HI560 Unit 2 Data Set Description Example

Where the Rows Came From: A Unit 2 Data Set Description of 21,906 MRI Exams From Two Imaging Sites

[Student Name]

Purdue University Global

HI560: Health Care Data Analytics

Unit 2 Data Set Description

[Instructor Name]

[Date]

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

Provenance

The extract was pulled on January 15, 2026, from Thistlewood Health's radiology information system, covering every MRI exam scheduled from January 1, 2024, through December 31, 2025. Each exam row was joined by accession number to the image archive's series log, which records each acquired sequence with its time stamp. The query selected exams by modality code and scheduled date, not by completion, so canceled and incomplete exams are included. The hospital suite contributes 12,470 rows and the Fennimore outpatient center 9,436.

Unit of Analysis

One row is one scheduled exam attempt, identified by accession number. A patient with two exams on different days appears twice; a patient whose exam stopped and was finished in a second slot the same day may appear once or twice depending on how the technologist handled the record, which the quality assessment examines. Patients are not the unit here, because the question concerns what happens to exams.

Fields by Purpose

Before analysis, the patient's chart number, name and birth date were stripped out and replaced with a study key and an age in whole years.

1. Exam descriptors: site, scanner identifier, procedure code, body region, contrast ordered, protocol name, scheduled slot length in minutes.

2. Patient descriptors: age, sex, patient class (outpatient, inpatient or emergency), height and weight where recorded, the booking question on claustrophobia (yes, no or not asked).

3. Process fields: scheduled time, check-in time, scan start and end time, sedation or anxiolytic ordered, anesthesia involvement.

4. Outcome fields: completion status (complete, incomplete, canceled), incompletion reason chosen by the technologist, number of series acquired, number required by the protocol.

5. Linkage fields: accession number, order number, prior attempt accession where the technologist recorded one.

What Is Missing

Missing data are reported by field and by site, because a gap concentrated at one site can look like a difference between sites.

1. Incompletion reason is blank on 22 percent of incomplete exams at the hospital and 9 percent at Fennimore. Where present, technologists choose among claustrophobia, pain, motion, equipment and other, and the boundary between motion and claustrophobia is applied differently at the two sites.

2. Weight is missing on 9 percent of hospital rows and 17 percent of Fennimore rows, so body size cannot be used for adjustment without losing many Fennimore exams.

3. The booking claustrophobia question reads not asked on 14 percent of hospital rows and 3 percent of Fennimore rows, because the hospital's referral clinics book some exams directly.

4. Scanner identifier is blank on 230 rows, all from a three-week period after a software upgrade.

5. Prior attempt accession is filled on only 41 percent of exams that were repeats, so repeat attempts must be found by matching patient, procedure and date.

The field that decides the outcome, completion status, is complete on every row, but the field that would explain it, the reason, is the least complete in the file.

Narrowing to the Analytic File

The later units compare adult outpatient first attempts that were scheduled without sedation. The file narrows in five steps.

1. Hospital: 12,470 rows. Remove 1,480 inpatient and emergency exams, leaving 10,990.

2. Remove 310 exams for patients under 18, leaving 10,680.

3. Remove 610 exams with sedation, an anxiolytic or anesthesia ordered in advance, leaving 10,070.

4. Remove 160 repeat attempts, leaving 9,910.

5. Remove 60 duplicate accession rows created by protocol changes, leaving 9,850.

Fennimore: 9,436 rows, no inpatient or emergency exams. Remove 640 exams for patients under 18, leaving 8,796. No exams had sedation ordered, because Fennimore has no nurse on site to give it. Remove 754 repeat attempts, leaving 8,042, and 52 duplicate rows, leaving 7,990.

The analytic file holds 17,840 exams: 9,850 at the hospital and 7,990 at Fennimore. Fennimore's larger number of repeat attempts deserves a note: when an exam stops there, the patient is often rebooked into a later slot the same week, which the hospital less often does.

How the Description Was Checked

Counts at each narrowing step were reproduced by a second analyst from the same query. Field definitions were confirmed with the imaging informatics lead, who also explained the scanner identifier gap. Describing data before analysis in this way follows the view that data quality depends on whether a data set suits a particular use, not on whether it is error-free in general (Wang & Strong, 1996), and the framework used in Unit 4 builds on that view (Kahn et al., 2016). A review of electronic record quality methods also found that completeness and correctness are the dimensions most often assessed, and that they must be judged against the intended reuse (Weiskopf & Weng, 2013).

Questions for the Data Owner

1. Can technologists record a limited study separately from a complete one?

2. Is the booking question asked the same way at both sites?

3. When an exam stops and resumes the same day, which accession carries the images?

4. Can the anxiolytic order be linked to the booking question answer?

References

Kahn, M. G., Callahan, T. J., Barnard, J., Bauck, A. E., Brown, J., Davidson, B. N., Estiri, H., Goerg, C., Holve, E., Johnson, S. G., Liaw, S., Hamilton-Lopez, M., Meeker, D., Ong, T. C., Ryan, P., Shang, N., Weiskopf, N. G., Weng, C., Zozus, M. N., & Schilling, L. (2016). A harmonized data quality assessment terminology and framework for the secondary use of electronic health record data. eGEMs, 4(1), 1244. https://doi.org/10.13063/2327-9214.1244

Wang, R. Y., & Strong, D. M. (1996). Beyond accuracy: What data quality means to data consumers. Journal of Management Information Systems, 12(4), 5-33. https://doi.org/10.1080/07421222.1996.11518099

Weiskopf, N. G., & Weng, C. (2013). Methods and dimensions of electronic health record data quality assessment: Enabling reuse for clinical research. Journal of the American Medical Informatics Association, 20(1), 144-151. https://doi.org/10.1136/amiajnl-2011-000681