Predict Discharge Disposition Research Articles

Comparison of Two Previously Validated Risk Assessment and Prediction Tool Models in Extended Length of Stay Patients Following Total Hip Arthroplasty

Previous studies have attempted to validate the risk assessment and prediction tool (RAPT) in primary total hip arthroplasty (THA) patients. The purpose of this study was to: (1) identify patients who had an extended length of stay (ELOS) following THA; and (2) compare the accuracy of two previously validated RAPT models. We retrospectively reviewed all primary THA patients from 2014 to 2021 who had a completed RAPT score. Youden's J computational analysis was used to determine the LOS where facility discharge was statistically more likely. Based on the cut-offs proposed by Oldmeadow and Dibra, patients were separated into high- (O: 1 to 5 versus D: 1 to 3), medium- (O: 6 to 9 versus D: 4 to 7), and low- (O: 10 to 12 versus D: 8 to 12) risk groups. We determined that a LOS of greater than two days resulted in a higher chance of facility discharge. In these patients (n = 717), the overall predictive accuracy (PA) of the RAPT was 79.8%. The Dibra model had a higher PA in the high-risk group (D: 68.2 versus O: 61.2% facility discharge). The Oldmeadow model had a higher PA in the medium-risk (O: 78.7 versus D: 61.4% home discharge) and low-risk (O: 97.0 versus D. 92.5% home discharge) groups. As institutions continue to optimize LOS, the RAPT may need to be defined in the context of a patient's hospital stay. In patients requiring a LOS of greater than two days, the originally established RAPT cut-offs may be more accurate in predicting discharge disposition.

Behavioral Assessment With the Coma Recovery Scale-Revised Is Safe and Feasible in Critically Ill Patients With Disorders of Consciousness.

Accurate classification of disorders of consciousness (DoC) is key in developing rehabilitation plans after brain injury. The Coma Recovery Scale-Revised (CRS-R) is a sensitive measure of consciousness validated in the rehabilitation phase of care. We tested the feasibility, safety, and impact of CRS-R-guided rehabilitation in the ICU for patients with DoC after acute hemorrhagic stroke. Retrospective cohort study. This single-center study was conducted in the neurocritical care unit at the University of Maryland Medical Center. We analyzed records from consecutive patients with subarachnoid hemorrhage (SAH) or intracerebral hemorrhage (ICH), who underwent serial CRS-R assessments during ICU admission from April 1, 2018, to December 31, 2021, where CRS-R less than 8 is vegetative state/unresponsive wakefulness syndrome (VS/UWS); CRS-R greater than or equal to 8 is a minimally conscious state (MCS). None. Outcomes included adverse events during CRS-R evaluations and associations between CRS-R and discharge disposition, therapy-based function, and mobility. We examined the utility of CRS-R compared with other therapist clinical assessment tools in predicting discharge disposition. Seventy-six patients (22 SAH, 54 ICH, median age = 59, 50% female) underwent 276 CRS-R sessions without adverse events. Discharge to acute rehabilitation occurred in 4.4% versus 41.9% of patients with a final CRS-R less than 8 and CRS-R greater than or equal to 8, respectively (odds ratio [OR] 13.4; 95% CI, 2.7-66.1; p < 0.001). Patients with MCS on final CRS-R completed more therapy sessions during hospitalization and had improved mobility and functional performance. Compared with other therapy assessment tools, the CRS-R had the best performance in predicting discharge disposition (area under the curve: 0.83; 95% CI, 0.72-0.94; p < 0.0001). Early neurorehabilitation guided by CRS-R appears to be feasible and safe in the ICU following hemorrhagic stroke complicated by DoC and may enhance access to inpatient rehabilitation, with the potential for lasting benefit on recovery. Further research is needed to assess generalizability and understand the impact on long-term outcomes.

Use of The Risk Assessment and Prediction Tool to Predict Same-day Discharge After Primary Hip and Knee Arthroplasty.

The Risk Assessment and Prediction Tool (RAPT) is a preoperative screening tool developed to predict discharge disposition after total hip arthroplasty (THA) and total knee arthroplasty (TKA), but its predictive value for same-day discharge (SDD) has not been investigated. The aims of this study were (1) to assess RAPT's ability to predict SDD after primary THA and TKA and (2) to determine a cutoff RAPT score that may recognize patients appropriate for SDD. Data were retrospectively collected from patients undergoing primary THA and TKA at a single tertiary care center between February 2020 and May 2021. A receiver operating characteristic curve was generated to choose a cutoff value to screen for SDD. Logistic regression analysis was done to identify factors including age, BMI, or RAPT score that may be associated with SDD. Three hundred sixty-one patients with preoperative RAPT scores were included in the analysis of whom 147 (42.6%) underwent SDD. A cutoff of ≥9 was identified for TKA and ≥11 for THA. RAPT had a predictive accuracy of only 66.7% for SDD, whereas the discharge plan documented in the preoperative note was 91.7% accurate. Although there is a positive association between RAPT and SDD, it is not a useful screening tool given its low predictive accuracy.

