12.1 Measuring Physician Documentation Performance
Key Takeaways
- Query response rate and agreement rate belong on a named-provider line, not only as a hospital average that can hide non-responders
- A sustained volume of non-responses is the engagement warning; query volume and agree or disagree counts are not engagement scores by themselves
- Unspecified ICD-10-CM code patterns by provider and service identify education targets when the record already contains the missing detail
- Trend physician metrics across months or quarters with raw counts and rates; a single-month rank list is not a performance system
- Education and case-level feedback are first-line CDI tactics; disciplinary, shaming, or privileging language is not a first-line method
12.1 Measuring Physician Documentation Performance
Quick Answer: Measure documentation performance at the provider level with query response rate, query agreement rate, unspecified-code patterns, and education completed, then trend those measures over time. A sustained volume of non-responses is the classic signal of weak engagement—not how many queries you sent, and not how often the provider agreed. Use the scorecard for feedback and education. Do not open with disciplinary framing, public shaming, or privileging threats as a first-line clinical documentation integrity (CDI) tactic.
Program dashboards in Domain IV tell you whether the CDI service is productive and whether hospital case mix index (CMI) and complication capture are moving. They do not tell you which attending, hospitalist, surgeon, or advanced practice professional is actually documenting. Physician documentation performance is a named-provider view of the same query and coding data. The Association of Clinical Documentation Integrity Specialists (ACDIS) inpatient Certified Clinical Documentation Specialist (CCDS) content outline asks you to identify methods for measuring physician performance related to documentation and to track and trend that performance over time. This independent OpenExamPrep chapter helps learners study those methods. It is not an ACDIS product, and OpenExamPrep does not claim ACDIS approval, partnership, official review, or exact equivalence with ACDIS materials.
Why the unit of analysis is the provider
A hospital can post a 90% query response rate while three hospitalists answer almost every query and two surgeons answer almost none. The average hides the problem. Build a line for each documenting professional who receives queries, using a stable identifier (national provider identifier or medical staff ID) and a service-line rollup such as hospital medicine, cardiology, or trauma. Include nurse practitioners (NPs) and physician assistants (PAs) when they are the professionals who enter the diagnostic statements you query. Do not invent a "physician only" rule that the record itself does not use.
Keep denominators honest. A consultant with four queries in a quarter cannot be ranked against a nocturnist with ninety. Report rates and raw counts together. Suppress or footnote cells with tiny denominators so leaders do not treat noise as a personality score.
Query response and agreement, calculated on purpose
Query response rate for a provider in a period is responses received divided by queries sent to that provider. Define "response" in policy before you calculate. A signed addendum, a documented answer on the query form, or a verbal answer that has been entered into the permanent health record all count if that is how your policy records a complete query. Silence, an opened-but-unanswered inbox item, and a "see note" that never addresses the question are non-responses.
Query agreement rate is usually agree responses divided by responses received. Some programs also report agree divided by queries sent. Those two fractions answer different questions. Agree divided by responses asks, "When this provider answers, how often do they accept the clarification?" Agree divided by queries sent mixes engagement with clinical agreement and will look artificially low for a provider who simply does not answer. Publish the formula on the report so two analysts do not compare unlike percentages.
Disagreement is not automatically a quality defect. A compliant query leaves independent judgment with the provider. If the indicators were weak, or the diagnosis is not clinically present, disagree can be the correct patient-care answer. A 100% agreement rate is not a CCDS success target. It can warn that queries are leading, that options are too narrow, or that providers click agree to clear a task. Track agreement over time as a conversation starter, not as a compliance trophy.
Non-response as lack of engagement
Query volume measures CDI workload and case-finding. Agree counts measure how often documentation changed after a response. Disagree counts measure clinical pushback. None of those, by themselves, measure whether the provider is participating in the query process.
A high or rising volume of non-responses—especially after you have confirmed delivery, a reasonable response window, and a working electronic pathway—is the engagement signal. Those queries never entered the provider's decision process. The record still lacks a complete diagnostic statement, coding still lacks a usable response, and education never had a chance to land.
Do not wait for a single missed query to declare a difficult physician. Trend non-responses across months. Look for clusters: an electronic health record pool that does not notify surgeons; night coverage that never sees daytime queries; a service that receives queries after discharge when the attending has already signed. Some apparent non-response is a workflow defect. Fix the pathway first. What remains after a working pathway is a true engagement gap, and that is what you bring to a medical director with cases in hand—not with a threat.
Unspecified-code patterns
International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) includes many unspecified codes. They are valid when the record truly lacks detail. They become a documentation-performance pattern when the record already contains the missing piece—laterality, acuity, type, organism, stage, or linkage—and the diagnostic statement still stops at unspecified.
Useful pattern views include unspecified principal diagnosis rates by attending on high-volume families (heart failure, pneumonia, anemia, malnutrition, respiratory failure); unspecified secondary codes that would have been a complication or comorbidity (CC) or major CC (MCC) if specified, when clinical indicators support specification; and unspecified diagnoses that later affect quality or present-on-admission (POA) reporting.
