HIM 680 Module 6 Data Quality Short Paper Example

Eight Hops From a Clinic Question to a Dashboard: Lineage and Quality of Food Screening Data at Cedar Prairie Health

[Student Name]

Southern New Hampshire University

HIM 680: Advanced Topics in HIM I

Module Six Short Paper

[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.

Eight Hops From a Clinic Question to a Dashboard: Lineage and Quality of Food Screening Data at Cedar Prairie Health

Data quality is usually measured at the end, in the report someone is reading. By then it is too late to see where a problem began. Lineage, the record of where data come from and how they change on the way, lets an organization find the step where quality is lost. Here the element traced at the composite Iowa system is a patient's answer to the first food screening question. It describes each step the answer takes, measures how much data survive each step and recommends fixes.

The Journey

A medical assistant asks the question during rooming and selects an answer in the clinic flowsheet; that is hop one. The flowsheet stores the answer under a local code, hop two. Record logic combines it with the second food question to set a positive flag, hop three. A clinician may review the flag and add a problem list entry, hop four. A social worker may send a referral, hop five. Each night, an extract copies selected clinical data to the staging area of the warehouse, hop six. A transformation step builds the population health tables, hop seven. The dashboard reads those tables and applies its own filters, hop eight.

Where Data Are Lost

Four hops lose or change data. At hop one, the flowsheet offers a free-text option labeled other, which 4% of medical assistants used, typing answers such as patient declined or see note. Those answers cannot be read by the record logic. At hop two, the local codes mean nothing outside Cedar Prairie, so the answers cannot be shared with the Medicaid health plan or compared with other systems. At hop six, the nightly extract was built before the screener existed and does not copy that flowsheet at all, so screening answers never reach the warehouse; the population health team has been building its count from a monthly manual report. At hop eight, the dashboard filters to patients attributed to a primary care clinician, which silently drops roughly one in five screened patients, mostly those seen only in urgent care or specialty clinics.

Measuring Quality at Each Hop

Kahn et al. (2016) sort quality problems under three headings, format, presence and believability, and they separate verification, checking data against internal rules, from validation, checking it against an outside reference. Table 1 applies those ideas hop by hop for one month of screening.

Table 1. Food Screen Answers Surviving Each Hop, One Month

Note. Hops 4, 5 and 7 are not shown because they do not change the count of screen answers.

Plausibility and Validation

Counts tell only part of the story. On plausibility, the share of positive screens ranged from 4% to 27% across clinics serving similar neighborhoods. Observation at two clinics showed why: at one, the question was read aloud in a hurried way at the end of rooming, and at the other, patients completed it privately on a tablet. Weiskopf and Weng (2013) note that comparison with an external source is one way to judge correctness. National surveys have found food insecurity in more than one in ten U.S. households every year for decades, so clinics reporting positive rates near 4% are very likely missing patients. For validation, the logic at hop three was checked against the rule published by Hager et al. (2010), positive if either question is answered often or sometimes true, and it matched.

Why No One Noticed

The most striking finding is that a complete loss at the warehouse step went unnoticed for two years. Each group saw only its own part of the path. The clinics saw screens being completed and assumed the data went somewhere. The warehouse team loaded what the extract specification listed and had never been told about a new flowsheet. The population health analyst, faced with an empty table, built a manual report and moved on. Nobody owned the whole journey, so nobody compared the 9,040 answers recorded with the zero that arrived. This is the strongest argument for lineage documentation: it gives one person, the steward, a map on which a missing step stands out, and it turns a silent loss into a measurable gap that can be checked every month.

Recommended Fixes, in Order

Fixes are ranked by how much usable data they recover for the effort involved. First, add the screener flowsheet to the nightly extract, which recovers all answers for the warehouse and ends the late manual report; the warehouse lead estimates two weeks of work. Second, remove the attribution filter from the screening view of the dashboard and show attributed patients as a separate measure, which recovers about a quarter of screened patients in reporting. Third, remap local answer codes to LOINC so the data can be shared, as the governance framework requires. Fourth, replace the free-text option with coded choices for declined and unable to answer, which turns the 4% into usable information. Finally, standardize how the question is asked through training and the tablet workflow, which addresses the plausibility problem. The DAMA body of knowledge treats lineage documentation as part of metadata management (DAMA International, 2017), so the traced path will be recorded in the data catalog as the basis for future change control.

Conclusion

Of 9,040 answers recorded in a month, none reached the warehouse by design and only three quarters reached the dashboard by workaround. Most of the loss happened in two technical steps far from the clinic, which no one had looked at because no one owned the whole path. Tracing a single answer made those steps visible and gives the stewards a map for keeping them fixed. The first check, comparing answers recorded with answers loaded, will now run automatically each morning.

References

DAMA International. (2017). DAMA-DMBOK: Data management body of knowledge (2nd ed.). Technics Publications.

Hager, E. R., Quigg, A. M., Black, M. M., Coleman, S. M., Heeren, T., Rose-Jacobs, R., Cook, J. T., de Cuba, S. A. E., Casey, P. H., Chilton, M., Cutts, D. B., Meyers, A. F., & Frank, D. A. (2010). Development and validity of a 2-item screen to identify families at risk for food insecurity. Pediatrics, 126(1), e26-e32. https://doi.org/10.1542/peds.2009-3146

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