1
|
Wu TC, Kim A, Tsai CT, Gao A, Ghuman T, Paul A, Castillo A, Cheng J, Adogwa O, Ngwenya LB, Foreman B, Wu DT. A Neurosurgical Readmissions Reduction Program in an Academic Hospital Leveraging Machine Learning, Workflow Analysis, and Simulation. Appl Clin Inform 2024; 15:479-488. [PMID: 38897230 PMCID: PMC11186699 DOI: 10.1055/s-0044-1787119] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/08/2023] [Accepted: 04/26/2024] [Indexed: 06/21/2024] Open
Abstract
BACKGROUND Predicting 30-day hospital readmissions is crucial for improving patient outcomes, optimizing resource allocation, and achieving financial savings. Existing studies reporting the development of machine learning (ML) models predictive of neurosurgical readmissions do not report factors related to clinical implementation. OBJECTIVES Train individual predictive models with good performance (area under the receiver operating characteristic curve or AUROC > 0.8), identify potential interventions through semi-structured interviews, and demonstrate estimated clinical and financial impact of these models. METHODS Electronic health records were utilized with five ML methodologies: gradient boosting, decision tree, random forest, ridge logistic regression, and linear support vector machine. Variables of interest were determined by domain experts and literature. The dataset was split divided 80% for training and validation and 20% for testing randomly. Clinical workflow analysis was conducted using semi-structured interviews to identify possible intervention points. Calibrated agent-based models (ABMs), based on a previous study with interventions, were applied to simulate reductions of the 30-day readmission rate and financial costs. RESULTS The dataset covered 12,334 neurosurgical intensive care unit (NSICU) admissions (11,029 patients); 1,903 spine surgery admissions (1,641 patients), and 2,208 traumatic brain injury (TBI) admissions (2,185 patients), with readmission rate of 13.13, 13.93, and 23.73%, respectively. The random forest model for NSICU achieved best performance with an AUROC score of 0.89, capturing potential patients effectively. Six interventions were identified through 12 semi-structured interviews targeting preoperative, inpatient stay, discharge phases, and follow-up phases. Calibrated ABMs simulated median readmission reduction rates and resulted in 13.13 to 10.12% (NSICU), 13.90 to 10.98% (spine surgery), and 23.64 to 21.20% (TBI). Approximately $1,300,614.28 in saving resulted from potential interventions. CONCLUSION This study reports the successful development and simulation of an ML-based approach for predicting and reducing 30-day hospital readmissions in neurosurgery. The intervention shows feasibility in improving patient outcomes and reducing financial losses.
Collapse
Affiliation(s)
- Tzu-Chun Wu
- Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
| | - Abraham Kim
- Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Medical Sciences Baccalaureate Program, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
| | - Ching-Tzu Tsai
- Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
| | - Andy Gao
- Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Medical Sciences Baccalaureate Program, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
| | - Taran Ghuman
- Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Medical Sciences Baccalaureate Program, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
| | - Anne Paul
- UCHealth, Cincinnati, Ohio, United States
| | | | - Joseph Cheng
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- UCHealth, Cincinnati, Ohio, United States
| | - Owoicho Adogwa
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- UCHealth, Cincinnati, Ohio, United States
| | - Laura B. Ngwenya
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Department of Neurology and Rehabilitation Medicine, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- UCHealth, Cincinnati, Ohio, United States
| | - Brandon Foreman
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Department of Neurology and Rehabilitation Medicine, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- UCHealth, Cincinnati, Ohio, United States
| | - Danny T.Y. Wu
- Department of Biomedical Informatics, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Neuroinformatics Laboratory, Department of Neurosurgery, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
- Medical Sciences Baccalaureate Program, College of Medicine, University of Cincinnati, Cincinnati, Ohio, United States
| |
Collapse
|
2
|
Lusk JB, Hoffman MN, Clark AG, Bae J, Luedke MW, Hammill BG. Association Between Neighborhood Socioeconomic Status and 30-Day Mortality and Readmission for Patients With Common Neurologic Conditions. Neurology 2023; 100:e1776-e1786. [PMID: 36792379 PMCID: PMC10136022 DOI: 10.1212/wnl.0000000000207094] [Citation(s) in RCA: 2] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 06/17/2022] [Accepted: 01/10/2023] [Indexed: 02/17/2023] Open
