NR-651 · Week 7 of 8 · Evaluation and measurement planning

NR-651 Week 7 Evaluation and Measurement Planning: How to Write It

The short answer

An evaluation plan is the promise a project makes about how it will be judged, and it has to be written before results exist or it is worth nothing. The seventh stage typically asks you to derive measures from your objectives, define each one so precisely that two people counting independently would agree, name where the data lives and who may access it, choose a comparison that makes movement interpretable, and add the balancing measures that will tell you whether the change harmed something else. Your section may print this as NR 651 or NR651; 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.

NR-651 Week 7 grading scale at Chamberlain, the criterion levels this assessment is scored on, from Chamberlain Tutors
How Chamberlain grades NR-651 Week 7, visualized by Chamberlain Tutors.

What NR-651 Week 7 asks for

On a 60-bed rehabilitation wing, everyone talks about sending residents back to the hospital as though it were one event with one meaning. An evaluation plan cannot afford that looseness. Does an observation stay count. Does a scheduled outpatient procedure that becomes an admission count. Does a resident who transfers on day 3 and again on day 26 count once or twice. Is the denominator admissions to the wing, discharges from it, or resident-days. Is the window 30 days from admission or 30 days from discharge. Six reasonable people will answer those six questions six different ways, and any pair of answers produces a different number from the same records. Measurement planning is the written work of settling all of it in advance.

The second demand is coverage across levels. Objectives written at process, output and outcome levels each need measurement, and a plan that measures only the outcome will be unable to explain a null result. If the return-to-acute rate does not move, the plan needs to be able to say whether the intervention was actually delivered, because a project that was never implemented and a project that was implemented and did not work are different findings with different conclusions. Fidelity measurement is what separates them, and it is the element most often missing from student evaluation plans.

Balancing measures are the third demand and the mark of maturity. Every change consumes something. A structured review before transfer consumes minutes at the least covered hour; a new document consumes attention that was going somewhere else. Naming one or two measures that would detect that cost, and committing to report them whatever they show, is what distinguishes an evaluation from an advertisement for the project.

Finally, this stage has to be honest about what a short project can establish. Small denominators, no randomization and a single site mean the design supports feasibility and direction rather than proof. Say so in the plan rather than in the discussion after the fact. Graders in graduate project courses consistently reward pre-stated limitations and penalize the same limitations when they arrive late as excuses.

Boundary setting: data access is governed, hours stay your record

Evaluation planning brings you closer to organizational data than any other stage, so the boundary has two parts. The first is the familiar one: practicum hours, hour logs, activity or encounter documentation, site attendance records and every preceptor, mentor or sponsor evaluation are your own record, produced by you, never drafted, reconstructed or estimated by anyone else.

The second part is about the data itself. What you may access, for what purpose, in what form and with what approvals is governed by your site and your program, not by convenience. Before you write a plan that depends on a data source, confirm you will be permitted to use it and say in the plan who authorized it. Work from aggregate or de-identified extracts wherever possible, take only the minimum necessary, and never move identifiable information out of a site system into a personal document or drive. A plan that names a data steward and an approval route reads as professional; a plan that assumes access reads as naive, and can become an incident.

De-identification governs everything you submit. No resident, patient, family member or staff member should be identifiable in your measurement plan, your sample data displays or your worked examples. Where a table would leave a single individual recognizable because the cell count is one, suppress or aggregate it and say that you have done so.

The NR-651 Week 7 method, step by step

Six moves for building a measurement plan that will survive contact with real records.

  1. Derive every measure from an objective, and check the reverse

    Each objective gets at least one measure, and every measure traces back to an objective. Orphan measures mean you are counting something for its convenience, and orphan objectives mean part of your project will end unevaluated.

  2. Write numerator, denominator, exclusions and window for each measure

    Four lines per measure, in plain language, with every ambiguous term defined. This is the most valuable page in the whole proposal and the one students most often skip. If two people could not count identically from your definition, keep editing it.

  3. Classify measures by type and check the set is complete

    Structure, process, outcome and balancing. A complete set usually has at least one fidelity or process measure, one outcome measure, and one balancing measure. If your set is all outcome, you cannot explain a null result. If it is all process, you never tested the aim.

  4. Name the data source, the collector, the cadence and the approval

    Where the number lives, who pulls it, how often, in what form, and who authorized access. Add what you will do when the field is blank or the record is inconsistent, because in post-acute records it will be, and a plan that ignores missing data produces a denominator argument later.

  5. Choose the comparison and defend it

    Baseline period against implementation period, a pilot unit against a comparable unit, or a series of points over time. Each has a rival explanation: seasonality, case mix drift, other initiatives, awareness created by the project itself. Name the main threat for the comparison you chose and say how you will address or acknowledge it.

  6. Pre-state the analysis and the reporting cadence

    What will be calculated, how it will be displayed, at what interval it will be reviewed and by whom. Committing in advance to how results will be read removes the temptation to select an analysis after seeing the numbers, and it gives your sponsor a rhythm they can hold you to.

A layout and word budget for an evaluation plan

Our frame for the evaluation component of a policy proposal, sized for roughly 1,300 to 1,700 words plus the measure specification table. It is our own outline rather than anything the university issues, and your week's rubric outranks it wherever they disagree.

