Length Of Stay Estimates Research Articles

Predicting the Length of Stay of Cardiac Patients Based on Pre-Operative Variables-Bayesian Models vs. Machine Learning Models.

Length of stay (LoS) prediction is deemed important for a medical institution's operational and logistical efficiency. Sound estimates of a patient's stay increase clinical preparedness and reduce aberrations. Various statistical methods and techniques are used to quantify and predict the LoS of a patient based on pre-operative clinical features. This study evaluates and compares the results of Bayesian (simple Bayesian regression and hierarchical Bayesian regression) models and machine learning (ML) regression models against multiple evaluation metrics for the problem of LoS prediction of cardiac patients admitted to Tabba Heart Institute, Karachi, Pakistan (THI) between 2015 and 2020. In addition, the study also presents the use of hierarchical Bayesian regression to account for data variability and skewness without homogenizing the data (by removing outliers). LoS estimates from the hierarchical Bayesian regression model resulted in a root mean squared error (RMSE) and mean absolute error (MAE) of 1.49 and 1.16, respectively. Simple Bayesian regression (without hierarchy) achieved an RMSE and MAE of 3.36 and 2.05, respectively. The average RMSE and MAE of ML models remained at 3.36 and 1.98, respectively.

Abstract P362: Length Of Stay Discrepancies Between Rural And Urban Hospitals - A Cross Sectional Analysis Of Cardiovascular Disease Health Inequity

Introduction: Rural communities have historically been shown to have greater obstacles to healthcare compared to their urban counterparts. These barriers manifest through multiple factors such as a reduced access to healthcare and public health services, lower availability of specialty care, and a larger number of uninsured. Compounded with a greater poverty rate, these disparities in minority populations have been shown to parallel their worse health outcomes and higher death rates. Our study aims to highlight cardiovascular disease health inequity between rural and urban populations through a cross-sectional analysis of the Length of Stay (LOS) with respect to cardiovascular disease admissions, specifically for unstable angina (UA), NSTEMI, and STEMI. Methods: This is a cross-sectional review for the LOS estimates in patients with cardiovascular disease in the state of New York, stratified by the type of community hospital. This review utilizes the 2020 National Inpatient Sample (NIS) for the state of New York. Cardiovascular disease was categorized using ICD-10 codes for STEMI (I21), NSTEMI (I21.4), and Unstable Angina (I20.0). Community hospital setting was categorized as either rural or urban, determined by the zip code of the primary inpatient hospital. Results: LOS was unique across both classification of cardiovascular disease ( p <0.001) and community hospital setting ( p <0.001). Overall, the LOS was less in urban-level community hospitals compared to their rural counterparts, 3.5 days vs 5.1 days ( p <0.001), as shown in Figure 1. Stratified by the cardiovascular disease classifications, admissions in rural hospitals had an overall increased LOS, 0.8 days for UA ( p <0.001), 1.4 days for NSTEMI ( p <0.001), and 2.7 days for STEMI ( p <0.001). Conclusion: The LOS of patients admitted for ACS was significantly higher across the board for rural hospitals. Furthermore, there was a stepwise increase in the comparative LOS between these two communities as the severity of their cardiovascular disease increased. Factors such as socioeconomic status, general access to healthcare and preventative services, insurance status, and availability of specialty care services are just some of the factors that contribute to rural health disparities.

Abstract P226: Length Of Stay Discrepancies Between Rural And Urban Hospital Settings For Hypertension Admissions

Introduction: Rural populations experience greater barriers to healthcare access compared to urban populations. This can be attributed to fewer public health services, greater uninsured rate, and greater burden of healthcare shortages in urban healthcare settings. Our study aims to quantify the disproportionate allocation in healthcare resources through analysis of Length of Stay (LOS) in-respect to hypertensive urgency and hypertensive urgency. Methods: This cross-sectional study reviewed the LOS estimates for patients with ICD-10 diagnoses for Hypertension Emergency or Hypertension Urgency within the state of New York, categorized by the type of community hospital; rural or urban. This study utilizes the 2020 National Inpatient Sample (NIS), specifically for urban and rural hospitals in New York. ICD-10 codes included I16.1 & I16.9 for Hypertensive Emergency and I16.0 for Hypertensive Urgency. Community hospital setting was categorized as either rural or urban based on the overlying zip code of the admitting inpatient hospital. Results: LOS was unique to the community setting (p<0.01), but not to hypertensive disease (p=0.11). In summary, the LOS was longer in rural-level inpatient hospital systems when compared to their urban counterparts in both Hypertensive Urgency, 4.4 days vs 3.1 days (p=0.01), and Hypertensive Emergency, 5.5 days vs 3.4 days (p<0.01). Conclusion: The LOS of patients admitted for both hypertensive urgency and emergency were significantly higher in rural hospitals. Furthermore, increased disease severity emphasizes the divide in LOS between the community healthcare settings. Redistribution of resources to rural communities can enhance access to preventative healthcare services, specialty healthcare, and potentially decrease overall community disease burden.

