Hospital competitive intensity and perioperative outcomes following lumbar spinal fusion

Patient frailty association with cerebral arteriovenous malformation microsurgical outcomes and development of custom risk stratification score: an analysis of 16,721 nationwide admissions

OBJECTIVE

Patient frailty is associated with poorer perioperative outcomes for several neurosurgical procedures. However, comparative accuracy between different frailty metrics for cerebral arteriovenous malformation (AVM) outcomes is poorly understood and existing frailty metrics studied in the literature are constrained by poor specificity to neurosurgery. This aim of this paper was to compare the predictive ability of 3 frailty scores for AVM microsurgical admissions and generate a custom risk stratification score.

METHODS

All adult AVM microsurgical admissions in the National (Nationwide) Inpatient Sample (2002–2017) were identified. Three frailty measures were analyzed: 5-factor modified frailty index (mFI-5; range 0–5), 11-factor modified frailty index (mFI-11; range 0–11), and Charlson Comorbidity Index (CCI) (range 0–29). Receiver operating characteristic curves were used to compare accuracy between metrics. The analyzed endpoints included in-hospital mortality, routine discharge, complications, length of stay (LOS), and hospitalization costs. Survey-weighted multivariate regression assessed frailty-outcome associations, adjusting for 13 confounders, including patient demographics, hospital characteristics, rupture status, hydrocephalus, epilepsy, and treatment modality. Subsequently, k-fold cross-validation and Akaike information criterion–based model selection were used to generate a custom 5-variable risk stratification score called the AVM-5. This score was validated in the main study population and a pseudoprospective cohort (2018–2019).

RESULTS

The authors analyzed 16,271 total AVM microsurgical admissions nationwide, with 21.0% being ruptured. The mFI-5, mFI-11, and CCI were all predictive of lower rates of routine discharge disposition, increased perioperative complications, and longer LOS (all p < 0.001). Their AVM-5 risk stratification score was calculated from 5 variables: age, hydrocephalus, paralysis, diabetes, and hypertension. The AVM-5 was predictive of decreased rates of routine hospital discharge (OR 0.26, p < 0.001) and increased perioperative complications (OR 2.42, p < 0.001), postoperative LOS (+49%, p < 0.001), total LOS (+47%, p < 0.001), and hospitalization costs (+22%, p < 0.001). This score outperformed age, mFI-5, mFI-11, and CCI for both ruptured and unruptured AVMs (area under the curve [AUC] 0.78, all p < 0.001). In a pseudoprospective cohort of 2005 admissions from 2018 to 2019, the AVM-5 remained significantly associated with all outcomes except for mortality and exhibited higher accuracy than all 3 earlier scores (AUC 0.79, all p < 0.001).

CONCLUSIONS

Patient frailty is predictive of poorer disposition and elevated complications, LOS, and costs for AVM microsurgical admissions. The authors’ custom AVM-5 risk score outperformed age, mFI-5, mFI-11, and CCI while using threefold less variables than the CCI. This score may complement existing AVM grading scales for optimization of surgical candidates and identification of patients at risk of postoperative medical and surgical morbidity.

Association of Patient Frailty With Vestibular Schwannoma Resection Outcomes and Machine Learning Development of a Vestibular Schwannoma Risk Stratification Score

BACKGROUND

Patient frailty is predictive of higher neurosurgical morbidity and mortality. However, existing frailty measures are hindered by lack of specificity to neurosurgery.

OBJECTIVE

To analyze the association between 3 risk stratification scores and outcomes for nationwide vestibular schwannoma (VS) resection admissions and develop a custom VS risk stratification score.

METHODS

We identified all VS resection admissions in the National Inpatient Sample (2002-2017). Three risk stratification scores were analyzed: modified Frailty Index-5, modified Frailty Index-11(mFI-11), and Charlson Comorbidity Index (CCI). Survey-weighted multivariate regression evaluated associations between frailty and inpatient outcomes, adjusting for patient demographics, hospital characteristics, and disease severity. Subsequently, we used k-fold cross validation and Akaike Information Criterion-based model selection to create a custom risk stratification score.

