Probability Of Length Of Stay Research Articles

Can machine learning models predict prolonged length of hospital stay following primary total knee arthroplasty based on a national patient cohort data?

The total length of stay (LOS) is one of the biggest determinators of overall care costs associated with total knee arthroplasty (TKA). An accurate prediction of LOS could aid in optimizing discharge strategy for patients in need and diminishing healthcare expenditure. The aim of this study was to predict LOS following TKA using machine learning models developed on a national-scale patient cohort. The ACS-NSQIP database was queried to acquire 267,966 TKA cases from 2013 to 2020. Four machine learning models-artificial neural network (ANN), random forest, histogram-based gradient boosting, and k-nearest neighbor were trained and tested on the dataset for the prediction of prolonged LOS (LOS exceeded the 75th of all values in the cohort). The model performance was assessed by discrimination (area under the receiver operating characteristic curve [AUC]), calibration, and clinical utility. ANN delivered the best performance among the four models. ANN distinguished prolonged LOS in the study cohort with an AUC of 0.71 and accurately predicted the probability of prolonged LOS for individual patients (calibration slope: 0.82; calibration intercept: 0.03; Brier score: 0.089). All models demonstrated clinical utility by generating positive net benefits in decision curve analyses. Operation time, pre-operative transfusion, pre-operative laboratory tests (hematocrit, platelet count, and white blood cell count), and BMI were the strongest predictors of prolonged LOS. ANN demonstrated modest discrimination capacity and excellent performance in calibration and clinical utility for the prediction of prolonged LOS following TKA. Clinical application of the machine learning models has the potential to improve care coordination and discharge planning for patients at high risk of extended hospitalization after surgery. Incorporating more relevant patient factors may further increase the models' prediction strength.

Does advanced maternal age explain the longer hospitalisation of mothers after childbirth?

Fertility postponement, which has comprised the most significant reproductive trend in developed countries over the last few decades, involves a number of social, personal and health consequences. The length of stay (LOS) in hospital following childbirth varies considerably between countries. Czechia, where the fertility postponement process has been particularly dynamic, has one of the longest mean LOS of the OECD member countries. We analyse the influence of the age of mothers on the LOS in hospital associated with childbirth. We employed anonymised individual data provided by the General Health Insurance Company of the Czech Republic on women who gave birth in 2014. Kaplan-Meier survival plots and binary logistic regression were employed to identify factors associated with long stays (> = 7 days for vaginal births, > = 9 days for CS births). The impact of the maternal age on the LOS is U-shaped. A higher risk of a longer hospitalisation period for young mothers was identified for both types of birth (OR = 1.58, 95% CI 1.33-1.87, p˂0.001 for age less than 20, OR = 1.31, 95% CI 1.20-1.44, p˂0.001 for age 20-24 compared to 30-34). The risk of a longer stay in hospital increases with the increasing age of the mother (OR = 1.23, 95% CI 1.13-1.35, p˂0.001 for age 35-39, OR = 2.05, 95% CI 1.73-2.44, p˂0.001 for age 40+ compared to 30-34), especially with concern to vaginal births. The probability of a long LOS increases significantly after the age of 35, especially in the case of vaginal births. Thus, the fertility postponement process with the significant change in the age structure of mothers contributes to the increase in health care costs associated with post-birth hospitalisation.

Dysphagia is associated with worse clinical outcomes in geriatric trauma patients

IntroductionDysphagia is associated with increased morbidity, mortality, and resource utilization in hospitalized patients, but studies on outcomes in geriatric trauma patients with dysphagia are limited. We hypothesized that geriatric trauma...

Malnutrition according to the Global Leadership Initiative on Malnutrition criteria is associated with in-hospital mortality and prolonged length of stay in patients with cirrhosis

