METHOD: Observational study of patients admitted to two geriatric clinics in the Stockholm Region of Sweden during the first wave of the COVID-19 pandemic from March 1st to June 15th 2020. The difference in incidence, risk factors and adverse outcomes for AKI between patients with or without COVID-19 were examined. Odds ratios (ORs) for AKI were obtained from logistic regressions. The hazard ratios (HRs) for the risk of in-hospital death were calculated from Cox proportional hazard regression models. RESULTS: We analyzed 316 older patients hospitalized for COVID-19 and 876 patients for non-COVID-19 diagnoses. The mean age was 8369 years, 57% were women, and mean baseline kidney function as depicted by estimated glomerular filtration rate (eGFR) was 62623 ml/min/1.73m 2 . AKI occurred in 92 (29%) of patients with COVID-19 vs. 159 (18%) without COVID-19. The severity of AKI was significantly worse in patients with COVID-19 compared with non-COVID patients. The odds for developing AKI were higher in patients with COVID-19 (adjusted OR, 1.70; 95% CI, 1.04-2.76), low baseline kidney function [4.19 (2.48-7.05), for eGFR 30  <60 ml/min/1.73m 2 , and 20.3 (9.95-41.3) for eGFR <30ml/min/1.73m 2 ], and higher C-reactive protein (CRP) level (OR 1.81(1.11-2.95)). The risk of in-hospital death was highest in patients with COVID-19 and AKI [adjusted HR 23.5, 95% CI (8.75-63.0)], followed by COVID-19 without AKI [9.10 (3.52-23.6)] and by patients without COVID-19 and with AKI [6.38 (2.28-17.9)] after adjusting for patient demographics, vital signs, baseline kidney function and medications and using non-COVID patients with no AKI as reference. CONCLUSION: Geriatric patients hospitalized with COVID-19 had a higher incidence of AKI compared with patients hospitalized with other diagnoses. AKI and COVID-19 were associated with in-hospital death. Optimal management of AKI may improve the outcome of COVID-19 in geriatric patients. MO359 ASSOCIATIONS BETWEEN RENAL FUNCTION TRAJECTORIES AFTER 3 MONTHS ACUTE KIDNEY INJURY AND LONG-TERM RENAL OUTCOMES Chien-Ning Hsu 1 , You-Lin Tain 2 1 Kaohsiung Chang Gung Memorial Hospital, Deaprtment of Pharmacy, Kaohsiung, Taiwan, R.O.C. and 2 Kaohsiung Chang Gung Memorial Hospital, Department of Pediatrics, Kaohsiung, Taiwan, R.O.C. BACKGROUND AND AIMS: Renal function recovery after acute kidney injury (AKI) is associated with patient outcomes. The study objectives were to assess the patterns of AKI recovery within 6 months following discharge for AKI and subsequent incidence of chronic dialysis. METHOD: A retrospective cohort of 234,867 hospitalized adult patients was examined for AKI between January 1, 2010, and December 31, 2017 in the largest healthcare delivery system in Taiwan. Renal function recovery at 3- and 6-month post discharge, incident chronic kidney disease and chronic dialysis initiation were analyzed over 7 years of follow-up. Renal recovery was defined by < 1.5 baseline SCr (prior to the hospitalization). Independent associations between renal function recovery patterns and renal outcomes was assessed by Cox proportional hazard model controlling for potential confounders, and subdistribution hazard ratio (SHR) with [95% CI] was analysed for competing risk of early death. RESULTS: Among 3 months AKI survivors (n=24,132), 14.28% (n=3,430) did not recovery back to baseline, and 16% of recovery did not sustain. Three distinct renal function recovery continuums at 6 months post hospital discharge were: persistent non-recovery (10.18%), non-recovery (14.33%), and recovery (75.5%). Comparing to survivors without AKI (n=50,387), the impact of renal recovery continuum on chronic dialysis initiation varied by patient’s baseline renal disease (SHR was 2.82 [95%CI, 2.42-3.28] in CKD, and 0.8 [95%CI, 0.27-2.38] for non-CKD. Persistent non-recovery was significantly associated with a greater increased risk of chronic dialysis than non- recovery in any patients with AKI. Comparing to patients with sustained AKI recovery, risk of CKD onset increased 5-fold in persistent non-recovery and 3-fold risk in non- recovery. CONCLUSION: The continuum of AKI recovery post 6 months is associated with increased risk of chronic dialysis, particularly in patients with baseline CKD. These study results suggested that patients ever with AKI should receive close renal function monitoring for post-discharge management. MO360 MACHINE LEARNING MODELS FOR PREDICTING ACUTE KIDNEY INJURY: A SYSTEMATIC REVIEW Iacopo Vagliano 1 , Nicholas Chesnaye 1,2,3 , Jan Hendrik Leopold 1 , Kitty J Jager 1,2,3 , Ameen Abu Hanna 1 , Martijn C. Schut 1 1 Amsterdam UMC, University of Amsterdam, Dept. of Medical Informatics, Amsterdam, The Netherlands, 2 ERA-EDTA Registry, Amsterdam, The Netherlands and 3 Amsterdam Public Health research Institute, Amsterdam, The Netherlands BACKGROUND AND AIMS: Acute kidney injury (AKI) has a substantial impact on global disease burden of Chronic Kidney Disease. To assist physicians with the timely diagnosis of AKI, several prognostic models have been developed to improve early recognition across various patient populations with varying degrees of predictive performance. In the prediction of AKI, machine learning (ML) techniques have been demonstrated to improve on the predictive ability of existing