Evaluation of Hospital-wide Readmission Risk Calculator to Predict 30-Day Readmission in Neurocritical Care Patients.
Multiple studies have been performed evaluating the most common causes for 30-day hospital readmission and the readmission rates for a variety of patients with neurologic disorders including patients with stroke, subarachnoid hemorrhage, and various types of neurosurgery. The range of 30-day readmission rates for neurological patients has been reported from 4.2% to 7.4% for spine surgery patients, (4) to 11.5% of mixed neurosurgical patients, (5) to 14.4% in stroke patients, (6) and to 11.4% for subarachnoid hemorrhage patients. (7) Despite these studies evaluating reasons for readmission, there is a lack of statistical models to predict readmission in this patient population.
There have been several models developed to predict general patient readmission risk. One of the largest and most broad tools currently in use is the pneumonic LACE (length of stay [LOS], acuity of admission, Charlson comorbidity index, and number of emergency department visits in the preceding 6 months) that was developed to predict readmission risk in medical and surgical patients. The LACE model was found to be accurate in quantifying the risk of death or unplanned readmissions, and its use is now very broad in multiple centers throughout the world to predict 30-day readmissions. The current readmission risk calculator tool in use at our institution was developed from a combination of previously validated models including the LACE model for predicting hospital readmissions and is used for all admitted patients.
Our own institute's model uses a general scoring system (Table 1) using criteria from the combination with local factors. (8-10) The scale weighs heaviest toward insurance type, end-stage liver disease, human immunodeficiency virus infection, and LOS and only 1 point for a specific neurological disease (history of stroke). There was anecdotal evidence regarding inaccuracy of the current hospital-wide readmission tool with neuroscience patients, which was observed by advanced registered nurse practitioners working at the bedside in the neuroscience intensive care unit (NSICU). The current readmission risk calculator was scoring the majority of the patients admitted to the NSICU as "low risk" for readmission, including those with a significantly higher severity of illness, thus prompting review of the accuracy of the readmission risk calculator in place at our institution for this patient population. The goal of this study was to retrospectively examine discharges from our neurocritical care unit to examine whether our institute's readmission risk score does a sufficient job in predicting readmission. We hypothesize that neurocritical care patient readmission differ from that of general patients for whom the readmission scale was devised and that it would be inaccurate in predicting readmissions in this specific patient population.
This study used a retrospective review of all patients 18 years or older, between February 2014 and February 2015, who were admitted both emergently or electively to the Mayo Clinic Florida NSICU, an academic tertiary care hospital. The research was approved by the Mayo Clinic Institutional Review Board as a minimal risk protocol. Per hospital protocol, all patients had a readmission score automatically calculated via the electronic health record (EHR), on admission. This score is calculated automatically via the EHR on all patients admitted to the hospital. The authors of the study reviewed the EHR (Cemer PowerChart) and, using a uniform data collection form, recorded the initial admission diagnosis, reason for admission, LOS, readmission prediction score, and whether the patient died or was discharged on the index admission. All patients were originally admitted to the NSICU upon their index admission and were discharged from the hospital from the neurology floor after being downgraded from the NSICU. For those patients who were readmitted to the NSICU, the time elapsed from discharge to readmission, reason for readmission, readmission diagnosis, and disposition were recorded.
Outcomes and Aims
The primary outcome was the difference in median readmission score between NSICU patients who were readmitted within 30 days and those who were not readmitted. The secondary outcomes compared was to evaluate the hospital-wide readmission risk tool and its ability to predict 30-day readmissions in this specific patient population, as well as describe the characteristics of patients who were readmitted compared with those who were not. Because this was a retrospective study of all admitted patients, with objective data and no follow-up data, there is a minimal risk for bias (selection, recall, confounding, or otherwise).
The continuous variables were presented as median values with their respective interquartile ranges (IQRs), and the categorical variables were reported as counts and proportions. Univariate analysis was performed to determine the associations of independent variables with readmission after initial discharge, and nonparametric statistical, Fisher, and Wilcoxon rank sum tests were used when applicable. Statistical significance was considered with a P value of less than .05. The statistics were performed by one of the authors (J.L.S.) in conjunction with a Mayo Clinic statistician. We used JMP 10 Pro statistical software for analysis from SAS (Cary, North Carolina). There was no funding for this study.