Access, socioeconomic environment, and death from COVID-19 in Nebraska.

Our study assesses whether factors related to healthcare access in the first year of the pandemic affect mortality and length of stay (LOS). Our cohort study examined hospitalized patients at Nebraska Medicine between April and October 2020 who were tested for SARS-CoV-2 and had a charted sepsis related diagnostic code. Multivariate logistic was used to analyze the odds of mortality and linear regression was used to calculate the parameter estimates of LOS associated with COVID-19 status, age, gender, race/ethnicity, median household income, admission month, and residential distance from definitive care. Among 475 admissions, the odds of mortality is greater among those with older age (OR: 1.04, 95% CI: 1.02–1.07) and residence in an area with low median household income (OR: 2.11, 95% CI: 0.52–8.57), however, the relationship between mortality and wealth was not statistically significant. Those with non-COVID-19 sepsis had longer LOS (Parameter Estimate: −5.11, adjusted 95% CI: −7.92 to −2.30). Distance from definitive care had trends toward worse outcomes (Parameter Estimate: 0.164, adjusted 95% CI: −1.39 to 1.97). Physical and social aspects of access to care are linked to poorer COVID-19 outcomes. Non-COVID-19 healthcare outcomes may be negatively impacted in the pandemic. Strategies to advance patient-centered outcomes in vulnerable populations should account for varied aspects (socioeconomic, residential setting, rural populations, racial, and ethnic factors). Indirect impacts of the pandemic on non-COVID-19 health outcomes require further study.

MedCDI: A technology-based documentation and quality improvement initiative in neurosurgery.

Background: Clinical documentation of patient care alters coding accuracy of Medicare Severity Diagnosis-Related Groups (MS-DRGs), expected mortality, and expected length of stay (LOS) which impact quality metrics. We aimed to determine if neurosurgical quality metrics could be improved by facilitating accurate documentation and subsequently developed a mobile application and educational video to target areas of opportunity.Methods: Vizient software was used to analyze MS-DRGs and expected LOS for sample of patients requiring surgery for spinal pathology, brain tumors, and subarachnoid hemorrhage (SAH) between January 2019 and August 2021. Chart reviews were conducted to discover variables missed by documenting provider and/or coder.Results: Review of 114 spinal surgeries, 20 brain tumors, and 53 SAH patients revealed at least one additional variable impacting LOS in 43% of spine, 75% of brain tumor, and 92% of SAH patients, with an average of 1 (1.25), 2 (1.75), and 3 (2.89) new variables, respectively. Recalculated expected LOS increased by an average of 0.86 days for spine, 3.08 for brain tumor, and 6.46 for SAH cases.Conclusion: Efforts to accurately document patient care can improve quality metrics such as expected LOS, mortality, and cost estimates. We determined several missing variables which impact quality metrics, showing opportunity exists in neurosurgical documentation improvement. Subsequently, we developed an educational video and mobile-supported application to specifically target these variables. To the best of our knowledge, this represents the first initiative to utilize the proven powers of mobile phones in health care toward the novel application of specifically improving neurosurgical quality metrics.

Variations in length of stay among survived very preterm infants admitted to Chinese neonatal intensive care units

BackgroundThis study aimed to describe length of stay (LOS) to discharge and site variations among very preterm infants (VPIs) admitted to 57 Chinese neonatal intensive care units (NICUs) and to investigate factors associated with LOS for VPIs.MethodsThis retrospective multicenter cohort study enrolled all infants < 32 weeks’ gestation and admitted to 57 NICUs which had participated in the Chinese Neonatal Network, within 7 days after birth in 2019. Exclusion criteria included major congenital anomalies, NICU deaths, discharge against medical advice, transfer to non-participating hospitals, and missing discharge date. Two multivariable linear models were used to estimate the association of infant characteristics and LOS.ResultsA total of 6580 infants were included in our study. The overall median LOS was 46 days [interquartile range (IQR): 35–60], and the median corrected gestational age at discharge was 36 weeks (IQR: 35–38). LOS and corrected gestational age at discharge increased with decreasing gestational age. The median corrected gestational age at discharge for infants at 24 weeks, 25 weeks, 26 weeks, 27–28 weeks, and 29–31 weeks were 41 weeks, 39 weeks, 38 weeks, 37 weeks and 36 weeks, respectively. Significant site variation of LOS was identified with observed median LOS from 33 to 71 days in different hospitals.ConclusionsThe study provided concurrent estimates of LOS for VPIs which survived in Chinese NICUs that could be used as references for medical staff and parents. Large variation of LOS independent of infant characteristics existed, indicating variation of care practices requiring further investigation and quality improvement.