Predictors of hospital length of stay and long-term care facility placement in aneurysmal subarachnoid hemorrhage

BackgroundAneurysmal subarachnoid hemorrhage (aSAH) is frequently associated with complications, extended hospital length of stay (LOS) and high health care related costs. We sought to determine predictors for hospital LOS and discharge disposition to a long-term care facility (LTCF) in aSAH patients. MethodsWe performed a retrospective study of a prospectively collected cohort of consecutive patients with aSAH admitted to an academic referral center from 2016 to 2021. Multiple linear regression was performed to identify predictors for hospital LOS. We then created a 10-point scoring system to predict discharge disposition to a LTCF. ResultsIn a cohort of 318 patients with confirmed aSAH, mean age was 57 years (SD 13.7), 61% were female and 70% were white. Hospital LOS was longer for survivors (median 19 days, IQR 14–25) than for non-survivors (median 5 days, IQR 2–8; p < 0.001). Main predictors for longer LOS for this cohort were ventriculoperitoneal shunt (VPS) requirement (p < 0.001), delayed cerebral ischemia (p = 0.026), and pneumonia (p = 0.014). The strongest predictor for LTCF disposition was age older than 60 years (OR 1.14, 95% CI 1.07–1.21; p < 0.001). LTCF score had high accuracy in predicting discharge disposition to a LTCF (area under the curve [AUC] 0.83; 95% CI 0.75–0.91). Forty-one percent of patients who were discharged to a LTCF had significant functional recovery at 3 months post-discharge. ConclusionsVPS requirement and aSAH related complications were associated with longer hospital LOS compared to other factors. LTCF score has high accuracy in predicting discharge disposition to a LTCF.

Nurse-Sensitive Indicators as Predictors of Trauma Patient Discharge Disposition.

About 3.5 million trauma patients are hospitalized every year, but 35%-40% require further care after discharge. Nurses' ability to affect discharge disposition by minimizing the occurrence of nurse-sensitive indicators (catheter-associated urinary tract infection [CAUTI], central line-associated bloodstream infection [CLABSI], and hospital-acquired pressure injury [HAPI]) is unknown. These indicators may serve as surrogate measures of quality nursing care. The purpose of this study was to determine whether nursing care, as represented by three nurse-sensitive indicators (CAUTI, CLABSI, and HAPI), predicts discharge disposition in trauma patients. This study was a secondary analysis of the 2021 National Trauma Data Bank. We performed logistic regression analyses to determine the predictive effects of CAUTI, CLABSI, and HAPI on discharge disposition, controlling for participant characteristics. A total of n =29,642 patients were included, of which n =21,469 (72%) were male, n =16,404 (64%) were White, with a mean (SD) age of 44 (14.5) and mean (SD) Injury Severity Score of 23.2 (12.5). We created four models to test nurse-sensitive indicators, both individually and compositely, as predictors. While CAUTI and HAPI increased the odds of discharge to further care by 1.4-1.5 and 2.1 times, respectively, CLABSI was not a statistically significant predictor. Both CAUTI and HAPI are statistically significant predictors of discharge to further care for patients after traumatic injury. High-quality nursing care to prevent iatrogenic complications can improve trauma patients' long-term outcomes.

Factors Predicting Discharge Disposition Following Head and Neck Free Flap Reconstruction.

