Home/Device Sandbox
Open access · vendor neutral · no standard exists for this · version 1.1, 7 September 2026

Infusion pumps on hospital networks

A scoring sheet for comparing up to three pumps on the one thing nobody tests before the purchase order: what happens when several hundred of them associate to your wireless network at the same time.

Why this exists

A hospital does not buy one infusion pump. It buys hundreds, sometimes thousands, and they all light up on the same network in the same few weeks. Almost every hard problem in this domain comes from that fact, and almost nothing in a vendor evaluation addresses it.

There is no standard for this, and there is no notified body to ask. The one international technical report specifically on wireless medical device networks, IEC/TR 80001-2-3:2012, was withdrawn on 5 March 2024 and nothing replaced it. The governing FDA guidance on RF wireless technology in medical devices was issued in August 2013 and is still current. NIST, ECRI and AAMI publish excellent material on infrastructure and on cybersecurity. None of it operates at this level.

So these criteria are one practitioner's, built from forty years of doing it. They are not a standard and they are not a substitute for one. What they are is a list of questions that turn out to matter, most of which are buried in a specification nobody reads until the deployment is already unhappy. Ask them before the purchase order rather than after.

Version 1.1, 7 September 2026. Section D was reworked after checking what the five US pump vendors actually deploy, in their own FDA filings rather than their brochures. Four of the five describe a server inside the hospital. Cloud no longer scores as a virtue by itself. Three new criteria ask what still runs in your building, what a fleet-wide push can break, and where the EHR interface terminates.

How to score

Rate each criterion 0 to 5 using the scale below. Each criterion also carries a weight from 1 to 5 reflecting how much it matters, shown as a small number beside it. Score = rating × weight, so a mediocre answer on something important still outranks a perfect answer on something trivial.

RatingMeaningUse it when
A rating of 1 is a finding, not a shrug. "Unclear from documentation" means you asked, and after reasonable effort you still cannot tell. Record it as 1 rather than guessing at 3. A pattern of 1s across a section is usually the most informative result on the page, and it is the thing you can put in front of the vendor.

Sections A, B, C, D and F are claims, answered from a datasheet. Section E is measurement, answered by running a test. Where a measurement contradicts a claim, the measurement wins and the claim scores zero. Read Section E first when comparing two vendors. A device scoring high on claims and low on measurement is the worse choice, and it will give the better demonstration.

What to require from the vendor before scoring

A rating of 1 only means something if you defined what documentation you expected. Request all of the following in writing, at the RFP stage, before any of it becomes a negotiation.

Put the list in the RFP. The response to it is itself a measurement, and it arrives before you have spent anything.

Test equipment for Section E

The test schema in Section E is adapted from a 2008 methodology that used a laptop and an AirMagnet card. That entire product architecture is gone. AirMagnet WiFi Analyzer PRO and Spectrum XT were discontinued and left support on 31 December 2025. The line passed from AirMagnet to Fluke Networks in 2009, to NETSCOUT in 2015, to NetAlly in 2019. Ekahau went from Ookla to Accenture, closing 17 June 2026. Both major survey vendors have been through roll-ups, which is worth knowing before a hospital standardizes on either.

The methodology survived. The equipment did not. Current equivalents:

ToolWhat it does hereUsed for
NetAlly AirCheck G3 Pro Handheld tri-band analyzer and survey. 2.4, 5 and 6 GHz, 802.11a through ax with Wi-Fi 7 visibility. The direct replacement for the laptop-and-card approach. E-1, E-2, E-5
NetAlly EtherScope nXG Portable wired and wireless analyzer with packet capture, iPerf and discovery. Captures the frames when you need to prove what actually happened. E-2, E-3, E-4, E-7
NetAlly AirMagnet Survey PRO The one surviving AirMagnet product. Design and site survey, 2.4, 5 and 6 GHz, Wi-Fi 7. Now collects with a handheld rather than a laptop adapter. E-1
Ekahau AI Pro + Sidekick 2 The other mainstream survey platform. Predictive design, validation survey and spectrum in one walk-through. E-1, E-9
NetAlly CyberScope Air Wireless vulnerability and security audit, rogue AP detection, on top of the G3 Pro test set. Security review
Portable spectrum analyzer, for example the NetAlly NXT series Non-Wi-Fi interference. The replacement for Spectrum XT. Finds the things Wi-Fi tools cannot see because they are not Wi-Fi. E-9
Wireshark with a monitor-mode adapter Frame-level analysis, free. Practical limit: decoding encrypted enterprise traffic needs the session keys, so plan your capture points before the test day. E-2, E-5
Hamina Cloud-based planning and design. Worth a look if you are designing rather than validating. E-1
Calibrated load bank of real devices No instrument substitutes for this. E-3 and E-4 need the actual planned number of pumps, or the closest you can borrow, powered and associated. E-3, E-4, E-6
Links go to the manufacturer rather than to a product page, because product URLs churn and these will outlive it. Named for identification only. No affiliation, no referral arrangement, and no payment of any kind. Two vendors dominate this category and both are named. If you use something better, tell me and it goes in the next version.

Test the interoperability claim yourself, free

Section E needs RF instruments. Section F needs a receiver, and that is harder to come by. A vendor's IHE Integration Statement names profiles. It does not tell you which transactions, which options, or against which revision of the Technical Framework. Statements in the field range from naming five profiles with no transaction numbers and “HL7 Not Applicable”, to naming every transaction explicitly. Same claimed profile, very different amounts of information.

There is a live public endpoint you can point a gateway at. Send the PCD-01, PCD-10 or PCD-04 your gateway already emits, unchanged, and it reports what actually arrived:

Delivery Stop does not mean the infusion stopped. It also fires on a rate change, a secondary start, a bolus and a VTBI change. A receiver reading the event code alone is wrong more often than it is right, and wrong silently. The response counts how many times that happened in your window.

Free, no sign-in, nothing to sign, nothing retained. You mint your own key in one command. Synthetic or simulator output only, never a real patient.

Devices under evaluation

Full document with sources
Total
Vendor A
0
0% of 990
Vendor B
0
0% of 990
Vendor C
0
0% of 990

Daniel Pettus. Forty years in medical device and health IT, contributor to IHE Patient Care Device interoperability standards. Retired from industry, now doing occasional paid advisory work. Offered open access and vendor neutral. The author has no financial interest in any infusion pump manufacturer whose product is on the US market, no affiliation with any test equipment manufacturer, and none of these criteria were written for or against a specific product. The wireless test protocol in Section E is adapted from a 2008 methodology by Todd C. Nessen, whose test schema remains, in this author's judgement, the soundest public description of how to measure a large infusion fleet's real footprint on a hospital network. Nothing you type here is transmitted anywhere; entries stay in your own browser. Please copy this, correct it, and argue with it. Corrections are more useful than agreement.