RESULTS

We analyzed 32 465 VS resection admissions. High frailty, as identified by the mFI-11 (odds ratio [OR] = 1.27, P = .021) and CCI (OR = 1.72, P < .001), predicted higher odds of perioperative complications. All 3 scores were also associated with lower routine discharge rates and elevated length of stay (LOS) and costs (all P < .05). Our custom VS-5 score (https://skullbaseresearch.shinyapps.io/vs-5_calculator/) featured 5 variables (age ≥60 years, hydrocephalus, preoperative cranial nerve palsies, diabetes mellitus, and hypertension) and was predictive of higher mortality (OR = 6.40, P = .001), decreased routine hospital discharge (OR = 0.28, P < .001), and elevated complications (OR = 1.59, P < .001), LOS (+48%, P < .001), and costs (+23%, P = .001). The VS-5 outperformed the modified Frailty Index-5, mFI-11, and CCI in predicting routine discharge (all P < .001), including in a pseudoprospective cohort (2018-2019) of 3885 admissions.

CONCLUSION

Patient frailty predicted poorer inpatient outcomes after VS surgery. Our custom VS-5 score outperformed earlier risk stratification scores.

The National Inpatient Sample: A Primer for Neurosurgical Big Data Research and Systematic Review

OBJECTIVE

The National Inpatient Sample - the largest all-payer inpatient database in the United States - is an important instrument for big data analysis of neurosurgical inquiries. However, earlier research has determined that many NIS studies are limited by common methodological pitfalls. In this study, we provide the first primer of NIS methodological procedures in the setting of neurosurgical research and review all published neurosurgical studies utilizing the NIS.

METHODS

We designed a protocol for neurosurgical big data research using the NIS, based on the authors' subject matter expertise, NIS documentation, and input and verification from the Healthcare Cost and Utilization Project. We subsequently used a comprehensive search strategy to identify all neurosurgical studies utilizing the NIS in the PubMed and MEDLINE, Embase, and Web of Science databases from inception to August 2021. Studies underwent qualitative categorization (years of the NIS studied, neurosurgical subspecialty, age group, and thematic focus of study objective) and analysis of longitudinal trends.

RESULTS

We identified a canonical, four-step protocol for NIS analysis: study population selection, defining additional clinical variables, identification and coding of outcomes, and statistical analysis. Methodological nuances discussed include identifying neurosurgery-specific admissions, addressing missing data, calculating additional severity and hospital-specific metrics, coding perioperative complications, and applying survey weights to make nationwide estimates. Inherent database limitations and common pitfalls of NIS studies discussed include lack of disease process-specific variables and data following the index admission, inability to calculate certain hospital-specific variables after 2011, performing state-level analyses, conflating hospitalization charges and costs, and not following proper statistical methodology for performing survey-weighted regression. In a systematic review, we identified 647 neurosurgical studies utilizing the NIS. While almost 60% of studies were published after 2015, <10% of studies analyzed NIS data after 2015. The average sample size of studies was 507,352 patients (standard deviation=2,739,900). Most studies analyzed cranial procedures (58.1%) and adults (68.1%). The most prevalent topic areas analyzed were surgical outcome trends (35.7%) and health policy and economics (17.8%), while patient disparities (9.4%) and surgeon or hospital volume (6.6%) were the least studied.

CONCLUSIONS

We present a standardized methodology to analyze the NIS, systematically review the state of the NIS neurosurgical literature, suggest potential future directions for neurosurgical big data inquiries, and outline recommendations to improve the design of future neurosurgical data instruments.

Brain tumor craniotomy outcomes for dual-eligible medicare and medicaid patients: a 10-year nationwide analysis

INTRODUCTION

Dual-eligible (DE) patients, simultaneous Medicare and Medicaid beneficiaries, have been shown to have poorer clinical outcomes while incurring higher resource utilization. However, neurosurgical oncology outcomes for DE patients are poorly characterized. Accordingly, we examined the impact of DE status on perioperative outcomes following glioma, meningioma, or metastasis resection.

METHODS

We identified all admissions undergoing a craniotomy for glioma, meningioma, or metastasis resection in the National Inpatient Sample from 2002 to 2011. Assessed outcomes included inpatient mortality, complications, discharge disposition, length of stay (LOS), and hospital costs. Multivariable regression adjusting for 13 patient, severity, and hospital characteristics assessed the association between DE status and outcomes, relative to four reference insurance groups (Medicare-only, Medicaid-only, private insurance, self-pay).

