A retrospective look at length of stay for pediatric psychiatric patients in an urban emergency department

Machine learning-enabled prediction of prolonged length of stay in hospital after surgery for tuberculosis spondylitis patients with unbalanced data: a novel approach using explainable artificial intelligence (XAI)

BACKGROUND

Tuberculosis spondylitis (TS), commonly known as Pott's disease, is a severe type of skeletal tuberculosis that typically requires surgical treatment. However, this treatment option has led to an increase in healthcare costs due to prolonged hospital stays (PLOS). Therefore, identifying risk factors associated with extended PLOS is necessary. In this research, we intended to develop an interpretable machine learning model that could predict extended PLOS, which can provide valuable insights for treatments and a web-based application was implemented.

METHODS

We obtained patient data from the spine surgery department at our hospital. Extended postoperative length of stay (PLOS) refers to a hospitalization duration equal to or exceeding the 75th percentile following spine surgery. To identify relevant variables, we employed several approaches, such as the least absolute shrinkage and selection operator (LASSO), recursive feature elimination (RFE) based on support vector machine classification (SVC), correlation analysis, and permutation importance value. Several models using implemented and some of them are ensembled using soft voting techniques. Models were constructed using grid search with nested cross-validation. The performance of each algorithm was assessed through various metrics, including the AUC value (area under the curve of receiver operating characteristics) and the Brier Score. Model interpretation involved utilizing methods such as Shapley additive explanations (SHAP), the Gini Impurity Index, permutation importance, and local interpretable model-agnostic explanations (LIME). Furthermore, to facilitate the practical application of the model, a web-based interface was developed and deployed.

RESULTS

The study included a cohort of 580 patients and 11 features include (CRP, transfusions, infusion volume, blood loss, X-ray bone bridge, X-ray osteophyte, CT-vertebral destruction, CT-paravertebral abscess, MRI-paravertebral abscess, MRI-epidural abscess, postoperative drainage) were selected. Most of the classifiers showed better performance, where the XGBoost model has a higher AUC value (0.86) and lower Brier Score (0.126). The XGBoost model was chosen as the optimal model. The results obtained from the calibration and decision curve analysis (DCA) plots demonstrate that XGBoost has achieved promising performance. After conducting tenfold cross-validation, the XGBoost model demonstrated a mean AUC of 0.85 ± 0.09. SHAP and LIME were used to display the variables' contributions to the predicted value. The stacked bar plots indicated that infusion volume was the primary contributor, as determined by Gini, permutation importance (PFI), and the LIME algorithm.

CONCLUSIONS

Our methods not only effectively predicted extended PLOS but also identified risk factors that can be utilized for future treatments. The XGBoost model developed in this study is easily accessible through the deployed web application and can aid in clinical research.

Factors Associated With Length of Stay in Emergency Departments for Pediatric Patients With Psychiatric Problems

OBJECTIVES

Length of stay (LOS) and boarding for pediatric psychiatric patients presenting in the emergency department (ED) have been understudied, despite evidence that children with psychiatric disorders experience longer LOS relative to those without. This investigation examined correlates of LOS and boarding among youth with psychiatric disorders presenting to the ED in a large, statewide database.

METHODS

Using the 2010 to 2013 Florida ED discharge database, generalized linear mixed models were used to examine for associations between LOS and patient and hospital characteristics among pediatric patients (<18 years) who presented with a primary psychiatric diagnosis (N = 44,328).

RESULTS

Patients had an overall mean ± SD ED LOS of 5.96 ± 8.64 hours. Depending on the definition used (ie, 12 or 6 hours), between 23% and 58% of transferred patients were boarded. Patient characteristics associated with a longer LOS included female sex, being 15 to 17 years old, Hispanic ethnicity, having Medicaid or VA/TriCare insurance, having impulse control problems, having mood or psychotic disorders, and exhibiting self-harm behaviors. Patient transfer, large hospital size, and rural designation were associated with longer LOS. Teaching hospital status and profit status were not significantly associated with LOS.

CONCLUSIONS

These data suggest that LOS for pediatric psychiatry patients in the ED varies significantly by psychiatric presentation, patient disposition, and hospital factors. Such findings have implications for quality of care, patient safety, and health outcomes.