SectionWhat belongs in itWord target
Evaluation questionsThe two or three questions the evaluation exists to answer, traceable to the aim and objectives.110 to 150
Measure setEach measure with type, numerator, denominator, exclusions and observation window in plain language.320 to 400
Data sources and accessWhere each number lives, who pulls it, at what cadence, under whose authorization, and how missing data is handled.200 to 250
Design and comparisonThe comparison chosen, the reason, and the main rival explanation it leaves open.200 to 250
Analysis planWhat will be calculated and displayed, stated before any data exists.150 to 200
Balancing and unintended effectsWhat the change might cost, how it would be detected, and the commitment to report it either way.150 to 190
Limitations and reportingWhat this design can and cannot establish, plus the review cadence and audience for results.170 to 220

Evidence craft for measurement planning

Prefer existing measures over invented ones. Where a national quality program, a payer program or a professional body already defines a measure for your outcome, adopt its specification and cite it. Borrowed definitions give you comparability, defensibility and a benchmark, and inventing a private definition forfeits all three.

Treat small numbers with visible caution. On a 60-bed wing over a short period, a shift of two events can move a percentage several points. Report counts alongside proportions, avoid comparing percentages built on single-digit denominators, and consider displaying data as a series over time rather than as two bars, because a series shows variation that a before-and-after pair conceals.

Plan a data quality check into the method. Say that a stated number of records will be re-abstracted independently and that discrepancies will be reconciled and reported. This one sentence answers the question a careful reader will otherwise ask, and it takes very little time to execute.

Measure fidelity, not just outcome. A simple count of how often the intervention was delivered as specified turns an ambiguous result into an interpretable one. Without it, every disappointing outcome has two explanations and you cannot choose between them.

State the limits of the design in the plan, not in the discussion. Single site, small denominator, no randomization, concurrent initiatives, and observation by people who know the project is running. Naming these in advance is analysis; naming them afterward reads as apology.

Five mistakes that cost points in this week's territory

  • Measures without denominators. A count of anything is uninterpretable until the reader knows out of how many and over what period.
  • Definitions loose enough to argue about. If a term in your numerator could be read two ways, the result will be disputed by whoever dislikes it.
  • No fidelity measure. Without one, a null outcome cannot be distinguished from an intervention that never happened.
  • No balancing measure. An evaluation that can only find good news is not an evaluation, and graduate readers say so.
  • Assumed data access. A plan that depends on a report you have not been authorized to receive is a plan with a hole in the middle.

Before you submit

  • Every objective has a measure and every measure has an objective
  • Each measure specifies numerator, denominator, exclusions and window
  • The set includes at least one fidelity measure and one balancing measure
  • Data sources name a collector, a cadence and an authorization
  • Missing and inconsistent data have a stated handling rule
  • The comparison is defended and its main rival explanation is named
  • Limitations appear in the plan rather than being saved for the results

Designing the evaluation this week?

Send the rubric out of Canvas with your aim, objectives and whatever you know about available data. A premium original draft comes back in 24 to 48 hours, with measures defined tightly enough that two abstractors would agree and limitations stated up front, and revisions run until the grade lands. Hours, logs and evaluations stay entirely your own record.

Questions students ask about this stage

I can only measure process, not outcome. Is that acceptable?
It is frequently the honest answer for a short project in a small setting, and it is acceptable when it is stated deliberately rather than discovered at the end. The move is to write the outcome measure anyway, define it fully, and say plainly that the observation period available cannot detect a change of the size the literature suggests at your volume. Then commit to the process and fidelity measures that a short window can support, and add the outcome measure to a sustainability plan with a stated review point after the project period. That structure earns credit twice: it shows you understand what your design can carry, and it leaves the organization with something to continue. A project that overclaims a tiny outcome movement as success is weaker than one that measures delivery well and says what would need to happen next to test the outcome properly.
How long should the measurement period be inside an eight-week session?
Long enough to accumulate a denominator that means something, which usually pushes measurement to the shortest defensible window and forces you to choose a high-frequency measure. Ask how many eligible events occur per week in your setting. If the answer is two, an outcome measure over four weeks will not tell you anything and a process measure over the same weeks might. Also account for the calendar honestly: approval, printing, training and go-live all consume time before a single measurable event occurs, and students routinely plan a measurement period that begins in a week the intervention will not yet exist. Write the timeline first, mark the go-live date, then count the weeks that remain. If the answer is uncomfortably few, say so in the limitations and design the measure set around what those weeks can support.
Do I need statistical testing, or is a simple comparison enough?
Follow your rubric first, and where it leaves the choice open, match the analysis to what the design can support. Many small single-site projects are best served by descriptive reporting and a display of data over time, because a significance test on a handful of events invites a conclusion the data cannot bear in either direction. If you do test, say in advance what you will test, name the test and why it fits the data type, and report the effect in real units alongside any p value. Above all, state before you look at the data how results will be read, including what would count as no change. Deciding on an analysis after seeing which comparison looks best is the error that most damages a project's credibility, and it is entirely avoidable by writing this section now.

Keep going

Online now