RESULTS

Of 195,725 total admissions analyzed, 3.0% were dual-eligible beneficiaries (n = 5933). DEs were younger than Medicare admissions (P < 0.001) but older than Medicaid, private, and self-pay admissions (P < 0.001). Relative to other insurance groups, DEs also exhibited higher severity of illness, risk of mortality, and Charlson Comorbidity Index scores as well as treatment at low-volume hospitals (all P < 0.001). DEs had lower mortality than self-pay admissions (odds ratio [OR] 0.47, P = 0.017). Compared to Medicare, Medicaid, private, and self-pay admissions, DEs had lower rates of discharge disposition (OR 0.53, 0.50, 0.34, and 0.27, respectively, all P < 0.001). DEs also had higher complications (OR 1.23 and 1.20, respectively, both P < 0.05) and LOS (β = 1.06 and 1.13, respectively, both P < 0.01) than Medicare and private insurance beneficiaries. Differences in discharge disposition remained significant for all three tumor subtypes, but only glioma DE admissions continued to exhibit higher complications and LOS.

CONCLUSIONS

DEs undergoing definitive craniotomy for brain tumor had higher rates of unfavorable discharge disposition compared to all other insurance groups and, especially for glioma surgery, had higher inpatient complication rates and LOS. Practice and policy reforms to improve outcomes for this vulnerable clinical population are warranted.

The Impact of Interhospital Competition on Treatment Strategy and Outcomes for Unruptured Intracranial Aneurysms

BACKGROUND

Interhospital competition has been shown to affect surgical outcomes and expenditures. However, interhospital competition's impact on neurosurgery is poorly characterized.

OBJECTIVE

To assess how interhospital competition is associated with treatment strategy and outcomes for unruptured intracranial aneurysms (UIAs).

METHODS

We identified all elective UIA admissions in the National Inpatient Sample from 2002 to 2011. Competitive intensity of each hospital market was quantified using the validated Herfindahl-Hirschman Index (HHI), with lower values denoting higher competition. We then obtained nationwide HHI values for 2012 to 2016 from the Health Care Cost Project. Outcomes included treatment modality (clipping, coiling, or nonoperative management), inpatient mortality, disposition, complications, length of stay (LOS), and costs. Multivariate regression assessed the association between HHI and outcomes, controlling for patient demographics, severity metrics, hospital characteristics, and treatment.

RESULTS

We studied 157 979 elective UIA admissions at 1435 hospitals from 2002 to 2011, with an increase in coiling admissions (13.4% to 33.7%) and decrease in clipping admissions (30.9% to 17.6%). Mean hospital HHI was 0.11 (range = 0.001-0.97). Competition decreased for 61.8% of hospitals from 2002 to 2011 and 68.1% of metropolitan localities from 2012 to 2016. Admissions in more competitive hospital markets exhibited increased odds of undergoing surgery (odds ratio [OR] = 1.37, P < .001), with preference toward coiling over clipping (OR = 1.27, P < .001). HHI was not associated with mortality, disposition, or LOS. However, increased interhospital competition was associated with more complications (OR = 1.09, P = .001) and greater hospital costs (β-coefficient = 1.06, P < .001).

CONCLUSION

For UIA patients, admission to hospitals in more competitive geographies was associated with increased rates of surgical intervention, coiling utilization, complications, and hospitalization costs.

Prediction of Major Complications and Readmission After Lumbar Spinal Fusion: A Machine Learning-Driven Approach

BACKGROUND

Given the significant cost and morbidity of patients undergoing lumbar fusion, accurate preoperative risk-stratification would be of great utility. We aim to develop a machine learning model for prediction of major complications and readmission after lumbar fusion. We also aim to identify the factors most important to performance of each tested model.

METHODS

We identified 38,788 adult patients who underwent lumbar fusion at any California hospital between 2015 and 2017. The primary outcome was major perioperative complication or readmission within 30 days. We build logistic regression and advanced machine learning models: XGBoost, AdaBoost, Gradient Boosting, and Random Forest. Discrimination and calibration were assessed using area under the receiver operating characteristic curve and Brier score, respectively.