Abstract
BACKGROUND AND OBJECTIVES Patients of low individual socioeconomic status (SES) are at a greater risk of unfavorable health outcomes. However, the association between neighborhood socioeconomic deprivation and health outcomes for patients with neurologic disorders has not been studied at the population level. Our objective was to determine the association between neighborhood socioeconomic deprivation and 30-day mortality and readmission after hospitalization for various neurologic conditions. METHODS This was a retrospective study of nationwide Medicare claims from 2017 to 2019. We included patients older than 65 years hospitalized for the following broad categories based on diagnosis-related groups (DRGs): multiple sclerosis and cerebellar ataxia (DRG 058-060); stroke (061-072); degenerative nervous system disorders (056-057); epilepsy (100-101); traumatic coma (082-087), and nontraumatic coma (080-081). The exposure of interest was neighborhood SES, measured by the area deprivation index (ADI), which uses socioeconomic indicators, such as educational attainment, unemployment, infrastructure access, and income, to estimate area-level socioeconomic deprivation at the level of census block groups. Patients were grouped into high, middle, and low neighborhood-level SES based on ADI percentiles. Adjustment covariates included age, comorbidity burden, race/ethnicity, individual SES, and sex. RESULTS After exclusions, 905,784 patients were included in the mortality analysis and 915,993 were included in the readmission analysis. After adjustment for age, sex, race/ethnicity, comorbidity burden, and individual SES, patients from low SES neighborhoods had higher 30-day mortality rates compared with patients from high SES neighborhoods for all disease categories except for multiple sclerosis: magnitudes of the effect ranged from an adjusted odds ratio of 2.46 (95% CI 1.60-3.78) for the nontraumatic coma group to 1.23 (95% CI 1.19-1.28) for the stroke group. After adjustment, no significant differences in readmission rates were observed for any of the groups. DISCUSSION Neighborhood SES is strongly associated with 30-day mortality for many common neurologic conditions even after accounting for baseline comorbidity burden and individual SES. Strategies to improve health equity should explicitly consider the effect of neighborhood environments on health outcomes.
Collapse
Affiliation(s)
- Jay B Lusk
- From the Duke University School of Medicine (J.B.L., B.G.H.); Duke University Fuqua School of Business (J.B.L.); Duke University Department of Population Health Sciences (M.N.H., A.G.C., B.G.H.); Duke University Health System (J.B.); Duke University Department of Medicine (J.B.); and Duke University Department of Neurology (M.W.L.), Durham, NC
| | - Molly N Hoffman
- From the Duke University School of Medicine (J.B.L., B.G.H.); Duke University Fuqua School of Business (J.B.L.); Duke University Department of Population Health Sciences (M.N.H., A.G.C., B.G.H.); Duke University Health System (J.B.); Duke University Department of Medicine (J.B.); and Duke University Department of Neurology (M.W.L.), Durham, NC
| | - Amy G Clark
- From the Duke University School of Medicine (J.B.L., B.G.H.); Duke University Fuqua School of Business (J.B.L.); Duke University Department of Population Health Sciences (M.N.H., A.G.C., B.G.H.); Duke University Health System (J.B.); Duke University Department of Medicine (J.B.); and Duke University Department of Neurology (M.W.L.), Durham, NC
| | - Jonathan Bae
- From the Duke University School of Medicine (J.B.L., B.G.H.); Duke University Fuqua School of Business (J.B.L.); Duke University Department of Population Health Sciences (M.N.H., A.G.C., B.G.H.); Duke University Health System (J.B.); Duke University Department of Medicine (J.B.); and Duke University Department of Neurology (M.W.L.), Durham, NC
| | - Matthew W Luedke
- From the Duke University School of Medicine (J.B.L., B.G.H.); Duke University Fuqua School of Business (J.B.L.); Duke University Department of Population Health Sciences (M.N.H., A.G.C., B.G.H.); Duke University Health System (J.B.); Duke University Department of Medicine (J.B.); and Duke University Department of Neurology (M.W.L.), Durham, NC
| | - Bradley G Hammill
- From the Duke University School of Medicine (J.B.L., B.G.H.); Duke University Fuqua School of Business (J.B.L.); Duke University Department of Population Health Sciences (M.N.H., A.G.C., B.G.H.); Duke University Health System (J.B.); Duke University Department of Medicine (J.B.); and Duke University Department of Neurology (M.W.L.), Durham, NC.
