The Standard

About this Standard. Cirtiq published this Standard and builds a system designed to meet it. The Standard names no product, and any system that meets it qualifies. We welcome comment from clinicians, hospitals, technology developers, insurers and regulators, and we intend to transfer stewardship of the Standard to an independent body.

The definition

Why it matters now

AI tools now draft clinical orders from the conversation in the exam room, and their use is growing quickly. The developers of these tools build them around a clinician who reviews and signs each order. That review remains essential.

When a system is right almost all the time, however, human review alone becomes a thinner safeguard. Researchers have documented this pattern in medicine and in aviation alike. It reflects human nature, not carelessness. An independent check strengthens clinical review by bringing a clinician only the orders that need attention.

Every industry that automated its core action eventually separated the system that acts from the system that checks. That separation is what allowed the automation to scale safely.

Terms

Order that would change a patient's care.

An order for a medication, laboratory test, imaging study, procedure, referral or other clinical intervention. The term does not cover orders that only record, schedule or communicate information.

Drafting software.

Software that generates or proposes an order for a clinician to review and sign, including ambient AI tools.

Platform.

The electronic health record or other system through which an order is entered, transmitted or carried out.

Hold.

A pause placed on an order before it takes effect. The order stays paused until the person named in the hospital's policy resolves it.

Time-critical order.

An order placed in an emergency, or under a hospital protocol that requires immediate action.

Requirements

1. Coverage. Every order that would change a patient's care is checked before it takes effect. This applies whether a clinician wrote the order or software drafted it.¹
2. The record. The check uses the patient's record at the hospital and the patient's history from other organizations, retrieved when the order is placed through the national health information exchange (TEFCA). The system records which sources it checked.
3. Independence. The verification is developed and operated by a party that is organizationally and financially independent of:
   • the clinician who wrote the order;
   • the developer of any software that drafted it;
   • the platform that carries it;
   • the insurer who pays for the care.
   The hospital sets the policy and resolves holds, but it does not build or control the verification itself.
4. What the check does. At a minimum, the check does three things:
   • (a) Verifies facts. It confirms that the facts an order relies on are supported by the record. These include diagnoses, results, allergies and current medications, and the check includes more recent results held elsewhere.
   • (b) Detects conflicts. It identifies conflicts between the order and the patient's current condition and care, including combinations that no coded rule covers.
   • (c) Recognizes deliberate decisions. Before holding an order, it distinguishes a documented, deliberate clinical decision from a likely error.
5. Written criteria. Orders are cleared or held on written criteria set by the hospital's policy. The criteria are available to the clinicians they apply to.
6. Resolution. A hold is resolved by the clinician or authority that the hospital's policy names for that kind of finding. That person records the outcome and the reason. The policy sets a time limit for resolving holds and escalates any hold not resolved within it, so no order waits indefinitely.
7. The record of verification. Every check, hold and resolution is kept in a record that cannot be altered after the fact. The record is retained for the period required for medical records. It is available to the hospital and, as the hospital authorizes and the law permits, to its accreditors, its insurers and a federally listed Patient Safety Organization.
8. Performance. The verifier's own performance is measured and reported to the hospital at least quarterly. The report covers:
   • how often orders are held;
   • how often holds are overridden;
   • errors confirmed as caught;
   • errors later found that the check missed.
   Performance is also monitored across patient populations.
9. Identification. Orders drafted by software are identified as such when they are signed, including the software and its version.

¹ Time-critical orders proceed without delay and are checked immediately afterward. ↩

What the Standard does not do

• It does not replace clinical judgment. Clinicians make every decision about care.
• It does not replace existing safety checks. Pharmacist review, electronic health record alerts and other safeguards remain in place, and the Standard adds to them.
• It does not prescribe a technology. Any method that meets the requirements qualifies.
• It does not set a hospital's policy. It requires only that the policy be written, followed and recorded.