RESULTS

There were 4470 major complications (11.5%). The XGBoost algorithm demonstrates the highest discrimination of the machine learning models, outperforming regression. The variables most important to XGBoost performance include angina pectoris, metastatic cancer, teaching hospital status, history of concussion, comorbidity burden, and workers' compensation insurance. Teaching hospital status and concussion history were not found to be important for regression.

CONCLUSIONS

We report a machine learning algorithm for prediction of major complications and readmission after lumbar fusion that outperforms logistic regression. Notably, the predictors most important for XGBoost differed from those for regression. The superior performance of XGBoost may be due to the ability of advanced machine learning methods to capture relationships between variables that regression is unable to detect. This tool may identify and address potentially modifiable risk factors, helping risk-stratify patients and decrease complication rates.

The impact of hospital safety-net status on inpatient outcomes for brain tumor craniotomy: a 10-year nationwide analysis

Background

Outcome disparities have been documented at safety-net hospitals (SNHs), which disproportionately serve vulnerable patient populations. Using a nationwide retrospective cohort, we assessed inpatient outcomes following brain tumor craniotomy at SNHs in the United States.

Methods

We identified all craniotomy procedures in the National Inpatient Sample from 2002–2011 for brain tumors: glioma, metastasis, meningioma, and vestibular schwannoma. Safety-net burden was calculated as the number of Medicaid plus uninsured admissions divided by total admissions. Hospitals in the top quartile of burden were defined as SNHs. The association between SNH status and in-hospital mortality, discharge disposition, complications, hospital-acquired conditions (HACs), length of stay (LOS), and costs were assessed. Multivariate regression adjusted for patient, hospital, and severity characteristics.

Results

304,719 admissions were analyzed. The most common subtype was glioma (43.8%). Of 1,206 unique hospitals, 242 were SNHs. SNH admissions were more likely to be non-white (P < .001), low income (P < .001), and have higher severity scores (P = .034). Mortality rates were higher at SNHs for metastasis admissions (odds ratio [OR] = 1.48, P = .025), and SNHs had higher complication rates for meningioma (OR = 1.34, P = .003) and all tumor types combined (OR = 1.17, P = .034). However, there were no differences at SNHs for discharge disposition or HACs. LOS and hospital costs were elevated at SNHs for all subtypes, culminating in a 10% and 9% increase in LOS and costs for the overall population, respectively (all P < .001).

Conclusions

SNHs demonstrated poorer inpatient outcomes for brain tumor craniotomy. Further analyses of the differences observed and potential interventions to ameliorate interhospital disparities are warranted.

The value of artificial neural networks for predicting length of stay, discharge disposition, and inpatient costs after anatomic and reverse shoulder arthroplasty

HYPOTHESIS/PURPOSE

The objective is to develop and validate an artificial intelligence model, specifically an artificial neural network (ANN), to predict length of stay (LOS), discharge disposition, and inpatient charges for primary anatomic total (aTSA), reverse total (rTSA), and hemi- (HSA) shoulder arthroplasty to establish internal validity in predicting patient-specific value metrics.

METHODS

Using data from the National Inpatient Sample between 2003 and 2014, 4 different ANN models to predict LOS, discharge disposition, and inpatient costs using 39 preoperative variables were developed based on diagnosis and arthroplasty type: primary chronic/degenerative aTSA, primary chronic/degenerative rTSA, primary traumatic/acute rTSA, and primary acute/traumatic HSA. Models were also combined into diagnosis type only. Outcome metrics included accuracy and area under the curve (AUC) for a receiver operating characteristic curve.

RESULTS

A total of 111,147 patients undergoing primary shoulder replacement were included. The machine learning algorithm predicting the overall chronic/degenerative conditions model (aTSA, rTSA) achieved accuracies of 76.5%, 91.8%, and 73.1% for total cost, LOS, and disposition, respectively; AUCs were 0.75, 0.89, and 0.77 for total cost, LOS, and disposition, respectively. The overall acute/traumatic conditions model (rTSA, HSA) had accuracies of 70.3%, 79.1%, and 72.0% and AUCs of 0.72, 0.78, and 0.79 for total cost, LOS, and discharge disposition, respectively.

