HCI 5073 Module 2 Data Quality and Interoperability Analysis Example

Fit for the Heat Season? A Data Quality and Interoperability Analysis of Emergency Department Syndromic Feeds Using the Harmonized Conformance, Completeness and Plausibility Framework

Student Name

American College of Education

HCI5073: Public Health Informatics

Module 2 Assignment

Instructor Name

October 12, 2026

Why Data Quality Decides What Surveillance Can See

The first module of this project found that a composite Midwestern county's syndromic surveillance of heat-related illness is fast and dependable but misses about a third of heat cases, most of them among older adults. That finding raises a narrower question for this paper: how much of the shortfall comes from the data themselves, and how much from the way systems exchange them? A surveillance query can only find what the records contain, in a form it can read, at the moment it runs. If a field is empty, coded in an unexpected way or arrives days late, the query fails silently and the dashboard looks calmer than the emergency departments actually are.

This analysis examines the emergency department feed from the county's four hospitals to the state syndromic platform during the most recent warm season, May through September. It applies a harmonized data quality framework to the fields that heat surveillance depends on and then traces the interoperability steps where information is lost or delayed.

The Data Source and Its Standards

Each visit generates a series of electronic messages from the hospital's electronic health record, sent in the Health Level Seven version 2 format that public health agencies have long used for syndromic data. A registration message opens the visit, and update messages follow as triage notes, diagnoses and the final disposition are added. Each message carries a facility identifier, a visit identifier, patient age, sex and residential ZIP code, the chief complaint as typed at triage, and diagnosis codes drawn from the International Classification of Diseases, Tenth Revision, Clinical Modification. The state platform assembles the messages for a visit into one record and runs the heat query against the combined text and codes.

Two features of this arrangement matter for quality. First, the most informative field for heat illness, the chief complaint, is free text written under time pressure, so its value depends on individual habits rather than a code set. Second, diagnosis codes often arrive in a later message, sometimes days after the visit, which means a record can change after the dashboard has already counted, or failed to count, it. Hughes et al. (2020), reviewing 115 emergency department syndromic systems in 15 countries and territories, found that systems vary widely in which fields they capture and how quickly, which is why quality has to be judged locally rather than assumed from the national program.

A Framework for Judging Fitness for Use

Kahn et al. (2016) harmonized the many terms used to describe data quality in electronic health records into three categories. Conformance asks whether values follow the expected format, allowed values and relationships between fields. Completeness asks whether expected values are present. Plausibility asks whether values are believable, either on their own or compared with other data or over time. The framework also separates verification, which checks data against internal rules and expectations, from validation, which compares data with an external standard such as a chart review. Its central idea is fitness for use: the same data can be good enough for one purpose and poor for another. This paper therefore asks whether the feed is fit for detecting and counting heat illness during hot weather, not whether it is good in general.

Conformance

Conformance was mostly strong. Across 61,480 visits from May through September, more than 99% of records had a valid facility identifier, a sex value from the allowed list and an age within the expected range. Two problems stood out. One hospital sent residential ZIP codes with a trailing four-digit extension in a field expected to hold five characters, and the state platform truncated some of them incorrectly, placing about 400 visits in the wrong ZIP code for most of June. A second hospital occasionally sent diagnosis codes without the decimal point, which the platform read correctly, but the county's own export script did not, so those codes disappeared from local analyses until the script was fixed in July. Neither issue was visible on the dashboard; both were found only because an analyst compared counts across sources.

Completeness

Completeness varied by field and by hospital. Chief complaint text was present in nearly every record at all four facilities. Diagnosis codes were a different story. Within 24 hours of the visit, 91% of records from the two largest hospitals carried at least one diagnosis code, but the figure was 58% at a smaller hospital whose coding staff work only on weekdays, and weekend visits there often waited three or four days. Because the dashboard is read each morning, those gaps fell exactly on the days when outreach decisions are made. Triage notes, which add detail such as time spent outdoors, were sent by only two of the four hospitals. Disposition, showing whether a patient was admitted, was 96% complete after a week but under half complete on the day after the visit.