ObjectivesMalnutrition is prevalent and negatively affects patients with cirrhosis, but a generally accepted consensus pertaining to its diagnosis is lacking. Recently, a framework called the Global Leadership Initiative on Malnutrition (GLIM) has been proposed to diagnose malnutrition, but there is scant evidence regarding its validity. We aimed to investigate associations of malnutrition according to the GLIM criteria, as well as its individual indicator with in-hospital adverse outcomes. MethodsThis was a prospective, observational study of consecutively hospitalized patients with cirrhosis. The malnutrition diagnosis was built on a stepwise GLIM process with initial screening, followed by fulfillment of at least one phenotypic and one etiologic criterion. Patients were followed up for a combined endpoint of in-hospital mortality and prolonged length of stay (LOS). Covariates compromise malnutrition according to the GLIM criteria and its indicators in separation. Logistic regression analyses were implemented to determine predictive validity. ResultsA total of 387 cirrhotic patients were assessed. Malnutrition was diagnosed in 28.7% of patients according to the GLIM criteria, and increased the risk of in-hospital mortality and prolonged LOS by 2.166 and 1.767 times, respectively, adjusting for age, sex, biochemical parameters, and clinical scores of disease severity. When analyzing separate criteria, all constituents were independently associated with in-hospital adverse outcomes, adjusting for model for end-stage liver disease sodium score. ConclusionsMalnutrition according to the GLIM criteria was considerably prevalent among hospitalized patients with cirrhosis, and associated with approximately two times greater probability of in-hospital mortality and prolonged LOS. These diagnostic criteria may be implemented and disseminated during daily practice considering their predictive validity.

Individual components of post-hepatectomy care pathways have differential impacts on length of stay

BackgroundThe value of individual variable contributions to post-hepatectomy length of stay (LOS) are difficult to quantify within bundled care pathways. MethodsPoisson regression and marginal effects models for prolonged post-hepatectomy LOS (>25% median) included Kawaguchi-Gayet (KG) complexity, perioperative variables, and pathways (minimally-invasive = MIS; low-intermediate-risk = KGI/II; high-risk = KGIII; combination). ResultsMedian LOS was 2, 4, 5, and 5 days for MIS, KGI/II, KGIII and combination pathways (N = 978). Poisson regression identified age, intraoperative fluids, delayed diet tolerance, and combination cases as associated with increased LOS (p < 0.01). Marginal effects analysis demonstrated the following added probability of longer LOS: each year of age 0.03x, 250 mL intraoperative fluids 0.06x, each operative hour 0.2x, additional day before diet tolerance 0.4x, combination cases 0.7x. MIS was associated with 1.2x increased probability of shorter LOS. ConclusionsOptimizing intraoperative fluids, operative time, and postoperative diet, while favoring MIS approach when feasible, may maximize effects of post-hepatectomy care pathways to reduce LOS.

Individual Components of Post-Hepatectomy Care Pathways Have Differential Impacts on Length of Stay

Introduction: Optimization of perioperative factors contributing to post-hepatectomy length of stay (LOS) is an aim of current care pathways. However, individual variable contributions to LOS are difficult to separate. Method: Using a prospectively-maintained quality improvement database, hepatectomy patients (1/1/2017-7/1/2020) were included in Poisson regression and marginal effects analysis for LOS. Kawaguchi-Gayet (K-G) complexity classifications were used for comparing expected LOS among pathways (minimally-invasive=MIS; low-intermediate=K-G I/II; high=K-G III; combination operations). Prolonged LOS was >25% median LOS by pathways. “Diet tolerance” was defined as 24 hours of eating. Results: Among 978 consecutive patients, median LOS was 2, 4, 5, and 5 days for MIS, K-G I/II, K-G III and combination operations, respectively. Multivariable Poisson regression identified age, intraoperative fluids, delayed diet tolerance, and combination cases associated with increased LOS (all p<0.01). MIS pathway was associated with shorter LOS (p<0.001, Table 1). Marginal effects analysis demonstrated on average, each additional year of age contributed 0.03x increased LOS; each 250mL of intraoperative IV fluids contributed 0.06x; each additional operative hour contributed 0.2x; each additional day to diet tolerance contributed 0.4x, while MIS approach had 1.2x increased probability of shorter LOS. Combination operations contributed 0.7x increased probably of longer LOS (Figure 1). Conclusions: Modifiable components of post-hepatectomy care pathways to potentially impact LOS include restricting intraoperative fluids, optimizing operative time, introducing a diet on postoperative day 1, and favoring MIS approaches when feasible. Intentionally highlighting these individual pathway components may maximize the overall effects of the comprehensive care bundle.Tabled 1Table 1: Multivariable Model for Length of Stay after Hepatectomy According to Level of Operative ComplexityVariableIRRp-value*95% CIAge (in years)1.006<0.0011.003-1.009Male sex1.0060.8710.94-1.08Intraoperative IV fluids (per 250 mL)1.0140.0111.003-1.02Total operative time (per 60 min)1.037<0.0011.17-1.06Estimated blood loss (in mL)1.000.6470.99-1.00Tolerated regular diet for 24 hours (in days)1.09<0.0011.07-1.10PathwayK-G I/IIK-G IIIMISCombination casesRef1.080.731.15Ref0.129<0.0010.004Ref0.98-1.190.64-0.821.04-1.26IRR: Incidence Rate Ratio; IV: intravenous; CI: confidence interval; MIS: minimally invasive surgery; K-G: Kawaguchi-Gayet*p-value set at alpha of 0.05 Open table in a new tab IRR: Incidence Rate Ratio; IV: intravenous; CI: confidence interval; MIS: minimally invasive surgery; K-G: Kawaguchi-Gayet *p-value set at alpha of 0.05

