Preventing Harm Through Reliable Clinical Workflows

A defensible workflow makes it difficult for an important clinical task to disappear. It gives the task an owner, a time appropriate to its urgency, a backup when the owner is unavailable, and a way to discover that the task was not completed. It also gives the patient a usable explanation of what to expect and when to seek help.

The workflow below is an original practice-design model informed by the federal SAFER guides and AHRQ diagnostic-safety resources. It is not a statutory checklist, a substitute for specialty guidelines, or a guarantee against malpractice. A practice must adapt it to its patient population, staffing, clinical services, and actual technology.

Before the encounter

Verify identity using the practice's approved process. Reconcile the reason for the visit with any new symptoms reported by scheduling staff, telephone messages, or a caregiver. Identify interpreter and accessibility needs. Make recent critical results, unresolved referrals, hospital discharge changes, and medication discrepancies visible to the clinician who will act on them.

Do not turn a risk flag into an automated diagnosis. A screening response, coded problem, or imported hospital label may need confirmation. The clinician needs enough context to decide whether the planned visit is still appropriate or the patient requires another level of care.

During assessment and decision-making

Record the important positive and negative findings that were actually assessed. Explain why the leading diagnosis fits, what alternative would materially change management, and what would prompt reassessment. More text is not necessarily more useful: a short explanation of why an important dangerous alternative was considered and addressed can be more informative than several pages of imported normal findings.

When a decision depends on a limitation, describe the limitation and the response. Examples include an incomplete examination by video, unavailable records, a patient unable to obtain a medicine, or a specialist appointment outside the clinically appropriate interval. Acknowledging a limitation does not resolve it; the plan should address the remaining risk.

Orders and results

The ordering pathway should identify the responsible clinician and the coverage arrangement. The SAFER Test Results guide recommends explicit follow-up responsibility and an accepted transfer when another clinician takes over. A result should move through receipt, clinical review, action, patient communication, and completion of the necessary next step. An electronic acknowledgment is evidence of one stage, not proof of all five. [P1]

Use an escalation mechanism appropriate to the clinical risk. The practice should be able to find tests that were ordered but not performed, specimens that did not yield results, amended reports, results sent to a departed clinician, and results arriving during leave. Specify who checks each queue and who responds when the expected recipient does not.

Referrals and transitions

A referral is not complete because an order was entered. Define what completion means: appointment made, consultation occurred, report returned, clinical recommendations reviewed, and necessary treatment undertaken. A referral declined by the receiving office, rejected by insurance, or postponed by the patient requires a clinical disposition when the underlying problem remains unresolved.

At discharge or handoff, identify the receiving clinician, medication changes, pending tests, warning symptoms, and who will take the next action. The SAFER Clinician Communication guide addresses reliable electronic communication in referrals, discharges, and team messaging. The specific process should distinguish a message sent from an accepted transfer of responsibility. [P2]

Medication safety

Review indication, dose, route, duration, relevant organ function, allergies, interactions, duplicate therapies, monitoring, and practical access when those matters affect safe use. Confirm that a stop order actually reached the relevant medication list, pharmacy, facility, or caregiver. A historical prescription in the chart is not proof that the patient is taking it; a prescription sent is not proof that it was obtained.

For higher-risk therapies, define who reviews monitoring results and what happens after a missed test or refill request. Build the process around the medicine and patient, rather than relying on a generic annual medication reconciliation flag. AHRQ's primary-care patient-engagement material supports medication understanding and communication as parts of safety improvement. [P3]

Before signing the note

Compare the narrative, diagnoses, orders, prescriptions, referrals, and after-visit instructions. Remove unsupported findings, distinguish current symptoms from history, and preserve uncertainty that matters clinically. If an AI draft was used, review its meaning as well as its grammar. A statement that a patient declined treatment, denied a dangerous symptom, or agreed to a plan can materially change later care.

When a workflow repeatedly produces the same error, correction of one note is only the immediate response. The recurring hazard also needs an owner, a technical or operational change, and follow-up to determine whether the change worked.

Between visits and during absence

Use an identified backup for time-sensitive tasks. Make coverage permissions functional before the absence starts. Staff should know how to reach clinical help and when a telephone or portal request requires urgent escalation. Do not rely on an absent clinician's personal inbox as the sole route for a high-risk matter.

Test downtime procedures before an outage. The SAFER Contingency Planning guide addresses preparation for partial or complete EHR unavailability. A usable local plan explains how to identify patients, obtain essential clinical information, communicate orders and results, and reconcile temporary records after restoration. [P4]

Measure completion and learning

Suggested local measures include unresolved critical results, overdue clinically important referrals, unacknowledged handoffs, medication discrepancies with consequences, and recurring AI or template errors. These are proposed operational measures, not published benchmarks. Define the denominator and clinical exclusions before comparing teams; otherwise, a low number can reflect missed detection or favorable case mix.

Review a small sample of completed tasks to determine whether the apparent closure represents the clinical work. An inbox can be empty while important follow-up remains undone. AHRQ's Measure Dx and Calibrate Dx resources provide approaches to learning from diagnostic events and evaluating diagnostic decision-making. [P5]

A practical ownership table

Sources

• P1: 2025 SAFER Test Results Reporting and Follow-Up Guide, especially practice 2.4.
• P2: 2025 SAFER Clinician Communication Guide.
• P3: AHRQ, Improving Patient Safety in Primary Care by Engaging Patients and Families.
• P4: 2025 SAFER Contingency Planning Guide.
• P5: AHRQ, Tools to Improve Diagnostic Safety.