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

ASN / Education & Meetings / Kidney Week /

Please note that you are viewing an archived section from 2021 and some content may be unavailable. To unlock all content for 2021, please visit the archives.

Abstract: PO2296

A New Epidemiologic Methodological Approach Using Machine Learning in Prevalence Estimation of CKD

Session Information

Category: CKD (Non-Dialysis)

  • 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention

Authors

  • Dauvergne, Maxime, Department of Dialysis AURA Nord Saint Ouen, Saint Ouen, France
  • Sigogne, Raphaël Rs, IQVIA, Data science and Advanced Analytics, La Défense, France
  • Maravic, Milka, Assistance Publique - Hopitaux de Paris, Paris, France
  • Urena Torres, Pablo A., Department of Dialysis AURA Nord Saint Ouen, Saint Ouen, France
Background

Prevalence of chronic kidney disease (CKD), a pandemic condition, is generally estimated at 5-10% of the general population and increases with ageing. With the emergence of artificial intelligence, machine learning approaches could identify patients with such condition, improve our understanding of health, and provide opportunities for intervention. The aim was to automatically identify people with CKD and estimate more precisely its prevalence, which remains a real challenge.

Methods

Two sources of data were used, LPD (Longitudinal Patients Data) and LRx (Lifelink Treatments Dynamics), including data of near 2.5 and 40 million subjects, respectively. LPD, a medical database, included 191,905 patients receiving medications usually defined as specific for CKD from July 1st, 2019 to June 30, 2020. Of these subjects, 1.9% had a firm diagnosis of CKD, dialysis, or kidney transplant status. These patients were followed by 1,210 general practitioners who participated in a permanent longitudinal observatory of ambulatory medicine prescriptions (LPD). LRx contained all anonymized medication dispensing in outpatient care database from a representative panel of 45% of all French metropolitan retail pharmacies. A machine learning algorithm using a gradient boosting model was trained from CKD patients identified in LPD (metrics performance - sensitivity: 68%, specificity: 99%, positive predictive value: 52%, negative predictive value: 99%, F1 score: 59%). The model was implemented in LRx to obtain the overall number of CKD patients in the period of interest. As we will underestimate the true number of CKD patients, rules-based algorithm focused on erythropoietin delivery for renal condition and keto-analog was applied on LRx. We calculated the raw number of CKD patients and extrapolated it and described demographic characteristics from November 1st, 2019 to October 31th 2020.

Results

In LRx, we numbered 269,183 CKD patients corresponding to an extrapolated number of 678,102 patients with 40.8% of women and the mean age was of 77.0 years (+11.1) in the period of interest. This corresponded to a prevalence of 1%.

Conclusion

A combined approach using machine learning and rules-based algorithm may be useful in identifying CKD patients who require careful management of their renal condition.

Funding

  • Private Foundation Support