Tomoyoshi Takenaka
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Tomoyoshi Takenaka has a total of 84 co-authors. Here are the top co-authors by number of shared publications:
Tomoharu Yoshizumi
Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan
27 shared publicationsMikihiro Kohno
Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1, Maidashi, Higashi-ku, Fukuoka, 812-8582, Japan
21 shared publicationsFumihiko Kinoshita
Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan
20 shared publicationsTaichi Nagano
Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan
19 shared publicationsKyoto Matsudo
Department of Surgery and Science, Graduate School of Medical Sciences, Kyushu University, 3-1-1, Maidashi, Higashi-ku, Fukuoka, 812-8582, Japan
17 shared publicationsMototsugu Shimokawa
Department of Biostatistics, Yamaguchi University Graduate School of Medicine, Yamaguchi, Japan
15 shared publications
Latest Publications
- 27 May 2026
- Published by International Institute of Anticancer Research in Anticancer Research
- Vol. 46  (6) , 3483-3491
Comparison of RATS Lobectomy and VATS Lobectomy for Early-stage Non-small Cell Lung Cancer
Robotic-assisted thoracoscopic surgery (RATS) for pulmonary resection of early-stage non-small cell lung cancer (NSCLC) has been increasingly adopted worldwide. To evaluate its safety and oncological efficacy, we compared the outcomes of RATS and video-assisted thoracoscopic surgery (VATS) lobectomy in patients with stage 0 or I NSCLC. Between August 2014 and December 2024, 330 patients with stage 0 or I NSCLC underwent pulmonary lobectomy via RATS (n=144) or VATS (n=186). RATS was performed using the da Vinci Xi system and VATS was performed using a conventional multiport technique. The perioperative and clinicopathological outcomes were retrospectively evaluated and analyzed after propensity score matching. The baseline characteristics were similar between the groups. The RATS group included more right upper lobectomies and fewer left upper lobectomies than the VATS group. Intraoperative blood loss was significantly lower in the RATS group. There were no significant differences in operative time, chest tube duration, postoperative complications, or length of hospital stay. Regarding lymph node (LN) dissection, the RATS group showed significantly higher numbers of dissected and total harvested LNs in the overall cohort. However, after propensity score matching, the LN yields were comparable between the groups. Both RATS and VATS lobectomies were safely performed for early-stage NSCLC. RATS achieved perioperative outcomes comparable to those achieved with VATS, with a reduced intraoperative blood loss. The lymph node harvest was similar after propensity score matching, suggesting that RATS provides oncological quality equivalent to that of VATS.
- 27 May 2026
- Published by International Institute of Anticancer Research in Anticancer Research
- Vol. 46  (6) , 3483-3491
Comparison of RATS Lobectomy and VATS Lobectomy for Early-stage Non-small Cell Lung Cancer
Robotic-assisted thoracoscopic surgery (RATS) for pulmonary resection of early-stage non-small cell lung cancer (NSCLC) has been increasingly adopted worldwide. To evaluate its safety and oncological efficacy, we compared the outcomes of RATS and video-assisted thoracoscopic surgery (VATS) lobectomy in patients with stage 0 or I NSCLC. Between August 2014 and December 2024, 330 patients with stage 0 or I NSCLC underwent pulmonary lobectomy via RATS (n=144) or VATS (n=186). RATS was performed using the da Vinci Xi system and VATS was performed using a conventional multiport technique. The perioperative and clinicopathological outcomes were retrospectively evaluated and analyzed after propensity score matching. The baseline characteristics were similar between the groups. The RATS group included more right upper lobectomies and fewer left upper lobectomies than the VATS group. Intraoperative blood loss was significantly lower in the RATS group. There were no significant differences in operative time, chest tube duration, postoperative complications, or length of hospital stay. Regarding lymph node (LN) dissection, the RATS group showed significantly higher numbers of dissected and total harvested LNs in the overall cohort. However, after propensity score matching, the LN yields were comparable between the groups. Both RATS and VATS lobectomies were safely performed for early-stage NSCLC. RATS achieved perioperative outcomes comparable to those achieved with VATS, with a reduced intraoperative blood loss. The lymph node harvest was similar after propensity score matching, suggesting that RATS provides oncological quality equivalent to that of VATS.