Patients undergoing head and neck free flap reconstruction (HNFFR) may have significant change to their baseline functional status requiring inpatient rehabilitation (IPR) after discharge. We sought to identify patient/procedure characteristics predictive of discharge destination. Patients undergoing elective HNFFR between July 2017 and July 2022 were reviewed for discharge destination. Those discharged to IPR versus home were compared across patient/procedure characteristics and physical/occupational therapy metrics. Significance was assessed via bivariate and multivariable analyses. Of the 531 patients, 102 (19.2%) required IPR postoperatively. Patients discharged to IPR versus home were significantly older (70.1 [11.6] vs. 64.1 [13.1] years; p < 0.001) and more likely to lack family assistance (26.5% vs. 8.6%; p < 0.001), require baseline assistance for activities of daily living (ADLs) (31.4% vs. 9.8%; p < 0.001), have baseline cognitive dysfunction (15.7% vs. 6.1%; p = 0.001), were more likely to have neoplasm as the surgical indication for HNFFR (89.2% vs. 80.0%; p = 0.033) and more likely to have a tracheostomy postop (62.7% vs. 51.7%), and had a significantly longer length of stay (11.2 [8.0] vs. 6.8 [8.3] days; p < 0.001). There was no significant difference in gender, donor site, use of tube feeds, and use of assistive devices between the two groups. Following logistic regression, the strongest predictors of discharge to IPR include lack of family assistance (OR = 3.8; p < 0.001) and baseline assistance for ADLs (OR = 4.0, p < 0.001). Certain patient factors predict the need for discharge to rehab after HNFFR. Perioperative identification of these factors may facilitate patient counseling and discharge planning with potential to reduce hospital length of stay and further optimize patient care. III Laryngoscope, 134:2721-2725, 2024.

Machine learning prediction of hospital discharge disposition for inpatients with acute ischemic stroke following mechanical thrombectomy in the United States

Background and purposePredicting patient recovery and discharge disposition following mechanical thrombectomy remains a challenge in patients with ischemic stroke. Machine learning offers a promising prognostication approach assisting in personalized post-thrombectomy care plans and resource allocation. As a large national database, National Inpatient Sample (NIS), contain valuable insights amenable to data-mining. The study aimed to develop and evaluate ML models predicting hospital discharge disposition with a focus on demographic, socioeconomic and hospital characteristics. Materials and methodsThe NIS dataset (2006–2019) was used, including 4956 patients diagnosed with ischemic stroke who underwent thrombectomy. Demographics, hospital characteristics, and Elixhauser comorbidity indices were recorded. Feature extraction, processing, and selection were performed using Python, with Maximum Relevance - Minimum Redundancy (MRMR) applied for dimensionality reduction. ML models were developed and benchmarked prior to interpretation of the best model using Shapley Additive exPlanations (SHAP). ResultsThe multilayer perceptron model outperformed others and achieved an AUROC of 0.81, accuracy of 77 %, F1-score of 0.48, precision of 0.64, and recall of 0.54. SHAP analysis identified the most important features for predicting discharge disposition as dysphagia and dysarthria, NIHSS, age, primary payer (Medicare), cerebral edema, fluid and electrolyte disorders, complicated hypertension, primary payer (private insurance), intracranial hemorrhage, and thrombectomy alone. ConclusionMachine learning modeling of NIS database shows potential in predicting hospital discharge disposition for inpatients with acute ischemic stroke following mechanical thrombectomy in the NIS database. Insights gained from SHAP interpretation can inform targeted interventions and care plans, ultimately enhancing patient outcomes and resource allocation.

The Impact of Hospital Exposures Prior to Total Knee Arthroplasty on Postoperative Outcomes.

Total knee arthroplasty (TKA) procedures are expected to grow exponentially in the upcoming years, highlighting the importance of identifying preoperative risk factors that predispose patients to poor outcomes. The present study sought to determine if preoperative healthcare events (PHEs) influenced outcomes following TKA. This was a retrospective review of all patients who underwent TKA at a single institution from June 2011 to April 2022. Patients who had a PHE within 90 days of surgery, defined as an emergency department visit or hospital admission, were compared to patients with no history of PHE. Patients who underwent revision, nonelective, and/or bilateral TKA were excluded. Chi-squared analysis and independent sample t-tests were used to determine significant differences between demographic variables. All significant covariates were included in binary logistic regressions used to predict discharge disposition, 90-day readmission, and 1-year revision. Of the 10,869 patients who underwent TKA, 265 had ≥1 PHE. Patients who had a PHE were significantly more likely to require facility discharge (odds ratio [OR]: 1.662; P= .001) than patients who did not have a PHE. Any PHE predisposed patients to significantly higher 90-day readmission rates (OR: 2.173; P= .002). Patients with ≥2 PHEs were at a significantly higher risk of 1-year revision (OR: 5.870; P= .004) compared to patients without a PHE. Our results demonstrate that PHEs put patients at significantly greater risk of facility discharge, 90-day readmission, and 1-year revision. Moving forward, consideration of elective surgery scheduling in the context of a recent PHE may lead to improved postoperative outcomes. Retrospective Cohort Study.

