HIM 690 Module 6 Final Project Milestone Three Example

Final Project Milestone Three: Why the Flags Miss. Analysis and Integration for the Cimarron Heights Flag Accuracy Capstone

[Student Name]

Southern New Hampshire University

HIM 690: Health Information Management Capstone

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: Why the Flags Miss. Analysis and Integration for the Cimarron Heights Flag Accuracy Capstone

Milestone Two showed what happened: coded admission flags disagreed with the chart in about a quarter of records, pressure wound flags were wrong in two of five and their errors leaned heavily toward calling wounds pre-existing. This milestone asks why. It reports the logistic model for the third research question, the themes from eight interviews and a joint display that brings the two together, then returns to the three explanations raised in the Module Five discussion.

The Model

The model estimated the odds that a coded flag disagreed with the second reading. Predictors were a missing structured skin and risk check in the first 24 hours, the complication, admission from a facility, intensive care admission and coder experience under three years. With 75 disagreements and seven predictor terms, the model has about 11 events per term, above the threshold of roughly ten that Peduzzi et al. (1996) found necessary for stable estimates. The hospital's biostatistician reviewed the model and its fit.

Table 1. Adjusted Odds of Flag Disagreement (n = 317)

Note. Logistic regression; each odds ratio is adjusted for all other predictors.

Reading the Model

After adjustment, a missing first-day check remained strongly associated with disagreement: records without one had about three and a half times the odds of a wrong flag, compared with about four times in the unadjusted comparison. Pressure wounds carried higher odds than the other complications even with documentation accounted for, which suggests something specific to wounds beyond whether a check was done. Admission from a facility carried somewhat higher odds, with a confidence interval just reaching 1.0. Coder experience and intensive care admission showed no clear association. The hypothesis for the third research question is supported, with the caution that a cross-sectional study shows association, not cause.

Sensitivity Checks

Three checks tested whether the model's main result depended on choices made in the analysis. First, the model was rerun without the three records whose abstracted flags had been changed by the mid-study rule clarification; the odds ratio for a missing skin check was 3.3, essentially unchanged. Second, the model was rerun with unit included as a set of indicators instead of the single intensive care term; with nine units and only 75 events this model was strained, but the odds ratio for a missing check stayed above 2.5, and no unit stood out once documentation was considered. Third, the analysis was repeated for pressure wounds alone, the complication with the most errors; with 31 disagreements among 80 records the estimates were wide, but a missing check again carried the largest odds ratio, about 4. None of the checks changed the conclusion, which gives more confidence that the association is not an artifact of a single modeling choice.

A further question was whether the association might run backward, with wounds already present leading nurses to delay the check. Interview accounts made this unlikely: nurses described deferring checks because of workload, not because a wound was already seen.

Interview Themes

Four coders and four nurses, two of them on night shift, were interviewed for about 30 minutes each. Transcripts went through the stepwise thematic method of Braun and Clarke (2006), from repeated reading and initial codes to reviewed and named themes, and a second analyst reviewed the coding of two transcripts. Four themes emerged.

The first, the skin check waits for a quiet moment, came from nurses. During clusters of night admissions, the skin inspection is deferred so that medications and vital signs are done first, and it is sometimes completed the next day or not documented at all. The second, the safe answer is yes, came from coders. When a patient arrives from a facility and any note mentions skin, two coders described choosing present on admission because a no would mean the hospital caused the wound, and they felt unqualified to make that judgment. The third, no road back to the nurse, described coders' inability to query nurses; the query process reaches only physicians, who rarely address skin. The fourth, the skin lives somewhere else, explained that wound findings sit in a nursing flowsheet coders do not routinely open.

Joint Display

Fetters et al. (2013) describe joint displays as a way to integrate quantitative and qualitative results side by side. Table 2 links each statistical finding to the themes that help explain it.

Table 2. Joint Display of Statistical Findings and Interview Themes

Note. Themes from eight interviews; statistics from Table 1.

The Three Explanations Revisited

The Module Five discussion offered three explanations for wound over-reporting. The evidence supports two. Missing documentation is supported by the model and by nurses' accounts. Coder defaulting is supported by the interviews and by the facility association, and the absence of an experience effect suggests it is a shared habit rather than individual error. The third explanation, deep tissue injuries surfacing late, is weakened: of the 26 wound over-reports, only four involved deep tissue injuries, and in three of those the reviewers had already applied the rule for injuries documented at arrival. Clinical timing may explain a few errors but not the pattern.

Conclusion

Flags miss mainly because the first day's skin check is often absent and because coders, without a way to ask nurses, default to yes when uncertain. Those two findings point to fixes in nursing workflow and coding process at the same time, which Module Seven will turn into recommendations.

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