Five-Year Outcomes After Implementing a Pain, Agitation, and Delirium Protocol in a Mixed Intensive Care Unit.

Objective: Assess for continued improvements in patient outcomes after updating our institutional sedation and analgesia protocol to include recommendations from the 2013 Society of Critical Care Medicine (SCCM) Pain, Agitation, and Delirium (PAD) guidelines. Methods: Retrospective before-and-after study in a mixed medical/surgical intensive care unit (ICU) at an academic medical center. Mechanically ventilated adults admitted from September 1, 2011 through August 31, 2012 (pre-implementation) and October 1, 2012 through September 30, 2017 (post-implementation) were included. Measurements included number of mechanically ventilated patients, APACHE IV scores, age, type of patient (medical or surgical), admission diagnosis, ICU length of stay (LOS), hospital LOS, ventilator days, number of self-extubations, ICU mortality, ICU standardized mortality ratio, hospital mortality, hospital standardized mortality ratio, medication data including as needed (PRN) analgesic and sedative use, and analgesic and sedative infusions, and institutional savings. Results: Ventilator days (Pre-PAD = 4.0 vs. Year 5 post = 3.2, P < .0001), ICU LOS (Pre-PAD = 4.8 days vs. Year 5 post = 4.1 days, P = .0004) and hospital LOS (Pre-PAD = 14 days vs. Year 5 post = 12 days, P < .0001) decreased after protocol implementation. Hospital standardized mortality ratio (Pre-PAD = 0.69 vs. Year 5 post = 0.66) remained constant; while, APACHE IV scores (Pre-PAD = 77 vs. Year 5 post = 89, P < .0001) and number of intubated patients (Pre-PAD = 1146 vs. Year 5 post = 1468) increased over the study period. Using the decreased ICU and hospital LOS estimates, it is projected the institution saved $4.3 million over the 5 years since implementation. Conclusions: Implementation of an updated PAD protocol in a mixed medical/surgical ICU was associated with a significant decrease in ventilator time, ICU LOS, and hospital LOS without a change in the standardized mortality ratio over a five-year period. These favorable outcomes are associated with a significant cost savings for the institution.

Machine learning in the prediction of medical inpatient length of stay.

Length of stay (LOS) estimates are important for patients, doctors and hospital administrators. However, making accurate estimates of LOS can be difficult for medical patients. This review was conducted with the aim of identifying and assessing previous studies on the application of machine learning to the prediction of total hospital inpatient LOS for medical patients. A review of machine learning in the prediction of total hospital LOS for medical inpatients was conducted using the databases PubMed, EMBASE and Web of Science. Of the 673 publications returned by the initial search, 21 articles met inclusion criteria. Of these articles the most commonly represented medical specialty was cardiology. Studies were also identified that had specifically evaluated machine learning LOS prediction in patients with diabetes and tuberculosis. The performance of the machine learning models in the identified studies varied significantly depending on factors including differing input datasets and different LOS thresholds and outcome metrics. Common methodological shortcomings included a lack of reporting of patient demographics and lack of reporting of clinical details of included patients. The variable performance reported by the studies identified in this review supports the need for further research of the utility of machine learning in the prediction of total inpatient LOS in medical patients. Future studies should follow and report a more standardised methodology to better assess performance and to allow replication and validation. In particular, prospective validation studies and studies assessing the clinical impact of such machine learning models would be beneficial.