models that rely on more conventional statistical methods. ML is a broad term which refers to various types of models: Parametric models, such as linear or logistic regression use a pre-specified model form which is believed to fit the data, and its parameters are estimated. Non-parametric models, such as decision trees, random forests, and neural networks may have varying complexity (e.g. the depth of a classification tree model) based on the data. Deep learning neural network models exploit temporal or spatial arrangements in the data to deal with complex predictors. Given the rapid growth and development of ML methods and models for AKI prediction over the past years, in this systematic review, we aim to appraise the current state-of-the-art regarding ML models for the prediction of AKI. To this end, we focus on model performance, model development methods, model evaluation, and methodological limitations. METHOD: We searched the PubMed and ArXiv digital libraries, and selected studies that develop or validate an AKI-related multivariable ML prediction model. We extracted data using a data extraction form based on the TRIPOD (transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) and CHARMS (critical appraisal and data extraction for systematic reviews of prediction modelling studies) checklists. RESULTS: Overall, 2,875 titles were screened and thirty-four studies were included. Of those, thirteen studies focussed on intensive care, for which the US derived MIMIC dataset was commonly used; thirty-one studies both developed and validated a model; twenty-one studies used single-centre data. Non-parametric ML methods were used more often than regression and deep learning. Random forests was the most popular method, and often performed best in model comparisons. Deep learning was typically used (and also effective) when complex features were included (e.g., with text or time series). Internal validation was often applied, and the performance of ML models was usually compared against logistic regression. However, the simple training/test split was often used, which does not account for the variability of the training and test samples. Calibration, external validation, and interpretability of results were rarely considered. Comparisons of model performance against medical scores or clinicians were also rare. Reproducibility was limited, as data and code were usually unavailable. CONCLUSION: There is an increasing number of ML models for AKI, which are mostly developed in the intensive care environment largely due to the availability of the MIMIC dataset. Most studies are single-centre, and lack a prospective design. More complex models based on deep learning are emerging, with the potential to improve predictions for complex data, such as time-series, but with the disadvantage of being less interpretable. Future studies should pay attention to using calibration measures, external validation, and on improving model interpretability, in order to improve uptake in clinical practice. Finally, sharing data and code could improve reproducibility of study findings. MO361 INCIDENCE AND RISK FACTORS OF INFECTION AFTER AN EPISODE OF ACUTE KIDNEY INJURY DURING HOSPITALIZATION Ana Sanchez 1 , Alicia Cabrera 1 , Laura Salanova Villanueva 1 , Patricia Mu~ noz Ramos 1 , Pablo Ruano 1 , Borja Quiroga 1 1 HOSPITAL DE LA PRINCESA, NEPHROLOGY, MADRID, Spain BACKGROUND AND AIMS: Acute kidney injury (AKI) is a major risk factor for development and progression to chronic kidney disease (CKD). The aim of the present study is to assess the incidence of infections after an admission for AKI. METHOD: In this retrospective study all patients who developed AKI during hospitalization and were discharged from 2013 to 2014 were included. Factors associated to infections were evaluated. The mean follow-up after discharge was 39630 months. RESULTS: We included 1255 patients with a mean age of 75613 years, of which 692 (55%) were men. At baseline, 944 (75%) patients presented with hypertension, 379 (30%) with diabetes, 560 (44%) with hypercholesterolemia and 543 (43%) with CKD. Mean baseline creatinine was 1,361,8 mg/dl (glomerular filtration rate [eGFR] estimated by CKD-EPI was 55625 ml/min/1,73m2). The peak level of creatinine reached during AKI was 2,4761,97 mg/dl (eGFR 30618 ml/min/1,73m2). At discharge, creatinine was 1,62 mg/dL and eGFR 53627 ml/min/1,73m2. Seven hundred and seventy-three (62%) patients presented an eGFR inferior to 60 ml/min/ 1,73m2. During follow-up, 681(54%) patients presented an infectious event. Urinary tract infection was the most frequent infection (286 patients, 23%) followed by respiratory infection (214 patients, 17%). Factors associated with infection were age (p<0,001), hypertension (p=0,03), atrial fibrillation (p=0,014), functional dependence measured by Barthel index (p=0,03), previous diagnosis of CKD (p=0,01), baseline eGFR (p>0,001) and eGFR at discharge (p=0,002). Survival analysis using Kaplan-Meier demonstrated an existing association between eGFR inferior to 60 ml/min/1,73m2 and infections (LogRank 12,2, p<0,001, figure 1). Abstracts Nephrology Dialysis Transplantation i250 | Abstracts Downloaded from https://academic.oup.com/ndt/article/36/Supplement_1/gfab082.0015/6288975 by guest on 29 April 2023