We investigated the extent to which dialysis prescription is achieved in the United States and the reasons for failure to do so. Dialysis treatment data (blood urea nitrogen [BUN], duration [T], and dialyzer urea clearance [K]), for 297 patients treated in 48 dialysis units in the United States were used to calculate the urea kinetic modeling parameter, V (urea distribution). V was also calculated for 5,650 treatments of 357 patients in five representative units starting or continuing to kinetically model patients monthly over the past 2 years. V was estimated in patients from height, weight, and sex (Vest). If V differed from Vest by 20% or more, we concluded that a decrease in BUN was inconsistent with expected therapy and represented "nondelivery" of dialysis (ie, KT is decreased by approximately 20% or more). Half the US units studied had more than 35% nondelivery of therapy; more than 50% of treatments were not delivered in 44% of the units. The reasons for nondelivery could not be determined directly in the US study because we had limited contact with their dialysis programs. Common causes are treatment delivery errors (K and T), access problems, recirculation due to rapid blood flow, and dialyzers that deviate from the manufacturers' specifications. The long-term (five-unit) study showed the same percent of nondelivery of dialysis in start-up units as in the US study (40% to 50%). Long-term, nondelivery decreases to 10% to 20% in these units. For these units, nondelivery tends to occur for the same patients month to month; a decrease in nondelivery of treatment from start-up to lower long-term levels is due to more deliberate delivery of dialysis therapy. The persistence of a 10% to 20% shortfall in therapy stems from either access or clotting problems, many of which cannot be corrected. We conclude that widespread nondelivery of dialysis exists in US dialysis facilities and probably world-wide. If optimal (minimal) dialysis prescription is the therapy goal of a dialysis unit, the deliberate delivery of the desired treatment should be routinely checked using quantitative assessment techniques such as urea kinetic modeling. If this is not done, optimal treatment may not be achieved and inadequate treatment may result.