CONCLUSIONS: When using drugs mentioned above on hemodialysis patients, the likelihood of mental change should be considered. Hence, we recommend frequent check-ups upon administering drugs and discontinuing unnecessary drugs SP666 MACHINE LEARNING IN PREDICTION OF VULNERABLE OR RESILIENT END-STAGE RENAL DISEASE PATIENTS Jo~ ao Pedro Pedroso 1 , Maria Jo~ ao Valente 2 , Susana Rocha 2 , Susana Coimbra 3 , Cristina Catarino 2 , Petronila Rocha-Pereira 2 , Elsa Bronze-Da-Rocha 2 , M arcia Carvalho 4 , Jose ´ Gerardo Oliveira 5 , Jose ´ Madureira 6 , Jo~ ao Carlos Fernandes 6 , Maria Do Sameiro-Faria 7 , Vasco Miranda 8 , Lu  ıs Belo 2 , Alice Santos-Silva 2 1 Universidade do Porto, Porto, Portugal, 2 UCIBIO, REQUIMTE, Porto, Portugal, 3 CESPU, Gandra-Paredes, Portugal, 4 Universidade Fernando Pessoa, Porto, Portugal, 5 CHP, Porto, Portugal, 6 NefroServe, Barcelos, Portugal, 6 NefroServe, Viana do Castelo, Portugal, 7 CHF, Felgueiras, Portugal and 8 CHG, Gondomar, Portugal INTRODUCTION: End-stage renal disease (ESRD) patients on dialysis have higher mortality rate than general population. This higher risk has been associated with cardiovascular risk factors, uremia-related factors and with associated complications. In this work we propose an analysis of analytical and clinical data, using machine learning tools, to identify predictive biomarkers of mortality risk that could be used in the clinical setting to identify vulnerable ESRD patients. METHODS: The study included 497 Portuguese ESRD patients on dialysis. Demographic data, clinical history, dialysis, analytical data including cardiovascular risk factors, uremia-related factors and co-morbidities, were recorded at starting of the study. A 1-year follow-up was performed, identifying cases of death; patients treated by renal transplant or transferred for peritoneal dialysis along the follow-up, were excluded of the study. Machine learning tools were used in the analysis of data, including more than fifty parameters associated to each patient; each of them is classified as alive or deceased at the end of 1-year follow-up. We are in a supervised learning setup, where the learning task is to classify an unlabeled observation into one of these classes. We used the random forest (RF) classifier. An important step in the analysis concerns the selection of appropriate properties - features, in machine learning language - that should be used as explanatory variables. RFs provide a way for doing so, using the relative rank of each feature in decision nodes in a tree. After selecting the most important features, we have trained a RF classifier and assessed its quality by means of leave one out cross validation. RESULTS: Along the follow-up period, 56 (11.3%) patients died. The most important features for predicting this, determined by RFs, were interleukin-6, albumin, D-dimers, dose of erythropoietin stimulating agents, transferrin, creatinine, age, body mass index, hepcidin, C-reactive protein, ultrafiltration volume, tissue plasminogen activator, in this order of importance. RFs proved to be effective for learning, with accuracy estimated with cross validation above 88%, though with low sensitivity. CONCLUSIONS: The results obtained by using machine learning techniques are encouraging; however, further studies with larger number of ESRD patients are necessary to take full advantage of these methods. This work was supported by the Applied Molecular Biosciences Unit (UCIBIO), which is financed by national funds from FCT/MCTES (UID/MULTI/04378/2019) and North Portugal Regional Coordination and Development Commission (CCDR-N)/ NORTE2020/Portugal 2020 (Norte-01-0145-FEDER-000024). SP667 QUALITY OF LIFE IN PATIENTS ON THRICE WEEKLY DIALYSIS VERSUS TWICE WEEKLY DIALYSIS USING SF 36 SCORE G.k Prakash 1 , BADARINATH VELLABOINA 1 1 MANIPAL HOSPITALS, Bengaluru, India INTRODUCTION: Maintenance hemodialysis is a well-recognized modality of treating patients having end stage renal disease. In recent years, observational studies of twice weekly HD in Taiwan and China have shown a possible benefit of the slower decline of renal function and acceptable nutritional status. However, there is a lacuna of knowledge of the unique clinical aspects of patients dialyzed twice weekly in India. We attempted to study this aspect by comparing differences between patients dialyzed twice and thrice weekly. The primary objective of study is to compare quality of life of hemodialysis patients undergoing twice weekly hemodialysis and thrice weekly hemodialysis – using SF 36 score. The secondary objectives are to compare mortality rate, infection rate and hospitalization rate of hemodialysis patients undergoing twice weekly hemodialysis and thrice weekly hemodialysis. METHODS: Fifty stable outpatients on dialysis (25 patients on twice weekly hemodialysis and 25 patients on thrice weekly hemodialysis) above age of 18 years and dialysis vintage of more than 3 months were followed up for 1 year. Residual urine output, Adequacy of hemodialysis, Nutritional and CKD MBD parameters were analysed at baseline, 6 months and at 1 year. Quality of life using SF 36 score was done at baseline, at 6 months and at 1 year. Mortality rate, hospitalisation rate and infection rate between two groups were analysed at 1 year. RESULTS: The mean age of patients in both groups is 57 years in our study. In 50 patients on hemodialysis, 68% were males and 32% were females. Females are more in twice weekly hemodialysis compared to thrice weekly