- Open Access
- 1 December 2025
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 13, e71617
Identifying Key Variances in Clinical Pathways Associated With Prolonged Hospital Stays Using Machine Learning and ePath Real-World Data: Model Development and Validation Study
Background: Prolonged hospital stays can lead to inefficiencies in health care delivery and unnecessary consumption of medical resources. Objective: This study aimed to identify key clinical variances associated with prolonged length of stay (PLOS) in clinical pathways using a machine learning model trained on real-world data from the ePath system. Methods: We analyzed data from 480 patients with lung cancer (age: mean 68.3, SD 11.2 years; n=263, 54.8% men) who underwent video-assisted thoracoscopic surgery at a university hospital between 2019 and 2023. PLOS was defined as a hospital stay exceeding 9 days after video-assisted thoracoscopic surgery. The variables collected between admission and 4 days after surgery were examined, and those that showed a significant association with PLOS in univariate analyses (P<.01) were selected as predictors. Predictive models were developed using sparse linear regression methods (Lasso, ridge, and elastic net) and decision tree ensembles (random forest and extreme gradient boosting). The data were divided into derivation (earlier study period) and testing (later period) cohorts for temporal validation. The model performance was assessed using the area under the receiver operating characteristic curve, Brier score, and calibration plots. Counterfactual analysis was used to identify key clinical factors influencing PLOS. Results: A 3D heatmap illustrated the temporal relationships between clinical factors and PLOS based on patient demographics, comorbidities, functional status, surgical details, care processes, medications, and variances recorded from admission to 4 days after surgery. Among the 5 algorithms evaluated, the ridge regression model demonstrated the best performance in terms of both discrimination and calibration. Specifically, it achieved area under the receiver operating characteristic curve values of 0.84 and 0.82 and Brier scores of 0.16 and 0.17 in the derivation and test cohorts, respectively. In the final model, a range of variables, including blood tests, care, patient background, procedures, and clinical variances, were associated with PLOS. Among these, particular emphasis was placed on clinical variances. Counterfactual analysis using the ridge regression model identified 6 key variables strongly linked to PLOS. In order of impact, these were abnormal respiratory sounds, postoperative fever, arrhythmia, impaired ambulation, complications after drain removal, and pulmonary air leaks. Conclusions: A machine learning–based model using ePath data effectively identified critical variances in the clinical pathways associated with PLOS. This automated tool may enhance clinical decision-making and improve patient management.
- 1 December 2025
- Published by JMIR Publications Inc. in JMIR Medical Informatics
- Vol. 13, e71617
Open AccessIdentifying Key Variances in Clinical Pathways Associated With Prolonged Hospital Stays Using Machine Learning and ePath Real-World Data: Model Development and Validation Study
Background: Prolonged hospital stays can lead to inefficiencies in health care delivery and unnecessary consumption of medical resources. Objective: This study aimed to identify key clinical variances associated with prolonged length of stay (PLOS) in clinical pathways using a machine learning model trained on real-world data from the ePath system. Methods: We analyzed data from 480 patients with lung cancer (age: mean 68.3, SD 11.2 years; n=263, 54.8% men) who underwent video-assisted thoracoscopic surgery at a university hospital between 2019 and 2023. PLOS was defined as a hospital stay exceeding 9 days after video-assisted thoracoscopic surgery. The variables collected between admission and 4 days after surgery were examined, and those that showed a significant association with PLOS in univariate analyses (P<.01) were selected as predictors. Predictive models were developed using sparse linear regression methods (Lasso, ridge, and elastic net) and decision tree ensembles (random forest and extreme gradient boosting). The data were divided into derivation (earlier study period) and testing (later period) cohorts for temporal validation. The model performance was assessed using the area under the receiver operating characteristic curve, Brier score, and calibration plots. Counterfactual analysis was used to identify key clinical factors influencing PLOS. Results: A 3D heatmap illustrated the temporal relationships between clinical factors and PLOS based on patient demographics, comorbidities, functional status, surgical details, care processes, medications, and variances recorded from admission to 4 days after surgery. Among the 5 algorithms evaluated, the ridge regression model demonstrated the best performance in terms of both discrimination and calibration. Specifically, it achieved area under the receiver operating characteristic curve values of 0.84 and 0.82 and Brier scores of 0.16 and 0.17 in the derivation and test cohorts, respectively. In the final model, a range of variables, including blood tests, care, patient background, procedures, and clinical variances, were associated with PLOS. Among these, particular emphasis was placed on clinical variances. Counterfactual analysis using the ridge regression model identified 6 key variables strongly linked to PLOS. In order of impact, these were abnormal respiratory sounds, postoperative fever, arrhythmia, impaired ambulation, complications after drain removal, and pulmonary air leaks. Conclusions: A machine learning–based model using ePath data effectively identified critical variances in the clinical pathways associated with PLOS. This automated tool may enhance clinical decision-making and improve patient management.