Development of a Machine Learning–Based Prescriptive Tool to Address Racial Disparities in Access to Care After Penetrating Trauma

The use of artificial intelligence (AI) in clinical medicine risks perpetuating existing bias in care, such as disparities in access to postinjury rehabilitation services. To leverage a novel, interpretable AI-based technology to uncover racial disparities in access to postinjury rehabilitation care and create an AI-based prescriptive tool to address these disparities. This cohort study used data from the 2010-2016 American College of Surgeons Trauma Quality Improvement Program database for Black and White patients with a penetrating mechanism of injury. An interpretable AI methodology called optimal classification trees (OCTs) was applied in an 80:20 derivation/validation split to predict discharge disposition (home vs postacute care [PAC]). The interpretable nature of OCTs allowed for examination of the AI logic to identify racial disparities. A prescriptive mixed-integer optimization model using age, injury, and gender data was allowed to "fairness-flip" the recommended discharge destination for a subset of patients while minimizing the ratio of imbalance between Black and White patients. Three OCTs were developed to predict discharge disposition: the first 2 trees used unadjusted data (one without and one with the race variable), and the third tree used fairness-adjusted data. Disparities and the discriminative performance (C statistic) were compared among fairness-adjusted and unadjusted OCTs. A total of 52 468 patients were included; the median (IQR) age was 29 (22-40) years, 46 189 patients (88.0%) were male, 31 470 (60.0%) were Black, and 20 998 (40.0%) were White. A total of 3800 Black patients (12.1%) were discharged to PAC, compared with 4504 White patients (21.5%; P < .001). Examining the AI logic uncovered significant disparities in PAC discharge destination access, with race playing the second most important role. The prescriptive fairness adjustment recommended flipping the discharge destination of 4.5% of the patients, with the performance of the adjusted model increasing from a C statistic of 0.79 to 0.87. After fairness adjustment, disparities disappeared, and a similar percentage of Black and White patients (15.8% vs 15.8%; P = .87) had a recommended discharge to PAC. In this study, we developed an accurate, machine learning-based, fairness-adjusted model that can identify barriers to discharge to postacute care. Instead of accidentally encoding bias, interpretable AI methodologies are powerful tools to diagnose and remedy system-related bias in care, such as disparities in access to postinjury rehabilitation care.

Improved performance of machine learning models in predicting length of stay, discharge disposition, and inpatient mortality after total knee arthroplasty using patient-specific variables

BackgroundThis study aimed to compare the performance of ten predictive models using different machine learning (ML) algorithms and compare the performance of models developed using patient-specific vs. situational variables in predicting select outcomes after primary TKA.MethodsData from 2016 to 2017 from the National Inpatient Sample were used to identify 305,577 discharges undergoing primary TKA, which were included in the training, testing, and validation of 10 ML models. 15 predictive variables consisting of 8 patient-specific and 7 situational variables were utilized to predict length of stay (LOS), discharge disposition, and mortality. Using the best performing algorithms, models trained using either 8 patient-specific and 7 situational variables were then developed and compared.ResultsFor models developed using all 15 variables, Linear Support Vector Machine (LSVM) was the most responsive model for predicting LOS. LSVM and XGT Boost Tree were equivalently most responsive for predicting discharge disposition. LSVM and XGT Boost Linear were equivalently most responsive for predicting mortality. Decision List, CHAID, and LSVM were the most reliable models for predicting LOS and discharge disposition, while XGT Boost Tree, Decision List, LSVM, and CHAID were most reliable for mortality. Models developed using the 8 patient-specific variables outperformed those developed using the 7 situational variables, with few exceptions.ConclusionThis study revealed that performance of different models varied, ranging from poor to excellent, and demonstrated that models developed using patient-specific variables were typically better predictive of quality metrics after TKA than those developed employing situational variables.Level of EvidenceIII.

Predictors of discharge to home/community following inpatient-rehabilitation in a US national sample of Guillain-Barre-Syndrome patients.

