HIM 675 Module 7 Final Project Milestone Three Example

Final Project Milestone Three: How the Study Will Be Done. Methods for Testing Admission Flags in Tulsa

[Student Name]

Southern New Hampshire University

HIM 675: Research Methods and Evaluation

Final Project Milestone Three

[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 Three: How the Study Will Be Done. Methods for Testing Admission Flags in Tulsa

This milestone specifies how the study of present-on-admission flag accuracy at the composite Tulsa hospital will be carried out. It follows the conceptual framework from Milestone Two, which traces each flag from the condition at arrival to the value on the claim, and it answers the three research questions stated in Milestone One. Each section is written so that someone else could repeat the study.

Design

The study uses an explanatory sequential mixed methods design. The first and main phase is a retrospective, cross-sectional review of records in which coded flags are compared with blind re-abstraction. The second, smaller phase consists of interviews with coders and admitting nurses, chosen and focused using the first phase's results. Fetters et al. (2013) stress that the two phases must be integrated rather than run side by side; here, integration happens at two points. Phase one results decide which units and conditions to sample for interviews and which error patterns to ask about, and the discussion will interpret both sets of findings together.

Population and Sample

The population covers every adult discharged from the hospital during fiscal 2026, which ran from July 2025 through June 2026, with a secondary diagnosis code for one of four hospital-acquired conditions: stage 2 or deeper pressure injuries, urinary infections linked to a catheter, blood clots in the veins or injuries from falls. Records with the condition as the principal diagnosis are excluded, as are patients under 18 and records missing more than one day of documentation.

The sample size was calculated to estimate overall agreement with a margin of error of five percentage points at 95% confidence. Assuming agreement near 75%, consistent with earlier research, the formula multiplies the squared z value of 1.96 by 0.75 and by 0.25 and divides by the squared margin of 0.05, which yields about 289 records. Allowing roughly 10% for records that prove unusable, the study will sample 320, stratified as 80 per condition so that each condition can be described separately. Within each stratum, records will be selected by computer-generated random numbers from the list of eligible discharges. If a condition has fewer than 80 eligible records, all will be included.

Data Collection

Two credentialed coders who did not code the original records will abstract each sampled record with the tool tested in Module Six, blinded to the coded flag. A 10% random sample will be abstracted by both, with kappa calculated for each batch of 100 and a third reviewer resolving disagreements. Coded flags, coder identifiers, query records and admission data will be extracted separately by the data integrity analyst and merged with abstraction results only after abstraction is complete, using study numbers.

Variables

Table 1 lists the study's variables and how each is measured.

Table 1. Study Variables

Note. Values recorded as documentation insufficient or clinically undetermined are reported separately.

Analysis Plan

For RQ1, the share of flags matching re-abstraction, with its 95% confidence interval, will be given overall and per condition, and kappa between coded and re-abstracted flags will be calculated as a secondary measure. Because the sample is stratified equally, the overall estimate will be weighted by each condition's share of eligible discharges.

For RQ2, among disagreements, the numbers of over-reports and under-reports will be compared with an exact binomial test against equal proportions, overall and for pressure injuries, where the hypothesis predicts more over-reports.

For RQ3, logistic regression will model the odds of disagreement as a function of whether a structured assessment was documented, controlling for condition, admission source, unit type and coder experience. If about a quarter of records disagree, the model will have around 80 events for six predictor terms, roughly 13 events per variable. Peduzzi et al. (1996) found that logistic models become unreliable when events per variable fall much below ten, so the model is adequately supported, and coder experience will be dropped first if fewer disagreements occur. Results will be reported as adjusted odds ratios with 95% confidence intervals.

Interview Phase

After phase one, the researcher will interview four coders and four admitting nurses from the units and conditions with the most disagreement, using a semi-structured guide built around the error patterns found, for example asking nurses what happens to the skin assessment when admissions arrive in clusters. Interviews of about 30 minutes will be recorded with consent, transcribed and analyzed with thematic analysis, following the six phases described by Braun and Clarke (2006), from becoming familiar with the data through coding and theme development to writing up. A second analyst will review the coding of two transcripts.

Ethics and Privacy

The record review involves no contact with patients and poses minimal risk, so it will be submitted for expedited review by the hospital's institutional review board. Because obtaining authorization from more than 300 former patients would be impracticable, the study will request a waiver of HIPAA authorization, which federal guidance permits when the use of health information poses minimal privacy risk, the project would be impractical both without the waiver and without the data, and there is a plan to protect identifiers and destroy them at the earliest opportunity (U.S. Department of Health and Human Services, n.d.). Identifiers will be kept in a separate encrypted file on the hospital's network, linked only by study numbers, and destroyed when analysis is complete. Interview participants will give written consent, may stop at any time and will not be identified, and no supervisor will see individual responses.

Timeline

Review board submission is planned for the first month, record selection and abstraction for months two through four, analysis for month five, interviews for month six and integration and writing for month seven. If abstraction runs late, the interview phase will be shortened rather than the sample reduced.

Conclusion

The methods specify a design matched to a two-part question, a sample sized by calculation, a blinded and reliability-checked gold standard, an analysis plan matched to each research question and protections for patients and staff. Together with the first two milestones, they form the complete proposal presented in the final project.

References

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

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

U.S. Department of Health and Human Services. (n.d.). Research. Health information privacy. https://www.hhs.gov/hipaa/for-professionals/special-topics/research/index.html