Skip to main content

Two Staff Editing One Patient Record: Resolve Conflicts Without Losing Work

When two people edit the same patient record, the software should detect when a proposed change is based on an older relevant version and preserve enough context for an authorised person to resolve it. A conflict warning prevents neither mistakes nor lost work by itself. The useful outcome is that accepted changes remain intact, the second person's draft remains available where supported, and any clinical disagreement is reviewed rather than silently merged.

By Bractis · Updated · 9 min read

Indian clinics where reception staff, assistants and doctors update records concurrently or where a doctor uses more than one device.

Identify which record is actually being edited

Two people viewing the same patient does not necessarily mean they are changing the same object. A contact detail, a vital observation, a consultation note and an appointment can have different ownership and version boundaries. Ask the vendor to explain which changes can coexist and which require review.

A broad patient-level lock may block useful work unnecessarily. At the other extreme, treating every field as independent can miss relationships that matter clinically. The decision is about the application's data model and workflow, not a rule that either every screen must lock or every field must merge.

Distinguish three problems. A repeated request concerns whether one operation happened twice. A stale edit concerns whether someone saved against an older record. A clinical disagreement concerns what the record should say. A mechanism that solves one does not automatically resolve the others.

Understand what a version check can and cannot do

RFC 9110 defines If-Match as a request condition that can prevent a change from being applied against an unexpected representation. It is one technical mechanism for avoiding accidental overwrites. Its presence in an architecture description is not proof that every clinic form uses it correctly.

For evaluation, ask what happens when the relevant version changes after a user opens a form. The application should make the conflict understandable and avoid claiming that the rejected update is already saved. It also needs a supported way to preserve or recover the user's intended changes.

A version comparison cannot decide whether an allergy statement, medicine history or clinical conclusion is correct. That needs the appropriate clinician and evidence. Automatic combination is acceptable only where the workflow and data permit it; adding contradictory clinical statements together is not reconciliation.

The checks also need to apply to the actual write, not only the screen's initial load. A page that appeared current when opened may be stale by the time Save is pressed. Testing a fresh page alone will not reveal that gap.

Source: RFC 9110: HTTP Semantics

Use a two-user timeline to expose a silent overwrite

In a fictional trial, clinician A opens a draft labelled version 12 at 09:00. Clinician B, authorised to edit that same draft, makes a correction and saves version 13 at 09:02. At 09:03, A tries to save changes based on version 12. These simple numbers describe the scenario; they are not a required format for a vendor's version identifiers.

A poor result is that A's older copy replaces B's correction without a warning. Another poor result is a warning that destroys A's draft when the page refreshes. A useful result identifies the changed record, preserves the relevant proposed content and gives an authorised person a way to compare it with the current version.

After review, the team may intentionally create a new accepted version, or decide that A's change is no longer needed. Either outcome can be legitimate if it is attributable and understandable. The objective is not that every attempted edit eventually gets applied.

Repeat the trial when users change different objects, such as a contact detail and a separate consultation note. The application should protect the related records without unnecessarily forcing one worker to wait for all other activity on that patient.

On a narrow screen, scroll the table horizontally to read all columns.

Concurrent editing cases worth demonstrating
CaseRisk to examineAcceptance evidence
Two edits to the same draftA newer accepted change is overwritten.Conflict is detected and both intentions can be reviewed.
Contact update beside note editingAn unrelated save clears or blocks the doctor's work.Independent work remains intact within the supported model.
New observation during an open noteThe draft is replaced by an older screen snapshot.The observation and note retain their own authorship and timing.
Signed record changes while another tab is openAn old editable view bypasses current status.The actual write respects current permissions and record state.
User loses connection during conflict recoveryA comparison is mistaken for a confirmed save.The final accepted revision can be retrieved and inspected.
Patient or clinic switchA preserved draft is applied to the wrong context.Draft recovery remains bound to the original record and facility.

Make the comparison useful to the responsible person

A comparison should explain the relevant changed content, not merely display a long technical identifier. Show what the user started from, what is current and what they proposed where the product supports that view. Preserve author and timing context without asking reception to adjudicate a clinical issue.

Keep the user's next action precise. 'Reload' may mean discarding unsaved work, while 'review latest record' may mean opening a comparison that preserves it. Staff should know which is happening before they act. A warning with two ambiguous buttons can turn a protective check into a source of loss.

