Development and external validation of machine learning models for predicting glucose trajectories after continuous intravenous insulin therapy in hyperglycemic crises
Entwicklung und externe Validierung von Machine-Learning-Modellen zur Vorhersage von Glukoseverläufen nach kontinuierlicher intravenöser Insulintherapie bei hyperglykämischen Krisen
Journal of Diabetes Investigation · 11.11.2026
DOI: 10.1111/jdi.70489 ↗Auf einen Blick
Bei 191 Patient:innen mit hyperglykämischen Krisen sagte ein direktes Modell die erneute Glukosemessung nach CII stabil voraus. Ein ΔGlu-Modell ließ sich zwischen Zentren schlechter verallgemeinern.
Fragestellung
Lassen sich nach Beginn einer kontinuierlichen intravenösen Insulininfusion Glukosewerte und Glukoseabfälle mit Machine-Learning-Modellen vorhersagen?
Methode
Entwicklung und externe Validierung von 2 Gradient-Boosted-Decision-Tree-Modellen anhand von 574 post-CII-Blutzuckermessungen von 191 Patient:innen aus 5 japanischen Einrichtungen; externe Kohorte mit 29 Patient:innen.
Ergebnisse
Das 8-Feature-Modell zur direkten Vorhersage der Re-Test-Glukose zeigte interne Validierung R 2 = 0,853, MAE = 76,9 mg/dL, RMSE = 106,0 mg/dL; fünffach patientenbezogen R 2 = 0,803 ± 0,080; extern R 2 = 0,802, MAE = 78,0 mg/dL, RMSE = 101,0 mg/dL. Das ΔGlu-Modell war extern limitiert.
Einschränkungen
Die externe Generalisierbarkeit des ΔGlu-Modells war begrenzt; weitere prospektive Validierung wurde als notwendig genannt.
Schlussfolgerung der Autor:innen
Das direkte Re-Test-Glukosemodell zeigte stabile Leistung in interner und externer Validierung, während ΔGlu zwischen Institutionen schwerer zu verallgemeinern war.
Kernaussagen
- 574 post-CII-Blutzuckermessungen aus 191 Patient:innen
- Direktes Modell: extern R 2 = 0,802; MAE = 78,0 mg/dL
- ΔGlu-Modell: externe Vorhersage begrenzt
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift Journal of Diabetes Investigation, online erschienen am 11.11.2026. Verfasst von Yuichiro Iwamoto und 17 weiteren Autor:innen. Beteiligte Einrichtung: Department of Diabetes, Endocrinology and Metabolism, Kawasaki Medical School, Kurashiki, Japan. Die Arbeit nennt 27 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Background and aims Predicting short‐term blood glucose changes after initiation of continuous intravenous insulin infusion (CII) in hyperglycemic crises remains challenging because glucose trajectories are influenced by both patient‐specific metabolic severity and dynamically adjusted treatment. This study aimed to develop and externally validate machine learning models for estimating retest glucose levels and glucose‐lowering responses after CII initiation.
Methods
We analyzed 574 post‐CII blood glucose measurements from 191 patients with hyperglycemic crises across five Japanese institutions. An independent external validation cohort included 29 patients from another institution. Two complementary models were developed using gradient‐boosted decision tree algorithms: a direct retest glucose prediction model and a ΔGlu treatment‐response model. We evaluated model performance using internal validation, patient‐level fivefold cross‐validation, and external validation. ΔGlu was defined as the decrease in blood glucose between two consecutive reassessment points.
Results
The eight‐feature direct retest glucose model showed stable predictive performance in internal validation (R 2 = 0.853, MAE = 76.9 mg/dL, RMSE = 106.0 mg/dL), patient‐level fivefold cross‐validation (R 2 = 0.803 ± 0.080), and external validation (R 2 = 0.802, MAE = 78.0 mg/dL, RMSE = 101.0 mg/dL). Although the eight‐feature ΔGlu treatment‐response model showed moderate internal performance, its external generalizability for ΔGlu prediction was limited.
Conclusions
The direct retest glucose model showed stable performance across internal and external validation, whereas ΔGlu itself was more difficult to generalize across institutions. These complementary models may support objective monitoring of short‐term glucose trajectories during CII for hyperglycemic crises, although further prospective validation is warranted.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Diabetes, Endocrinology and Metabolism, Kawasaki Medical School, Kurashiki, Japan.
- Department of Pediatrics, Kindai University Faculty of Medicine, Osakasayama, Japan.
- Department of General Internal Medicine 1, Kawasaki Medical School, Okayama, Japan.
- Department of Internal Medicine, Murakami Memorial Hospital, Onomichi, Japan.
- Department of Internal Medicine, Kurashiki Sweet Hospital, Kurashiki, Japan.
- Department of Diabetes and Metabolism, Kurobe City Hospital, Kurobe, Japan.
ORCID-Profile: Yuichiro Iwamoto, Tomohiko Kimura, Yuichi Morimoto, Hideyuki Iwamoto, Yoshiro Fushimi, Takafumi Osaka, Masashi Shimoda, Shuhei Nakanishi, Kohei Kaku, Hideaki Kaneto
Zitieren
APA
Iwamoto, Y., Kimura, T., Morimoto, Y., Iwamoto, H., Fushimi, Y., Sanada, J., Isobe, H., Tatsumi, F., Kimura, Y., Kawasaki, F., Yamabe, M., Matsuki, M., Mohri, K., Osaka, T., Shimoda, M., Nakanishi, S., Kaku, K., & Kaneto, H. (2026). Development and external validation of machine learning models for predicting glucose trajectories after continuous intravenous insulin therapy in hyperglycemic crises. Journal of Diabetes Investigation. https://doi.org/10.1111/jdi.70489
Vancouver
Iwamoto Y, Kimura T, Morimoto Y, Iwamoto H, Fushimi Y, Sanada J, et al. Development and external validation of machine learning models for predicting glucose trajectories after continuous intravenous insulin therapy in hyperglycemic crises. Journal of Diabetes Investigation. 2026. doi:10.1111/jdi.70489
BibTeX
@article{Iwamoto2026,
author = {Iwamoto, Yuichiro and Kimura, Tomohiko and Morimoto, Yuichi and Iwamoto, Hideyuki and Fushimi, Yoshiro and Sanada, Junpei and Isobe, Hayato and Tatsumi, Fuminori and Kimura, Yukiko and Kawasaki, Fumiko and Yamabe, Mizuho and Matsuki, Michihiro and Mohri, Kensuke and Osaka, Takafumi and Shimoda, Masashi and Nakanishi, Shuhei and Kaku, Kohei and Kaneto, Hideaki},
title = {Development and external validation of machine learning models for predicting glucose trajectories after continuous intravenous insulin therapy in hyperglycemic crises},
journal = {Journal of Diabetes Investigation},
year = {2026},
doi = {10.1111/jdi.70489}
}
EUVIMED