The Analgesic Properties of a Music Intervention in the Post Anesthesia Care Unit

<h3>Introduction</h3> Twenty to 85% of surgical patients develop persistent postsurgical pain (PPP) following surgery. Listening to music has been shown to decrease pain scores and increase patient satisfaction in postoperative adult patients. <h3>Identification of the problem</h3> Under management of the severity and duration of acute postoperative pain in the post anesthesia care unit (PACU) is a significant risk factor for PPP and causes delays in discharge from the PACU. Evidence of the analgesic effects of music interventions suggest that listening to music has a positive effect on decreasing pain scores and increasing patient satisfaction in postoperative adult patients<i>.</i> <h3>The purpose of this study was to</h3> 1) determine whether a combined music pharmacological intervention was an effective multimodal approach to mitigate adult pain for patients undergoing joint replacement surgery in the PACU; and 2) measure the effect of music intervention on reducing PACU length of stay (LOS). <h3>Methodology</h3> A prospective, randomized control trial study design is proposed to determine whether a combined music pharmacologic intervention decreases discharge pain scores and decreases pain level delays to discharge from the PACU compared to pharmacological interventions alone. <h3>Results</h3> The study was completed December 2020. Dissemination of findings will occur by the end of 2021 months by submitting results to the Journal of Perianesthesia Nursing (JOPAN) for publication and presentation of findings at the national conference. <h3>Discussion</h3> Multiple lines of evidence demonstrate that music interventions are effective in decreasing a patient's perception of pain. <h3>Conclusion</h3> The lack of improvement in pain scores in the intervention group may be due, in part, to the low average pain scores participants in the study reported both upon admission and at discharge from the PACU. <h3>Implications for perianesthesia nurses and future research</h3> Further opportunities exist to test the effects of music intervention on patient and clinical outcomes immediately after surgery. Our results provide estimates for pain and PACU LOS to determine sample sizes for future randomized controlled trials. Future studies should also measure the benefits of music interventions with different surgical and anesthetic populations.

Hospital length of stay for COVID-19 patients: Data-driven methods for forward planning

BackgroundPredicting hospital length of stay (LoS) for patients with COVID-19 infection is essential to ensure that adequate bed capacity can be provided without unnecessarily restricting care for patients with other conditions. Here, we demonstrate the utility of three complementary methods for predicting LoS using UK national- and hospital-level data.MethodOn a national scale, relevant patients were identified from the COVID-19 Hospitalisation in England Surveillance System (CHESS) reports. An Accelerated Failure Time (AFT) survival model and a truncation corrected method (TC), both with underlying Weibull distributions, were fitted to the data to estimate LoS from hospital admission date to an outcome (death or discharge) and from hospital admission date to Intensive Care Unit (ICU) admission date. In a second approach we fit a multi-state (MS) survival model to data directly from the Manchester University NHS Foundation Trust (MFT). We develop a planning tool that uses LoS estimates from these models to predict bed occupancy.ResultsAll methods produced similar overall estimates of LoS for overall hospital stay, given a patient is not admitted to ICU (8.4, 9.1 and 8.0 days for AFT, TC and MS, respectively). Estimates differ more significantly between the local and national level when considering ICU. National estimates for ICU LoS from AFT and TC were 12.4 and 13.4 days, whereas in local data the MS method produced estimates of 18.9 days.ConclusionsGiven the complexity and partiality of different data sources and the rapidly evolving nature of the COVID-19 pandemic, it is most appropriate to use multiple analysis methods on multiple datasets. The AFT method accounts for censored cases, but does not allow for simultaneous consideration of different outcomes. The TC method does not include censored cases, instead correcting for truncation in the data, but does consider these different outcomes. The MS method can model complex pathways to different outcomes whilst accounting for censoring, but cannot handle non-random case missingness. Overall, we conclude that data-driven modelling approaches of LoS using these methods is useful in epidemic planning and management, and should be considered for widespread adoption throughout healthcare systems internationally where similar data resources exist.

Mixed-data deep learning in repeated predictions of general medicine length of stay: a derivation study.

The accurate prediction of likely discharges and estimates of length of stay (LOS) aid in effective hospital administration and help to prevent access block. Machine learning (ML) may be able to help with these tasks. For consecutive patients admitted under General Medicine at the Royal Adelaide Hospital over an 8-month period, daily ward round notes and relevant discrete data fields were collected from the electronic medical record. These data were then split into training and testing sets (7-month/1-month train/test split) prior to use in ML analyses aiming to predict discharge within the next 2days, discharge within the next 7days and an estimated date of discharge (EDD). Artificial neural networks and logistic regression were effective at predicting discharge within 48h of a given ward round note. These models achieved an area under the receiver operator curve (AUC) of 0.80 and 0.78, respectively. Prediction of discharge within 7days of a given note was less accurate, with artificial neural network returning an AUC of 0.68 and logistic regression an AUC of 0.61. The generation of an exact EDD remains inaccurate. This study has shown that repeated estimates of LOS using daily ward round notes and mixed-data inputs are effective in the prediction of general medicine discharges in the next 48h. Further research may seek to prospectively and externally validate models for prediction of upcoming discharge, as well as combination human-ML approaches for generating EDDs.

