HIM 360 Module 4 Project One Example

Project One: Care Problem or Data Problem? A Mortality Coding Review at Juniper Health

[Student Name]

Southern New Hampshire University

HIM 360: Coding and Classifications Systems II

Project One

[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.

Project One: Care Problem or Data Problem? A Mortality Coding Review at Juniper Health

Juniper Health's risk-adjusted mortality ratio of 1.14 means that 14% more patients died than a statistical model predicted for patients with the same recorded characteristics. The ratio affects public rankings, referrals and the board's confidence in clinical quality. Before leaders act on it, they need to know how much reflects care and how much reflects incomplete or inaccurate data. This project answers that question for the past twelve months.

How the Ratio Works

A mortality ratio divides observed deaths by expected deaths. Expected deaths come from a model that estimates each patient's risk of dying from age, sex, the principal diagnosis, secondary conditions present on admission and sometimes procedures and status codes. Pine et al. (2007) showed that models become considerably more accurate when present-on-admission indicators separate conditions patients arrived with from those that developed later. The model can only count what is coded: a condition missing from the claim, or wrongly marked as hospital-acquired, lowers expected deaths and raises the ratio.

Review Methods

Two reviewers, a credentialed coder and a documentation integrity nurse, independently examined all 120 inpatient deaths from the past year. For each case they read physician notes, nursing and dietitian documentation, consults, laboratory results and orders, then compared findings with the final codes. They looked for clinically supported conditions not coded, present-on-admission indicators that conflicted with the record, missing do-not-resuscitate and palliative care codes and principal diagnosis selection errors. Disagreements were resolved by the coding manager.

Finding 1: Conditions Documented but Not Coded

In 37 of 120 deaths, 31%, reviewers found at least one clinically significant condition documented by nurses, dietitians or consultants but not stated by the attending physician in a way coders could use. The most common were severe malnutrition identified by dietitians, chronic respiratory failure on home oxygen and encephalopathy described by neurology consultants. Official guidelines allow some details, such as pressure injury stage and body mass index, to be coded from other clinicians' documentation, but the underlying diagnosis must come from a provider. These gaps require physician documentation, not coding changes.

Finding 2: Present-on-Admission Errors

Among 1,064 secondary diagnoses in the 120 records, 96, or 9%, carried present-on-admission indicators that conflicted with the record, most often conditions marked N or U when emergency department notes showed them on arrival. Meddings et al. (2010) found that documentation and coding problems kept hospital-acquired catheter infections from being captured accurately; the reverse problem appears here, with conditions present at arrival wrongly counted as hospital-acquired, which lowers expected risk and can create false hospital-acquired condition flags.

Finding 3: Missing Status Codes

Of 86 patients with a documented do-not-resuscitate order, 15, or 17%, lacked Z66, do not resuscitate. Of 72 patients who received palliative care consultation, 16, or 22%, lacked Z51.5, encounter for palliative care. Several risk models use these codes, and their absence makes deaths among patients who chose comfort care appear unexpected.

Table 1. Review Findings, 120 Deaths

Note. Review by the author's team; disagreements resolved by the coding manager.

Finding 4: Principal Diagnosis Selection

In four cases, reviewers judged the principal diagnosis inconsistent with guidelines, for example a symptom sequenced first when a definitive diagnosis was documented. Principal diagnosis drives the risk model's baseline, so even a few errors matter. All four were reviewed with the coders involved and corrected.

What the Ratio Might Be

The analytics team reran the risk model after applying corrections that were supported by existing documentation: fixing present-on-admission indicators, adding missing Z66 and Z51.5 codes and correcting the four principal diagnoses. The ratio fell from 1.14 to about 1.07. Adding the uncoded conditions would require physician documentation and cannot be done retroactively without a compliant query; modeling them hypothetically suggested a further drop to about 1.03. Even under the most favorable assumptions, the ratio stays above 1.0.

Separating Data From Care

Because the ratio remains above expected, not all of the gap is a data problem. Reviewers flagged six deaths with possible care concerns, such as delayed recognition of deterioration, for referral to physician peer review. Rosen et al. (2012) found that safety indicators based on coded data often misclassify events, which is why chart review, not codes alone, should drive clinical conclusions. The review's purpose was accuracy, and accuracy revealed both documentation gaps and cases that deserve clinical attention.

What Physicians Said

The team shared de-identified examples with the hospitalist and critical care groups. Physicians were surprised that a dietitian's malnutrition assessment could not be coded unless they documented the diagnosis themselves, and several said they had assumed the dietitian note was enough. Others noted that do-not-resuscitate orders were entered as orders but rarely restated in notes. Their reactions suggest that simple education and templates could close many gaps, and that physicians support accuracy when they understand how the data are used.

Recommendations

Four actions follow. Route every inpatient death to a documentation integrity review before final coding, checking status codes, present-on-admission indicators and principal diagnosis. Train physicians on documenting conditions identified by dietitians and consultants, especially malnutrition and chronic respiratory failure, when they agree with the finding. Build a nursing flowsheet prompt that alerts the attending when a dietitian documents severe malnutrition. And refer care concerns identified in reviews to peer review routinely.

Limitations

The review examined only deaths, so it cannot show whether survivors' records have similar gaps that would also raise expected risk. The recalculated ratio used the hospital's internal replica of the ranking model, which may differ from the official version. And reviewers' judgments about clinical significance involve some subjectivity, reduced but not eliminated by dual review. O'Malley et al. (2005) noted that measuring coding accuracy requires attention to such sources of error in the review process itself.

Conclusion

About half of Juniper Health's excess mortality ratio appears to reflect correctable data problems: conflicting present-on-admission indicators, missing status codes and undocumented conditions. The rest points to cases worth clinical review. Accurate coding does not erase quality problems; it lets the organization see them clearly.

References

Meddings, J., Saint, S., & McMahon, L. F. (2010). Hospital-acquired catheter-associated urinary tract infection: Documentation and coding issues may reduce financial impact of Medicare's new payment policy. Infection Control & Hospital Epidemiology, 31(6), 627-633. https://doi.org/10.1086/652523

O'Malley, K. J., Cook, K. F., Price, M. D., Wildes, K. R., Hurdle, J. F., & Ashton, C. M. (2005). Measuring diagnoses: ICD code accuracy. Health Services Research, 40(5, Pt. 2), 1620-1639. https://doi.org/10.1111/j.1475-6773.2005.00444.x

Pine, M., Jordan, H. S., Elixhauser, A., Fry, D. E., Hoaglin, D. C., Jones, B., Meimban, R., Warner, D., & Gonzales, J. (2007). Enhancement of claims data to improve risk adjustment of hospital mortality. JAMA, 297(1), 71-76. https://doi.org/10.1001/jama.297.1.71

Rosen, A. K., Itani, K. M. F., Cevasco, M., Kaafarani, H. M. A., Hanchate, A., Shin, M., Shwartz, M., Loveland, S., Chen, Q., & Borzecki, A. (2012). Validating the patient safety indicators in the Veterans Health Administration: Do they accurately identify true safety events? Medical Care, 50(1), 74-85. https://doi.org/10.1097/MLR.0b013e3182293edf