This is a randomized, double-blind controlled trial to compare AI (model) with the physician on prescribing ESA dose to maintain hemoglobin near the therapeutic target, 11g/dl.

It has been known well for a long time that End stage renal disease (ESRD) patients usually need ESA to maintain their hemoglobin (Hb) to improve both mortality as well as quality of life. Later, several large-scale RCTs showing normalization of Hb in this population increased the risk of thromboembolic event, which leaded current guidelines to recommend the therapeutic goal for anemia in ESRD should be within a range (usually between 10-12 gm/dl) rather than above or below a certain level (or value). In addition to ESA dose, many factors contribute to the severity of anemia in this population, such as iron status, chronic blood loss, adequacy of dialysis, chronic inflammation, renal wasting, et al. To put all these factors together, maintaining the Hb levels within the target level is a challenge to physician. It was reported that only one third of Hb within the target at any given time. Our data showed, for those under maintenance hemodialysis and without any blood transfusion in observation period, near two third of patients' Hb level were within the target range.

For better anemia management, NKF-K/DOQI developed and published guidelines using protocol or algorithm for EPO prescription.6 Recently, artificial intelligence (AI) has been widely applied to medicine in the field of reducing human error, robotic surgical system and decision- making aid. Since then, there were several studies working on decision making programs to set up model of predicting the ESA dose needed for target Hb level. Artificial neural network (ANN) model is most commonly used for ESA dose-response prediction. However, the effectiveness of these kinds of AI is not confirmed clinically and the result not satisfactory. In other words, it is still inconclusive whether contemporary AI has any role in decision making aid when prescribing ESA dose for dialysis patient, a typical trial and error which AI supposed should be very helpful with.

Based on the mentioned above, we are going to conduct this clinical try to test the hypothesis that AI is not inferior to physician in prescribing ESA dose for hemodialysis patients to maintain hemoglobin(Hb) level to meet the target

1. Main inclusion criteria:

1. signed informed consent
2. Age older than 20 year old.
3. End stage renal disease under regular hemodialysis with three times per week and duration of each session at least 4 hours
4. Having at least consecutive 6 months data of both Hb and biochemical study during dialysis before enrollment and at least one Hb level within the range of 11-12 g/dl
5. Having at least one prescription of ESA supplement to keep Hb within the range of 10-12 g/dl in the past 6 months before the enrollment
6. Having received ESA of the same brand at least 6 months before the enrollment

2. Main exclusion criteria:

1. Ever receiving blood transfusion in the past 12 months
2. Active bleeding with blood loss more than250cc in 3 months before the enrollment
3. Active infection or malignancy
4. Study subject can not follow with the study protocol 3. End of Study

The eligible subject will not be allowed to continue the study once informed consent is withdrawn or event happening meets the following criteria:

1. Taking extra ESAs, Androgens and iron-chelating agents
2. Undergoing surgery with massive haemorrhage, or blood transfusion
3. Receiving systemic chemotherapy or radiotherapy or immunosuppressive therapies.
4. The end of study will be 6 months later after the randomization of last study subject or failed prescription rate higher than 5%, which comes earlier.