Elsevier

World Neurosurgery

Volume 104, August 2017, Pages 24-38
World Neurosurgery

Original Article
The Impact of Race on Discharge Disposition and Length of Hospitalization After Craniotomy for Brain Tumor

https://doi.org/10.1016/j.wneu.2017.04.061Get rights and content

Background

Racial disparities exist in health care, frequently resulting in unfavorable outcomes for minority patients. Here, we use guided machine learning (ML) ensembles to model the impact of race on discharge disposition and length of stay (LOS) after brain tumor surgery from the Healthcare Cost and Utilization Project National Inpatient Sample.

Methods

We performed a retrospective cohort study of 41,222 patients who underwent craniotomies for brain tumors from 2002 to 2011 and were registered in the National Inpatient Sample. Twenty-six ML algorithms were trained on prehospitalization variables to predict non–home discharge and extended LOS (>7 days) after brain tumor resection, and the most predictive algorithms combined to create ensemble models. Partial dependence analysis was performed to measure the independent impact of race on the ensembles.

Results

The guided ML ensembles predicted non–home disposition (area under the curve, 0.796) and extended LOS (area under the curve, 0.824) with good discrimination. Partial dependence analysis showed that black race increases the risk of non–home discharge and extended LOS over white race by 6.9% and 6.5%, respectively. Other, nonblack race increases the risk of extended LOS over white race by 6.0%. The impact of race on these outcomes is not seen when analyzing the general inpatient or general operative population.

Conclusions

Minority race independently increases the risk of extended LOS and black race increases the risk of non–home discharge in patients undergoing brain tumor resection, a finding not mimicked in the general inpatient or operative population. Recognition of the influence of race on discharge and LOS could generate interventions that may improve outcomes in this population.

Introduction

It has been well established that racial disparities exist in health care, frequently resulting in unfavorable outcomes for minority patients.1, 2 Research has shown that these disparities also exist in outcomes for patients undergoing treatment for brain tumor.3 The persistence of racial disparities in the treatment of neurosurgical patients is directly at odds with one of the key aspects of the Institute of Medicine's definition of high-quality care: equitable delivery of care.4

In our resource-limited health care system, increasing emphasis has been placed on improving quality. This factor has been made particularly manifest in the introduction of reimbursement schema that incentivize high-quality rather than high-quantity care. Quality alone is an abstract concept, and providers interested in improving quality of care require quantifiable metrics by which their efforts can be assessed. Decreasing racial disparities in specific postsurgical outcomes represents a concrete opportunity for neurosurgeons to improve quality of care for their patients.

Here, we use machine learning (ML) techniques to build a guided ML ensemble to predict 2 postoperative outcomes for patients undergoing craniotomy for brain tumor (CFBT): discharge disposition and length of hospital stay (length of stay [LOS]). We then interrogate the predictive models to investigate the independent impact of race on these outcomes. Previous work has shown the existence of racial disparities for these outcomes after CFBT: Curry et al.5 showed that black patients have more non–home discharge after CFBT, and Dasenbrock et al.6 reported that black or Hispanic patients were more likely to have extended LOS (>7 days). We expand on this work in 3 important ways: 1) we use a novel guided ML ensemble technique to validate findings ascertained using more traditional statistical methods; 2) we compare findings in the National Inpatient Sample (NIS) CFBT population with the NIS population as a whole to determine whether our findings are systemwide or specific to the CFBT population; and 3) we explore the interplay between discharge disposition and LOS.

Section snippets

Database

We used the NIS in-hospital discharge database for 2002–2011. The NIS is the largest all-payer inpatient database publicly available in the United States, containing approximately 80 million hospital stays from approximately 1000 hospitals, sampled to approximate a 20% stratified sample of U.S. hospitals.7 The NIS is compiled and maintained by the Agency for Healthcare Research and Quality (Rockville, Maryland, USA). This publicly available, de-identified database was considered exempt from

Patient Characteristics

A total of 41,222 admissions for CFBT were reviewed for analysis. Of these admissions, 25,406 resulted in discharge to home and 15,705 admissions did not. A total of 111 admissions had no or unknown discharge disposition recorded and were excluded from the study. Black patients were more likely than white or other nonblack minorities to have non–home discharge after CFBT (P < 0.001). (For additional patient characteristics, see Table A.2).

A total of 27,314 admissions lasted ≤7 days and 13,907

Discussion

Using guided ML ensembles, we validated the findings of previous studies that minority race is an independent risk factor for non–home discharge and extended LOS in ensembles that predict discharge disposition and LOS after CFBT. Bivariate analysis showed that minority patients are more likely to have multiple features that put them at higher risk for poor postoperative outcomes, including nonelective surgery, preoperative paralysis, preoperative electrolyte or fluid abnormalities, and

Conclusions

Minority race independently increases the risk of extended LOS and increases the risk of non–home discharge in patients undergoing brain tumor resection, a finding not mimicked in the general inpatient or operative population. Recognition of the influence of race on discharge and LOS could generate interventions, such as increasing education for neurosurgical providers on the existence of racial disparities or early social work or case management assessment of minority patients at hospital

References (26)

  • L. Breiman

    Random forests

    ML

    (2001)
  • J.H. Friedman

    Greedy function approximation: a gradient boosting machine

    Ann Stat

    (2001)
  • A.M. Darcy et al.

    ML and the profession of medicine

    JAMA

    (2016)
  • Cited by (47)

    View all citing articles on Scopus

    Conflict of interest statement: D.S.A. is a data scientist at DataRobot, Inc., the company whose machine learning software we use. W.E.M. is married to D.S.A. The other authors have no conflicts of interests, ethical violations, or financial disclosures. This work was supported by the Vanderbilt Medical Scholars Program and the National Institutes of Health (UL1 RR 024975 CTSA grant).

    View full text