- 12 August 2025
- Published by Springer Nature in International Journal of Clinical Oncology
- Vol. 30  (10) , 1972-1981
Clinical and biological impact of SNAT7 in lung adenocarcinoma: implications for prognosis and treatment
Glutamine metabolism plays a crucial role in cancer cell proliferation and modulates the tumour microenvironment. High expression of glutamine transporters is associated with poor prognosis in non-small cell lung cancer. SNAT7, encoded by SLC38A7, facilitates glutamine transport from lysosomes. However, its function and clinical significance in lung adenocarcinoma remain unclear. Immunohistochemistry (IHC) was performed with samples from 373 patients with completely resected lung adenocarcinoma, and the association between SNAT7 expression, clinicopathological features, and prognosis was examined. In addition, the biological findings were investigated in lung adenocarcinoma cell lines. Based on IHC analysis, we classified patients into high (n = 226, 60.6%) and low SNAT7 expression (n = 147, 39.4%) groups. High SNAT7 expression was substantially associated with male sex, smoking status, high maximum standardised uptake value, advanced pathological stage, and pleural, lymphatic, and vascular invasion, compared to low SNAT7 expression. Patients with high SNAT7 expression demonstrated substantially worse recurrence-free survival (RFS) and overall survival rates. Multivariable analysis revealed that high SNAT7 expression was an independent prognostic factor for RFS. In addition, SLC38A7 knockdown induced a decrease in proliferation with G1 arrest in lung adenocarcinoma cell lines. Our findings demonstrate that SNAT7 plays a pivotal role in promoting tumour malignancy and is significantly associated with poor prognosis in lung adenocarcinoma. These findings suggest that SNAT7 is a potential therapeutic target in lung adenocarcinoma.
- 12 August 2025
- Published by Springer Nature in International Journal of Clinical Oncology
- Vol. 30  (10) , 1972-1981
Clinical and biological impact of SNAT7 in lung adenocarcinoma: implications for prognosis and treatment
Glutamine metabolism plays a crucial role in cancer cell proliferation and modulates the tumour microenvironment. High expression of glutamine transporters is associated with poor prognosis in non-small cell lung cancer. SNAT7, encoded by SLC38A7, facilitates glutamine transport from lysosomes. However, its function and clinical significance in lung adenocarcinoma remain unclear. Immunohistochemistry (IHC) was performed with samples from 373 patients with completely resected lung adenocarcinoma, and the association between SNAT7 expression, clinicopathological features, and prognosis was examined. In addition, the biological findings were investigated in lung adenocarcinoma cell lines. Based on IHC analysis, we classified patients into high (n = 226, 60.6%) and low SNAT7 expression (n = 147, 39.4%) groups. High SNAT7 expression was substantially associated with male sex, smoking status, high maximum standardised uptake value, advanced pathological stage, and pleural, lymphatic, and vascular invasion, compared to low SNAT7 expression. Patients with high SNAT7 expression demonstrated substantially worse recurrence-free survival (RFS) and overall survival rates. Multivariable analysis revealed that high SNAT7 expression was an independent prognostic factor for RFS. In addition, SLC38A7 knockdown induced a decrease in proliferation with G1 arrest in lung adenocarcinoma cell lines. Our findings demonstrate that SNAT7 plays a pivotal role in promoting tumour malignancy and is significantly associated with poor prognosis in lung adenocarcinoma. These findings suggest that SNAT7 is a potential therapeutic target in lung adenocarcinoma.