Guillain-Barre-Syndrome (GBS), an autoimmune polyneuropathy causing acute flaccid paralysis, is a rare condition with1-2 cases per 100,000 annually (approximately 5000 cases/year) in the United States (US). There is a paucity of published data regarding patient outcomes in association with discharge destinations following inpatient-rehabilitation (IR) in this patient population, thus this study. To analyze IR efficacy, and possible predictors of discharge to home/community in a US-national-sample of GBS patients. Retrospective-observational-cohort study of 1304 GBS patients admitted to IR comparing discharge disposition destinations (community/home, skilled-nursing-facility [SNF], or return to acute-care) by demographic (age, gender) and clinical variables (length-of-stay [LOS], case-mix-index [CMI], and Functional-Independence-Measure [FIM] score changes). Multinomial-logistic-regression and discriminant-function-analysis were performed to determine model fit in predicting discharge destination. 81.8% were discharged to home/community- average LOS 19-days, total-FIM-gain 43.2; 9.8% discharged to SNFs- average LOS 27.5-days, total-FIM-gain 27.2; and 8.4% discharged to acute-care- average LOS 15.4-days and total-FIM-gain 16.5, (F = 176, p < .001). Stepwise-linear-regression for prediction of community discharge showed change in FIM-Bed/chair/wheelchair-Transfers was the most significant predictor (Wald = 42.2; p < .001), followed by CMI (Wald = 26.9; p < .001), change in FIM-walking/wheelchair (Wald = 14.9; p < .001), and age (Wald = 9.5; p < .002). Using discriminant-function-analysis to test model validity for predicting discharge disposition, FIM-change for Bed/chair/wheelchair Transfers, Walking, and Self-Care as predictors resulted in a classification rate of 78.1%, 92% of variance explained, and Eigenvalue of .53 (p < .001). Total-FIM scores improved in all groups, and most patients were discharged to home/community suggesting IR efficacy. The ability to transfer bed/chair/wheelchair was the most important predictive factor associated with discharge destination.

Evaluation of Emergency Department Surge Spaces During the COVID-19 Crisis

Background: In response to the surge of COVID-19 patients in March and April 2020, our emergency department in New York City implemented a novel triage screening process to identify low acuity patients who could be diverted from the emergency department to a surge space. Patients who were (1) &le; 70 years old; (2) triaged as Emergency Severity Index category 4 or 5, and (3) not requiring use of a stretcher (4) displaying mild COVID-19 symptoms were diverted to one of two surge spaces [West Surge Space (WSS) or Ambulatory Clinic Surge Space (ACSS)], where a complete set of vital signs were taken and an exam by a medical provider was conducted. Methods: From March 12 to April 20, 2020, we analyzed the total volume of patients diverted to the surge space and their discharge disposition (i.e., admitted to hospital, transfer to main emergency department, left against medical advice, discharged from a surge space, or expired). A receiver operator curve (ROC) was calculated to predict discharge disposition based on triage diversion to either surge space. Results: There were 6,241 patients triaged during the study period, in which one-quarter [n = 1,753 (26%)] were diverted to a surge space. The average number of patients per day was 141.8 (standard deviation [SD] 54.5) in the main emergency department and 39.8 (SD 29.9) in the surge spaces. Of the patients triaged and diverted to a surge space, a large proportion were successfully discharged from those areas without entering the main emergency department (84%, n = 866 in WSS and 94%, n = 745 in ACSS). Very few signed out against medical advice (3%, n = 26 in WSS and 2%, n = 12 in ACSS), transferred to the main emergency department (1%, n = 8 in WSS and 0.4%, n = 3 in ACSS), or admitted (13%, n = 130 in WSS and 4%, n = 30 in ACSS]. No patients expired in either surge space. The ROC analysis showed that the model had excellent discrimination power, with an area under the curve of 0.97 for WSS and 0.96 for ACSS. Conclusion: A novel triage screening process successfully diverted 26% of low acuity patients with mild symptoms of COVID-19 to surge spaces. Nearly 92% of these patients were successfully discharged without entering the main emergency department footprint.

Predictors of discharge disposition and mortality following hospitalization with SARS-CoV-2 infection.