692 A Machine Learning Approach to Predict the Postoperative Length of Stay After Coronary Artery Bypass Grafting Using Preoperative Characteristics

Abstract Aim Lengthy hospital length of stay (LOS) has a direct impact on healthcare costs. We aimed to design predictive models of prolonged LOS after coronary artery bypass grafting (CABG) with only preoperative characteristics and machine learning (ML) strategies. Method In a single centre retrospective analysis, 2,082 consecutive patients underwent first-time elective/urgent CABG: 1,262 has a short postoperative LOS (≤ 6 days) while the remaining 820 had a long LOS (&amp;gt; 6 days). 70/30 training/testing ratio and resampling methods were used, and cross-validation was conducted. Results The two groups differ significantly in terms of pre-operative variables: short LOS patients were younger (p &amp;lt; 0.01), more frequently male (p &amp;lt; 0.01) with lower BMI (p &amp;lt; 0.01) and better angina class (p &amp;lt; 0.01) and NYHA class (p &amp;lt; 0.01). Moreover, they had lower incidence of hypertension (p = 0.04), COPD (p &amp;lt; 0.01) and PVD (p &amp;lt; 0.01). The Logistic Euroscore was also better in this group (median 0.02 vs 0.03, p &amp;lt; 0.01). The predictive abilities of the ML models were as follows: logistic regression: Area under the Curve (AUC) = 0.71, accuracy = 0.69; Generalized additive model: AUC= 0.7, Accuracy = 0.68; Random Decision Forest: AUC = 0.7, Accuracy = 0.68; Naïve Bayes AUC = 0.63, Accuracy 0.58. Conclusions Developing a reliable predictive model with only pre-operative variables proved to be difficult, but several preoperative characteristics have a significant impact on the probability of prolonged LOS after CABG. Larger studies are needed to investigate the possibility of developing a reliable predictive model that would help to improve surgical planning.

A Novel Scoring System to Predict Length of Stay After Anterior Cervical Discectomy and Fusion.

The movement toward reducing healthcare expenditures has led to an increased volume of outpatient anterior cervical diskectomy and fusions (ACDFs). Appropriateness for outpatient surgery can be gauged based on the duration of recovery each patient will likely need. Patients undergoing 1- or 2-level ACDFs were retrospectively identified at a single Level I spine surgery referral institution. Length of stay (LOS) was categorized binarily as either less than two midnights or two or more midnights. The data were split into training (80%) and test (20%) sets. Two multivariate regressions and three machine learning models were developed to predict a probability of LOS ≥ 2 based on preoperative patient characteristics. Using each model, coefficients were computed for each risk factor based on the training data set and used to create a calculatable ACDF Predictive Scoring System (APSS). Performance of each APSS was then evaluated on a subsample of the data set withheld from training. Decision curve analysis was done to evaluate benefit across probability thresholds for the best performing model. In the final analysis, 1,516 patients had a LOS <2 and 643 had a LOS ≥2. Patient characteristics used for predictive modeling were American Society of Anesthesiologists score, age, body mass index, sex, procedure type, history of chronic pulmonary disease, depression, diabetes, hypertension, and hypothyroidism. The best performing APSS was modeled after a lasso regression. When applied to the withheld test data set, the APSS-lasso had an area under the curve from the receiver operating characteristic curve of 0.68, with a specificity of 0.78 and a sensitivity of 0.49. The calculated APSS scores ranged between 0 and 45 and corresponded to a probability of LOS ≥2 between 4% and 97%. Using classic statistics and machine learning, this scoring system provides a platform for stratifying patients undergoing ACDF into an inpatient or outpatient surgical setting.

