The second stage of an evaluation term is usually specification: writing down, at a level of detail that would let a stranger reproduce your figures, exactly what each measure counts. That means numerator, denominator, eligibility, exclusions, timing rule and data source for every measure you intend to report, including the process and balancing measures students most often skip. The discipline is not bureaucratic. A measure that is loosely specified will produce a different number every time somebody runs it, and a result nobody can reproduce cannot support a sustainability decision. Your section may print this as NR 709A or NR709A; it is the same course. Chamberlain publishes no syllabi outside Canvas. The placement here is our teaching judgment from the course's catalog arc; your section's rubric decides what your week actually asks.
What NR-709A Week 2 asks for
What happens when a measure meets the data that is supposed to hold it? A school-based health center that had implemented an asthma action plan step pulled its first post-implementation numbers and found the completion rate had barely moved. Then somebody asked what was in the denominator. It turned out to be every child with an asthma diagnosis code ever entered at the site, which included several who had graduated, moved out of the district, or been coded once during an urgent visit for something else and never seen again. The measure the team thought they were reporting was completion among children currently receiving asthma care at this site. The measure the report was actually producing was something else entirely, and the difference was large enough to hide a real improvement.
That is what this stage is for. In the planning courses of a project sequence, measures are chosen. Here they are specified, which is a different and more demanding act, and it can only be done properly once you have seen the data. A specification names the population that goes into the denominator and the rule that determines whether each case lands in the numerator, plus the timing that governs both. It also names what is excluded and why, and exclusions are where most reproducibility is lost.
The deliverable at this depth is usually a measurement or analysis plan, sometimes as a section of a larger evaluation document and sometimes as a standalone piece with a table. Expect to be scored on completeness and on the honesty of the limitations you attach. A specification that quietly omits the fact that one measure depends on a free-text field nobody fills reliably is a specification that will embarrass its author at presentation.
The boundary that governs this manual. Clinical hours, hour logs, encounter counts, site records, preceptor evaluations and signatures are your own and your site's, and they are never drafted, reconstructed or estimated with outside help, nor should anyone perform or document clinical activity on your behalf. What can be supported is the written layer: specifying measures, structuring an analysis plan, presenting results, and writing about work you genuinely did. Any real encounter that appears in your writing is de-identified first, and the site is described by type and volume rather than by name.
The NR-709A Week 2 method, step by step
Seven moves for specifying a measure so that two people would produce the same number.
-
Write the denominator as an eligibility rule, not a group name
Patients aged twelve and over with a visit of a stated type during the measurement period, excluding those with a documented reason for exclusion. Every clause in that sentence has to be retrievable from a field somebody can point to.
-
Write the numerator as an event with a location and a clock
Where the evidence of the event lives, in which field or document, and within what time relative to the denominator event. Same day, same visit and within thirty days produce different numbers from the same underlying care.
-
Test the specification against ten real cases before you finalize it
Take ten de-identified records, apply the rule by hand, and see whether you can classify each one without a judgment call. Every case that requires a decision is a specification gap, and finding ten of them now saves the evaluation later.
-
List exclusions with a reason attached to each
Clinical exclusion, data quality exclusion, out-of-scope population. Then count how many cases each removes, because an exclusion that quietly removes a third of the denominator has changed what you are measuring.
-
Specify a process measure that shows whether the change happened
Outcome measures move slowly and for many reasons. A process measure tracking whether the step was actually performed is what lets you tell a failed idea apart from an idea that was never delivered, and it is the measure most likely to be missing from a student plan.
-
Specify a balancing measure and mean it
Ask what your change could plausibly have made worse: visit length, a downstream referral queue, time spent by a role you did not consult, another screening that got displaced. Choose one that is genuinely at risk rather than one that is safely irrelevant.
-
Record the provenance of every measure in one table
Source system, report name or query, who runs it, refresh schedule, and the date range you pulled. This table is what makes your evaluation reproducible and it takes twenty minutes to maintain and days to reconstruct.
A layout and word budget for a measurement specification
Our frame for a measurement plan document, sized for roughly 1,200 to 1,500 words plus the specification table. It is our own outline rather than anything the university issues, and your week's rubric outranks it wherever the two disagree.
| Section | What belongs in it | Word target |
|---|---|---|
| What each measure is for | The decision each measure informs, so a reader can tell why these and not others. | 120 to 150 |
| Primary outcome specification | Denominator rule, numerator event, timing, exclusions and source, written to reproduce. | 230 to 280 |
| Process measure specification | The delivery step being counted, how its performance is evidenced, and how completely that field is used. | 190 to 230 |
| Balancing measure specification | The plausible harm being watched, why it is plausible, and how it will be detected. | 170 to 210 |
| Provenance table narrative | Where each measure comes from, who can regenerate it, and how stable the report specification is. | 160 to 200 |
| Known data quality problems | Fields with poor completion, changes in coding practice, and any period where collection was interrupted. | 180 to 220 |
| Analysis approach | How the measures will be compared and displayed, stated plainly and without borrowed trial vocabulary. | 150 to 190 |
Evidence craft for measurement writing
Prefer an existing validated specification where one exists. National quality measures and professional society definitions come with published specifications, and adopting one lets you compare your site against something outside itself. Cite the specification with its steward and version, and say plainly where you departed from it and why, because undocumented departure is the thing that makes a comparison misleading.
Report completeness alongside performance. If the field your numerator depends on is completed in most encounters but not all, that proportion belongs in the paper. A performance figure calculated on a partially populated field is not wrong so much as unreadable without knowing how partial it was.
Do not import a measure's reputation. A measure used in a published study of a large integrated system may be impossible at a two-clinician site, and forcing it produces a number built on a field nobody uses. Choose measures your site can actually produce, and note in the paper where feasibility drove the choice.
Freeze the specification and date the freeze. Write the date you locked each definition and treat later changes as documented amendments with reasons. Specifications that keep evolving until the results look reasonable are the practical equivalent of choosing an outcome after seeing the data, and doctoral readers ask about it.
Keep de-identification inside the analysis workflow. Use aggregate or de-identified extracts wherever the site allows, keep any linkage separate from your academic document, and never place identifiers in a paper, a slide or an appendix. Describe the site by type and volume, and where a case illustrates a specification problem, strip it to the structural detail that makes the point.
Five mistakes that cost points in this week's territory
- A denominator described by name. Eligible patients is not a specification; a retrievable eligibility rule is.
- No process measure. Without one, a flat outcome cannot be distinguished from a change that was never delivered.
- A decorative balancing measure. Choosing something your change could not possibly affect satisfies the row and tells the site nothing.
- Silent exclusions. Exclusions that appear in the query and not in the paper make the result impossible to interpret and easy to distrust.
- Trial vocabulary in an analysis plan. Power, arms and randomization describe a design you are not using and undercut the quality improvement framing.
Before you submit
- Each measure has a denominator rule, a numerator event, a timing rule and a source
- The specification was tested against real cases and the ambiguous ones resolved
- Every exclusion carries a reason and an approximate count
- A process measure exists that shows whether the change was delivered
- The balancing measure is one your change could genuinely worsen
- Field completeness is reported wherever a numerator depends on documentation
- Any departure from a published specification is stated with its reason
- The date each definition was frozen appears in the document
- No identifiers appear anywhere in the paper or its appendices
Writing the measurement plan for NR-709A?
Send the rubric and your draft measure list out of Canvas. A premium original draft comes back in 24 to 48 hours with denominators written as rules, exclusions counted, and process and balancing measures specified rather than named, and revisions run until the grade lands.