A total of N = 340 patients were admitted to our NSICU within the study period. The median age was 65 years (IQR, 51-77), with the majority being white (N = 254, 75%) and female (N = 180,53%). More than half of the patients were admitted because of an emergency (63%) and had a median LOS of 5 days (IQR, 2-10). The overall median initial readmission score was 8 (IQR, 4-10), and most patients were given a low risk for readmission (n = 224, 66%). Acute ischemic stroke was the most prevalent diagnosis at admission (26%), followed by neoplasm (14%) and hemorrhagic stroke (12%). The immediate mortality/hospice rate was 16.8% (n = 57).
The analysis excluded patients who died or were discharged to hospice on initial admission (n = 57) and patients who did not have readmission scores (n = 4), leaving us with 279 eligible patients. Only 38 (13.6%) were readmitted within 30 days of discharge. There was no difference in the age, sex, or readmission score between the 2 groups. Most readmitted patients had an initial low risk score for readmission (number and percentage). After univariate analysis, emergent admission, the diagnosis of neoplasia, and LOS were associated with the risk of readmission.
The median time between readmission was 9 days (IQR, 2-18), and median LOS after readmission was 4 days (IQR, 3-6). Emergent readmissions were seen in 97% of the patients, and only 1 patient was readmitted for a planned surgical procedure. The median new readmission score was 12 (IQR, 9-15), with 27.0% having a low risk score (compared with 65.8% on their first admission). Almost half of the patients were scored with an intermediate risk for readmission (n = 17,46%), and I patient did not have a readmission score calculated. The most common diagnosis at readmission was neoplasia (n = 8, 21%), but all infectious etiologies accounted for 15 readmissions (39%).
Preventing hospital readmissions is a Medicare quality benchmark and a challenge for hospitals across the country. Several centers have developed readmission score predictors to promptly identify patients at risk for readmission and potentially prevent their return. These predictor models have not been standardized, and there are currently no specific models for neurologically ill patients. Because Medicare payment is directly tied to readmission, having an accurate scoring model is essential.
In analyzing the current readmission tool at our institution, we found that it underestimated the risk of readmission in our neurocritical care patient readmission. Although our overall readmission rates reflected national trends, 65.8% of the patients who were readmitted were labeled "low risk" for readmission. The current readmission tool does not adequately address the highest risks of readmission (emergent initial admission and central nervous system malignancy) or the most common reason for readmission (infection). The single strongest risk factor for all patients admitted to our NSICU was emergent initial admission, which only receives 3 points on the readmission scale.
There are several limitations to our study including that this is a retrospective single-center study with our institute's unique risk calculator looking at a very small patient population of only NSICU patients. An additional limitation is heterogeneity within the patient population, particularly in emergent and elective admissions. Despite this limitation, there was importance identified with reviewing all admissions to the NSICU as future risk scales should be institution specific based on their own demographic. Different institutes have different emergent/elective ratios; thus, these results may not be applicable. This scale may differ greatly than other centers, but the objective was to not only critically analyze this scale but also identify that neurologically critically ill patients should be assessed differently than general patient populations and that hospitals should customize neurological-specific readmission risk scales based on local data and practice.
The findings of this study have importance given the impact on patient outcomes, as well as Medicare penalties involved with readmissions, and the use of nonvalidated data sets for readmission risk stratification within other institutions. This further reinforces the need for nurses to continually seek validation of clinical tools and advocate for the benefit of patients. It is clear that a significant fiscal pressure for actions is placed upon institutions and there is evidence to suggest that additional investment of resources benefits the use of these tools. Although this initially requires review of institutional demographics and data, the development of specific tools would likely create more accurate methods for identifying readmission risk.
The reason to identify high-risk patients remains to concisely direct additional resources and measures to prevent readmission, because predicting readmission is pointless unless care teams can develop interventions to reduce readmissions. However, the role of identifying and preventing readmissions can be supported by the nursing staff, including bedside registered nurses and advanced practice nurses. As the primary advocates for patients, it remains necessary to create environments in which advanced practice nursing is empowered to question the validity of current practice. The development of a readmission risk calculator specific to neuroscience patients would provide the bedside nurse an accurate scoring tool that identifies patients at a high risk for hospital readmission. The tool would enable bedside nurses, advanced practice nurses, and physicians to be alerted if a patient is at a high risk for readmission and the specific characteristics of the patient that make him/ her high risk. This study is a very small subset of NSICU patients and provides an example of nurses questioning practice and identifying a quality gap to ultimately improve patient care. Further studies are needed to be able to generalize the results. In the future, studies should be performed that include all neurologic, such as neurology, NSICU, and neurosurgery, patients to be able to generalize the results for this population.