Low Serum Albumin Correlates With Adverse Events Following Surgery for Male Urinary Incontinence: Analysis of the American College of Surgeons National Surgical Quality Improvement Project

OBJECTIVETo investigate the incidence and risk factors associated with artificial urinary sphincter (AUS) and male urethral sling placement (MUS), revision, and removal. METHODSWe identified CPT codes of patients undergoing AUS and sling placement, revision, and removal in the American College of Surgeons National Surgery Quality Improvement Program database. Bivariate analysis was used to compare preoperative parameters against adverse events of interest (Length of stay (LOS) >1, readmission, reoperation, other postoperative complications, and death). Variables that were significant or neared significance (P <.1) in the univariate analysis were entered into multivariable logistic regression models. Multivariable models were used to estimate the probability of adverse events. RESULTSAbout 2792 patients underwent surgical treatment for stress urinary incontinence in the American College of Surgeons National Surgery Quality Improvement Program database from 2008 to 2016. Increased length of stay was the most common adverse event (12.7%), followed by other postoperative complications (4.9%), readmission (4%), reoperation (2.3%), and death (0.3%). We noted an association between perioperative adverse events and preoperative hypoalbuminemia. Patients with preoperative hypoalbuminemia compared with patients with normal preoperative serum albumin had an increase predicted probability of LOS >1 day (42% vs 10%), readmission (10% vs 4%), reoperation (6% vs 2%), other postoperative complications (18% vs 4%) after adjusting for other factors. CONCLUSIONSurgical treatment for stress urinary incontinence is well tolerated with acceptable levels of perioperative adverse events. Low serum albumin (<3.5 ng/dL) was associated with perioperative adverse events. These data may affect preoperative decision making and direct future quality improve efforts at the highest risk patients to help minimize perioperative morbidity and mortality.

Association of Wearable Activity Monitors With Assessment of Daily Ambulation and Length of Stay Among Patients Undergoing Major Surgery

Early postoperative ambulation is vital to minimizing length of stay (LOS), but few hospitals objectively measure ambulation to predict outcomes. Wearable activity monitors have the potential to transform assessment of postoperative ambulation, but key implementation data, including whether digitally monitored step count can identify patients at risk for poor efficiency outcomes, are lacking. To define the distribution of digitally measured daily step counts after major inpatient surgical procedures, to assess the accuracy of physician assessment and ordering of ambulation, and to quantify the association of digitally measured step count with LOS. Prospective cohort study at Cedars-Sinai Medical Center, an urban tertiary referral center. Participants were patients undergoing 8 inpatient operations (lung lobectomy, gastric bypass, hip replacement, robotic cystectomy, open colectomy, abdominal hysterectomy, sleeve gastrectomy, and laparoscopic colectomy) from July 11, 2016, to August 30, 2017. Use of activity monitors to measure daily postoperative step count. Operation-specific daily step count, daily step count by physician orders and assessment, and a prolonged LOS (>70th percentile for each operation). Among 100 patients (53% female), the mean (SD) age was 53 (18) years, and the median LOS was 4 days (interquartile range, 3-6 days). There was a statistically significant increase in daily step count with successive postoperative days in aggregate (r = 0.55; 95% bootstrapped CI, 0.47-0.62; P < .001) and across individual operations. Ninety-five percent (356 of 373) of daily ambulation orders were "ambulate with assistance," although daily step counts ranged from 0 to 7698 steps (0-5.5 km) under this order. Physician estimation of ambulation was predictive of the median step count (r = 0.66; 95% bootstrapped CI, 0.59-0.72; P < .001), although there was substantial variation within each assessment category. For example, daily step counts ranged from 0 to 1803 steps (0-1.3 km) in the "out of bed to chair" category. Higher step count on postoperative day 1 was associated with lower odds of prolonged LOS from 0 to 1000 steps (odds ratio [OR], 0.63; 95% CI, 0.45-0.84; P = .003), with no further decrease in odds after 1000 steps (OR, 0.99; 95% CI, 0.75-1.30; P = .80). In this study, digitally measured step count up to 1000 steps on postoperative day 1 was associated with lower probability of a prolonged LOS. Wearable activity monitors improved the accuracy of assessment of daily step count over the current standard of care, providing an opportunity to identify patients at risk for poor efficiency outcomes.

