ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: TH-OR097

Use of Machine Learning to Inform Decision Making and Optimal Renal Replacement Therapy

Session Information

  • Home Dialysis
    November 07, 2019 | Location: 143, Walter E. Washington Convention Center
    Abstract Time: 06:18 PM - 06:30 PM

Category: Dialysis

  • 704 Dialysis: Vascular Access

Authors

  • Son, Jung Hoon, pulseData, New York, New York, United States
  • Wang, Xiaoyan, pulseData, New York, New York, United States
  • Fielding, Ollie, pulseData, New York, New York, United States
  • Lee, Edward M., pulseData, New York, New York, United States
  • Kipers, Chris, pulseData, New York, New York, United States
  • Wiener, Lauren Alexandra, pulseData, New York, New York, United States
  • Liu, Frank, The Rogosin Institute, New York, New York, United States
  • Bohmart, Andrew, The Rogosin Institute, New York, New York, United States
  • Silberzweig, Jeffrey I., The Rogosin Institute, New York, New York, United States

Group or Team Name

  • The PEAK team
Background

We deployed a machine learning (ML) model to identify patients at risk of requiring RRT to support clinical care decision making in a multidisciplinary care (MDC) team. We compare the difference in optimal renal replacement therapy (RRT) starts pre and post implementation. To the knowledge of the authors, this is the first live application of ML to inform transition workflows.

Methods

An EHR database of 110,998 patients was used to create an ML model to predict progression to an eGFR <10 or RRT start in the next six months (see Kidney Week 2018 SA-PO953). The system calculates weekly risk scores for non-dialysis patients with an eGFR <35. For high risk patients an alert is sent to the patient’s nephrologist suggesting prompt referral to the PEAK MDC team. The team reviews high risk patients and provides education to inform their decision making. Optimal dialysis starts were defined as outpatient starts with access via AV fistula, AV graft, or peritoneal dialysis catheter.

Results

Since deployment of the ML model in October 2018, 54% of patients enrolled in PEAK had an optimal dialysis start. This is almost three times the national average of 20% (USRDS 2018 data) and 14% better than the 47.3% rate prior to use of the ML model. PEAK home dialysis rates have increased 20% vs. before deployment (24% vs 20%), and is now eight times the NYC average 24% vs 2.5%. PEAK members also received pre-emptive transplants at a rate five times the NYC average 12.5% vs 2.5%.
PEAK patients with optimal starts had significantly greater provider interactions, as measured by unique appointment days prior to dialysis, than non-optimal starts (3.9 vs. 2.5 appointments, p<0.0001, unequal variances t-test). Optimal start patients are also associated with earlier enrollment, defined as the time from the first PEAK appointment to dialysis (329 vs. 179 days, p<0.02, unequal variances t-test).

Conclusion

The PEAK MDC-pulseData partnership has improved optimal dialysis starts and home dialysis modality rates by 14% and 22% respectively. Enrollment to the PEAK program has increased by 22% since Oct. 2018. Our results demonstrate that purpose-built AI tools used by an MDC team can increase optimal RRT outcomes.

Funding

  • Commercial Support