Inaccuracies in Postoperative Inpatient Stays Assumed in the Valuation of Surgical RVUs.

The aim of this study was to assess the accuracy of inpatient postoperative visits assumed in the valuation of surgical relative value units (RVUs). Medicare reimburses physicians based on the number of RVUs assigned to a service. For surgical procedures with a 10- or 90-day global period, the RVU valuation is based, in part, on a presumed number of inpatient postoperative visits whether or not those visits occur. The Centers for Medicare and Medicaid Services (CMS) have recently proposed changing all surgical procedures to a 0-day global period. We combined 2017 National Surgical Quality Improvement (NSQIP) data with physician time and RVU files from CMS. We then compared the number of inpatient postoperative visits assumed in the valuation to actual length of stay (LOS) information from the surgical registry. The analysis included 10 specialties and 601 distinct current procedural terminology codes. The number of patient observations underlying NSQIP LOS estimates ranged from 50 to 57,904. Eighty-three percent of procedures had median NSQIP LOS values that were shorter than the values assumed in the global period. These differences varied by specialty, with the largest discrepancy in neurosurgery. Procedures in this sample were last reviewed, on average, in 2000, with procedures reviewed more recently having more accurate valuations with respect to LOS. The number of postoperative visits assumed in the valuation of surgical RVUs is grossly inaccurate. Holding all else equal, removing global periods from surgical RVUs would dramatically reduce surgeon compensation.

558 Relationship Between Patient Characteristics and Number of Procedures as Well as Length of Stay for Patients Surviving Severe Burn Injuries: Analysis of the American Burn Association National Burn Repository

Abstract Introduction Published information on the relationship between patient characteristics such as total body surface area (TBSA) of burn on number of procedures and length of stay (LOS) is not widely available in the United States. Clinical expertise assumes a “rule of thumb” of 1 day stay per percentage TBSA, but deviations based on burn and patient characteristics is rarely explored. The American Burn Association NBR version 8.0 (2002–2011) was analyzed to understand the relationship between key patient and burn characteristics for surviving, severe (TBSA 10–60%) burn patients and number of procedures or LOS. Methods Outcomes include the number of procedures autograft, debridement, and excision procedures and LOS. Independent variables considered were TBSA, TBSA of partial-thickness and full-thickness (FT) burn, age (linear, squared and cubed to account for non-linearity), hospital-acquired infection (HAI), other infection, inhalation injury, female gender and diabetes status. Statistical regression models were developed to control for the independent variables and predict the number of procedures and LOS based on such characteristics. Results Among 21,175 surviving burn patients (TBSA 10–60%), the mean age was 33 years old, and the mean TBSA was 19.9%. Number of excision and autografting procedures increased with TBSA. All independent variables were retained in the LOS model. After adjusting for gender, age and comorbidities, predicted LOS for adults (18+) was 16.4, 29.5, 42.7 and 56.0 days for 10%, 20%, 30% and 40% TBSA respectively. Similarly, predicted LOS for pediatrics (age&amp;lt; 18) was 12.9, 26.0, 28.6 and 55.4 days for each TBSA group, respectively. Conclusions When considering all independent variables, the LOS per percent TBSA is estimated at approximately 1.12 days for adults and 1.01 for pediatrics. However, when considering patient (age, comorbidity status) and burn (burn depth, TBSA) characteristics, the observed LOS could vary by 66% more, as seen with detailed investigations into trends for patients with TBSA 20%. Using the predictive equations from this study, burn centers can generate tailored rule-of-thumb estimates for LOS/%TBSA that better reflect the influence of factors beyond burn center practice patterns. Applicability of Research to Practice

PSU45 THE IMPACT OF TIME-DEPENDENT BIAS ON EVALUATION OF INPATIENT COST AND LENGTH OF STAY IN HOSPITAL-ASSOCIATED VENOUS THROMBOEMBOLISM AMONG ARTHROPLASTY PATIENTS

