Comparing continuous renal replacement therapy systems for procurement starts with a defined set of measurable performance parameters, not a broad product rating. The most useful comparison separates machine-level accuracy from clinical outcomes, because published device studies typically report effluent ratios, downtime, filter life and alarm behavior rather than mortality differences. Procurement teams should also treat regulatory and standards documents as minimum verification baselines. The U.S. pediatric CRRT classification under 21 CFR § 876.5861 lists specific performance testing categories such as fluid flow accuracy, ultrafiltration, clearance rates and mechanical hemolysis, while a Chinese health industry standard sets maximum permissible errors for pumps, pressure monitors and scales [5][6]. Together these establish the technical thresholds a candidate system must meet before any institutional preference for interface design or workflow is considered. The goal is a documented, repeatable evaluation method that leaves less room for unverifiable vendor assertions.

Dose delivery and downtime as primary comparator metrics

Delivered dose is one of the strongest starting points for head-to-head comparison because it can be extracted from device data logs under consistent definitions. In a multicenter comparison of the PrisMax system against the prior Prismaflex platform, the mean delivered-to-prescribed effluent ratio was 0.92 ± 0.15 for PrisMax compared with 0.85 ± 0.21 for Prismaflex, a difference the authors reported as statistically significant [2]. Delivered effluent dose was 18.16 ± 12.93 mL/kg/h versus 10.95 ± 10.96 mL/kg/h for the two platforms. These figures describe machine-recorded delivery, not patient outcomes, and procurement teams should not treat them as evidence of clinical superiority. However, they are useful for benchmarking how accurately a device translates a prescription into treatment.

Downtime functions as a companion metric because interruptions directly limit achieved therapy time. In the same PrisMax comparison, downtime was reported as 27 minutes less than for the older device, though the source does not specify whether this analysis adjusted for treatment duration or clinical setting differences [2]. A separate ACUsmart pilot reported that net treatment duration averaged 92.37% of total treatment duration and that mean prescribed versus delivered dose was 26.33 versus 24.10 mL/kg/h [3]. Procurement teams should request both metrics from every candidate vendor, defined in the same way, rather than accepting a single favorable number. A ten percent gap between prescribed and delivered dose may reflect alarm burden, access problems or pump calibration drift, and the device log should be able to reveal which of these factors dominates.

Sensor and pump tolerances in standards and calibration documents

Procurement specifications should include quantitative accuracy limits for the components that most often fail or drift. The Chinese standard WS/T 836—2024 specifies a maximum permissible error of ±10% for blood pump flow and for dialysis, replacement and effluent pumps, ±5% for heparin syringe pump rate, ±1.3 kPa for arterial or venous pressure monitoring, and ±1% for weighing scales [6]. These tolerances are not product-specific endorsements; they are recommended inspection thresholds for institutions that manage CRRT devices. A U.S. purchaser can use the same numeric values as internal acceptance criteria, while recognizing that they come from a different regulatory jurisdiction and are not a substitute for FDA-cleared labeling.

Functional alarm testing matters as much as steady-state accuracy because a machine that pumps accurately but fails to detect air or blood leaks creates immediate patient risk. The standard describes testing the bubble sensor by simulating air in the venous chamber and confirming both an audible-visual alarm and clamping of the return line; it describes leak detection by simulating red cells in the waste line [6]. It also requires a fully charged backup battery to sustain circulation for at least 15 minutes without heating. Procurement evaluations should reproduce these tests on demonstration units and document that the alarm state is visible, audible, and recoverable within the institution's staffing model. This treats alarms as verifiable engineering functions rather than generic safety assurances.

Minimum documentation to request from every vendor

Request data log samples showing prescribed and delivered effluent dose, treatment time, downtime and filter life under definitions that match the published studies being used as benchmarks [2][3]. Ask whether alarm events, anticoagulation mode and access type can be exported per session, since these variables affect delivered dose. Require the maximum permissible error specifications for each pump, pressure monitor and scale, and ask the vendor to identify which regulatory or calibration standard the machine was tested against [6]. Confirm that bubble and leak sensor functions, backup battery duration and electrical safety testing align with the applicable medical electrical equipment standard for extracorporeal blood treatment devices. Finally, obtain the model-specific labeling summary of in vitro performance and fluid flow accuracy rather than marketing material [5].

What device pilots can and cannot establish for procurement

Published device evaluations are useful but narrower than many buyers assume. The PrisMax multinational pilot prospectively collected data from seven intensive care units in six countries and included 305 filters, while historical data from 4,247 filters in the same centers provided comparison points [4]. The study reported treatment modes and anticoagulation patterns, including regional citrate use in 188 of 305 filters. These denominators explain the sample structure, but the source does not report randomization, blinding, or adjustment for patient severity. A procurement analyst should therefore treat any between-device performance gap as observational evidence from a specific set of centers.

The ACUsmart evaluation followed a similar logic: a multicenter pilot designed to assess functionality, usability, feasibility and reasons for treatment withdrawal, with treatment data recorded by the device and downloaded from logs [3]. Users rated the machine as practical and easy to use, while noting that a few components needed slight improvement. Such statements are subjective usability feedback from named investigators, not independent usability validation. For procurement, pilot data support a shortlist, not a purchase decision. Institutions with different patient case mix, anticoagulation practice or nurse staffing should expect delivered-dose ratios and downtime to differ from published pilot values.

Regulatory verification baselines and labeling requirements

The U.S. pediatric CRRT device classification supplies a concrete set of premarket performance testing expectations that procurement teams can mirror in vendor due diligence. Under 21 CFR § 876.5861, performance data must include fluid flow accuracy, functional testing of sensors, pumps and scales, ultrafiltration, blood and dialysate pressure drop, clearance rates, sieving coefficients, mechanical hemolysis, structural integrity, blood compartment integrity and chemical analysis of the dialyzer membrane [5]. Software verification, validation and hazard analysis are also specified. A buyer can request summary evidence for each category without needing to interpret raw engineering data, while documenting any category the vendor declines to share.

Labeling expectations are equally specific. The rule lists required elements for the delivery system, hemodialyzer and blood tubing set, including blood flow and maximum dialysate rate for each model, a summary of in vitro performance data, priming volume, tubing length, compatible delivery systems, and maximum operating pressures for transducer protectors [5]. Procurement should treat these items as minimum disclosure obligations. If a vendor cannot produce the required labeling content, that gap is a direct procurement risk indicator. Conversely, presence of compliant labeling does not mean the device matched a competitor's pilot results; it means the device met the regulatory floor for its intended classification.

A structured demonstration and local verification workflow

A concrete evaluation can be organized around three layers: regulatory benchmark, published device data, and local bench reproduction. The first layer documents that the candidate meets the applicable classification testing categories or inspection tolerances. The second layer records published delivered-dose and downtime figures with source dates and denominators. The third layer, conducted on demonstration hardware, reproduces the pump accuracy points at low, middle and high flow settings, such as 50, 100 and 200 mL/min for the blood pump, with three measurements per point and the mean error calculated against the set value [6]. The same logic applies to dialysate and replacement pumps at three points across their nominal range.

A hypothetical institution could then compare two platforms using identical test conditions: same environmental temperature and humidity range, same supply voltage specification, and a documented maintenance state before testing. This is a suggestion for internal procedure, not a report of a real evaluation. The comparison should not produce a single winner without weighting criteria in advance, because one device may show tighter weighing accuracy while another shows shorter alarm recovery time. Procurement committees can assign those weights based on local nursing workflow, but the published sources do not provide an objective weighting model. The value of the workflow is that disagreements return to measured values rather than feature descriptions.