The SARS-CoV-2 pandemic has overwhelmed hospital capacity, prioritizing the need to understand factors associated with type of discharge disposition. Characterization of disposition associated factors following SARS-CoV-2. Retrospective study of SARS-CoV-2 positive patients from March 7th, 2020, to May 4th, 2022, requiring hospitalization. Midwest academic health-system. Patients above the age 18 years admitted with PCR + SARS-CoV-2. None. Discharge to home versus PAC (inpatient rehabilitation facility (IRF), skilled-nursing facility (SNF), long-term acute care (LTACH)), or died/hospice while hospitalized (DH). We identified 62,279 SARS-CoV-2 PCR+ patients; 6,248 required hospitalizations, of whom 4611(73.8%) were discharged home, 985 (15.8%) to PAC and 652 (10.4%) died in hospital (DH). Patients discharged to PAC had a higher median age (75.7 years, IQR: 65.6-85.1) compared to those discharged home (57.0 years, IQR: 38.2-69.9), and had longer mean length of stay (LOS) 14.7 days, SD: 14.0) compared to discharge home (5.8 days, SD: 5.9). Older age (RRR:1.04, 95% CI:1.041-1.055), and higher Elixhauser comorbidity index [EI] (RRR:1.19, 95% CI:1.168-1.218) were associated with higher rate of discharge to PAC versus home. Older age (RRR:1.069, 95% CI:1.060-1.077) and higher EI (RRR:1.09, 95% CI:1.071-1.126) were associated with more frequent DH versus home. Blacks, Asians, and Hispanics were less likely to be discharged to PAC (RRR, 0.64 CI 0.47-0.88), (RRR 0.48 CI 0.34-0.67) and (RRR 0.586 CI 0.352-0.975). Having alpha variant was associated with less frequent PAC discharge versus home (RRR 0.589 CI 0.444-780). The relative risks for DH were lower with a higher platelet count 0.998 (CI 0.99-0.99) and albumin levels 0.342 (CI 0.26-0.45), and higher with increased CRP (RRR 1.006 CI 1.004-1.007) and D-Dimer (RRR 1.070 CI 1.039-1.101). Increased albumin had lower risk to PAC discharge (RRR 0.630 CI 0.497-0.798. An increase in D-Dimer (RRR1.033 CI 1.002-1.064) and CRP (RRR1.002 CI1.001-1.004) was associated with higher risk of PAC discharge. A breakthrough (BT) infection was associated with lower likelihood of DH and PAC. Older age, higher EI, CRP and D-Dimer are associated with PAC and DH discharges following hospitalization with COVID-19 infection. BT infection reduces the likelihood of being discharged to PAC and DH.

322 Distributed Vector Representations of Neurosurgical Admission Notes Predict Discharge Disposition

INTRODUCTION: Natural language processing (NLP) aims to extract information from unstructured language input for use in machine learning, without the need for laborious manual review of text to retrieve data of potential interest. An NLP-based model utilizing term frequency-inverse document frequency (TF-IDF) vectoral representations of pre-operative clinical notes has previously been shown to predict discharge disposition following meningioma resection. The TF-IDF vectorization scheme is based on word counts across documents but does not account for word order or semantic information. An alternative algorithm, paragraph vectorization, utilizes an unsupervised neural network to learn document vector representations. METHODS: We analyzed 7,122 admission notes. We trained a neural network to derive paragraph vector representations of notes and trained a classifier to predict home vs non-home discharge disposition based on these representations. Out-of-sample performance was assessed on a held-out testing set. We repeated the analysis using TF-IDF document representations to compare the two approaches. Performance was quantified using area under the receiver operating characteristic curve (ROC-AUC). RESULTS: Model performance increased with vector dimensionality. A 200-dimensional distributed document vectorization was highly predictive of nonhome discharge with an ROC-AUC of 0.82 ± 0.01 (mean ± standard error over 5-fold nested cross-validation). When the dimensionality of the embedding vector space was relatively low (e.g. &lt;100), learned paragraph vectors predicted discharge disposition better than TF-IDF vectors. CONCLUSIONS: Natural language processing of admission notes via neural network-learned distributed vectoral representations is a viable strategy of information extraction from neurosurgical clinical documentation. This embedding scheme overcomes some of the limitations of TF-IDF and encodes more information in lower-dimensional embedding vector spaces. It holds promise as a method of using free text to train machine learning algorithms for prediction of clinically relevant outcomes.

Validation and Generalizability of Machine Learning Models for the Prediction of Discharge Disposition Following Revision Total Knee Arthroplasty

BackgroundPostoperative discharge to facilities account for over 33% of the $ 2.7 billion revision total knee arthroplasty (TKA)-associated annual expenditures and are associated with increased complications when compared to home discharges. Prior studies predicting discharge disposition using advanced machine learning (ML) have been limited due to a lack of generalizability and validation. This study aimed to establish ML model generalizability by externally validating its prediction for nonhome discharge following revision TKA using national and institutional databases. MethodsThe national and institutional cohorts comprised 52,533 and 1,628 patients, respectively, with 20.6 and 19.4% nonhome discharge rates. Five ML models were trained and internally validated (five-fold cross-validation) on a large national dataset. Subsequently, external validation was performed on our institutional dataset. Model performance was assessed using discrimination, calibration, and clinical utility. Global predictor importance plots and local surrogate models were used for interpretation. ResultsThe strongest predictors of nonhome discharge were patient age, body mass index, and surgical indication. The area under the receiver operating characteristic curve increased from internal to external validation and ranged between 0.77 and 0.79. Artificial neural network was the best predictive model for identifying patients at risk for nonhome discharge (area under the receiver operating characteristic curve = 0.78), and also the most accurate (calibration slope = 0.93, intercept = 0.02, and Brier score = 0.12). ConclusionAll five ML models demonstrated good-to-excellent discrimination, calibration, and clinical utility on external validation, with artificial neural network being the best model for predicting discharge disposition following revision TKA. Our findings establish the generalizability of ML models developed using data from a national database. The integration of these predictive models into clinical workflow may assist in optimizing discharge planning, bed management, and cost containment associated with revision TKA.