| |
Collapse
|
3
|
Chen N, Li R, Wang E, Hu D, Tang Z. [Outcomes of patients experiencing cardiovascular adverse events within 1 year following craniotomy for intracranial aneurysm clipping: a retrospective cohort study]. NAN FANG YI KE DA XUE XUE BAO = JOURNAL OF SOUTHERN MEDICAL UNIVERSITY 2022; 42:1095-1099. [PMID: 35869776 DOI: 10.12122/j.issn.1673-4254.2022.07.20] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Subscribe] [Scholar Register] [Indexed: 11/24/2022]
Abstract
OBJECTIVE To investigate the impact of postoperative serious cardiovascular adverse events (CAE) on outcomes of patients undergoing craniotomy for intracranial aneurysm clipping. METHODS This retrospective cohort study was conducted among the patients undergoing craniotomy for intracranial aneurysm clipping during the period from December, 2016 to December, 2017, who were divided into CAE group and non-CAE group according to the occurrence of Clavien-Dindo grade ≥II CAEs after the surgery. The perioperative clinical characteristics of the patients, complications and neurological functions during hospitalization, and mortality and neurological functions at 1 year postoperatively were evaluated. The primary outcome was mortality within 1 year after the surgery. The secondary outcomes were Glasgow outcome scale (GOS) score at 1 year, lengths of postoperative hospital and intensive care unit (ICU) stay, and Glasgow coma scale (GCS) score at discharge. RESULTS A total of 361 patients were enrolled in the final analysis, including 20 (5.5%) patients in CAE group and 341 in the non-CAE group. No significant differences were found in the patients' demographic characteristics, clinical history, or other postoperative adverse events between the two groups. The 1-year mortality was significantly higher in CAE group than in the non-CAE group (20.0% vs 5.6%, P=0.01). Logistics regression analysis showed that when adjusted for age, gender, emergency hospitalization, subarachnoid hemorrhage, volume of bleeding, duration of operation, aneurysm location, and preoperative history of cardiovascular disease, postoperative CAEs of Clavien-Dindo grade≥II was independently correlated with 1-year mortality rate of the patients with an adjusted odds ratio of 3.670 (95% CI: 1.037-12.992, P=0.04). The patients with CEA also had a lower GOS score at 1 year after surgery than those without CEA (P=0.002). No significant differences were found in the occurrence of other adverse events, postoperative hospital stay, ICU stay, or GCS scores at discharge between the two groups (P > 0.05). CONCLUSION Postoperative CAEs may be a risk factor for increased 1-year mortality and disability in patients undergoing craniotomy for intracranial aneurysms.
Collapse
Affiliation(s)
- N Chen
- Department of Anesthesiology, Xiangya Hospital of Central South University, Changsha 410008, China
| | - R Li
- Department of Anesthesiology, Hunan Provincial People's Hospital, Changsha 410005, China
| | - E Wang
- Department of Anesthesiology, Xiangya Hospital of Central South University, Changsha 410008, China.,National Clinical Research Center for Geriatric Disorders (Xiangya Hospital), Changsha 410008, China
| | - D Hu
- School of Life Sciences, Central South University, Changsha 410008, China
| | - Z Tang
- Department of Anesthesiology, Xiangya Hospital of Central South University, Changsha 410008, China.,National Clinical Research Center for Geriatric Disorders (Xiangya Hospital), Changsha 410008, China
| |
Collapse
|
4
|
Zachrison KS, Li S, Reeves MJ, Adeoye O, Camargo CA, Schwamm LH, Hsia RY. Strategy for reliable identification of ischaemic stroke, thrombolytics and thrombectomy in large administrative databases. Stroke Vasc Neurol 2020; 6:194-200. [PMID: 33177162 PMCID: PMC8258073 DOI: 10.1136/svn-2020-000533] [Citation(s) in RCA: 14] [Impact Index Per Article: 3.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/23/2020] [Revised: 09/28/2020] [Accepted: 10/02/2020] [Indexed: 12/14/2022] Open
Abstract
