The following was taken from the methods slide, (ML) replaced it with the original footnote that went along with the graph

Patients on dialysis > 90 days in initial prevalent cross-sections

USA shows the highest mean upper ferritin levels while JPN shows both the smallest mean upper and lower ferritin levels

Each facility’s SD in Hb values (facility-level Hb SD) was calculated for a prevalent cross-section of patients with end-stage renal disease for more than 180 days in each facility at the start of participation in each DOPPS study phase.

The histogram indicates the percentage of facilities having facility-level Hb SD of various values across 904 facilities in DOPPS 1, 2, and 3.

All included facilities had at least 10 patients with a reported Hb value within the prevalent cross-section.

Linear regression showed no significant relationship between number of facility patients with a Hb measurement and facility-level Hb SD (slope=0.003 g/dL per additional patient; p=0.1).

HRs of mortality were estimated using Cox regression in 2 different samples of hemodialysis patients (Pat) with end-stage renal disease (ESRD) for more than 180 days.

(A) Dialysis Outcomes and Practice Patterns Study data phases 1-3 (n = 861 facilities and 25,417 patients) adjusted for age, sex, black race, years with ESRD, body mass index, 14 classes of comorbidity, facility mean Hb, and facility mean erythropoiesis-stimulating agent dose; models were stratified by country and study phase and accounted for facility clustering effects; study follow-up time: mean, 1.71 years, median, 1.82 years. The relationship of facility-level Hb SD with mortality did not differ meaningfully at different patient Hb (interaction p=0.3) in a sensitivity analysis excluding facility mean Hb.