We explored the impact of time-dependent bias through a cohort study which estimated the impact of hospital-acquired venous thromboembolism (HA-VTE) on inpatient cost and length of stay (LOS) among patients receiving total hip or knee arthroplasty. The impact of HA-VTE on inpatient cost and LOS was estimated using three different approaches with Veterans Health Administration data. First a naïve approach ignored the time-varying nature of the exposure by defining exposed subjects as those with an HA-VTE at any time in the hospital stay. Second, the time-varying exposure was addressed by restricting the cohort to subjects whose hospital stays crossed two calendar months and exposure was defined as the occurrence of HA-VTE at the beginning of the 2nd month; this enabled isolation of costs in the period before versus after HA-VTE. Third, exposure incidence density sampling was used to account for time-dependent bias. In the naïve analysis, HA-VTE increased inpatient costs by $21,551 (95% CI $16,862-$26,239) and LOS by 6.57 (95% CI 4.83-8.32) days. After correcting for time-dependent bias in the restricted cohort, these estimates were $17,113 (95% CI $3,372-$30,853) and 4.74 (95% CI -3.33-12.82) days, respectively. In the exposure incidence density sampling analysis, findings were much closer to the naïve analysis at $20,552 (95% CI $15,822-$25,283) and 6.32 (95% CI 4.58-8.07) days. Compared to estimation approaches that attempted to overcome time-dependent bias, naïve approaches led to estimates of the cost attributable to HA-VTE that were 4.86% to 25.93% higher and estimates of the LOS attributable to HA-VTE that were 3.94% to 38.64% higher. This study reveals that ignoring time-dependent bias results in the potential for considerable inflation of inpatient cost and LOS estimates attributable to HA-VTE. Researchers should take the time-dependent nature of an exposure into consideration when estimating the impact of time-varying exposures on health utilization outcomes.

A History Embedded Accelerated Failure Time Model to Estimate Nursing Home Length of Stay

With its aging population, the United States is under increasing pressure to provide long-term care (LTC) coverage to its citizens and to manage their chronic health conditions. However, the research on LTC transition and utilization modeling remains in its infancy; needless to mention LTC resource allocation and transition pathway optimization. In this paper, we developed a parametric survival model to characterize nursing home (NH) length of stay (LOS), which incorporates information on transition history. In addition, the model addressed issues such as recurrent events and competing risks. To study the effect of covariates on LOS and to ensure the flexibility of the model in evaluating operational-level interventions, we elected to develop an accelerated failure time parametric survival model. We fit the model to care transition data collected from a large cohort of older adults receiving coordinated care in a Midwestern United States urban area. Through our study, we drew the following major conclusions: 1) transition history is a significant factor and a potential predictor of an individual’s LOS in NH; 2) significance of frailty terms indicates that LOS estimates based on data with recurrent transition events can be significantly biased if not accounted for explicitly; and 3) the same clinical covariate can have opposite effects on NH LOS, depending on the destination care setting. Finally, we identified better-suited baseline hazard functions and frailty terms in each survival model from several representative candidates. Findings from our model can aid in operational-level NH care transition and utilization policy development. This paper also serves as the basis for extension into network-wide LTC transition models and utilization simulators. Note to Practitioners —This paper was motivated by the complex problem of characterizing care transitions in an LTC network, based on integrated electronic health records. It exemplifies the beginning of a series of rigors investigations on modeling multisite care transition networks. This line of research has great potential to lead to the development of intelligent decision support systems for LTC capacity planning and transition management, and eventually, empower care organizations to make data-informed management decisions. Furthermore, the empirical aspect of this paper was made possible by the availability of a comprehensive care transition longitudinal data set, at the Regenstrief Institute, constructed from a multiyear follow-up study on an elderly cohort. This cohort was prospectively followed as it moved along a Midwestern United States urban LTC system. This problem is of great interest to population-level LTC policy makers and management of LTC providers. The majority of the existing studies on LTC delivery have relied on cost-effectiveness analysis based on aggregate spending data. Our study, on the other hand, focuses on characterizing time-to-transition event for patients who reside at NH. Such a shift in focus in modeling at an increased level of granularity enhances our ability to evaluate the impact of operational-level interventions on various outcomes of LTC delivery. Finally, the novelty of our study arises from exploring useful transition history patterns to incorporate into the parametric survival model and estimating resource use from censored data. Our research findings highlight the significance of: 1) identifying the most informative time window for incorporation of transition history as a LOS predictor in NH and 2) accounting for bias in transition time estimation arising from an individual tendency to stay longer or shorter in NH through frailty terms incorporated in the parametric survival model.

Estimating Local Costs Associated With Clostridium difficile Infection Using Machine Learning and Electronic Medical Records.

