Seminars Article
Hospital length of stay following radical cystectomy for muscle-invasive bladder cancer: Development and validation of a population-based prediction model

https://doi.org/10.1016/j.urolonc.2018.10.024Get rights and content

Abstract

Objective

Length of hospital stay for patients following radical cystectomy is an important determinant for improved quality of care. We sought to develop and validate a predictive model for length of hospital stay following radical cystectomy.

Methods

Patients aged 66 to 90 years diagnosed with clinical stage T2-4a muscle-invasive bladder cancer who underwent radical cystectomy were included from January 1, 2002 through December 31, 2011 using the Surveillance, Epidemiology, and End Results (SEER)-Medicare data. Linear regression analyses were used to develop and validate a predictive model for length of hospital stay.

Results

A total of 2,448 patients met inclusion criteria. After random assignment, 1,224 patients were included in the discovery cohort and 1,224 patients included in the validation cohort. The cohorts were well balanced with no significant difference in any of the preoperative variables. A best model was developed using marital status, Surveillance, Epidemiology, and End Results (SEER) region, clinical stage, Charlson comorbidity index, logarithm of hospital cystectomy volume, and use of neoadjuvant chemotherapy in a backward selection to predict the length of stay. There was robust internal validation (sum square error (SSE): 258.1 vs. predicted sum of squares (PRESS): 264.0 at SLS = 0.10), consistent with the external validation (average square error (ASE): discovery (0.248) vs. validation (0.258)) cohort. The strength of the model in predicting length of stay for the entire cohort was (R2 = 0.048).

Conclusion

In this large population-based study, we developed and validated a model to predict length of hospital stay following radical cystectomy. Identification of at-risk patients for prolonged hospital stay may aid in targeted interventions to reduce length of stay, improve quality of care, and decrease healthcare costs.

Introduction

Nearly 81,190 bladder cancer cases and 17,240 associated deaths are estimated in the United States in 2018 [1]. Radical cystectomy with extended pelvic lymphadenectomy is recommended for patients with muscle-invasive bladder cancer [2]. Despite these longstanding guidelines, radical cystectomy is markedly underused as historically only 21% of patients with muscle-invasive disease are offered this potentially curative surgery [3]. Given concerns regarding the non-negligible morbidity and mortality associated with this complex surgery, along with patients more often being elderly with increased comorbidities, identifying determinants to improve outcomes are needed. One of the potential surrogates for improved quality of care defined by the Center for Medicare and Medicaid Services (CMS) includes length of stay. Prolonged hospital stays were historically associated with increased adverse events, subsequent utilization of healthcare resources, as well as associated cost of care, particularly following major cancer surgeries [4]. While some factors, such as patient age and racial background, were observed as independent predictors of prolonged hospital stays, there is no comprehensive model to accurately predict length of stay for patients following radical cystectomy [5], [6]. The purpose of the present study was to develop and validate a predictive model which may identify at-risk patients to develop targeted interventions to reduce length of stay.

Section snippets

2.1. Data source

We used the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked data from the National Cancer Institute [7]. SEER database contains information on patient demographics, cancer characteristics (stage, grade, histology), and patient follow-up information. The Medicare database contains information on inpatient and outpatient claims. The study was deemed exempt by the Institutional Review Board at The University of Texas Medical Branch at Galveston.

2.2. Ascertainment of study cohort

Using SEER-Medicare, we included

Results

Patient demographics in the discovery and validation cohorts are summarized in Table 1. Overall, 1,224 (50.0%) and 1,224 (50.0%) of the 2,448 patients who underwent radical cystectomy for muscle-invasive bladder cancer were included in the discovery and validation cohorts, respectively. Patients were similar across all demographic and clinical predictors except for census region. Patients in the validation cohort were more often from the West (41.8% vs. 40.6%) and South (25.5% vs. 23.9%) than

Discussion

We developed and validated a predictive model for length of hospital stay following radical cystectomy. The model performed well as shown by the robust internal and external validation enhancing generalizability and applicability of the model. With the need to improve quality and outcomes of care at decreased costs, models such as the one developed and validated in the present may identify potential at-risk patients and measures to decrease length of stay following surgery.

Our study has several

Conclusions

In this large population-based study, we developed and validated a model to predict length of hospital stay following radical cystectomy. Patients who are unmarried, increased comorbidities, advanced clinical stage, received neoadjuvant chemotherapy, and underwent radical cystectomy at lower-volume hospitals in the West census region of the United States were at greater risk for prolonged hospital stay following surgery. Refinement of the model with inclusion of additional confounders may

Conflicts of interest

None.

Acknowledgments

This study was conducted with the support of a Department of Defense Peer Reviewed Cancer Research Program (PRCRP), Career Development Award (W81XWH1710576), and the Herzog Foundation (SBW). This study was conducted with the support of the Institute for Translational Sciences at the University of Texas Medical Branch, supported in part by a Clinical and Translational Science Award (UL1 cTR001439 and 1TL1TR00144003) from the National Center for Advancing Translational Sciences (MDR). These

References (19)

There are more references available in the full text version of this article.

This is part of the Seminar series guest edited by Dr. Schmitz-Drager under the supervision of Dr. Droller.

View full text