IMPROVED PREDICTION OF HOSPITALIZATION AND MORTALITY RISK BY ADDITION OF PATIENT REPORTED OUTCOME MEASURES (PROM) INTO RISK ALGORITHMS AMONG DIALYSIS PATIENTS

Abstract BACKGROUND AND AIMS Health-Related Quality of Life (HRQOL) is a key outcome in dialysis as it stands at the vertex of the outcomes pyramid. The Kidney Disease Quality of Life (KDQOL™-36) questionnaire has been specifically developed for patients affected by kidney disease and it combines both generic and disease-specific scales capturing unique facets of patients’ adaptation to the physical and mental burden of renal disease and dialysis treatment. We sought to evaluate the association between HRQOL scores and the prevalence of adverse outcomes such as mortality and hospitalization events among prevalent dialysis patients. METHOD This is a retrospective, observational analysis of patient-reported outcomes measures (PROMs) collected in NephroCare clinics in Spain in 2019. Clinical and socio-demographic patients’ information were abstracted from EuCliD® (the European Clinical Database) and merged to PROMs HRQOL was assessed by KDQOL™-36 questionnaire and its scores for physical and mental composite (PCS and MCS) as well as for specific-disease dimensions (effects, burden and symptoms) were computed. A renal factor score (RFS) was calculated through principal component analysis to summarize the score of these latter. Mortality and hospitalizations events in the year following the survey administration date were collected and the possible association of these events with HRQOL scores was then explored. Cox and logistic regression models evaluated the relationship between KDQOL™-36 scores and survival and hospitalization events, respectively. A hierarchical model building strategy was adopted to test the incremental discrimination provided by HRQOL scores. First, two outcome risk scores (ORS) were constructed to predict the likelihood of hospitalization and mortality within 12 months given traditional clinical and socio-demographic confounders. Second, KDQOL™-36 scores were added to ORS in the fully adjusted model. We assessed both individual ORS-adjusted coefficient estimates for the KDQOL-36 scales and the KDQOL-36 associated increment in model ROC-AUC. RESULTS We enrolled 1838 prevalent dialysis patients in the present analysis with a response rate of 80.6% (study flowchart is reported in Figure 1). Mean age was 68.8 ± 14.4, 60% were men, 65% presented arteriovenous fistula as vascular access and 66% were on HDF modality. We observed a graded relationship between almost all KDQOL™-36 scores and adverse patient-centred outcomes (average KDQOL-36™ scores are reported in Figure 2). These associations were robust to adjustment for potential confounders. ROC-AUC describe the discrimination accuracy of models predicting hospitalization and mortality events, where both improved after including KDQOL-36™ scores beyond traditional clinical and socio-demographic factors. CONCLUSION We observed a monotonic, clinically important association denoting increased hospitalization and mortality risk among patients with worse KDQOL-36™ scores. The association was robust to adjustment for potential confounders. The KDQOL-36™ score captures unique facets of health, which are not sufficiently accounted for by traditional clinical and socio-demographic correlates of hospitalization and survival. Careful assessment of HRQOL in routine practice can uncover important health trajectories otherwise overlooked by standard clinical care.