower agreement, reflecting inherent ambiguity in cluster overlap. These findings support use of automated tools as reliable, objective ground truth for large-scale analyses and for training future markerless or video-based gait assessment algorithms aimed at expanding gait evaluation beyond specialized motion laboratories.
- Open Access
A Phenotypic Paradigm for Cerebral Palsy Genetics
Disease-causing genetic variants can be found in a subset of individuals with cerebral palsy (CP), with variants deemed causal of CP having been published for at least 515 genes. We develop a statistical approach that treats CP as a phenotypic feature for which some genetic disorders confer an increased risk. Based on comprehensive literature curation we show that the null hypothesis of no CP association can be rejected for only 89 of the 515 genes. We applied these findings to the analysis of a cohort of 460 children diagnosed with CP in the Shriner Children's network. We identified pathogenic or likely pathogenic (P/LP) variants in 60 genes in 15.8% of the children. Only 16 of the 60 genes had significant evidence for CP association in our literature analysis. Our results suggest that a stratified approach to attributing causality to genetic variants in CP could support precision genomic medicine for affected individuals.
Open AccessA Phenotypic Paradigm for Cerebral Palsy Genetics
Disease-causing genetic variants can be found in a subset of individuals with cerebral palsy (CP), with variants deemed causal of CP having been published for at least 515 genes. We develop a statistical approach that treats CP as a phenotypic feature for which some genetic disorders confer an increased risk. Based on comprehensive literature curation we show that the null hypothesis of no CP association can be rejected for only 89 of the 515 genes. We applied these findings to the analysis of a cohort of 460 children diagnosed with CP in the Shriner Children's network. We identified pathogenic or likely pathogenic (P/LP) variants in 60 genes in 15.8% of the children. Only 16 of the 60 genes had significant evidence for CP association in our literature analysis. Our results suggest that a stratified approach to attributing causality to genetic variants in CP could support precision genomic medicine for affected individuals.
- Open Access
- 16 December 2025
- Published by Elsevier in Journal of Orthopaedics
- Vol. 73, 227-233
Comparison of baseline outcomes between surgical and nonoperative management in youth with lower extremity torsional abnormalities
- 16 December 2025
- Published by Elsevier in Journal of Orthopaedics
- Vol. 73, 227-233
Open AccessComparison of baseline outcomes between surgical and nonoperative management in youth with lower extremity torsional abnormalities
- Open Access
- 12 March 2025
- Published by SAGE Publications in Journal of Children's Orthopaedics
- Vol. 19 (2) , 151-157
Factors associated with flexed knee gait in unilateral cerebral palsy
Purpose: A flexed knee gait is a common gait in children with unilateral cerebral palsy. In children without knee contracture, hamstring spasticity is commonly considered a major contributor to a flexed knee gait. We hypothesized that the popliteal angle would not correlate to a flexed knee gait. Methods: This retrospective study included 109 children with unilateral cerebral palsy who had undergone complete 3D gait analysis. Children who had previous surgery or knee flexion contracture were excluded. Children were divided into three groups based on knee position during stance as determined by 3D gait analysis: flexion (FK, 47), hyperextension (HK, 42), and normal (NK, 20). Results: There were no significant correlations between popliteal angle and dynamic peak knee extension in stance or at initial contact. Similarly, peak dorsiflexion during the stance phase did not correlate with dynamic peak knee extension in stance (all p > 0.05). Significant differences were observed in foot position during stance between FK and HK groups, as well as in quick stretch dorsiflexion with the knee extended between HK and NK groups. Conclusion: A flexed knee gait in children with unilateral cerebral palsy does not always correlate with the popliteal angle or dynamic ankle position in gait. These factors may contribute but are insufficient to explain all observed differences. A flexed knee gait likely involves a complex interplay of motor control, strength, spasticity, and lever arm dynamics, indicating that interventions at a single level may not fully improve dynamic knee extension.
