HI560 Unit 6 Population Comparison Example

Like With Like: A Unit 6 Population Comparison Standardizing Two Sites' Incomplete MRI Rates by Exam Group, Age and Slot Length

[Student Name]

Purdue University Global

HI560: Health Care Data Analytics

Unit 6 Population Comparison

[Instructor Name]

[Date]

Thistlewood Health, its sites, every cell count and every rate are composite; the sources cited are real.

The Crude Gap and Why It Is Not Enough

Unit 5 found that 4.69 percent of Fennimore exams and 3.64 percent of hospital exams ended incomplete, a crude difference of 1.05 points. Unit 3 showed three ways the sites' exams differ: brain and neck studies make up a larger share of Fennimore's work, its patients are slightly younger, and close to 64 percent of its bookings use the shortest slot length. The first two describe what arrives at the door. The third is a choice the site makes.

Method

Direct standardization was chosen over a regression model because the imaging director can check the arithmetic cell by cell. Exams were cross-classified into nine cells by exam group (head-first neuro, torso, feet-first extremity) and age band (under 40, 40 to 64, 65 and older). Each site's rate within each cell was then applied to the combined population's mix of cells, giving the rate each site would have if both served the same exams and patients (Rothman et al., 2008). A logistic regression with the same terms gave a similar answer and is noted for completeness.

The Stratified Table

Each line gives the combined number of exams in the cell, then the hospital rate and the Fennimore rate.

1. Head-first neuro, under 40: 1,510 exams; hospital 4.9 percent, Fennimore 5.9.

2. Head-first neuro, 40 to 64: 2,940; 5.4 and 6.4.

3. Head-first neuro, 65 and older: 1,680; 5.8 and 6.9.

4. Torso, under 40: 1,600; 3.1 and 3.7.

5. Torso, 40 to 64: 4,140; 3.4 and 4.0.

6. Torso, 65 and older: 2,620; 3.8 and 4.5.

7. Extremity, under 40: 1,070; 1.4 and 1.6.

8. Extremity, 40 to 64: 1,630; 1.6 and 1.9.

9. Extremity, 65 and older: 650; 1.9 and 2.3.

Fennimore's rate is higher in all nine cells, but the gaps inside cells are smaller than the crude gap, because so much of Fennimore's volume falls in the head-first cells where every site struggles most. Two cells hold fewer than ten incomplete exams at one site, extremity exams aged 65 and older at Fennimore with 5 and extremity under 40 at Fennimore with 7, and their rates should be read loosely.

Every cell favors the hospital, yet each cell's gap is smaller than the crude one, which is the signature of a difference partly made by case mix.

Standardized Rates

Standardized for exam group and age, the hospital's rate is 3.79 percent and Fennimore's is 4.48 percent, a difference of 0.70 points, 95 percent interval 0.10 to 1.29. Adding slot length as a third factor, with two levels, 30 minutes or less and longer, gives 18 cells and a difference of 0.38 points, interval minus 0.24 to 1.00.

What Should Be Adjusted

The core question is not computational. Exam group and age describe the exam and the patient before either reaches the scanner, so adjusting for them is uncontroversial: no one would want Fennimore judged worse simply because the neurology clinic sends it brain studies.

Slot length is different. Fennimore books short slots as a scheduling policy to increase throughput, and a technologist working against a 30-minute slot may have less time to coach an anxious patient or pause for a break. Adjusting for slot length removes whatever part of Fennimore's excess the short slots themselves cause. That is the right thing to do if the question is whether Fennimore's staff perform worse with comparable exams under comparable time; it is the wrong thing to do if the question is whether Fennimore, as it is run, completes fewer exams.

The paper therefore reports both figures and says which answers which question. The figure of 0.70 points describes how the two sites differ for comparable exams and patients, including the effect of their scheduling choices. The figure of 0.38 points describes how they differ when exams also had comparable time. For the imaging director's question about Fennimore as it operates, 0.70 is the headline. The difference between the two, about a third of a point, is the share plausibly attributable to slot length, and it suggests a lever the site controls.

The Bridge From Crude to Adjusted

1. Crude difference: 1.05 points.

2. Minus 0.35 for exam group and age: 0.70 points.

3. Minus 0.32 for slot length: 0.38 points.

Limits

Three factors could not be measured. Weight is missing on 17 percent of Fennimore exams, so body size, which affects comfort in the bore, was left out. Anxiety itself is captured only by a booking question asked unevenly across sites. And the analytic file excludes exams scheduled with sedation, which happen only at the hospital, so any patients most likely to struggle who were sedated in advance are absent from the hospital's side. Dewey et al. (2007) found that claustrophobic reactions vary with exam type and patient characteristics in ways a single file seldom captures fully, and pooled studies put claustrophobia among the most frequent reasons scans stop (Munn et al., 2015), so the unmeasured share could be meaningful.

References

Dewey, M., Schink, T., & Dewey, C. F. (2007). Claustrophobia during magnetic resonance imaging: Cohort study in over 55,000 patients. Journal of Magnetic Resonance Imaging, 26(5), 1322-1327. https://doi.org/10.1002/jmri.21147

Munn, Z., Moola, S., Lisy, K., Riitano, D., & Murphy, F. (2015). Claustrophobia in magnetic resonance imaging: A systematic review and meta-analysis. Radiography, 21(2), e59-e63. https://doi.org/10.1016/j.radi.2014.12.004

Rothman, K. J., Greenland, S., & Lash, T. L. (2008). Modern epidemiology (3rd ed.). Lippincott Williams & Wilkins.