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-PO042

Using an Ensemble Model to Improve ESA Prescription in Hemodialysis (HD)

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science

Authors

  • Chiu, Yi-Wen, Kaohsiung Medical University Chung Ho Memorial Hospital, Kaohsiung, Taiwan
  • Lin, Ming-Yen, Kaohsiung Medical University Chung Ho Memorial Hospital, Kaohsiung, Taiwan
  • Yen, Hong-Ren, National Sun Yat-sen University, Kaohsiung, Taiwan
  • Hsu, Chan, National Sun Yat-sen University, Kaohsiung, Taiwan
  • Ku, Chan-Tung, National Sun Yat-sen University, Kaohsiung, Taiwan
  • Kang, Yihuang, National Sun Yat-sen University, Kaohsiung, Taiwan
Background

How the ensemble model trained by new indexes can improve the ESA prescription is unclear.

Methods

We have compiled a dataset from 2013-2020, including bi-monthly ESA prescriptions, Iron doses, Hb, and relevant data of HD patients in our hospital. We divided all records into six categories, status 1 to 6, based on subsequent hemoglobin(sHb), using cut-off values of 10.8 and 11.2 and direction of change (not decreasing: 1-3 vs. decreasing:4-6). Indexes I, II, and III were used to evaluate how the ESA dose deviations of model recommendations theoretically impacted sHb. Index I assesses how model deviations keep sHb near 11 g/dL (positive index), while II and III are away from 11 g/dL (negative index). (Figure 1) Besides the traditional ESA-prescription algorithm (TEA, Figure 2), we have trained various models, including a meta-learning model (MLM).

Results

25,632 records of 315 ESKD patients were included, with 71.9% of the Hb levels between 10-12mg/dL. Besides TEA, we selected a generalized linear mixed-effect tree model (GLMM) and a weighted random forests model (WRF) for sHb and ESA dose prediction. The overall results are shown in Table 1. Our WRF and MLM can almost preserve the proper prescriptions (Index I), and avoid those improvable or unacceptable recommendations (Index II & III). The MLM can theoretically improve 65.4% and 43.8% of prescriptions with sHb >11.5 and <10.5, respectively. (undetermined)

Conclusion

The ensemble model trained by new indexes can theoretically recommend more proper ESA dosages.

Total
n=25,632
sHb between 10.8-11.2
Status 2, n=2,727; Status 5, n=2,550
sHb below 10.8
Status 1, n=3,859; Status 6, n=5,767
sHb above 11.2
Status 3, n=7,345; Status 4, n=3,384
 
Index
(Score assigned to record)
I
(1)
Undetermined
(0)
Unaccepted
(-2)
II
(-1)
II
unaccepted
(-2)
Undetermined
(0)
III
(-1)
III
unaccepted
(-2)
undetermined
(0)
Total score
Model/Status252525161616343434PositiveNegative
TEA, %91.474.56.221.82.43.72.225.50.31.797.572.82.40.20.30.0397.699.81530-434
GLMM, %71.857.018.627.19.615.81.213.11.32.897.585.111.24.74.93.488.995.3789-581
WRF, %95.397.53.62.30.90.24.29.400.0395.890.63.71.90.2096.198.12518-312
MLM,%95.597.73.22.21.30.23.37.600.0396.792.43.21.60.3096.598.42510-244

2-week prescripiton as example

Funding

  • Clinical Revenue Support