- 27 April 2025
- Published by Springer Nature in International Journal of Clinical Oncology
- Vol. 30  (7) , 1319-1330
CD155 expression and co-expression with PD-L1 are not associated with poor prognosis in patients with stage II and III lung adenocarcinoma undergoing surgical resection
Background CD155 has been identified as a ligand for T-cell immunoreceptor with Ig and ITIM domains. Herein, we investigated the relationship between the expressions of CD155 and programmed cell death-ligand 1 (PD-L1) and clinical outcomes in patients with surgically resected lung adenocarcinoma. Methods This study included 426 patients diagnosed with pathological stage (pStage) I–III lung adenocarcinoma who underwent surgery at Kyushu University Hospital. The number of tumor cells expressing CD155 and PD-L1 was assessed by immunohistochemistry, and the clinical significance of CD155 expression and CD155/PD-L1 co-expression in prognosis was investigated. Results Among the enrolled cohort, 320 (75.1%), 60 (14.1%), and 46 (10.8%) patients were diagnosed with pStage I, II, and III, respectively. Tissues from 112 patients (26.3%) were classified as having high CD155 expression. Co-expression of CD155 and PD-L1 was observed in 44 patients (10.3%). The High CD155 and CD155/PD-L1 co-expression groups had significantly poorer prognosis in pStage I–III lung adenocarcinoma. However, subgroup analysis revealed that the clinical significance of both CD155 expression and CD155/PD-L1 co-expression differed widely between patients with pStage I and II–III. Multivariate Cox proportional hazards regression analyses showed that high CD155 expression and CD155/PD-L1 co-expression were not independent poor prognostic factors in pStage II-III lung adenocarcinoma. Conclusion Our findings suggest that neither CD155 expression or CD155/PD-L1 co-expression are associated with poor prognosis in pStage II-III lung adenocarcinoma.
- 27 April 2025
- Published by Springer Nature in International Journal of Clinical Oncology
- Vol. 30  (7) , 1319-1330
CD155 expression and co-expression with PD-L1 are not associated with poor prognosis in patients with stage II and III lung adenocarcinoma undergoing surgical resection
Background CD155 has been identified as a ligand for T-cell immunoreceptor with Ig and ITIM domains. Herein, we investigated the relationship between the expressions of CD155 and programmed cell death-ligand 1 (PD-L1) and clinical outcomes in patients with surgically resected lung adenocarcinoma. Methods This study included 426 patients diagnosed with pathological stage (pStage) I–III lung adenocarcinoma who underwent surgery at Kyushu University Hospital. The number of tumor cells expressing CD155 and PD-L1 was assessed by immunohistochemistry, and the clinical significance of CD155 expression and CD155/PD-L1 co-expression in prognosis was investigated. Results Among the enrolled cohort, 320 (75.1%), 60 (14.1%), and 46 (10.8%) patients were diagnosed with pStage I, II, and III, respectively. Tissues from 112 patients (26.3%) were classified as having high CD155 expression. Co-expression of CD155 and PD-L1 was observed in 44 patients (10.3%). The High CD155 and CD155/PD-L1 co-expression groups had significantly poorer prognosis in pStage I–III lung adenocarcinoma. However, subgroup analysis revealed that the clinical significance of both CD155 expression and CD155/PD-L1 co-expression differed widely between patients with pStage I and II–III. Multivariate Cox proportional hazards regression analyses showed that high CD155 expression and CD155/PD-L1 co-expression were not independent poor prognostic factors in pStage II-III lung adenocarcinoma. Conclusion Our findings suggest that neither CD155 expression or CD155/PD-L1 co-expression are associated with poor prognosis in pStage II-III lung adenocarcinoma.
- 17 April 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (8) , 6119-6120
ASO Visual Abstract: DNA Polymerase Delta 2 Activates Cell Cycle in Lung Adenocarcinoma, Leading to High Malignancy and Poor Prognosis
- 17 April 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (8) , 6119-6120
ASO Visual Abstract: DNA Polymerase Delta 2 Activates Cell Cycle in Lung Adenocarcinoma, Leading to High Malignancy and Poor Prognosis
- Open Access
- 14 April 2025
- Published by Springer Nature in Research Square
CD155 expression and co-expression with PD-L1 are not associated with poor prognosis in patients with stage II and III lung adenocarcinoma undergoing surgical resection
Background: CD155 has been identified as a ligand for T-cell immunoreceptor with Ig and ITIM domains. Herein, we investigated the relationship between the expressions of CD155 and programmed cell death-ligand 1 (PD-L1) and clinical outcomes in patients with surgically resected lung adenocarcinoma. Methods: This study included 426 patients diagnosed with pathological stage (pStage) I to III lung adenocarcinoma who underwent surgery at Kyushu University Hospital. The number of tumor cells expressing CD155 and PD-L1 was assessed by immunohistochemistry, and the clinical significance of CD155 expression and CD155/PD-L1 co-expression in prognosis was investigated. Results: Among the enrolled cohort, 320 (75.1%), 60 (14.1%), and 46 (10.8%) patients were diagnosed with pStage I, II, and III, respectively. Tissues from 112 patients (26.3%) were classified as having high CD155 expression. Co-expression of CD155 and PD-L1 was observed in 44 patients (10.3%). The High CD155 and CD155/PD-L1 co-expression groups had significantly poorer prognosis in pStage I to III lung adenocarcinoma. However, subgroup analysis revealed that the clinical significance of both CD155 expression and CD155/PD-L1 co-expression differed widely between patients with pStage I and II-III. Multivariate Cox proportional hazards regression analyses showed that high CD155 expression and CD155/PD-L1 co-expression were not independent poor prognostic factors in pStage II-III lung adenocarcinoma. Conclusion: Our findings suggest that neither CD155 expression or CD155/PD-L1 co-expression are associated with poor prognosis in pStage II-III lung adenocarcinoma.