Effects of admission (adm) source, time, and provider on inpatient (inpt) oncology (onc) outcomes at the Cleveland Clinic Foundation (CCF).

140 Background: The quality of care transfers is known to influence clinical outcomes. In an inpt onc setting at a major tertiary care referral center, patient (pt) adm originate from many different areas and times. A detailed evaluation of onc adm by source of transfer, admission time, and provider type, may identify opportunities to improve inpt clinical outcomes. Methods: We retrospectively reviewed all adm to the inpt solid tumor onc service from July - December 2014 from CCF regional hospital emergency departments (ED), outside hospital (OSH) ED, OSH inpt services, and CCF outpt clinics. Pts were excluded if the adm was planned or if admitted from the CCF Main Campus ED. Data collected included pt and encounter characteristics and provider type (house-staff or nocturnal hospitalist). Clinical outcomes, including activation of the adult medical emergency team (AMET), ICU transfers, length of stay (LOS), and in-hospital mortality were compared using chi-squared test; ECOG PS and LOS with the Kruskal-Wallis tests and Wilcoxon rank sum test. Results: A total of 413 unique pt admissions were reviewed. 213 were included after exclusion criteria were applied. The probability of AMET activation, mortality, and LOS differed by origin of transfer. Pts admitted from CCF regional EDs had the lowest median LOS and no deaths. OSH int transfers demonstrated significantly higher mortality vs other origins of transfer. Pts whose first orders were placed after 5pm had no significant differences in AMET activation, ICU transfers, LOS, or mortality vs daytime adm. There were no differences in adverse outcomes by the type of admitting provider. Conclusions: Onc inpts transferred from an outside healthcare setting were at highest risk for adverse outcomes (AMETs, increased LOS, and mortality) include those originating from OSH inpt services. Process and communication interventions focused on transfers from outside inpt facilities may improve safety and outcomes in this population. [Table: see text]

Variations and Determinants of Mortality and Length of Stay of Very Low Birth Weight and Very Low for Gestational Age Infants in Seven European Countries.

The EuroHOPE very low birth weight and very low for gestational age infants study aimed to measure and explain variation in mortality and length of stay (LoS) in the populations of seven European nations (Finland, Hungary, Italy (only the province of Rome), the Netherlands, Norway, Scotland and Sweden). Data were linked from birth, hospital discharge and mortality registries. For each infant basic clinical and demographic information, infant mortality and LoS at 1 year were retrieved. In addition, socio-economic variables at the regional level were used. Results based on 16,087 infants confirm that gestational age and Apgar score at 5 min are important determinants of both mortality and LoS. In most countries, infants admitted or transferred to third-level hospitals showed lower probability of death and longer LoS. In the meta-analyses, the combined estimates show that being male, multiple births, presence of malformations, per capita income and low population density are significant risk factors for death. It is essential that national policies improve the quality of administrative datasets and address systemic problems in assigning identification numbers at birth. European policy should aim at improving the comparability of data across jurisdictions.

Abstract 261: Predicting Inpatient Hospital Length of Stay for Acute Health Failure Patients

