ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

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

email@asn-online.org

202-640-4660

The Latest on X

Kidney Week

Abstract: TH-PO046

Impact of Nationwide Utilization of a Machine Learning Model to Identify Home Therapy Candidates

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science

Authors

  • Monaghan, Caitlin, Fresenius Medical Care Holdings Inc, Waltham, Massachusetts, United States
  • Hanson, Maria, Fresenius Medical Care Holdings Inc, Waltham, Massachusetts, United States
  • Han, Hao, Fresenius Medical Care Holdings Inc, Waltham, Massachusetts, United States
  • Willetts, Joanna, Fresenius Medical Care Holdings Inc, Waltham, Massachusetts, United States
  • Usvyat, Len A., Fresenius Medical Care Holdings Inc, Waltham, Massachusetts, United States
Background

Home therapy (HT) modalities offer excellent treatment options for dialysis patients but are underutilized. As part of a larger effort to provide support to patients, there is a team of Kidney Care Advocates (KCAs) who provide education on treatment modality options with a goal of empowering patients to choose a home modality when appropriate. Towards this, a machine learning (ML) predictive model was built to identify which in center hemodialysis (ICHD) patients would be good candidates to target for HT.

Methods

Patient (n=298552) data collected from 2016-2019 were used to develop an XGBoost ML model to predict the likelihood of an ICHD patient switching to peritoneal dialysis or home hemodialysis in the next 90 days and staying on that HT for at least 90 days. The data were split into training (70%) and validation (30%) datasets, maintaining the 2.5% positive prevalence in both. Data sources included ICHD treatments, labs, hospitalizations, missed treatments, demographics (age, marital status, and vintage), comorbidities, clinical assessments, clinic home program information, and Zillow housing data based on zip code, resulting in 2475 variables. KCAs contacted ICHD patients identified by the model to offer education and services. From January-November 2020, KCAs completed an assessment when a patient was contacted (n=26055), which included questions on if the patient was interested in HT and how the patient was referred to the KCA.

Results

The model achieved good performance with an area under the curve of 0.87 for the validation dataset. Using a threshold set to the positive class prevalence (0.025), recall was 0.77 and precision was 0.09. The top predictors used by the model were dialysis vintage, previously expressing interest in or being referred to HT, age, and dining with one’s spouse.

During the time examined, KCAs reported contacting 20734 patients. Of those, the model was the sole referral source for 5382 patients and 1219 of them expressed interest in HT. In 2020-2021, 487 ICHD patients referred to KCAs by only the ML model switched to HT.

Conclusion

Through close collaboration with KCAs, an ML model was successfully used to identify ICHD patients who were good candidates for HT. Use of this model had measurable impact with hundreds of ICHD patients switching to HT who otherwise might not have been referred to a KCA.

Funding

  • Commercial Support – Fresenius Medical Care