Hemodialysis, Hemoglobin
Conditions
Brief summary
The global prevalence of end-stage renal disease (ESRD) was estimated ranging 5.3 to 9.7 million patients in 2017 and is a major healthcare cost driver in both developed and developing economies. The majority of patients with ESRD, approximately 90%, received in-center maintenance hemodialysis (HD). Although HD patients are under the close supervision of a nephrologist, they are vulnerable to anemia and substantial hemoglobin (Hb) variability, which are controversially associated with poor clinical outcomes, such as all-cause mortality. The contemporary narrow target hemoglobin level recommended in the KDIGO and KDOQI guidelines, despite the ongoing debates, poses a crucial challenge in maintaining the optimal hemoglobin level in HD patients. The Big Data Center at China Medical University Hospital (CMUH) has developed a tool, Hb Scope APP, that can use the color of the HD tubing to predict real-time Hb levels by leveraging the smartphone's camera capacity and machine learning (ML) technology. The performance of the Hb Scope ML algorithm in predicting Hb \> 10 g/dL can reach an accuracy of 0.93 and an AUROC of 0.99 in the testing dataset. This opens an opportunity to establish a vibrant digital ecosystem for automatic anemia management. Innovative ML tools must be appropriately regulated before these algorithms are adopted into clinical practice. Therefore, in the current validation study, we propose to do a multicenter validation trial for validating whether the Hb predicted by Hb Scope APP can achieve an area under the receiver operating curve (AUROC) of at least 0.80 in the adult HD populations from CMUH, Asia University Hospital (AUH) in Taiwan, and SEHA Kidney Care (SKC)-Central in the United Arab Emirates.
Interventions
OTHERHb Scope APP
The study intervention is the Hb Scope APP, a camera-based real-time Hb estimation tool using a data collection function and a machine learning-based algorithm to analyze the image of HD tubing and output the estimated Hb value. Our proposed Hb Scope APP can provide a non-invasive and real-time estimation of Hb level for detecting the alarming Hb status (i.e., \<=10 g/dL) during every HD session. The software is for information management purposes only and is not intended for diagnostic use.
Sponsors
National Science and Technology Council
SEHA Kidney Care
Asia University Hospital
China Medical University Hospital
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
20 Years to 90 Years
Healthy volunteers
No
Inclusion criteria
* Patients receiving regular maintenance HD in the three participating sites; AND * Aged 20-90 years old; AND * Received regular HD through an arteriovenous (AV) fistula or AV graft access; AND * Use the Fresenius HD machine 4008/4008S and 5008/5008S.
Exclusion criteria
* Do not have last Hb level from labs within 1 month prior; OR * Do not receive blood drawn for the true Hb level or do not have the true Hb level on the date of image taking by Hb Scope APP; OR * Do not agree to participate.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Agreement between the lab-derived and the Hb Scope-derived Hb level at the dichotomous cutoff of > 10 g/dL. | 2 months | The targeted area under the receiver operating curve (AUROC) is 0.80. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Agreement between the lab-derived and the Hb Scope-derived Hb level at the dichotomous cutoff of < 8 g/dL. | 2 months | The targeted area under the receiver operating curve (AUROC) is 0.80. |
| Agreement between the lab-derived and the Hb Scope-derived Hb level at the dichotomous cutoff of > 12 g/dL. | 2 months | The targeted area under the receiver operating curve (AUROC) is 0.80. |
| Consistency between the lab-derived and Hb Scope-derived Hb continuous value. | 2 months | The targeted correlation coefficient is 0.70. |
Countries
Taiwan, United Arab Emirates
Outcome results
None listed