- 14 April 2025
- Published by Springer Nature in Research Square
Open AccessCD155 expression and co-expression with PD-L1 are not associated with poor prognosis in patients with stage II and III lung adenocarcinoma undergoing surgical resection
Background: CD155 has been identified as a ligand for T-cell immunoreceptor with Ig and ITIM domains. Herein, we investigated the relationship between the expressions of CD155 and programmed cell death-ligand 1 (PD-L1) and clinical outcomes in patients with surgically resected lung adenocarcinoma. Methods: This study included 426 patients diagnosed with pathological stage (pStage) I to III lung adenocarcinoma who underwent surgery at Kyushu University Hospital. The number of tumor cells expressing CD155 and PD-L1 was assessed by immunohistochemistry, and the clinical significance of CD155 expression and CD155/PD-L1 co-expression in prognosis was investigated. Results: Among the enrolled cohort, 320 (75.1%), 60 (14.1%), and 46 (10.8%) patients were diagnosed with pStage I, II, and III, respectively. Tissues from 112 patients (26.3%) were classified as having high CD155 expression. Co-expression of CD155 and PD-L1 was observed in 44 patients (10.3%). The High CD155 and CD155/PD-L1 co-expression groups had significantly poorer prognosis in pStage I to III lung adenocarcinoma. However, subgroup analysis revealed that the clinical significance of both CD155 expression and CD155/PD-L1 co-expression differed widely between patients with pStage I and II-III. Multivariate Cox proportional hazards regression analyses showed that high CD155 expression and CD155/PD-L1 co-expression were not independent poor prognostic factors in pStage II-III lung adenocarcinoma. Conclusion: Our findings suggest that neither CD155 expression or CD155/PD-L1 co-expression are associated with poor prognosis in pStage II-III lung adenocarcinoma.
- 24 March 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (6) , 4524-4525
ASO Author Reflections: Clinical and Biological Significance of POLD2 in Lung Adenocarcinoma
- 24 March 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (6) , 4524-4525
ASO Author Reflections: Clinical and Biological Significance of POLD2 in Lung Adenocarcinoma
- 17 March 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (6) , 4487-4496
DNA Polymerase Delta 2 Activates Cell Cycle in Lung Adenocarcinoma, Leading to High Malignancy and Poor Prognosis
DNA polymerase delta 2 (POLD2) plays a crucial role in DNA repair and replication. POLD2 is upregulated and related to poor prognosis in several types of c
- 17 March 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (6) , 4487-4496
DNA Polymerase Delta 2 Activates Cell Cycle in Lung Adenocarcinoma, Leading to High Malignancy and Poor Prognosis
DNA polymerase delta 2 (POLD2) plays a crucial role in DNA repair and replication. POLD2 is upregulated and related to poor prognosis in several types of c
- 17 March 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (6) , 4207-4208
ASO Visual Abstract: Significance of Tumor Microvasculature in the Tumor Microenvironment in Adenocarcinoma with EGFR Common Mutation
- 17 March 2025
- Published by Springer Nature in Annals of Surgical Oncology
- Vol. 32  (6) , 4207-4208
ASO Visual Abstract: Significance of Tumor Microvasculature in the Tumor Microenvironment in Adenocarcinoma with EGFR Common Mutation
- Open Access
- 22 January 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Identifying Key Variances in Clinical Pathways Associated With Prolonged Hospital Stays Using Machine Learning and ePath Real-World Data: Model Development and Validation Study (Preprint)
BACKGROUND Prolonged hospital stays can lead to inefficiencies in health care delivery and unnecessary consumption of medical resources. OBJECTIVE This study aimed to identify key clinical variances associated with prolonged length of stay (PLOS) in clinical pathways using a machine learning model trained on real-world data from the ePath system. METHODS We analyzed data from 480 patients with lung cancer (age: mean 68.3, SD 11.2 years; n=263, 54.8% men) who underwent video-assisted thoracoscopic surgery at a university hospital between 2019 and 2023. PLOS was defined as a hospital stay exceeding 9 days after video-assisted thoracoscopic surgery. The variables collected between admission and 4 days after surgery were examined, and those that showed a significant association with PLOS in univariate analyses (P<.01) were selected as predictors. Predictive models were developed using sparse linear regression methods (Lasso, ridge, and elastic net) and decision tree ensembles (random forest and extreme gradient