Source synthesis, not an empirical claim-file study. Note quality can be reviewed through four observable fields: what was observed, which record supports it, what action was taken, and who owns the next step by when.
Research question and scope
This protocol asks whether a claim activity note lets another reviewer reconstruct the event and locate its source without contacting the author. The unit is a note entry associated with a claim event and a timestamp.
The audit does not evaluate whether the underlying claim decision was correct. It also does not score writing style. The focus is provenance, separation of observation from inference, and ownership of the next action.
Population and note export
Define the population from a fixed note export across a stated period and team. Record the systems included, note types, event categories, and the rule for excluding system-generated entries. Include late notes, corrected notes, and notes with no next step.
Sample across event types rather than only long notes. A short document-receipt note can be complete, while a long narrative can contain no source and no owner. Length is not the measure.
Coding frame
Code each note for an identifiable observation, a named source, a stated action, and a next owner with a date. Add a flag for inference presented as fact and a flag for backdating or ambiguous timing. Define what counts as a source: a call log, an email, a portal record, a report, or a recorded statement.
A note can be complete for source and still fail the next-step measure. Keep the fields separate. A composite quality score hides which part of the note is weak and gives a coach nothing specific to fix.
Sampling and reviewer agreement
Stratify by author group, event type, claim segment, and note length band. Protect small groups by reporting aggregates rather than individual results. Record the sample seed and every replacement.
Two reviewers should code a calibration set and compare each field. Disagreement about whether a sentence is an observation or an inference is useful evidence that the codebook needs examples. Preserve first-pass codes and adjudicate with the written rule.
Analysis and limitations
Report the share of notes with a source, the share with a dated next owner, and the count of inference flags by event type. Inspect examples behind each result and remove identifying detail before sharing.
A note can be terse and complete, or detailed and unusable. The audit cannot prove that a sourced note led to a better claim outcome, and it should not be used to compare individual employees without a fair, authorized performance process. System design, templates, and workload can all shape note quality.
Privacy, reporting, and replication
Notes may contain medical, financial, and family information. Use a restricted extract, redact names and claim identifiers in examples, and apply the carrier's retention and access rules.
Retain the export query, coding frame, calibration set, sample seed, reviewer notes, and adjudications. A second reviewer should be able to reproduce the field-level counts and the redacted examples.
Method note
We translated internal-control and privacy guidance into a note-level coding frame, then added a fact-and-inference distinction. No claim notes or personnel records were reviewed.
Sources
Common questions
Is a longer note a better note?
No. The measures are source, observation, action, and next owner. Length does not substitute for any of them.
How should a note with no next step be coded?
As missing a next owner or date, even if the event was fully described. The gap is separate from the factual content.
Can this method be used for coaching?
Only within an authorized performance process and with aggregate safeguards. The audit is designed to improve the note system, not to rank people.
Apply the brief to a real queue
Bring the source records, current procedure, and known exceptions. We will help turn the evidence questions into a bounded support role and review plan.
Discuss your workflow