Abstract
Background and Aims
End Stage Kidney Disease (ESKD) patients have high hospitalization rates. We have developed and deployed a predictive model to identify in-centre haemodialysis (HD) patients at an increased risk for all-cause hospitalization within the next 12 months. The model was used in a pilot called Dialysis Hospitalization Reduction Program (DHRP) to identify patients predicted to be at risk of >=6 hospital admissions and provide additional interdisciplinary team intervention. We investigated the impact of the DHRP on hospitalization rates in HD patients.
Method
We used data from 45 clinics in South Alabama/Florida Panhandle regions of the United States who participated in DHRP pilot starting January 2016. The predictive model used more than 200 variables to stratify patients as high risk (>=6 admissions), medium high risk (>=3 admissions) and medium low risk (>=1 admission) and low risk (<1 admission). For patients identified at high risk of hospitalization, social workers assessed psychosocial barriers and offered additional psychosocial intervention to target those barriers. Dietitians utilized a high risk assessment looking at weight, nutrition, and access to food and supplements. Resident nurses assessed high risk patients focusing on anaemia, adequacy, access, blood pressure, fluid management, prior hospitalizations, glycaemic control and risk of skin ulcers and blood stream infection Data from patients at the participating clinics was collected and yearly hospital admission and day rates per patient year were calculated 2 years prior to (2014, 2015) and 3 years after (2016-2018) pilot start. Comparison clinics were chosen from neighbouring regions in South and North Florida (43 and 45 clinics respectively).
Results
Over the study period the number of patients ranged from 4661 to 5672 in the DHRP pilot clinics, 5416 to 5947 in South Florida control clinics, and 6087 to 7596 in North Florida control clinics. Hospitalization rates in pilot clinics during the first year of the DHRP remined similar to the rates during the two years preceding the pilot start. In the second and third years of the DHRP, pilot clinics showed reductions in hospital admission and day rates. At control clinics in both regions the hospital admissions and day rates showed increasing trends while DHRP clinics showed decreasing trends over the study period (Figures 1a and 1b).
Conclusion
These findings suggest predictive model risk directed interdisciplinary team interventions associate with lower hospitalization rates in HD patients, compared to controls. Further studies are needed to confirm these results.
MO031EFFECTS OF PREDICTIVE MODELLING RISK DIRECTED INTERVENTIONS ON HOSPITALIZATION RATES IN HAEMODIALYSIS PATIENTS