What drafting software can contribute

Two things strengthen independent verification. Most drafting software can already provide them:

• a mark on each order identifying the software and version that drafted it;
• the passage of the conversation that the order came from.

With these, a check can also catch orders that were discussed but never drafted. The accuracy of drafting software can then be measured over time, which gives its developers evidence of how well their tools perform.

Where independent checks already exist

In other industries where an action is automated, regulators have written the separation of acting and checking into rule:

• Nuclear power. Federal design criteria for reactors include one titled "Separation of protection and control systems": "The protection system shall be separated from control systems to the extent that failure of any single control system component … leaves intact a system satisfying all reliability, redundancy, and independence requirements of the protection system." (10 CFR Part 50, Appendix A, Criterion 24)
• Securities trading. Brokers must use automated controls that prevent erroneous orders "by rejecting orders that exceed appropriate price or size parameters, on an order-by-order basis." These controls must be under the broker's "direct and exclusive control," not the control of the customer sending the orders. (SEC Market Access Rule, 17 CFR 240.15c3-5)
• Aviation. Airliners must carry a collision avoidance system, a requirement set by Congress in 1987. The system operates independently of both the autopilot and air traffic control.

Healthcare's own accreditors point the same way. Joint Commission guidance issued with the Coalition for Health AI states that "organizations should engage in confidential, blinded reporting of AI-related safety events to an independent organization that can share information to the field."

Medicine has done this before

In 1986, anesthesiologists adopted national monitoring standards. By 1989, continuous pulse oximetry during anesthesia had become the standard of care across the profession. At Harvard's teaching hospitals, malpractice premiums for anesthesia faculty fell by two-thirds between 1986 and 1991.

After a 2009 study found that it cut surgical deaths and complications by more than a third, the surgical safety checklist spread the same way: through hospitals, professional societies and accreditors.

Neither practice required new law.

Using the Standard

License. This Standard is published under a Creative Commons Attribution 4.0 license (CC BY 4.0). You may copy and adapt it for any purpose, including legislation, regulation, contracts, procurement and accreditation standards.

How to cite. Cirtiq. Independent Clinical Order Verification: The Standard. Version 1.0, September 2026. cirtiq.com/standard.

Comments and versions. Send comments to standard@cirtiq.com. Every revision is published with its version number, date and a summary of changes.

Sources

1. U.S. Nuclear Regulatory Commission, 10 CFR Part 50, Appendix A, General Design Criterion 24. ecfr.gov/current/title-10/chapter-I/part-50/appendix-Appendix A to Part 50
2. U.S. Securities and Exchange Commission, Market Access Rule, 17 CFR 240.15c3-5. ecfr.gov/current/title-17/chapter-II/part-240/section-240.15c3-5
3. Federal Aviation Administration, Collision Avoidance Systems (implementing Public Law 100-223, 1987). federalregister.gov/documents/2001/11/01/01-27340/collision-avoidance-systems
4. The Joint Commission and Coalition for Health AI, Responsible Use of AI in Healthcare, September 2025. JC-CHAI_RUAIH_Guidance.pdf
5. Anesthesia Patient Safety Foundation, "ASA 1986 Monitoring Standards Launched New Era of Care, Improved Patient Safety." apsf.org/article/asa-1986-monitoring-standards-launched-new-era-of-care-improved-patient-safety
6. Haynes AB et al., "A Surgical Safety Checklist to Reduce Morbidity and Mortality in a Global Population," New England Journal of Medicine 2009;360:491-499. nejm.org/doi/full/10.1056/NEJMsa0810119
7. Dratsch T et al., "Automation Bias in Mammography," Radiology 2023;307(4):e222176. doi.org/10.1148/radiol.222176
8. Federal Aviation Administration, Operational Use of Flight Path Management Systems, 2013. faa.gov/sites/faa.gov/files/aircraft/air_cert/design_approvals/human_factors/OUFPMS_Report.pdf

Contact

contact@cirtiq.com