CONCLUSION

Our ANN demonstrated fair to good accuracy and reliability for predicting inpatient cost, LOS, and discharge disposition in shoulder arthroplasty for both chronic/degenerative and acute/traumatic conditions. Machine learning has the potential to preoperatively predict costs, LOS, and disposition using patient-specific data for expectation management between health care providers, patients, and payers.

Interhospital competition and hospital charges and costs for patients undergoing cranial neurosurgery

OBJECTIVE

Research has documented significant growth in neurosurgical expenditures and practice consolidation. The authors evaluated the relationship between interhospital competition and inpatient charges or costs in patients undergoing cranial neurosurgery.

METHODS

The authors identified all admissions in 2006 and 2009 from the National Inpatient Sample. Admissions were classified into 5 subspecialties: cerebrovascular, tumor, CSF diversion, neurotrauma, or functional. Hospital-specific interhospital competition levels were quantified using the Herfindahl-Hirschman Index (HHI), an economic metric ranging continuously from 0 (significant competition) to 1 (monopoly). Inpatient charges (hospital billing) were multiplied with reported cost-to-charge ratios to calculate costs (actual resource use). Multivariate regressions were used to assess the association between HHI and inpatient charges or costs separately, controlling for 17 patient, hospital, severity, and economic factors. The reported β-coefficients reflect percentage changes in charges or costs (e.g., β-coefficient = 1.06 denotes a +6% change). All results correspond to a standardized -0.1 change in HHI (increase in competition).

RESULTS

In total, 472,938 nationwide admissions for cranial neurosurgery treated at 896 unique hospitals met inclusion criteria. Hospital HHIs ranged from 0.099 to 0.724 (mean 0.298 ± 0.105). Hospitals in more competitive markets had greater charge/cost markups (β-coefficient = 1.10, p < 0.001) and area wage indices (β-coefficient = 1.04, p < 0.001). Between 2006 and 2009, average neurosurgical charges and costs rose significantly ($62,098 to $77,812, p < 0.001; $21,385 to $22,389, p < 0.001, respectively). Increased interhospital competition was associated with greater charges for all admissions (β-coefficient = 1.07, p < 0.001) as well as cerebrovascular (β-coefficient = 1.08, p < 0.001), tumor (β-coefficient = 1.05, p = 0.039), CSF diversion (β-coefficient = 1.08, p < 0.001), neurotrauma (β-coefficient = 1.07, p < 0.001), and functional neurosurgery (β-coefficient = 1.11, p = 0.037) admissions. However, no significant associations were observed between HHI and costs, except for CSF diversion surgery (β-coefficient = 1.03, p = 0.021). Increased competition was not associated with important clinical outcomes, such as inpatient mortality, favorable discharge disposition, or complication rates, except for lower mortality for brain tumors (OR 0.78, p = 0.026), but was related to greater length of stay for all admissions (β-coefficient = 1.06, p < 0.001). For a sensitivity analysis adjusting for outcomes, all findings for charges and costs remained the same.

CONCLUSIONS

Hospitals in more competitive markets exhibited higher charges for admissions of patients undergoing an in-hospital cranial procedure. Despite this, interhospital competition was not associated with increased inpatient costs except for CSF diversion surgery. There was no corresponding improvement in outcomes with increased competition, with the exception of a potential survival benefit for brain tumor surgery.

The association of hospital competition with inpatient costs of stroke: Evidence from China

The main purpose of this article is to analyze the association between hospital competition and stroke inpatient costs. Stroke is selected as the representative of a class of diseases characterized by asymmetric information and lack of autonomy of service choice. A total of 98,061 observations are selected from a medical record dataset in the Sichuan Province. The fixed radius approach of 15 miles and Herfindahl-Hirschman Index (HHI) are employed to define the hospital market and measure the competition intensity, respectively. The log-linear multivariate regression model is used to analyze the association between competition and stroke inpatient costs. The results show that every 10% increase in competition (0.1 unit decrease of HHI value) associated with an average 2.38% decrease in stroke inpatient total costs. We also explore the relationship between competition and sub-group costs of stroke inpatient, finding that hospitals facing more competition incur lower treatment, drug, and consumable costs. Further analysis shows that for-profit, private, and low-level hospitals are more sensitive when facing changes in market competition degree. Our study offers empirical evidence to support the introduction of pro-competition in China's new round of national health reform and provide implications for other countries facing similar health care challenges.