HIM 675 Module 9 Final Project Example

Final Project: Present on Arrival? A Research Proposal to Measure and Explain Errors in Admission Flags at a Tulsa Teaching Hospital

[Student Name]

Southern New Hampshire University

HIM 675: Research Methods and Evaluation

Final Project

[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: Present on Arrival? A Research Proposal to Measure and Explain Errors in Admission Flags at a Tulsa Teaching Hospital

Abstract

Present-on-admission flags decide whether a condition on an inpatient claim counts as a hospital complication. At the composite Tulsa teaching hospital in this proposal, an audit found more than a third of coded pressure injuries flagged as present on admission without supporting documentation. This study will measure how often coded flags agree with blinded re-abstraction for four hospital-acquired conditions, whether errors lean toward over- or under-reporting and whether the absence of a structured admission assessment predicts disagreement. An explanatory sequential design pairs a review of 320 stratified records from fiscal 2026 with interviews of eight coders and nurses. Results will guide documentation templates, coding practice and the accuracy of the hospital's reported safety data.

Problem and Questions

Nobody at the hospital can yet say how often these flags are right, which way they miss or which documentation habits travel with the misses. If flags are wrong, the hospital's safety rates, regulatory reports and some payments are wrong with them. The study asks three questions. What proportion of coded flags for pressure injuries, catheter-associated urinary infections, venous thromboembolism and fall injuries agree with re-abstraction? Among disagreements, do over-reports outnumber under-reports, as predicted for pressure injuries? And is the lack of a structured skin and risk assessment within 24 hours associated with greater odds of disagreement once condition, admission source, unit and coder experience are taken into account?

Rationale From the Literature

Earlier research makes the case that the flag is consequential and fallible. In a statewide audit, coded flags matched the chart in about three of every four records, with errors in both directions and different patterns by hospital type (Goldman et al., 2011). Applying flags to one system's patient safety indicators lowered nearly every rate, sometimes sharply, showing how much the flag changes what a hospital appears to achieve (Bahl et al., 2008). In the vocabulary Weiskopf and Weng (2013) offer for record quality, a flag error is usually a failure of completeness at admission that becomes a failure of correctness on the claim. What remains unknown, and what this study adds, is current, condition-specific evidence for nursing-sensitive conditions and a test of the documentation link.

Framework

The study follows each flag along four steps: the patient's condition on arrival, what clinicians record in the first day, what the coder sees and decides and the value that reaches the claim. The outcome compares the last step with the first as reconstructed by re-abstraction. Documentation at the second step is the main explanatory variable, and admission source, unit, condition and coder experience act as controls. The framework's blind spot, conditions present at arrival that no one ever recorded, is acknowledged as a limit on detecting under-reporting.

Design and Sample

The explanatory sequential design runs a quantitative record review first and uses its results to choose interviewees and topics, integrating both strands in the interpretation as Fetters et al. (2013) recommend. Eligible records are fiscal 2026 adult discharges with one of the four conditions as a secondary diagnosis. To estimate agreement within five percentage points at 95% confidence, assuming about 75% agreement, roughly 289 records are needed; 320 will be drawn at random, 80 per condition, to allow for unusable records and to describe each condition.

Instrument and Data Collection

Each sampled chart will be re-read by one of two credentialed coders new to those records with a structured tool, blinded to the coded flag. The tool's rules were reviewed by a wound care nurse and two coding experts, and two pilot rounds raised interrater kappa from about 0.80 to about 0.89, within the range McHugh (2012) describes as strong. One record in ten will be double-abstracted throughout, with a third reviewer settling disagreements and abstraction pausing if kappa drops below 0.80.

Analysis

Agreement will be reported as a weighted proportion with a 95% confidence interval, overall and by condition. An exact binomial test will compare over- and under-reports. Logistic regression will estimate adjusted odds of disagreement for records without a 24-hour assessment; with about 80 expected disagreements and six predictor terms, the model meets the guidance of roughly ten events per variable from Peduzzi et al. (1996). Interviews with four coders and four nurses from the units with most disagreement will be analyzed thematically, following Braun and Clarke (2006), and a second analyst will check coding on two transcripts.

Ethics and Privacy

Because reading records poses minimal risk, the study will seek expedited review board approval with a waiver of HIPAA authorization, since contacting several hundred former patients is impracticable and identifiers will be protected. Study numbers replace record numbers, the linking key sits in a separate encrypted file and identifiers are destroyed after analysis. Abstractors collect only the tool's fields. Interviewees give written consent and are never identified to supervisors. A plain-language summary of findings will be posted for the community whose records were used.

Timeline and Budget

The study runs seven months: board review in month one, abstraction in months two to four, analysis in month five, interviews in month six and integration and writing in month seven. Table 1 lists the budget, which consists mainly of staff time.

Table 1. Study Budget

Note. Staff time is charged at fully loaded internal rates; software is already licensed.

Dissemination and Use

Findings will go first to the quality, coding and nursing leaders who must act on them, then to the review board and the community summary, and finally to a regional health information conference. If missing admission assessments prove to be a strong predictor, the hospital's planned redesign of nursing templates can make the assessment a required step, and coding policy can direct queries when it is absent. The capstone will carry out this plan.

Conclusion

This proposal asks a narrow question with broad consequences: whether a one-character field on the claim tells the truth about harm in the hospital. It answers with a design matched to the question, a tested gold standard, an analysis plan sized to the data and protections for the patients whose records make the study possible.

References

Bahl, V., Thompson, M. A., Kau, T.-Y., Hu, H. M., & Campbell, D. A. (2008). Do the AHRQ patient safety indicators flag conditions that are present at the time of hospital admission? Medical Care, 46(5), 516-522. https://doi.org/10.1097/MLR.0b013e31815f537f

Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77-101. https://doi.org/10.1191/1478088706qp063oa

Fetters, M. D., Curry, L. A., & Creswell, J. W. (2013). Achieving integration in mixed methods designs: Principles and practices. Health Services Research, 48(6pt2), 2134-2156. https://doi.org/10.1111/1475-6773.12117

Goldman, L. E., Chu, P. W., Osmond, D., & Bindman, A. (2011). The accuracy of present-on-admission reporting in administrative data. Health Services Research, 46(6pt1), 1946-1962. https://doi.org/10.1111/j.1475-6773.2011.01300.x

McHugh, M. L. (2012). Interrater reliability: The kappa statistic. Biochemia Medica, 22(3), 276-282. https://doi.org/10.11613/BM.2012.031

Peduzzi, P., Concato, J., Kemper, E., Holford, T. R., & Feinstein, A. R. (1996). A simulation study of the number of events per variable in logistic regression analysis. Journal of Clinical Epidemiology, 49(12), 1373-1379. https://doi.org/10.1016/S0895-4356(96)00236-3

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