TY - JOUR
T1 - Long-short-term memory machine learning of longitudinal clinical data accurately predicts acute kidney injury onset in COVID-19
T2 - a two-center study
AU - Lu, Justin Y.
AU - Zhu, Joanna
AU - Zhu, Jocelyn
AU - Duong, Tim Q.
N1 - Publisher Copyright:
© 2022 The Author(s)
PY - 2022/9
Y1 - 2022/9
N2 - Objectives: This study used the long-short-term memory (LSTM) artificial intelligence method to model multiple time points of clinical laboratory data, along with demographics and comorbidities, to predict hospital-acquired acute kidney injury (AKI) onset in patients with COVID-19. Methods: Montefiore Health System data consisted of 1982 AKI and 2857 non-AKI (NAKI) hospitalized patients with COVID-19, and Stony Brook Hospital validation data consisted of 308 AKI and 721 NAKI hospitalized patients with COVID-19. Demographic, comorbidities, and longitudinal (3 days before AKI onset) laboratory tests were analyzed. LSTM was used to predict AKI with fivefold cross-validation (80%/20% for training/validation). Results: The top predictors of AKI onset were glomerular filtration rate, lactate dehydrogenase, alanine aminotransferase, aspartate aminotransferase, and C-reactive protein. Longitudinal data yielded marked improvement in prediction accuracy over individual time points. The inclusion of comorbidities and demographics further improves prediction accuracy. The best model yielded an area under the curve, accuracy, sensitivity, and specificity to be 0.965 ± 0.003, 89.57 ± 1.64%, 0.95 ± 0.03, and 0.84 ± 0.05, respectively, for the Montefiore validation dataset, and 0.86 ± 0.01, 83.66 ± 2.53%, 0.66 ± 0.10, 0.89 ± 0.03, respectively, for the Stony Brook Hospital validation dataset. Conclusion: LSTM model of longitudinal clinical data accurately predicted AKI onset in patients with COVID-19. This approach could help heighten awareness of AKI complications and identify patients for early interventions to prevent long-term renal complications.
AB - Objectives: This study used the long-short-term memory (LSTM) artificial intelligence method to model multiple time points of clinical laboratory data, along with demographics and comorbidities, to predict hospital-acquired acute kidney injury (AKI) onset in patients with COVID-19. Methods: Montefiore Health System data consisted of 1982 AKI and 2857 non-AKI (NAKI) hospitalized patients with COVID-19, and Stony Brook Hospital validation data consisted of 308 AKI and 721 NAKI hospitalized patients with COVID-19. Demographic, comorbidities, and longitudinal (3 days before AKI onset) laboratory tests were analyzed. LSTM was used to predict AKI with fivefold cross-validation (80%/20% for training/validation). Results: The top predictors of AKI onset were glomerular filtration rate, lactate dehydrogenase, alanine aminotransferase, aspartate aminotransferase, and C-reactive protein. Longitudinal data yielded marked improvement in prediction accuracy over individual time points. The inclusion of comorbidities and demographics further improves prediction accuracy. The best model yielded an area under the curve, accuracy, sensitivity, and specificity to be 0.965 ± 0.003, 89.57 ± 1.64%, 0.95 ± 0.03, and 0.84 ± 0.05, respectively, for the Montefiore validation dataset, and 0.86 ± 0.01, 83.66 ± 2.53%, 0.66 ± 0.10, 0.89 ± 0.03, respectively, for the Stony Brook Hospital validation dataset. Conclusion: LSTM model of longitudinal clinical data accurately predicted AKI onset in patients with COVID-19. This approach could help heighten awareness of AKI complications and identify patients for early interventions to prevent long-term renal complications.
KW - Acute kidney injury
KW - Artificial intelligence
KW - D-dimer
KW - Long-short-term memory
KW - Multiorgan failure
KW - SARS-CoV-2
UR - http://www.scopus.com/inward/record.url?scp=85135368125&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85135368125&partnerID=8YFLogxK
U2 - 10.1016/j.ijid.2022.07.034
DO - 10.1016/j.ijid.2022.07.034
M3 - Article
C2 - 35872094
AN - SCOPUS:85135368125
SN - 1201-9712
VL - 122
SP - 802
EP - 810
JO - International Journal of Infectious Diseases
JF - International Journal of Infectious Diseases
ER -