A Collaborative Behavioral Model for Psychiatric Visits in a Pediatric Emergency Department

OBJECTIVES

Psychiatric pediatric emergency department (PED) visits are associated with extended length of stay (LOS) compared with nonpsychiatric visits. To address delays in definitive care, our PED incorporated a Collaborative Behavioral Model for patients with a psychiatric complaint. We evaluated LOS and 6-month readmission by comparing patients who received a psychiatric consultation through the model with a control group of nonpsychiatric patients. In addition, we assessed the utilization of laboratory tests and imaging and the impact of testing on final patient disposition.

METHODS

A matched-pair retrospective chart review of psychiatric PED visits was conducted for visits from January 1, 2009, to December 31, 2013. Patients who received a psychiatric diagnosis and consultation through the Collaborative Behavioral Model were included. Participants were matched with a control group of similar patients who did not receive a psychiatric diagnosis.

RESULTS

One hundred forty-seven patients met all the inclusion criteria. There was no statistical difference in median LOS for patients who received a psychiatric consultation versus nonpsychiatric visits (204 vs 200 minutes; P = 0.35). The 6-month PED readmission rate for psychiatric visits was significantly less than that in the control group (18% vs 32%; P = 0.003). Of the laboratory and imaging studies performed, 1 abnormal laboratory test (1%) and no imaging studies resulted in a change in patient management.

CONCLUSIONS

Collaborative models, with specialized psychiatric components, can better coordinate the care of mental health visits by providing focused interventions that promote efficient disposition. In addition, laboratory and imaging studies resulted in few changes in management or final disposition.

Factors associated with boarding and length of stay for pediatric mental health emergency visits

OBJECTIVE

To determine demographic and clinical risk factors associated with boarding (length of stay ≥24 h) for pediatric mental health emergency department (ED) visits.

METHODS

This is a retrospective cross-sectional analysis of mental health visits identified by diagnosis codes for children 5-18 years old presenting to a tertiary pediatric ED in 2016. We performed multivariate logistic regression to identify demographic and clinical factors associated with boarding.

RESULTS

There were 1746 mental health visits and 386 (22%) visits had length of stay ≥24 h. In the multivariate logistic regression model, factors associated with boarding included: private insurance (OR 1.59, 95% CI 1.15, 2.19) and having both private and public insurance (OR 1.68, 95% CI 1.16, 2.43) relative to public insurance; presentation during a school month (OR 2.17, 95% CI 1.30, 3.63); physical or chemical restraint use (OR 4.80, 95% CI 2.61, 8.84); comorbid autism or developmental delay (OR 1.82, 95% CI 1.35, 2.46); prior psychiatric hospitalization (OR 2.55, 95% CI 1.93, 3.36); and reasons for presentation of agitation, aggression, or homicidal ideation (OR 2.76, 95% CI 1.40, 5.45), depression, self-injury, or suicidal ideation (OR 2.79, 95% CI 1.45, 5.40), and bipolar, mania, or psychosis (OR 5.78, 95% CI 2.36, 14.09) relative to anxiety.

CONCLUSIONS

Insurance status, presentation month, restraint use, autism or developmental delay comorbidity, prior psychiatric hospitalization, and reason for presentation are associated with pediatric mental health ED boarding. Resources should be directed to improve the mental health care system for children with identified risk factors for boarding.

The Impact of Behavioral Health Patients on a Pediatric Emergency Department's Length of Stay and Left Without Being Seen

OBJECTIVE

Emergency department (ED) boarding of admitted patients negatively impacts ED length of stay (LOS). Behavioral health (BH) patients are often challenging to safely discharge. We examined the association between daily BH census and non-BH LOS and left without being seen (LWBS) rates.

METHODS

Retrospective analysis of BH and non-BH patients at a high-volume tertiary care pediatric ED from December 2014 to June 2016 examined the association between BH patients and non-BH LOS and LWBS rates. Behavioral health patients were identified by presence of social work assessment and BH chief complaint and/or final diagnosis. Data were analyzed using 1-sample test of proportions, Student t test, Spearman and Pearson correlations, logistic regression, and odds ratios with 95% confidence intervals.