- 12 March 2025
- Published by SAGE Publications in Journal of Children's Orthopaedics
- Vol. 19 (2) , 151-157
Open AccessFactors associated with flexed knee gait in unilateral cerebral palsy
Purpose: A flexed knee gait is a common gait in children with unilateral cerebral palsy. In children without knee contracture, hamstring spasticity is commonly considered a major contributor to a flexed knee gait. We hypothesized that the popliteal angle would not correlate to a flexed knee gait. Methods: This retrospective study included 109 children with unilateral cerebral palsy who had undergone complete 3D gait analysis. Children who had previous surgery or knee flexion contracture were excluded. Children were divided into three groups based on knee position during stance as determined by 3D gait analysis: flexion (FK, 47), hyperextension (HK, 42), and normal (NK, 20). Results: There were no significant correlations between popliteal angle and dynamic peak knee extension in stance or at initial contact. Similarly, peak dorsiflexion during the stance phase did not correlate with dynamic peak knee extension in stance (all p > 0.05). Significant differences were observed in foot position during stance between FK and HK groups, as well as in quick stretch dorsiflexion with the knee extended between HK and NK groups. Conclusion: A flexed knee gait in children with unilateral cerebral palsy does not always correlate with the popliteal angle or dynamic ankle position in gait. These factors may contribute but are insufficient to explain all observed differences. A flexed knee gait likely involves a complex interplay of motor control, strength, spasticity, and lever arm dynamics, indicating that interventions at a single level may not fully improve dynamic knee extension.
- 12 March 2025
- Published by Wolters Kluwer Health in Journal of Pediatric Orthopaedics
- Vol. 45 (6) , 306-311
Validation of Parental Reports in Assessing Idiopathic Toe Walking Using Quantitative In-Shoe Device Measurements
Background: Toe walking is prevalent among children, affecting 5% to 24% of the pediatric population. Clinicians rely on parental reports of frequency of toe walking to guide clinical decision making and outcomes assessment. However, recall accuracy and differing environments challenge the reliability of parental reports. This study aims to validate parental reports against quantitative in-shoe device measurements (NURVV/RUN). Methods: Twenty children with persistent idiopathic toe walking (ITWp) (mean age: 9.6y; 13 males, 7 females) from 8 pediatric orthopaedic specialty care sites participated in this multicenter study. Parents assessed toe walking frequency using a 6-point scale, while children wore NURVV/RUN insoles for 8 hours/day over 7 days. Insole sensors recorded foot strike patterns (rearfoot, midfoot, and forefoot), which were classified using the same severity scale. Agreement between parental reports and NURVV data was assessed using weighted Kappa statistics (PPP<0.005) for isolated forefoot contact. Conclusion: Parental reports of toe walking prevalence in their children are relatively accurate, supporting their use in clinical management. However, quantitative in-shoe devices provide a more objective and quantitative understanding of ITWp frequency and have the potential to guide clinical decision-making and outcomes assessment children with ITWp. Level of Evidence: Level II—diagnostic study. See instructions to authors for a complete description of levels of evidence.
- 12 March 2025
- Published by Wolters Kluwer Health in Journal of Pediatric Orthopaedics
- Vol. 45 (6) , 306-311
Validation of Parental Reports in Assessing Idiopathic Toe Walking Using Quantitative In-Shoe Device Measurements
Background: Toe walking is prevalent among children, affecting 5% to 24% of the pediatric population. Clinicians rely on parental reports of frequency of toe walking to guide clinical decision making and outcomes assessment. However, recall accuracy and differing environments challenge the reliability of parental reports. This study aims to validate parental reports against quantitative in-shoe device measurements (NURVV/RUN). Methods: Twenty children with persistent idiopathic toe walking (ITWp) (mean age: 9.6y; 13 males, 7 females) from 8 pediatric orthopaedic specialty care sites participated in this multicenter study. Parents assessed toe walking frequency using a 6-point scale, while children wore NURVV/RUN insoles for 8 hours/day over 7 days. Insole sensors recorded foot strike patterns (rearfoot, midfoot, and forefoot), which were classified using the same severity scale. Agreement between parental reports and NURVV data was assessed using weighted Kappa statistics (PPP<0.005) for isolated forefoot contact. Conclusion: Parental reports of toe walking prevalence in their children are relatively accurate, supporting their use in clinical management. However, quantitative in-shoe devices provide a more objective and quantitative understanding of ITWp frequency and have the potential to guide clinical decision-making and outcomes assessment children with ITWp. Level of Evidence: Level II—diagnostic study. See instructions to authors for a complete description of levels of evidence.