Pair every unspecified-code report with a "was specificity knowable?" check. If the workup never identified an organism, unspecified pneumonia is not a physician failure. If cultures and the infectious-disease note already name a specific organism and the discharge summary still says only "pneumonia," you have an education and query target. Do not auto-launch a leading query that names a desired code on every unspecified code. Pattern first, then write compliant queries on the cases that meet query criteria.
Education tracking is a metric, not a folder of slide decks
If you educate and never remeasure, you cannot tell whether physician performance changed. Log, at minimum: date, topic, audience (named providers or service), format (huddle, case review, brief on the unit), and owner. Then watch the same providers' response rate, non-response volume, agreement rate, and unspecified-code rate in the following two to three measurement cycles.
Education that never appears in the log cannot be correlated with outcomes. Education that is only "offered to the department" cannot explain why one hospitalist improved and another did not. Named attendance plus named metrics is the method. A later Domain V chapter covers how to build the teaching tools; this section treats education as a measured intervention tied to documentation performance.
Track over time; do not open with discipline
The CCDS content outline asks you to trend individual physician performance over time. A one-month rank list is a snapshot. Snapshots punish vacation months, trauma-call spikes, and new-hire orientation. Use a rolling three-month window or consecutive quarters, and annotate operational shocks such as an electronic health record go-live, a query-platform change, or a service expansion.
First-line CDI tactics are case-level feedback, brief education, query-process repair, and a medical-director conversation that still treats the provider as a clinician. Disciplinary framing—peer-review "delinquency" as the opening move, public leaderboards of worst documenters, blocking operating-room time, or threatening credentials because a query was unanswered—is not a first-line CDI method. It poisons later queries, raises pressure to click agree, and confuses documentation integrity with punishment. If a medical staff process is eventually needed, that is an institutional decision after education and process support have been tried and documented, not a CDI opening move.
| Measure | How to compute it | What a movement usually means | First-line use |
|---|---|---|---|
| Query response rate | Responses ÷ queries sent | A falling rate often means delayed or ignored queries | Confirm delivery and the response window; then coach |
| Non-response volume | Count of queries with no recorded answer | A rising count is the engagement warning | Case review with the provider or service chief |
| Query agreement rate | Agree ÷ responses (state the formula on the report) | Very high or very low both deserve review | Check query quality and clinical validity |
| Unspecified-code pattern | Unspecified codes ÷ eligible encounters, by provider | Rising unspecified with available detail is a teaching target | Education, then compliant queries |
| Education completion | Named attendance, topic, date, owner | No log means you cannot attribute later change | Tie topics to the metrics above and remeasure |
Exam-style scenarios
Scenario: the 92% hospital that is not fine. Leadership celebrates a facility query response rate of 92%. The physician report shows two orthopedic surgeons with 18 and 22 unanswered queries and response rates in the 40% range. The correct CDI reading is that the average is not the story. Non-response volume on those two lines is the engagement problem. Fix notification and timing first; then review cases with the service chief. Do not start with a credentials threat.
Scenario: perfect agreement. One hospitalist agrees with 40 of 40 answered queries and still has 15 unanswered queries. Treating 100% agreement as excellence misses both issues: unanswered work never reached a decision, and perfect agreement on answered queries should trigger a query-quality review, not a gold star.
Scenario: unspecified heart failure. A cardiology group's unspecified heart-failure principal diagnosis rate sits well above peer services, while echocardiography and the history already distinguish systolic from diastolic failure. That pattern is an education target plus compliant queries on cases with indicators. It is not a reason to drop a leading multiple-choice list that only offers the desired codes.
Scenario: the three-query consultant. A spreadsheet ranks a consulting endocrinologist last because two of three queries went unanswered. With a denominator of three, that rank is not a performance system. Report the raw counts, withhold ranking, and look at the hospitalists who generate the bulk of queries.
How to use these measures while you study
On application items, ask which metric answers the question being asked. Volume of queries generated measures CDI case-finding. Volume of agree or disagree responses describes answers that came back. Volume of non-responses describes engagement. Unspecified-code rates describe specificity patterns. Education logs describe whether an intervention even occurred. Analysis items often hinge on time: one month of data is a snapshot; several cycles after education is a trend. Recall items may ask you to name the methods—response, agreement, unspecified patterns, education tracking—and to reject disciplinary framing as the first CDI move.
A hospitalist service posts a high query volume, a high share of agree responses among returned queries, and a smaller share of disagree responses. Which additional finding most clearly indicates a lack of physician engagement with the query process?
A CDI analyst finds that one surgical service has a persistently high rate of unspecified ICD-10-CM principal diagnoses, and the records often already contain laterality, acuity, or organism detail. What is the most appropriate first-line use of that pattern?
Two trauma surgeons have unanswered concurrent queries after the electronic pathway and response window have been confirmed to work. What is the most appropriate first-line CDI tactic?
Which statement about query agreement rate as a physician documentation metric is most accurate?