RESULTS

A total of 143,141 patients were seen, 3% (n = 4351) for BH presentations. Median LOS for discharged non-BH patients was 128 minutes compared with 446 minutes for BH patients. Daily LOS and bed hold hours were significantly longer for BH than for non-BH patients (P < 0.0001 for each analysis). After adjusting for ED census, daily BH census was significantly associated with increasing LWBS rates and non-BH LOS.

CONCLUSIONS

Behavioral health census and bed hold hours were significantly associated with increased LOS and LWBS rates and with our inability to meet desired LOS and LWBS rates. These associations support the existence of a threshold where the ED has reached capacity and is no longer able to absorb BH patients. Improving BH facility access may help improve overall pediatric ED patient care.

Determinants of Pediatric Psychiatry Length of Stay in 2 Urban Emergency Departments

OBJECTIVES

Pediatric mental illness poses a significant burden with an overall prevalence of approximately 10%. Increasingly, children with mental disorders seek care in the emergency department (ED). However, the ED is not an ideal setting. Pediatric mental health patients receive limited treatment and experience significantly longer length of stay (LOS) than other patients seen in the ED. This study examines patient and hospital factors associated with LOS and prolonged LOS (PLOS).

METHODS

This is a retrospective chart review of patients between the ages of 3 and 17 presenting at 2 participating urban EDs with a psychiatric diagnosis from May 2010 to May 2012.

RESULTS

This study includes 939 patients with an average age of 14.1 years and a median LOS of 295 minutes. The diagnosis was the strongest predictor of LOS and PLOS. Patients with a psychotic disorder or suicide attempt or ideation experienced a longer LOS, 35% and 55% increases, respectively, and an increased odds of PLOS (odds ratio, 3.07 and 8.36, respectively). Patient sex, previous history of self-harm, and the daily census were associated with both a longer LOS and PLOS. Ethnicity, site of admission, and year of admission were only associated with LOS.

CONCLUSIONS

Diagnosis-specific management factors are the primary determinant of LOS. However, some patient characteristics and hospital operational factors are also associated with LOS. Organizational reforms and an evaluation of the required human and material resources are necessary to improve access to and availability of pediatric mental health care.

The Impact of Mental Health Services in a Pediatric Emergency Department: The Implications of Having Trained Psychiatric Professionals

OBJECTIVES

This study assessed improvement in the emergency department (ED) length of stay and costs after implementation of an ED program which added board-certified psychiatrists and trained psychiatric social workers to the pediatric ED.

METHODS

A retrospective medical record and administrative data review were conducted for all pediatric psychiatric visits of children aged 5 to 18 years who were seen and discharged from the Greenville Memorial Hospital ED between January 1, 2007, and June 31, 2013. These subjects were diagnosed by the ED physician at the time of the visit using codes ranging from 290.0 to 319.0 based on the International Statistical Classification of Diseases and Related Health Problems, Ninth Revision codes.

RESULTS

The mean (SD) age of children in the postprogram period (14.3 ± 3.1) was younger than during the preprogram period (14.9 ± 3.1) (P < 0.001) with the greatest increase in the 11- to 15-year age group (42% vs 35%, respectively). Patients in the postprogram period were significantly more likely to be discharged to a psychiatric hospital than during the pre-program period (18% vs 9%, respectively). After the initiation of the program, ED length of stay decreased significantly from 14.7 to 12.1 hours (P < 0.001) and costs per visit decreased slightly from US $602 to US $588 (this difference was not statistically significant).

CONCLUSIONS

Although this model of care has significant costs associated with it, the efficiency of care for psychiatric pediatric patients in the ED improved after targeted training of ED staff and provision of these specialized services within the ED.

Mental Health Utilization in a Pediatric Emergency Department

OBJECTIVE

Mental health complaints are frequent in the pediatric emergency department (PED). The objective of this study was to describe trends over time in PED utilization for mental health care at in a single pediatric tertiary care hospital. It is our hypothesis that the resources used by this patient population are high and that mental health-related visits have increased over the most recent decade.