- Open Access
- 16 November 2022
- Published by SAGE Publications in Journal of Children's Orthopaedics
- Vol. 16 (6) , 454-460
The incidence of avascular necrosis in children with cerebral palsy after hip containment surgery
Purpose: To establish the rate of avascular necrosis after hip reconstruction surgery in children with cerebral palsy and to identify risk factors that influence the development of avascular necrosis in this population. Methods: An institutional review board–approved retrospective review was conducted on children with cerebral palsy who underwent hip containment surgery at a single institution. Radiographs were evaluated at three time points. The Reimer’s migration percentage, neck shaft angle, epiphyseal shaft angle, acetabular index, center edge angle, and acetabular angle were measured. The presence of avascular necrosis was evaluated and graded by the Bucholz/Ogden and the Kalamchi/MacEwen classification systems. Multivariate logistic regression was performed to identify risk factors associated with the development of avascular necrosis. Results: A total of 154 children with cerebral palsy underwent hip containment surgery on 223 hips. Twenty-nine children (18.8%) underwent both pelvic and femoral procedures; 36 children (23.4%) had only femoral procedures; 47 children (30.5%) had femoral and soft tissue; and 42 children (27.3%) had pelvic, femoral, and soft tissue procedures. Using the Bucholz and Ogden or the Kalamchi classifications, the rate of avascular necrosis was 24.7% (38/154). Of the variables evaluated, preoperative Reimers was found to be significant predictors of avascular necrosis. The rate of avascular necrosis was 26.7% for Gross Motor Functional Classification System level III, 24.1% for Gross Motor Functional Classification System level IV, and 27.3% for Gross Motor Functional Classification System level V. Conclusion: The overall rate of avascular necrosis in children undergoing hip containment surgery was 26.7%. Together, age at surgery, open reduction, previous surgery, preoperative Reimers, and estimated blood loss contributed to the development of postoperative avascular necrosis; however, only preoperative Reimers significantly contributed to the development of avascular necrosis in children with cerebral palsy undergoing hip containment procedures.
- 16 November 2022
- Published by SAGE Publications in Journal of Children's Orthopaedics
- Vol. 16 (6) , 454-460
Open AccessThe incidence of avascular necrosis in children with cerebral palsy after hip containment surgery
Purpose: To establish the rate of avascular necrosis after hip reconstruction surgery in children with cerebral palsy and to identify risk factors that influence the development of avascular necrosis in this population. Methods: An institutional review board–approved retrospective review was conducted on children with cerebral palsy who underwent hip containment surgery at a single institution. Radiographs were evaluated at three time points. The Reimer’s migration percentage, neck shaft angle, epiphyseal shaft angle, acetabular index, center edge angle, and acetabular angle were measured. The presence of avascular necrosis was evaluated and graded by the Bucholz/Ogden and the Kalamchi/MacEwen classification systems. Multivariate logistic regression was performed to identify risk factors associated with the development of avascular necrosis. Results: A total of 154 children with cerebral palsy underwent hip containment surgery on 223 hips. Twenty-nine children (18.8%) underwent both pelvic and femoral procedures; 36 children (23.4%) had only femoral procedures; 47 children (30.5%) had femoral and soft tissue; and 42 children (27.3%) had pelvic, femoral, and soft tissue procedures. Using the Bucholz and Ogden or the Kalamchi classifications, the rate of avascular necrosis was 24.7% (38/154). Of the variables evaluated, preoperative Reimers was found to be significant predictors of avascular necrosis. The rate of avascular necrosis was 26.7% for Gross Motor Functional Classification System level III, 24.1% for Gross Motor Functional Classification System level IV, and 27.3% for Gross Motor Functional Classification System level V. Conclusion: The overall rate of avascular necrosis in children undergoing hip containment surgery was 26.7%. Together, age at surgery, open reduction, previous surgery, preoperative Reimers, and estimated blood loss contributed to the development of postoperative avascular necrosis; however, only preoperative Reimers significantly contributed to the development of avascular necrosis in children with cerebral palsy undergoing hip containment procedures.