boosting). The data were divided into derivation (earlier study period) and testing (later period) cohorts for temporal validation. The model performance was assessed using the area under the receiver operating characteristic curve, Brier score, and calibration plots. Counterfactual analysis was used to identify key clinical factors influencing PLOS. RESULTS A 3D heatmap illustrated the temporal relationships between clinical factors and PLOS based on patient demographics, comorbidities, functional status, surgical details, care processes, medications, and variances recorded from admission to 4 days after surgery. Among the 5 algorithms evaluated, the ridge regression model demonstrated the best performance in terms of both discrimination and calibration. Specifically, it achieved area under the receiver operating characteristic curve values of 0.84 and 0.82 and Brier scores of 0.16 and 0.17 in the derivation and test cohorts, respectively. In the final model, a range of variables, including blood tests, care, patient background, procedures, and clinical variances, were associated with PLOS. Among these, particular emphasis was placed on clinical variances. Counterfactual analysis using the ridge regression model identified 6 key variables strongly linked to PLOS. In order of impact, these were abnormal respiratory sounds, postoperative fever, arrhythmia, impaired ambulation, complications after drain removal, and pulmonary air leaks. CONCLUSIONS A machine learning–based model using ePath data effectively identified critical variances in the clinical pathways associated with PLOS. This automated tool may enhance clinical decision-making and improve patient management.
- 22 January 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessIdentifying Key Variances in Clinical Pathways Associated With Prolonged Hospital Stays Using Machine Learning and ePath Real-World Data: Model Development and Validation Study (Preprint)
BACKGROUND Prolonged hospital stays can lead to inefficiencies in health care delivery and unnecessary consumption of medical resources. OBJECTIVE This study aimed to identify key clinical variances associated with prolonged length of stay (PLOS) in clinical pathways using a machine learning model trained on real-world data from the ePath system. METHODS We analyzed data from 480 patients with lung cancer (age: mean 68.3, SD 11.2 years; n=263, 54.8% men) who underwent video-assisted thoracoscopic surgery at a university hospital between 2019 and 2023. PLOS was defined as a hospital stay exceeding 9 days after video-assisted thoracoscopic surgery. The variables collected between admission and 4 days after surgery were examined, and those that showed a significant association with PLOS in univariate analyses (P<.01) were selected as predictors. Predictive models were developed using sparse linear regression methods (Lasso, ridge, and elastic net) and decision tree ensembles (random forest and extreme gradient boosting). The data were divided into derivation (earlier study period) and testing (later period) cohorts for temporal validation. The model performance was assessed using the area under the receiver operating characteristic curve, Brier score, and calibration plots. Counterfactual analysis was used to identify key clinical factors influencing PLOS. RESULTS A 3D heatmap illustrated the temporal relationships between clinical factors and PLOS based on patient demographics, comorbidities, functional status, surgical details, care processes, medications, and variances recorded from admission to 4 days after surgery. Among the 5 algorithms evaluated, the ridge regression model demonstrated the best performance in terms of both discrimination and calibration. Specifically, it achieved area under the receiver operating characteristic curve values of 0.84 and 0.82 and Brier scores of 0.16 and 0.17 in the derivation and test cohorts, respectively. In the final model, a range of variables, including blood tests, care, patient background, procedures, and clinical variances, were associated with PLOS. Among these, particular emphasis was placed on clinical variances. Counterfactual analysis using the ridge regression model identified 6 key variables strongly linked to PLOS. In order of impact, these were abnormal respiratory sounds, postoperative fever, arrhythmia, impaired ambulation, complications after drain removal, and pulmonary air leaks. CONCLUSIONS A machine learning–based model using ePath data effectively identified critical variances in the clinical pathways associated with PLOS. This automated tool may enhance clinical decision-making and improve patient management.