HIM 550 Module 3 Final Project Milestone One Example

Final Project Milestone One: Can We Trust the Discharge Data? A Quality Assessment at Cypress Hollow Medical Center

[Student Name]

Southern New Hampshire University

HIM 550: Data Management and Data Quality

Final Project Milestone One

[Instructor Name]

[Date]

The organization, setting and figures below are a composite written as a model document. No real employer, client, colleague or patient is described.

Final Project Milestone One: Can We Trust the Discharge Data? A Quality Assessment at Cypress Hollow Medical Center

Cypress Hollow Medical Center wants to know whether discharging patients earlier in the day would shorten length of stay and free beds for the emergency department. Before that question can be answered, the data used to answer it must be tested. This milestone assesses the quality of the hospital's inpatient discharge data set for the six months from January through June, describes the methods used, reports the findings field by field and ranks the problems by how much they would distort the analysis planned for the next milestone.

Purpose and Scope

The assessment has one purpose: to decide whether each field needed for a discharge timing analysis is fit for that use. It therefore examines nine fields rather than the full record: encounter number, admission date and time, the registration discharge time, the nursing departure time, admission source, admission type, nursing unit, discharge disposition and principal diagnosis. Fields that the analysis does not need, such as payer and attending physician, were left for later work. Limiting the scope keeps the assessment focused on the question leaders are asking.

Data Set and Methods

The extract drew 9,840 inpatient discharges from the data warehouse, excluding observation stays and newborns. Each field was profiled for missing values, allowed values and ranges, and then compared with a second source where one existed: the nursing departure time against the registration discharge time, unit codes against the bed management system and principal diagnosis against a sample of 120 records reviewed by a credentialed coder.

The checks follow the harmonized framework of Kahn et al. (2016), which groups data quality into conformance, completeness and plausibility and asks each to be judged against the intended use. Weiskopf and Weng (2013) add concordance between sources and currency, and both were useful here. Hripcsak and Albers (2013) warned that record data reflect the process of care as much as the patient's condition, which meant each timestamp had to be traced to the workflow step that creates it.

Summary of Findings

Table 1 summarizes the results for each field. Four fields are fit for the analysis as they stand, three need correction or a substitute and two carry limitations that must be stated with any results.

Table 1. Quality Assessment of Nine Discharge Fields, January to June (n = 9,840)

Note. Prepared by the author from a warehouse extract and a coder review sample.

Discharge Time: Valid but Unfit

The registration discharge time passes every conformance and completeness check. It is always present, always a valid date and time and never earlier than admission. It fails on relevance. Comparing it with the nursing departure time on the 8,954 records with both values showed a median gap of 3.4 hours, with a longer gap on the medical units, where clerks close encounters in a batch after lunch. By departure time, 37% of patients left before noon; by registration time, only 29% did. A timing analysis built on the registration field would misstate both the problem and any change in it.

Admission Source and Nursing Unit

Admission source is blank on 1,181 records, or 12%, and the warehouse loads those blanks as Other, which makes Other the third largest category in every report. The blanks are not random. They cluster on nights and weekends in the emergency department, when registrars are fewest, so any comparison of patients admitted from home and from other facilities would be biased toward daytime admissions.

Nursing unit codes changed on March 14 when two medical units were reorganized into three. Records before that date use the old codes, and one old code now corresponds to two new units. A crosswalk built with the bed management team assigns 94% of the affected records by room number; the remainder are grouped as a combined unit for the first quarter.

Plausibility and Concordance Checks

Thirty-seven records show a nursing departure time earlier than the admission time. Each traced to manual entry on a downtime form after a system outage in February, and each was corrected against the paper form. Discharge disposition conflicted with case management notes in 11 records, all involving patients whose planned transfer to a skilled nursing facility changed at the last moment.

Principal diagnosis agreed with the coder's review in 111 of 120 sampled records, about 93%. The nine disagreements involved sequencing choices rather than wrong conditions. A systematic review by Burns et al. (2012) found wide variation in the accuracy of routine discharge coding across studies, so a local sample was worth the effort, and diagnosis groupings in the analysis will be broad enough that sequencing differences matter little.

Why the Problems Occur

Each major problem begins in a workflow, not in the warehouse. The discharge time reflects a billing step performed when clerks have time. Admission source is optional at registration and is skipped during busy hours. Unit codes changed without an update to warehouse documentation, and the loading rule that turns blanks into Other was written years ago without a record of why. These causes matter for the recommendations milestone, because a cleaning rule applied to the extract would hide each problem without preventing it.

Priorities for the Analysis

The problems are ranked by their effect on the planned analysis. First, the analysis will use nursing departure time, with rules for the 9% of records where it is missing: deaths and transfers will be excluded, and patients leaving against advice will be analyzed separately. Second, the unit crosswalk will be applied before any unit comparison. Third, results by admission source will be reported with the missing group shown rather than folded into Other. Fourth, the coder sample will be noted as a limitation. Each rule will be written down so the analysis can be repeated next quarter.

Conclusion

The discharge data set is usable for a timing analysis, but only after replacing the field leaders have trusted for years. The departure time, a unit crosswalk and honest reporting of missing admission sources make the next milestone's analysis defensible. The same findings also point to the fixes that the final data management plan must address at the point where the data are created, beginning with the registration screen and the nightly load.

References

Burns, E. M., Rigby, E., Mamidanna, R., Bottle, A., Aylin, P., Ziprin, P., & Faiz, O. D. (2012). Systematic review of discharge coding accuracy. Journal of Public Health, 34(1), 138-148. https://doi.org/10.1093/pubmed/fdr054

Hripcsak, G., & Albers, D. J. (2013). Next-generation phenotyping of electronic health records. Journal of the American Medical Informatics Association, 20(1), 117-121. https://doi.org/10.1136/amiajnl-2012-001145

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.-T., 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), Article 1244. https://doi.org/10.13063/2327-9214.1244

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