hemodialysis, but the difference is not statistically significant. The most common cause of ESRD is diabetic nephropathy (50%). There was no statistically significant difference in comorbid conditions (IHD, Diabetes mellitus, Hypertension) between two groups, even though all comorbid conditions are less in twice weekly hemodialysis group compared to thrice weekly hemodialysis group. There is no significant difference between insurance coverage, BMI, IDWG and vintage on dialysis between two groups of patients. Residual urine output volume was statistically significant between two groups at baseline, 6 months and 12 months. Parameters for adequacy of dialysis, nutritional and CKD MBD parameters were not statistically significant between two groups. At baseline, 6th month and 12th month of study there is no statistically significant difference between physical and mental components of SF 36 scores in our study even though scores of both physical and mental components are lower in twice weekly hemodialysis group compared to thrice weekly hemodialysis group. There is no statistically significant difference in access related issues, CVS related issues and hospitalisation rate between both groups. Infections are more significant in thrice weekly hemodialysis group compared to twice weekly hemodialysis group. Mortality is more in twice weekly hemodialysis group than thrice weekly hemodialysis group, but difference is not statistically significant. CONCLUSIONS: Twice weekly hemodialysis dosenot compromise health related quality of life compared to thrice weekly hemodialysis. Access related complications, cardiovascular events, infection, hospitalisation and mortality rates in twice weekly hemodialysis patients with significant residual kidney function is on par with thrice weekly hemodialysis patients. SP668 HEMODIAFILTRATION IMPROVED THE OUTCOMES IN DIALYSIS PATIENTS COMPARED TO CONVENTIONAL HEMODIALYSIS - SINGLE CENTER EXPERIENCE Nikolina Smokovska 1 , Risto Grozdanovski 1 , Angel Oncevski 1 , Sasho Gelev 1 , Elena Babalj Banskolieva 1 , Gjurovska Katerina Spaseska 1 , Marko Ilievski 1 , Goce Spasovski 2 1 Special hospital for nephrology and dialysis DIAMED, Skopje, Republic of Macedonia and 2 University of Skopje Medical Faculty, Skopje, Republic of Macedonia INTRODUCTION: For decades, end-stage renal disease (ESRD) patients had the chance only to undergo chronic intermittent hemodialysis (HD). But recently, successful efforts were made to improve the clearance of uremic toxins and decrease the high risk of morbidity and mortality. The development of online hemodiafiltration (OL-HDF) has resulted in markedly enhanced clearance of middle to large toxin’s molecules better than high-flux HD. Few clinical trials pointed out the benefits of OL- HDF in patients overall survival. Despite that OL-HDF can also be considered as cost- effective compared to high-flux HD, but it is unfortunately not widely available for every patient. Primary objective of our study was to show the superiority of OL-HDF compared to high-flux HD in ESRD patients on routine renal replacement treatment. Secondary objective was to identify which variables are showing improvements in OL- HDF vs high-flux HD. METHODS: In this retrospective, intra-patient comparison, single arm study, 31 dialysis patients were studied during 2 years’ dialysis treatment as follows: 12 months on high-flux HD, then 12 months on OL-HDF. Demographic, clinical and laboratory variables in study cohort were collected from both treatment regimens. Further, statistical analysis and intra-patient comparisons were made. RESULTS: Our study showed that switching from high-flux HD to OL-HDF treatment regimen improved several hemodialysis outcomes. In 74.2% of patients eKT/V was significantly improved (p=0.006); which was in line with the significant increase of blood flow. The levels of phosphorus were in reference values in 64.5% of the patients but in 35.5% it significantly changed, from which in 54.5% decreased (p=0.009). The usage of anti-coagulants was significantly decreased in 42% of patients (p=0.006), in 9.7% significantly increased and in the rest of the cohort there were not significant changes. Overall consumption of EPO in the study cohort was decreased, in 32.3% was significantly decreased (p=0.017), in 45.2% remained similar in the two regimens and in only 22.5% of patient its need was significantly increased. CONCLUSIONS: This study showed that online hemodiafiltration is superior treatment regimen compared to high-flux HD in patients on chronic hemodialysis. Significant improvements were noticed in the most important HD treatment outcomes who will potentially result in the improvement of quality of life. Further long-term analysis is needed to show the benefits in decreasing the risk of morbidity and mortality in OL-HDF patients. SP669 THE RELATIONSHIP BETWEEN SERUM BICARBONATE, PH LEVEL AND SLEEP QUALITY IN HEMODIALYSIS PATIENTS: A CROSS-SECTIONAL STUDY FROM TURKEY Yavuz Yigit 1 , Erkan Sengul 1 , Aysun Sengul 1 , Didem Eroglu 1 , Zeynep Ozturk 1 1 Derince Education and Research Hospital, _ Izmit, Turkey and 1 Derince Education and Research Hospital, Izmit, Turkey INTRODUCTION: We investigate the relationship of blood pH and bicarbonate levels with sleep disorders in patients with end-stage renal disease. i580 | Abstracts Abstracts Nephrology Dialysis Transplantation Downloaded from https://academic.oup.com/ndt/article/34/Supplement_1/gfz103.SP666/5515454 by guest on 20 September 2023