Background Administrative data are frequently used in stroke research. Ensuring accurate identification of patients who had an ischaemic stroke, and those receiving thrombolysis and endovascular thrombectomy (EVT) is critical to ensure representativeness and generalisability. We examined differences in patient samples based on mode of identification, and propose a strategy for future patient and procedure identification in large administrative databases. Methods We used non-public administrative data from the state of California to identify all patients who had an ischaemic stroke discharged from an emergency department (ED) or inpatient hospitalisation from 2010 to 2017 based on International Classification of Disease (ICD-9) (2010–2015), ICD-10 (2015–2017) and Medicare Severity-Diagnosis-related Group (MS-DRG) discharge codes. We identified patients with interhospital transfers, patients receiving thrombolytics and patients treated with EVT based on ICD, Current Procedural Terminology (CPT) and MS-DRG codes. We determined what proportion of these transfers and procedures would have been identified with ICD versus MS-DRG discharge codes. Results Of 365 099 ischaemic stroke encounters, most (87.70%) had both a stroke-related ICD-9 or ICD-10 code and stroke-related MS-DRG code; 12.28% had only an ICD-9 or ICD-10 code and 0.02% had only an MS-DRG code. Nearly all transfers (99.99%) were identified using ICD codes. We identified 32 433 thrombolytic-treated patients (8.9% of total) using ICD, CPT and MS-DRG codes; the combination of ICD and CPT codes identified nearly all (98%). We identified 7691 patients treated with EVT (2.1% of total) using ICD and MS-DRG codes; both MS-DRG and ICD-9/ICD-10 codes were necessary because ICD codes alone missed 13.2% of EVTs. CPT codes only pertain to outpatient/ED patients and are not useful for EVT identification. Conclusions ICD-9/ICD-10 diagnosis codes capture nearly all ischaemic stroke encounters and transfers, while the combination of ICD-9/ICD-10 and CPT codes are adequate for identifying thrombolytic treatment in administrative datasets. However, MS-DRG codes are necessary in addition to ICD codes for identifying EVT, likely due to favourable reimbursement for EVT-related MS-DRG codes incentivising accurate coding.
Collapse
Affiliation(s)
- Kori S Zachrison
- Department of Emergency Medicine, Massachusetts General Hospital, Boston, Massachusetts, USA .,Department of Emergency Medicine, Harvard Medical School, Boston, Massachusetts, USA
| | - Sijia Li
- Department of Emergency Medicine, Massachusetts General Hospital, Boston, Massachusetts, USA
| | - Mathew J Reeves
- Department of Epidemiology and Biostatistics, Michigan State University, East Lansing, Michigan, USA
| | | | - Carlos A Camargo
- Department of Emergency Medicine, Massachusetts General Hospital, Boston, Massachusetts, USA
| | - Lee H Schwamm
- Massachusetts General Hospital, Boston, Massachusetts, USA
| | - Renee Y Hsia
- Department of Emergency Medicine, University of California San Francisco, San Francisco, California, USA
| |
Collapse
|
5
|
Kwon M, Lekoubou A, Bishu KG, Ovbiagele B. Association of seizure co-morbidity with early hospital readmission among traumatic brain injury patients. Brain Inj 2020; 34:1625-1629. [PMID: 33017194 DOI: 10.1080/02699052.2020.1825808] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/23/2022]
Abstract
OBJECTIVE To assess the frequency of seizure co-morbidity and its independent association with 30-day readmission rate among patients hospitalized with traumatic brain injury (TBI) in the United States. METHODS The data source was the 2014 Nationwide Readmission Database. We included adults (Age ≥18 years) with a primary discharge diagnosis of TBI, identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes 800.0, 801.9, 803.0, 804.9, 850.0-854.1, and 959.01. Seizures were diagnosed using the ICD-9-CM codes of 345.x and 780.39. Overall and across pre-specified groups 30-readmission rate was computed. Logistic regression analysis was used to identify independent predictors of 30-day readmission. RESULTS Among 76,062 unweighted adults discharged with a diagnosis of TBI, 7,776 (10.14%) had a secondary discharge diagnosis of seizures.A total of 1,751 (2.3%) patients with a primary discharge diagnosis of TBI were readmitted within 30 days. On multivariate logistic analysis, patients discharged with a secondary diagnosis of seizures were 18% more likely to be readmitted within 30 days compared to those without seizures (OR 1.18, 95% CI: 1.01-1.39, P = .42). CONCLUSION One in 10 patients hospitalized with TBI in the US have a co-morbid seizure disorder. Seizure co-morbidity conferred 18% greater odds of being readmitted within 30 days.
Collapse
Affiliation(s)
| | - Alain Lekoubou
- Department of Neurology, Penn State University , Hershey, PA, USA.,Department of Public Health Sciences, Division of Epidemiology, Penn State University , Hershey, PA, USA
| | - Kinfe G Bishu
- Department of Medicine, Medical University of South Carolina , Charleston, SC, USA.,Health Equity and Rural Outreach Innovation Center (HEROIC), Ralph H. Johnson VA Medical Center , Charleston, SC, USA
| | - Bruce Ovbiagele
- Department of Neurology, University of California , San Francisco, CA, USA
| |
Collapse
|