Association between frailty status and complications in patients undergoing surgical excision of malignant esophageal neoplasms.

Research within the last decade highlights the patients' frailty status as an important predictor of esophageal cancer outcomes, but the literature evaluating frailty's role in these patients remains limited. We evaluated the role of frailty in patients undergoing resection of malignant esophageal neoplasms. We used the Nationwide Readmissions Database from 2016 and 2017 to identify patients who underwent excision of a malignant esophageal neoplasm. Patient frailty was queried using the Johns Hopkins Adjusted Clinical Groups frailty-defining diagnosis indicator. Propensity score matching identified 289 frail patients and 281 non-frail patients. Mann-Whitney U testing was performed and receiver operating characteristic (ROC) curves were created, following the creation of logistic regression models for predicting discharge disposition. The area under the curve (AUC) served as a proxy for model performance. Frail patients had significantly more nonroutine discharges, longer inpatient lengths of stay, higher costs, more acute infections, posthemorrhagic anemia and deep vein thrombosis, and greater mortality (P<0.05). No significant differences were found between the 2 cohorts with respect to readmission rates, pulmonary embolism or dysphagia. Predictive models for patient discharge disposition demonstrated that frailty status in combination with age resulted in better ROC curves (AUC: 0.652) compared to models using age alone (AUC: 0.601). Frailty was found to be significantly correlated with higher rates of inpatient medical complications following esophagectomy. The inclusion of patient frailty status in predictive models for discharge disposition resulted in a better predictive capacity compared to those using age alone.

The influence of frailty on perioperative outcomes in patients undergoing surgical resection of liver metastases: a nationwide readmissions database study.

Liver metastases arise frequently from primary colorectal, pancreatic, and breast cancers. Research has highlighted the patient's frailty status as an important predictor of outcomes, but the literature evaluating the role of frailty in patients with secondary metastatic disease of the liver remains limited. Using predictive analytics, we evaluated the role of frailty in patients who underwent hepatectomy for liver metastases. We used the Nationwide Readmissions Database from 2016-2017 to identify patients who underwent resection of a secondary malignant neoplasm of the liver. Patient frailty was evaluated using the Johns Hopkins Adjusted Clinical Groups (JHACG) frailty-defining diagnosis indicator. Propensity score matching was performed and Mann-Whitney U testing was used to analyze complication rates. Receiver operating characteristic (ROC) curves were created following creation of logistic regression models for predicting discharge disposition. Frail patients reported significantly higher rates of nonroutine discharges, longer inpatient stays, greater costs, higher rates of acute infection, posthemorrhagic anemia, urinary tract infection (UTI), deep vein thrombosis (DVT), wound dehiscence and readmission, and greater mortality (P<0.05). Predictive models for patient discharge disposition, DVT and UTI demonstrated that the use of frailty status and age improved the area under the ROC curves significantly compared to models using age alone. Frailty was found to be significantly correlated with higher rates of medical complications during inpatient stay following hepatectomy in patients with liver metastasis. The inclusion of patient frailty status in predictive models improved their predictive capacity compared to those using age alone.

The effect of increased emergency department demand on throughput times and disposition status for pediatric psychiatric patients