When the disagreement is clinical, the responsible clinician should resolve it through the appropriate record process. Do not change a signed historical statement merely to satisfy a technical save request. A correction or amendment may be needed, with its own attribution and reason.

If one operator lacks the authority to complete the change, route it to someone who has it. Sharing the doctor's login or temporarily granting broad access removes accountability and does not resolve the underlying version problem.

Count conflicts separately from lost records

Use an incident register that distinguishes detected conflicts, resolved conflicts, discarded proposals and unresolved work. A product that detects more stale edits may be exposing a problem that another product silently ignores. The raw number of warnings is therefore not a reliable quality ranking.

In a fictional pilot with 60 intended saves, 55 are accepted immediately and five are blocked for review. Three blocked proposals are reapplied after reviewing independent changes; one is reconciled through agreement; one remains unresolved. There are 59 accepted operations and one open item, not 65 successful saves.

Five divided by 60 is about 8.3%, the proportion of intended saves initially needing conflict review. It is not a rate of lost patient records. Keep retries and review actions out of that denominator unless you are separately measuring support effort.

Look for a pattern: frequent collisions on one field, a broad save that includes stale hidden data, two users unintentionally sharing one encounter, or insufficient separation between a current note and a longitudinal summary. Use that evidence to change the workflow or ask the vendor for a specific correction.

Test the workflow without manufacturing a live clinical dispute

Use synthetic records and explicitly authorised test users with realistic roles. Demonstrate at least one simultaneous edit, one signed-state change and one interruption during reconciliation. A single screenshot saying 'conflict protected' is not a substitute for the resulting record.

Ask what happens to a preserved proposal after logout, browser closure or device replacement. The answer may depend on how drafts are stored. Record that limit so a doctor does not leave an unresolved proposal on a device assuming another operator can retrieve it.

Retain the test's expected outcomes and repeat the relevant cases after material changes to the form or permission model. Acceptance belongs to a specific workflow; a successful note-editor test does not prove that billing, booking and every other module use identical controls.

A contact correction should not erase a longer history

A fictional receptionist updates a patient's phone number while the doctor adds a detailed history. The vendor trial checks the final contact record and clinical draft separately. The clinic rejects a design that submits an old whole-patient snapshot from either screen and silently restores older data. It also asks for a realistic same-note conflict so the independent-field test is not mistaken for complete concurrency protection.

Test concurrent record editing

Use these prompts in your clinic’s own notes. Record the result, responsible person and next action for each item.

Record boundaries
Identify the objects your staff change at the same time and who owns each decision.
Collision trial
Use two authorised synthetic users to save from an older and newer version, then inspect the result.
Review usability
Check whether proposed content survives and whether the responsible person can understand the difference.
Resolution measure
Count intended operations, initial conflicts, accepted resolutions and outstanding items separately.

Questions doctors ask

Does a vendor saying 'we use If-Match' prove the clinic cannot lose edits?

No. It describes a possible technical control. Test the actual form, write path, draft preservation, permissions and conflict-resolution process.

Should the software automatically merge every change?

No. Some independent changes can coexist, but conflicting clinical information requires the appropriate professional's review. Combining statements is not the same as resolving their meaning.

Does a conflict warning mean data has already been lost?

Not necessarily. It may show that the software prevented an overwrite. Verify the accepted record, the proposed draft and what the user must do next.

Is refreshing the page a safe way to resolve a conflict?

Only after understanding what will happen to the proposed work. Use the supported comparison or recovery process; a refresh may discard an unsaved proposal.

How should a clinic measure editing conflicts?

Track initial conflicts, final decisions and unresolved work against clearly defined intended operations. Do not count repeated attempts as additional patients or assume every warning represents loss.

Sources and scope

Updated on . Original collaboration model and fictional version/count examples. Protocol guidance does not establish a particular Bractis implementation, clinical correctness or universal conflict-free editing. No independent clinical or legal review is claimed.

  • RFC 9110: HTTP Semantics

    Protocol semantics for retries, conditional requests and response codes; clinic scenarios are original.

Try the workflow
with Bractis.

Ask Bractis to demonstrate two authorised users editing one synthetic note and, separately, two independent parts of one patient record. Inspect accepted versions, preserved proposals and the final resolution.

Request a clinic demo