METHODS

This was a retrospective study of all pediatric mental health presentations to the PED from January 2009 to July 2013 at a single pediatric hospital. All patients aged 1 to 19 years with an International Classification of Diseases, Ninth Revision code of 291, 292, 295 to 309, and 311 to 314 were included. Data collected included demographic data, medications received, restraint use, suicidality, length of stay (LOS), charges incurred, final disposition, and daily PED operation variables. Trends over time in presentation, charges, and LOS were analyzed using multiple mixed effects regression models after adjusting for potential patient and PED level confounding variables and clustering of multiple visits within patients.

RESULTS

A total of 732 PED visits from 2009 to 2013 were identified representing 646 unique patients. The average age was 13.8 years, and 53% were male. Approximately 25% of the patients expressed suicidal ideation, and 44% of those had attempted suicide before arrival. Behavioral or chemical restraints were used in 33% of patients during their PED visit. There were statistically significant increases in annual visits, LOS, and charges over this period (P < 0.05). Increased charges were significantly associated with longer LOS (P = 0.0062). Charges (P = 0.46) and LOS (P = 0.62) were not significantly different between suicidal and nonsuicidal patients. Approximately 21% of patients were admitted or transferred to another facility.

CONCLUSIONS

In this single-center study, we found evidence that the resources required to care for pediatric patients with mental health complaints have increased significantly over time both by increased number of annual visits and an increasing LOS. Further research is necessary to determine if our data are consistent with national trends to further our understanding of the problem and improve resource allocation.

CARES: improving the care and disposition of psychiatric patients in the pediatric emergency department

OBJECTIVES

Pediatric psychiatric emergencies are a nationwide crisis and have contributed to an increase in behavioral health emergency department (ED) visits. A collaborative response to this crisis was the creation of the Child & Adolescent Rapid Emergency Stabilization (CARES) program. The objective of this study is to determine how the CARES unit influenced length of stay (LOS) and costs for psychiatric patients in the pediatric ED.

METHODS

A retrospective review of ED patients was conducted. Children presenting 1 year before CARES, October 13, 2006, to October 13, 2007 (pre-CARES), were compared with 1 year after, October 15, 2007, to October 15, 2008 (post-CARES). The study population includes all patients presenting to the ED with psychiatric illnesses. The primary outcome is mean LOS in hours. Supplemental analyses of total charges, payments, and costs were performed.

RESULTS

There were 1719 visits before and 1867 visits after CARES, with 1190 and 1273 unique patient visits, respectively. Children in both groups had similar age, gender, and ethnicity. Pre-CARES had a mean LOS of 19.7 hours (SD, 32.6), whereas post-CARES had 10.8 hours (SD, 19.9) (P < 0.0001). Evaluating only unique visits, the difference remained highly significant. Post-CARES, compared with pre-CARES, the average charge per patient decreased by $905 (P < 0.0001), average payment decreased by $111 (P < 0.06), and average total cost decreased by $569 or 38.7% (P < 0.0001). The total cost savings the year after CARES opened was $1,019,168.55.

Children's Mental Health Visits to the Emergency Department: Factors Affecting Wait Times and Length of Stay

Objective. This study explores the association of patient and emergency department (ED) mental health visit characteristics with wait time and length of stay (LOS). Methods. We examined data from 580 ED mental health visits made to two urban EDs by children aged ≤18 years from April 1, 2004, to March 31, 2006. Logistic regressions identified characteristics associated with wait time and LOS using hazard ratios (HR) with 95% confidence intervals (CIs). Results. Sex (male: HR = 1.48, 95% CI = 1.20-1.84), ED type (pediatric ED: HR = 5.91, 95% CI = 4.16-8.39), and triage level (Canadian Triage and Acuity Scale (CTAS) 2: HR = 3.62, 95% CI = 2.24-5.85) were statistically significant predictors of wait time. ED type (pediatric ED: HR = 1.71, 95% CI = 1.18-2.46), triage level (CTAS 5: HR = 2.00, 95% CI = 1.15-3.48), number of consultations (HR = 0.46, 95% CI = 0.31-0.69), and number of laboratory investigations (HR = 0.75, 95% CI = 0.66-0.85) predicted LOS. Conclusions. Based on our results, quality improvement initiatives to reduce ED waits and LOS for pediatric mental health visits may consider monitoring triage processes and the availability, access, and/or time to receipt of specialty consultations.