NUR 650 Module 2 Risk Tool Paper Example

Predicting Readmission: What Risk Scores Can and Cannot Tell a Heart Failure Care Team

[Student Name]

Southern New Hampshire University

NUR 650: Care Coordination and Outcomes Management

Module Two Risk Tool Paper

[Instructor Name]

[Date]

Predicting Readmission: What Risk Scores Can and Cannot Tell a Heart Failure Care Team

Transitional care programs cost money and nurse time, so hospitals want to direct them to the patients most likely to be readmitted. Risk prediction tools promise to identify those patients from data already in the record. In the composite Brookhaven hospital, the quality committee has proposed using the LACE index to sort heart failure patients into high and low risk, with only high-risk patients receiving home visits and structured phone follow-up. Before adopting it, the clinical nurse leader was asked to evaluate whether the tool is good enough for that purpose. This paper explains how readmission tools are judged, reviews the evidence on their performance, reports a local test and argues that a risk score should inform the decision about who receives intensive support, not make it alone.

How Prediction Tools Are Judged

Two properties matter most. Discrimination is the tool's ability to separate patients who will be readmitted from those who will not, usually reported as a C statistic, which is the probability that a randomly chosen readmitted patient has a higher score than a randomly chosen patient who was not readmitted. A C statistic of 0.5 is no better than a coin toss, 0.7 is usually described as acceptable and 0.8 or higher as strong. Calibration is whether predicted risk matches observed risk, for example whether patients predicted to have a 30% chance of readmission are actually readmitted about 30% of the time. A tool can discriminate reasonably yet be poorly calibrated in a new population, which is why local testing matters. Neither property says whether a readmission was preventable, which is what a care coordination program really needs to know.

What the Evidence Shows

Kansagara et al. (2011) systematically reviewed readmission risk prediction models and found 26 unique models tested in a range of populations. Most had poor discriminative ability, with C statistics in the range of roughly 0.56 to 0.72 in the majority of studies, and only a few included variables for social determinants, functional status or illness severity, which clinicians believe drive many readmissions. Models built from administrative data could be used at scale but tended to perform worse than those using detailed clinical information. The authors concluded that most models were not ready for widespread use to target interventions.

The LACE index, developed in Ontario (van Walraven et al., 2010), uses four items available at discharge: length of stay, acuity of the admission, comorbidity measured by the Charlson index and emergency department visits in the previous six months. It predicted death or unplanned readmission within 30 days with a C statistic of about 0.68 in validation. The HOSPITAL score (Donzé et al., 2013) uses seven items, including hemoglobin and sodium at discharge, discharge from an oncology service, procedures during the stay, an urgent admission, prior admissions in the past year and a stay of five days or longer. It was designed to predict potentially avoidable readmissions and reached a C statistic of about 0.71. Both tools are simple and use routine data; neither includes social or functional factors.

A Local Test

Because tools often perform differently outside the populations where they were developed, the clinical nurse leader applied both scores retrospectively to 400 heart failure discharges from the previous year, of whom 96 (24%) were readmitted within 30 days. LACE produced a C statistic of 0.63 and the HOSPITAL score 0.66, both lower than in the original studies. Using a LACE cutoff of 10 or higher as the committee proposed would have flagged 61% of patients as high risk, which exceeds the capacity of the two transitional care nurses, and would still have missed 29 of the 96 patients who were readmitted. Chart review of those 29 found that 17 lived alone, had no reliable transportation or could not afford their medications, factors neither score captures.

Recommendation

The scores are useful but not sufficient. The paper recommends a two-step approach. First, calculate the HOSPITAL score automatically at discharge, since it performed slightly better locally and is designed around avoidable readmissions. Second, have the discharging nurse complete a five-item social and functional screen covering living situation, transportation, medication affordability, ability to weigh daily and a caregiver's availability. Patients with a high score, a positive screen or a nurse's documented concern receive the full transitional care bundle; others receive a standard follow-up call. This combines the consistency of a score with the information nurses gather at the bedside, and it directs scarce resources to patients whose risks the scores miss.

How the Score Will Be Used in Practice

A score is only as useful as the workflow around it. The HOSPITAL score will be calculated by the electronic record when the discharge order is entered and displayed on the discharge checklist beside the social and functional screen, so the nurse sees both together. If the nurse's judgment differs from the score, for example a low-scoring patient who seems confused about their medications, the nurse can assign the full bundle with a one-line reason; these overrides will be reviewed monthly to learn what the score misses. Transitional care nurses will receive the list of assigned patients each morning, ranked so that the highest-risk patients are contacted first if the day's volume exceeds capacity. Patients will be told that they are receiving extra follow-up because they may benefit from it, not because they have been labeled high risk.

Fairness and Monitoring

Using any tool to allocate services raises fairness questions. A score built only on prior utilization may under-identify patients who have had poor access to care and therefore few prior visits, a pattern that can disadvantage low-income and rural patients. Brookhaven will review quarterly who is and is not assigned to the full bundle by age, race, payer and county of residence, and compare readmission rates across groups. The score and screen will be re-tested annually against actual readmissions, and the cutoffs adjusted if the population or capacity changes.

Conclusion

Readmission risk tools answer a narrow question with modest accuracy. The evidence and the local test both show that LACE and the HOSPITAL score separate high- and low-risk patients only moderately and miss many patients whose risks are social. Used alongside a brief screen and nurse judgment, and monitored for fairness and accuracy, a score can help Brookhaven direct its transitional care nurses where they are most needed without mistaking a number for a full picture of the patient.

References

Donzé, J., Aujesky, D., Williams, D., & Schnipper, J. L. (2013). Potentially avoidable 30-day hospital readmissions in medical patients: Derivation and validation of a prediction model. JAMA Internal Medicine, 173(8), 632-638. https://doi.org/10.1001/jamainternmed.2013.3023

Kansagara, D., Englander, H., Salanitro, A., Kagen, D., Theobald, C., Freeman, M., & Kripalani, S. (2011). Risk prediction models for hospital readmission: A systematic review. JAMA, 306(15), 1688-1698. https://doi.org/10.1001/jama.2011.1515

van Walraven, C., Dhalla, I. A., Bell, C., Etchells, E., Stiell, I. G., Zarnke, K., Austin, P. C., & Forster, A. J. (2010). Derivation and validation of an index to predict early death or unplanned readmission after discharge from hospital to the community. Canadian Medical Association Journal, 182(6), 551-557. https://doi.org/10.1503/cmaj.091117