(1.) Coughlin DG, Kumar MA, Patel NN, Hoffman RL, Kasner SE. Preventing early bouncebacks to the neurointensive care unit: a retrospective analysis and quality improvement pilot. Neurocrit Care. 2018;28(2): 175-183. doi:10.1007/ s12028-017-0446-z
(2.) Hansen LO, Young RS, Hinami K, Leung A, Williams MV. Interventions to reduce 30-day rehospitalization: a systematic review. Ann Intern Med. 2011; 155(8):520-528.
(3.) Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med. 2009;360(14): 1418-1428.
(4.) Bernatz JT, Anderson PA. Thirty-day readmission rates in spine surgery: systematic review and meta-analysis. Neurosurg Focus. 2015;39(4):E7.
(5.) Shah MN, Stoev IT, Sanford DE, et al. Are readmission rates on a neurosurgical service indicators of quality of care? J Neurosurg. 2013;119(4): 1043-1049.
(6.) Lichtman JH, Leifheit-Limson EC, Jones SB, Wang Y, Goldstein LB. Preventable readmissions within 30 days of ischemic stroke among Medicare beneficiaries. Stroke. 2013;44(12): 3429-3435.
(7.) Greenberg JK, Washington CW, Guniganti R, Dacey RGJr, Derdeyn CP, Zipfel GJ. Causes of 30-day readmission after aneurysmal subarachnoid hemorrhage. J Neurosurg. 2016; 124(3)743-749. doi: 10.3171/2015.2.JNS142771
(8.) van Walraven C, Dhalla IA, Bell C, et al. Derivation and validation of an index to predict early death or unplanned readmission after discharge from hospital to the community. CMAJ. 2010;182(6):551-557.
(9.) Silverstein MD, Qin H, Mercer SQ, Fong J, HaydarZ. Risk factors for 30-day hospital readmission in patients >65 years of age. Proc (Bayl Univ Med Cent). 2008:21 (4):363-372.
(10.) Hasan O, Meltzer DO, Shaykevich SA, et al. Hospital readmission in general medicine patients: a prediction model. J Gen Intern Med. 2010;25(3):211-219.
Questions or comments about this article may be directed to Sarah Peacock, MSN ACNP-BC, at firstname.lastname@example.org. She is an Acute Care Nurse Practitioner, Department of Critical Care Medicine, Mayo Clinic, Jacksonville, FL.
Jason Siegel, MD, Departments of Critical Care and Neurology, Mayo Clinic, Jacksonville, FL.
Emily Harmer, ARNP, Department of Critical Care, Mayo Clinic, Jacksonville, FL.
David Alejos, MD, Department of Internal Medicine, Jacobi Medical Center, Albert Einstein College of Medicine, New York, NY.
W. David Freeman, MD, Departments of Critical Care, Neurology, and Neurosurgery, Mayo Clinic, Jacksonville, FL. Author Contributions: All authors contributed to the conception, design, image acquisition, and writing of the manuscript. They all had final approval of the submitted manuscript.
The authors declare no conflicts of interest.
DOI: 10.109 7/JNN.0000000000000410
TABLE 1. Cerner Readmission Risks3 Medicare, Medicaid, or self-pay = +4 Poor health literacy = +1 Lack of social support or inability to do self-care = +1 >Previous admits prior 12 mo = +1/admit within 12 mo, +2/admit within 30 d >7 Medications = +2 Emergent admission = +3 Hx DM, Ml, CVA, PVD = +1 each Hx CHF, COPD, liver dz, cancer, substance abuse, or depression = +2 each Altered cognition, dialysis = +3 each Hx ESLD, HIV = +4 each Hx metastatic cancer = +6 Current LOS = +1/d (max, +5) Scores: 0-7, low risk (~7% risk); 8-10, intermediate risk (-8-12%); >11, high risk. Abbreviations: CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; CVA, cerebrovascular accident; DM, diabetes mellitus; dz, disease; HIV, human immunodeficiency virus; Hx, history; LOS, length of stay; max, maximum; M1, myocardial infarction; PVD, peripheral vascular disease. (a) Modified from length of stay, acuity of the admission, comorbidity of the patient and emergency department use in the duration of 6 months before admission Index and general internal medicine multicenter study.
|Printer friendly Cite/link Email Feedback|
|Author:||Peacock, Sarah; Siegel, Jason; Harmer, Emily; Alejos, David; Freeman, W. David|
|Publication:||Journal of Neuroscience Nursing|
|Date:||Feb 1, 2019|
|Previous Article:||Electroencephalographic Patterns During Common Nursing Interventions in Neurointensive Care: A Descriptive Pilot Study.|
|Next Article:||Evaluation of Hospital-wide Readmission Risk Calculator to Predict 30-Day Readmission in Neurocritical Care Patients.|