Plausibility

Plausibility checks compared the feed with what should be expected. Daily visit totals from each hospital were compared with the hospitals' own census reports; the feed matched within 2% on most days, but one hospital's totals dropped by about a fifth for three days in August because of an interface outage that nobody reported. Heat-flagged visits were compared with daily heat index values; flagged visits rose on hot days as expected, which supports the query's validity, but a handful of heat-flagged visits in cool weather came from patients working in industrial settings, a pattern the query cannot distinguish from weather-related illness. Harduar Morano and Watkins (2017), analyzing Florida emergency visits, hospitalizations and deaths, found that using all diagnosis fields and multiple data sources may improve the sensitivity of heat illness surveillance and that man-made heat exposures appear in these records as well. The county's query currently reads only the first three diagnosis fields, so it is likely missing heat codes listed lower in the record.

Interoperability Gaps

Several losses occur between systems rather than within them. Visits by county residents to hospitals across the state line never enter the state platform, because the two states do not share syndromic data routinely, and the county estimates from past hospital discharge data that one in twelve of its residents' emergency visits happen out of state. When a hospital changed electronic health record vendors in April, the new system mapped its chief complaint field to a different message segment, and for five weeks the complaint text arrived empty until the interface was corrected. Update messages carrying late diagnoses are matched to earlier messages by visit identifier, but when a patient was transferred between facilities, the two visits were not linked, so a transfer could be counted twice or missed. Each of these problems is a failure of agreement between systems, not of any single system, and none generates an automatic alert.

What the Findings Mean for Heat Surveillance

Taken together, the findings suggest that a meaningful share of the sensitivity gap identified in Module 1 is a data problem rather than a query problem. Late diagnosis codes at one hospital, a restricted search of diagnosis fields and missing triage notes all remove the very information that would identify heat illness in an older adult whose chief complaint reads weakness or fall. Vaidyanathan et al. (2024) showed how national syndromic data allowed CDC to observe prolonged peaks in heat-related emergency visits during the record-setting summer of 2023 and to issue public health alerts, which demonstrates what timely, complete data make possible. At the county level, the same capability depends on the quality of four hospitals' feeds on the mornings that matter.

The county can act on several of these problems without new technology. Automated completeness reports by facility and day, sent to hospital interface staff each week, would expose gaps like the weekend coding delay and the silent outage. Expanding the query to all diagnosis fields is a simple change. Requesting triage notes from the two remaining hospitals and negotiating a data-sharing agreement with the neighboring state are harder steps that later modules will consider.

Conclusion

The emergency department feed is highly conformant and mostly complete for basic fields, but it is less fit for heat surveillance than the dashboard suggests. Delayed diagnosis codes, missing triage notes, limited diagnosis fields, unreported outages and cross-border and vendor-change gaps all reduce what the query can find, and they reduce it most for the older adults at highest risk. Measuring quality routinely, by field and by facility, is the first step toward a surveillance system whose counts can be trusted on the hottest days.

References

Harduar Morano, L., & Watkins, S. (2017). Evaluation of diagnostic codes in morbidity and mortality data sources for heat-related illness surveillance. Public Health Reports, 132(3), 326-335. https://doi.org/10.1177/0033354917699826

Hughes, H. E., Edeghere, O., O'Brien, S. J., Vivancos, R., & Elliot, A. J. (2020). Emergency department syndromic surveillance systems: A systematic review. BMC Public Health, 20(1), 1891. https://doi.org/10.1186/s12889-020-09949-y

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), 18. https://doi.org/10.13063/2327-9214.1244

Vaidyanathan, A., Gates, A., Brown, C., Prezzato, E., & Bernstein, A. (2024). Heat-related emergency department visits: United States, May-September 2023. MMWR. Morbidity and Mortality Weekly Report, 73(15), 324-329. https://doi.org/10.15585/mmwr.mm7315a1