BACKGROUND Reported per-patient costs of Clostridium difficile infection (CDI) vary by 2 orders of magnitude among different hospitals, implying that infection control officers need precise, local analyses to guide rational decision making between interventions. OBJECTIVE We sought to comprehensively estimate changes in length of stay (LOS) attributable to CDI at a single urban tertiary-care facility using only data automatically extractable from the electronic medical record (EMR). METHODS We performed a retrospective cohort study of 171,938 visits spanning a 7-year period. In total, 23,968 variables were extracted from EMR data recorded within 24 hours of admission to train elastic-net regularized logistic regression models for propensity score matching. To address time-dependent bias (reverse causation), we separately stratified comparisons by time of infection, and we fit multistate models. RESULTS The estimated difference in median LOS for propensity-matched cohorts varied from 3.1 days (95% CI, 2.2-3.9) to 10.1 days (95% CI, 7.3-12.2) depending on the case definition; however, dependency of the estimate on time to infection was observed. Stratification by time to first positive toxin assay, excluding probable community-acquired infections, showed a minimum excess LOS of 3.1 days (95% CI, 1.7-4.4). Under the same case definition, the multistate model averaged an excess LOS of 3.3 days (95% CI, 2.6-4.0). CONCLUSIONS In this study, 2 independent time-to-infection adjusted methods converged on similar excess LOS estimates. Changes in LOS can be extrapolated to marginal dollar costs by multiplying by average costs of an inpatient day. Infection control officers can leverage automatically extractable EMR data to estimate costs of CDI at their own institutions. Infect Control Hosp Epidemiol. 2017;38:1478-1486.

Calculating hospital length of stay using the Hospital Episode Statistics; a comparison of methodologies

BackgroundAccurate calculation of hospital length of stay (LOS) from the English Hospital Episode Statistics (HES) is important for a wide range of audit and research purposes. The two methodologies which are commonly used to achieve this differ in their accuracy and complexity. We compare these methods and make recommendations on when each is most appropriate.MethodsWe calculated LOS using continuous inpatient spells (CIPS), which link care spanning across multiple hospitals, and spells, which do not, for six conditions with short (dyspepsia or other stomach function, ENT infection), medium (dehydration and gastroenteritis, perforated or bleeding ulcer), and long (stroke, fractured proximal femur) average LOS. We examined how inter-area comparisons (i.e. benchmarking) and temporal trends differed. We defined a classification system for spells and explored the causes of differences.ResultsStroke LOS was 16.5 days using CIPS but 24% (95% CI: 23, 24) lower, at 12.6 days, using spells. Smaller differences existed for shorter-LOS conditions including dehydration and gastroenteritis (4.5 vs. 4.2 days) and ENT infection (0.9 vs. 0.8 days). Typical patient pathways differed markedly between areas and have evolved over time. One area had the third shortest stroke LOS (out of 151) using spells but the fourth longest using CIPS. These issues were most profound for stroke and fractured proximal femur, as patients were frequently transferred to a separate hospital for rehabilitation, however important disparities also existed for conditions with simpler secondary care pathways (e.g. ENT infections, dehydration and gastroenteritis).ConclusionsSpell-based LOS is widely used by researchers and national reporting organisations, including the Health and Social Care Information Centre, however it can substantially underestimate the time patients spend in hospital. A widespread shift to a CIPS methodology is required to improve the quality of LOS estimates and the robustness of research and benchmarking findings. This is vital when investigating clinical areas with typically long, complex patient pathways. Researchers should ensure that their LOS calculation methodology is fully described and explicitly acknowledge weaknesses when appropriate.

Not all patients meet the 1 day per percent burn rule: A simple method for predicting hospital length of stay in patients with burn

IntroductionHospital length of stay (LOS) is utilized to estimate resource utilization and quality of care. In burns the LOS estimation is 1day per percent total body surface area burn (1day/%TBSA). Our purpose was to evaluate the 1day/%TBSA burn rule and develop simple accurate formulas to predict LOS. MethodsThe American Burn Association National Burn Repository (NBR) from 2000 to 2013 was utilized to collate data on patients >18 years. We divided 106,543 records in half, utilizing one set to develop a model (training set) and the other to test the model (test set). We calculated the difference between observed and predicted LOS for all patients, and then examined the effect of inhalation injury and age using a linear regression model containing TBSA, age, inhalation injury and all two-way interactions. We compared predictive performance of the linear regression model to the 1day/%TBSA rule. Finally, we developed and validated three simple formulas to more accurately predict LOS than the 1day/% TBSA rule. ResultsLOS was significantly associated with patient age, TBSA, inhalation injury, and all two-way interactions. For patients <40 years without inhalation injury the main effect of TBSA was 0.71. For each decade increase in age, LOS increased by 0.74days/TBSA burn; inhalation injury added 1.70days. LOS was highly variable among patients with similar burn size, age and inhalation injury due to concomitant trauma, complications, and comorbidities. We developed 3 formulas to estimate patient LOS: (1) inhalation injury present, regardless of age (2) no inhalation injury and ≥40 years old (3) no inhalation injury and <40 years old. ConclusionsTraditional LOS estimates of 1day/%TBSA burn rule is biased, underestimating LOS, particularly for patients >40 years with inhalation injury. The following formulas applied at admission can accurately estimate hospital LOS, improve prediction over 1day/%TBSA, and provide results comparable to complicated models:I.Patients with inhalation injury: LOS=(age10)+(1.5×TBSA).II.Patients without inhalation injury <40 years of age, LOS=1day/%TBSA.III.Patients without inhalation injury ≥40 years, LOS=((age/10)−2)+TBSA.