ObjectivesEmergency department (ED) crowding has been shown to increase throughput measures of length of stay (LOS), wait time, and boarding time. Psychiatric utilization of the ED has increased, particularly among younger patients. This investigation quantifies the effect of ED demand on throughput times and discharge disposition for pediatric psychiatric patients in the ED. MethodsUsing electronic medical record data from 1,151,396 ED visits in eight North Carolina EDs from January 1, 2018, through December 31, 2020, we identified 14,092 pediatric psychiatric visits. Measures of ED daily demand rates included overall occupancy as well as daily proportion of non-psychiatric pediatric patients, adult psychiatric patients, and pediatric psychiatric patients. Controlling for patient-level factors such as age, sex, race, insurance, and triage acuity, we used linear regression to predict throughput times and logistic regression to predict disposition status. We estimated effects of ED demand by academic versus community hospital status due to ED and inpatient resource differences. ResultsMost ED demand measures had insignificant or only very small associations with throughput measures for pediatric psychiatric patients. Notable exceptions were that a one percentage point increase in the proportion of non-psychiatric pediatric ED visits increased boarding times at community sites by 1.06 hours (95% CI: 0.20–1.92), while a one percentage point increase in the proportion of pediatric psychiatric ED visits increased LOS by 3.64 hours (95% CI: 2.04–5.23) at the academic site. We found that ED demand had a minimal effect on disposition status, with small increases in demand rates favoring <1 percentage point increases in the likelihood of discharge. Instead, patient-level factors played a much stronger role in predicting discharge disposition. ConclusionsED demand has a meaningful effect on throughput times, but a minimal effect on disposition status. Further research is needed to validate these findings across other state and healthcare systems.

Simple and actionable preoperative prediction of postoperative healthcare needs of single-level lumbar fusion patients.

Preoperative prediction of a patient's postoperative healthcare utilization is challenging, and limited guidance currently exists. The objective of the present study was to assess the capability of individual risk-related patient characteristics, which are available preoperatively, that may predict discharge disposition prior to lumbar fusion. In total, 1066 consecutive patients who underwent single-level, posterior-only lumbar fusion at a university health system were enrolled. Patients were prospectively asked 4 nondemographic questions from the Risk Assessment and Prediction Tool during preoperative office visits to evaluate key risk-related characteristics: baseline walking ability, use of a gait assistive device, reliance on community supports (e.g., Meals on Wheels), and availability of a postoperative home caretaker. The primary outcome was discharge disposition (home vs skilled nursing facility/acute rehabilitation). Logistic regression was performed to analyze the ability of each risk-related characteristic to predict likelihood of home discharge. Regression analysis demonstrated that improved baseline walking ability (OR 3.17), ambulation without a gait assistive device (OR 3.13), and availability of a postoperative home caretaker (OR 1.99) each significantly predicted an increased likelihood of home discharge (all p < 0.0001). However, reliance on community supports did not significantly predict discharge disposition (p = 0.94). Patient mobility and the availability of a postoperative caretaker, when determined preoperatively, strongly predict a patient's healthcare utilization in the setting of single-level, posterior lumbar fusion. These findings may help surgeons to streamline preoperative clinic workflow and support the patients at highest risk in a targeted fashion.

Activity Measure for Post-acute Care Mobility Scoring System: Comparison of Nursing and Physical Therapy Evaluation for Primary Hip and Knee Arthroplasty Patients.

Optimizing resource utilization after total joint arthroplasty (TJA) has become increasingly vital. The Activity Measure for Post-acute Care (AM-PAC) "6-clicks" scoring system is a validated, physical therapist (PT)-administered metric of patient basic mobility and predicts discharge disposition. This study aimed to determine whether the use of AM-PAC scoring by nurses in the postoperative period could (1) substitute for AM-PAC scoring by therapists and (2) predict 90-day outcomes in TJA patients. We retrospectively reviewed all primary TJAs conducted by two surgeons at a single institution from 2019 to 2021. Patients underwent postoperative AM-PAC evaluation by nursing and physical therapy within 24 hours of surgery, and specific timing of nursing and PT scores was determined. Inter-rater reliability between therapy and nursing scores was analyzed. Multiple regression was used to determine the association between AM-PAC scores and readmissions, complications, length of stay, and nonhome discharge. In total, 1,119 patients were included. Agreement testing between therapy and nursing scores was weak for all six AM-PAC components, with a Spearman correlation of 0.437. Nursing scores were typically conducted earlier than therapist scores (204.0 ± 249.9 minutes versus 523.5 ± 449.4 minutes; P < 0.001). Therapy and nursing scores were not notable predictors for 90-day complications or readmissions. However, higher therapy and nursing scores were predictors of less than 2-day hospitalization (odds ratio [OR] 0.63, P < 0.001; OR 0.88, P < 0.001) and fewer nonhome discharges (OR 0.62, P < 0.001; OR 0.84, P < 0.001). Although nursing-driven mobility assessments could potentially improve efficiency of patient discharge and control costs, nursing AM-PAC scoring did not serve as an appropriate substitute for PT scoring in patients undergoing primary total hip and knee arthroplasty at our institution.