Chronic kidney disease (CKD) is both a consequence of progressive chronic disease and an important comorbidity with prognostic significance in its own right. Ultimately, progressive CKD can lead to endstage renal disease (ESRD) and dialysis. ESRD is both highly costly to the health care system and associated with significant morbidity and mortality rates for affected patients. Unfortunately, some patients progress more rapidly to ESRD, sometimes without the benefit of evidence-based interventions designed to slow the rate of progression of the disease and/or better plan for dialysis when that need occurs. Existing models to predict progression rate and care interventions can be optimized to improve health outcomes for this population using better analytic tools, data, care management and electronic health record tools.

This project merges select claims data with electronic health record information to develop an improved prediction model for patients at risk for more rapid progression of their CKD. Those patients thus identified will be evaluated by a care management team supported by a nephrologist and recommendations will be shared with the care team including their primary care physician (PCP). Care management interventions will include health behavior changes to optimize comorbid conditions like hypertension and diabetes, which when poorly controlled can lead to acceleration of CKD progression.  Recommendations for lab monitoring, medication management, and formal nephrology consultation when indicated, will be shared with the PCP.

Immediate and short term goals of the project include better educated patients, reduction in use of medications that can accelerate kidney disease, increased use of nephro-protective medications, and increased use of appropriate nephrology consultations. Longer term outcomes will be reduction in the rate of progression of CKD and delay in onset of ESRD and dialysis. If dialysis is eventually needed, we will be looking for improved planning and patient preparation so that it can be initiated electively and as an outpatient, avoiding expensive hospitalizations and use of catheters that increase the risk of infection.

 

Article: Using Machine Learning to Create New Models of Care