Costs and Mortality Associated With Multidrug-Resistant Healthcare-Associated Acinetobacter Infections.

BACKGROUND Our objective was to estimate the per-infection and cumulative mortality and cost burden of multidrug-resistant (MDR) Acinetobacter healthcare-associated infections (HAIs) in the United States using data from published studies. METHODS We identified studies that estimated the excess cost, length of stay (LOS), or mortality attributable to MDR Acinetobacter HAIs. We generated estimates of the cost per HAI using 3 methods: (1) overall cost estimates, (2) multiplying LOS estimates by a cost per inpatient-day ($4,350) from the payer perspective, and (3) multiplying LOS estimates by a cost per inpatient-day from the hospital ($2,030) perspective. We deflated our estimates for time-dependent bias using an adjustment factor derived from studies that estimated attributable LOS using both time-fixed methods and either multistate models (70.4% decrease) or matching patients with and without HAIs using the timing of infection (47.4% decrease). Finally, we used the incidence rate of MDR Acinetobacter HAIs to generate cumulative incidence, cost, and mortality associated with these infections. RESULTS Our estimates of the cost per infection were $129,917 (method 1), $72,025 (method 2), and $33,510 (method 3). The pooled relative risk of mortality was 4.51 (95% CI, 1.10-32.65), which yielded a mortality rate of 10.6% (95% CI, 2.5%-29.4%). With an incidence rate of 0.141 (95% CI, 0.136-0.161) per 1,000 patient-days at risk, we estimated an annual cumulative incidence of 12,524 (95% CI, 11,509-13,625) in the United States. CONCLUSION The estimates presented here are relevant to understanding the expenditures and lives that could be saved by preventing MDR Acinetobacter HAIs. Infect Control Hosp Epidemiol 2016;1-7.

Economic (gross cost) analysis of systematically implementing a programme of advance care planning in three Irish nursing homes.

BackgroundAlthough advance care planning (ACP) and the use of advanced care directives (ACD) and end-of-life care plans are associated with a reduction in inappropriate hospitalisation, there is little evidence supporting the economic benefits of such programmes. We assessed the economic impact (gross savings) of the Let Me Decide (LMD) ACP programme in Ireland, specifically the impact on hospitalisations, bed days and location of resident deaths, before and after systematic implementation of the LMD-ACP combined with a palliative care education programme.MethodsThe LMD-ACP was introduced into three long-term care (LTC) facilities in Southern Ireland and outcomes were compared pre and post implementation. In addition, 90 staff were trained in a palliative care educational programme. Economic analysis including probabilistic sensitivity analysis was performed.ResultsThe uptake of an ACD or end-of-life care post-implementation rose from 25 to 76 %. Post implementation, there were statistically significant decreases in hospitalisation rates from baseline (hospitalisation incidents declined from 27.8 to 14.6 %, z = 3.96, p < 0.001; inpatient hospital days reduced from 0.54 to 0.36 %, z = 8.85, p < 0.001). The percentage of hospital deaths also decreased from 22.9 to 8.4 %, z = 3.22, p = 0.001. However, length of stay (LOS) increased marginally (7–9 days). Economic analysis suggested a cost-reduction related to reduced hospitalisations ranging between €10 and €17.8 million/annum and reduction in ambulance transfers, estimated at €0.4 million/annum if these results were extrapolated nationally. When unit costs and LOS estimates were varied in scenario analyses, the expected cost reduction owing to reduced hospitalisations, ranged from €17.7 to €42.4 million nationally.ConclusionsImplementation of the LMD-ACP (ACD/end-of-life care plans combined with palliative care education) programme resulted in reduced rates of hospitalisation. Despite an increase in LOS, likely reflecting more complex care needs of admitted residents, gross costs were reduced and scenario analysis projected large annual savings if these results were extrapolated to the wider LTC population in Ireland.