NR-652 Week 3 Interim Data Analysis: How to Write It

What NR-652 Week 3 asks for

On a transitional care unit running a new post-discharge follow-up call, the first interim pull produced a completion figure that nobody believed. Half the calls were showing as not completed, and the unit knew they were happening. The cause turned out to be mundane and instructive: two staff were documenting the call in a progress note while the audit was counting a checkbox, so real work was invisible to the measure. That discovery is worth more than any number in the pull, and it is exactly what an interim analysis exists to catch. Data quality is not a footnote at this stage; it is the first section.

The second thing this stage asks for is an honest relationship with uncertainty. Three or four weeks of events in a small setting produce numbers that move for reasons having nothing to do with your intervention: a single high-acuity admission, a holiday week, one staff member on leave. Reading movement in such data as effect is the classic capstone error, and reading the absence of movement as failure is the same error wearing different clothes. The written task is to describe what you see, name the plausible explanations including chance, and say what would need to be true by the end for a conclusion to be defensible.

Third, an interim analysis is a decision document. Its purpose is not to entertain the reader with early numbers but to decide something: continue as planned, adjust a specific element, fix the measurement, or escalate a problem to the sponsor. Each decision belongs in writing with a reason and a date, both because your final report will need the trail and because a sponsor who learns about a problem in week seven will not forget that you knew in week three.

Deliverables here often combine a short analytic write-up with a display of data to date and a statement of adjustments. Where your section runs a discussion, post final copy and be careful with interim claims; contributions do not reopen after submission in Canvas, and an early claim of success is uncomfortable to walk back in public four weeks later.

Boundary setting: real data, real access, your own record

Two boundaries govern this stage. The first is the practicum record: hours, hour logs, activity or encounter documentation, attendance verification and any preceptor, mentor, sponsor or faculty evaluation are yours, produced by you, and never drafted, reconstructed or estimated with outside help. Interim pressure does not change that.

The second is data governance. Work only with data you are authorized to access, take the minimum necessary, and keep identifiable information inside the systems that are meant to hold it. De-identify before anything travels into a document, a spreadsheet on a personal device or a submitted assignment. Where a table cell would be small enough to identify one resident or one staff member, suppress or aggregate it and say so. Nothing in a capstone is worth a privacy incident, and graduate readers notice when a student handles this correctly.

There is also an integrity line specific to analysis. Numbers reported must be the numbers found. Excluding an inconvenient week, quietly changing a definition to make a figure look better, or filling a gap with an estimate are all forms of fabrication even when the underlying project is genuine. If data is missing, say what is missing and why. Help with an interim analysis means help with structure, display, cautious interpretation and clear writing; it never means supplying or improving the numbers themselves.

The NR-652 Week 3 method, step by step

Six moves for an interim look that will not embarrass you in week eight.

1. Freeze the definitions, then pull

   Copy the measure specifications from the approved plan into the top of your working file and count against them exactly. Any definitional question that arises during the pull gets resolved once, written down, and applied to every period including the baseline.

2. Audit the data before you interpret it

   Re-abstract a small sample independently, compare, and reconcile. Look specifically for the failure that hit the transitional care unit: real work happening in a place the measure does not look. Report the agreement rate you found rather than asserting that the data is sound.

3. Display the series before you compute a comparison

   Plot the measure by week or by small batch across baseline and implementation. A series shows whether the process was stable before you touched it, whether anything shifted after go-live, and how much ordinary variation exists. Two averages side by side conceal all three.

4. Read the pattern with the right vocabulary

   Distinguish ordinary variation from a signal, using an attributed improvement or statistical process control framework rather than your impression of the chart. A single high or low point is usually noise; a run of points on one side of the centre line is a candidate signal worth writing about.

5. List every rival explanation before you claim any effect

   Seasonality, case mix, staffing, a concurrent initiative, awareness created by the project, and measurement change. Write them out and say which you can rule out and which you cannot. This paragraph will migrate almost intact into your discussion section later.

6. Close with decisions, each with a reason and a date

   Continue, adjust one named element, repair the measure, or escalate. State what you expect the decision to change and when you will look again. An interim analysis without decisions is a status report wearing an analytic heading.

A layout and word budget for an interim analysis

Our frame for a mid-project analytic write-up in a capstone practicum, sized for roughly 1,100 to 1,400 words plus the data display. It is our own outline rather than anything the university issues, and your week's rubric outranks it wherever they disagree.

Evidence craft for interim reporting

Give counts, not only proportions, and give the period. In a setting with a handful of eligible events per week, percentages swing violently and mean little on their own. Four of seven this week and five of six last week is information; 57 percent then 83 percent is a story about noise.

Attribute the analytic method. If you use rules for identifying a signal in a time series, name the framework you took them from and apply them as written. Borrowing the vocabulary of process control without its rules is the kind of decorative usage that graduate graders mark down.

Report missing data explicitly. How many eligible events had no usable record, what you did about them, and which direction their absence would bias the result if they differ from the rest. Silent exclusion is the most common and least defensible handling.

Keep interim language conditional. Consistent with, not yet distinguishable from ordinary variation, and would need a further stated number of weeks to assess. Definite verbs at an interim look are a promise you may not be able to keep in the final report.

Document the analysis decisions as you make them. Which cases were included, which were excluded and why, how ties or ambiguous records were treated. Written now, this becomes your methods section. Reconstructed later, it becomes an argument with yourself.

Five mistakes that cost points in this week's territory

• Declaring success at week three. An early favourable number is the least reliable number the project will produce, and claiming it damages everything you write afterwards.
• Skipping the data quality check. The most valuable finding of most interim looks is a measurement defect, and it is invisible without re-abstraction.
• Two bars instead of a series. Before-and-after averages hide stability, trend and variation, which is where the actual interpretation lives.
• Changing a definition to improve a number. Even with an innocent motive this destroys comparability with the baseline and reads as manipulation.
• An interim look with no decision. If nothing follows from the analysis, the analysis has not done the work the stage asks for.

Before you submit

• Measure definitions are copied from the approved plan and applied to every period
• A re-abstraction check is reported with its agreement rate
• Counts and denominators appear beside any proportion
• Data is displayed as a series across baseline and implementation
• Rival explanations are listed with what can and cannot be ruled out
• Missing data is quantified and its likely direction of bias stated
• Each decision carries a reason, an owner and a date for the next look

Questions students ask about this stage

Keep going