Background: Hospital length of stay (LOS) is a primary driver of costs of care for hospitalized Acute Health Failure (AHF). A predictive model of LOS associated with AHF hospitalized patients would be beneficial in identifying potential high risk patients that may benefit from more intensive interventions. Objective: To develop predictive models for the length of inpatient hospital stay based on information available upon admission for AHF. Methods: A retrospective study using the MarketScan Hospital Drug Database identified patients with an index admission event with an acute heart failure primary discharge diagnosis (ICD-9 code 428.XX) and diagnosis reference group (DRG),(DRG 127, 291,292 or 293) for hospital admissions between 2006 and 2010. Index hospitalization was identified as the first AHF admission with no admissions in the preceding 90 days. Patient demographics, admission source, hospital characteristics (total number of beds, teaching status, urban-rural), geographic location, Charlson Comorbidity Score (CCI) and select comorbidities were identified within the index admission and served as potential input variables to model the index admission LOS. Variable selection was informed by univariate negative binomial regression models and regression tree models. Final predictive models took the form of full and stepwise negative binomial regression models (predicting LOS) and logistic regression models to predict short (≤ 2 days, approximately 20th percentile) and long (≥8 days, approximately 80th percentile) LOS. Results: 23,629 distinct admissions were identified with mean LOS 5.5 days. The three strongest predictors of LOS in both the univariate models and the regression tree model were CCI score, nephropathy and cardiac dysrhythmia. The negative binomial model shows that comorbid cardiac dysrhythmias, anemia, nephropathy, thromboembolic disease, COPD and high CCI score were associated with increased LOS. For the logistic regression models predicting the probability of long LOS (concordance=67.3-67.5%), the most important variables were cardiac dysrhythmia, index year, and CCI score. For the logistic regression models predicting the probability of short LOS (concordance=66.3-66.5%), patients with cardiac dysrhythmia, anemia, increasing age and high CCI score were less likely to have LOS≤2 days. Conclusion: Across multiple modeling methods and different outcome variable definitions, it was consistently observed that AHF patients with comorbid cardiac dysrhythmia, nephropathy, anemia, higher CCI scores and older patients tended to have increased lengths of inpatient stay for their index hospitalization. Clinicians may use this information in identifying patients with higher risk of longer hospitalizations to optimize inpatient healthcare decisions and cost management strategies.

41 UNDERSTANDING THE COMPARATIVE EFFECTIVENESS OF LAPAROSCOPIC RADICAL NEPHRECTOMY FOR PATIENTS WITH KIDNEY CANCER

INTRODUCTION AND OBJECTIVES: Although laparoscopic radical nephrectomy (LRN) is now a preferred treatment for many patients with kidney cancer, there is limited population-based data evaluating its comparative effectiveness relative to open radical nephrectomy (ORN). In this context, we compared recent trends in hospital-based outcomes among a nationally-representative sample of Medicare beneficiaries undergoing LRN vs. ORN. METHODS: Using linked SEER-Medicare data, we identified a population-based sample of patients treated with LRN or ORN for non-urothelial, non-metastatic kidney cancer from 2000 through 2005. For each patient, we measured an array of demographic (e.g., age), clinical (e.g., Charlson comorbidity index), and cancer-specific (e.g., tumor size and stage) characteristics. Next, we defined the following outcomes: ICU admission; prolonged length of stay (LOS); readmission within 30 days; and in-hospital mortality. We then fit multivariable logistic regression models to estimate the association between type of surgery and each clinical outcome, adjusting for measurable patient characteristics and year of surgery. RESULTS: Among our analytic sample, 2,108 (26%) and 5,895 (74%) patients underwent treatment with LRN and ORN, respectively. Patients treated with LRN were more likely to be White, female, of higher socioeconomic position, and to have a tumor size 4 cm (all p-values 0.05). After adjusting for demographic, clinical and tumor characteristics, the probability of ICU admission and prolonged LOS was 42.4% and 46.6% lower, respectively, for patients undergoing LRN vs. ORN (Figure, p 0.001). Conversely, although still uncommon, the adjusted probability of in-hospital mortality was 52% (p 0.041) higher for patients treated with a laparoscopic vs. open approach (Figure). CONCLUSIONS: In this population-based sample, patients treated with LRN had a lower likelihood of ICU admission and prolonged LOS, findings that support the convalescence benefits of laparoscopy. That in-hospital mortality was higher among patients treated with LRN suggests a potentially unanticipated consequence of this minimally-invasive technique and highlights the need for long-term monitoring following diffusion of new surgical technologies. Source of Funding: Edwin Beer Research Fellowship

Concurrent Prediction of Hospital Mortality and Length of Stay from Risk Factors on Admission

To develop a method for predicting concurrently both hospital survival and length of stay (LOS) for seriously ill or injured patients, with particular attention to the competing risks of death or discharge alive as determinants of LOS. Previously collected 1995-1996 registry data on 2,646 cases of injured patients from three trauma centers in Maine. Time intervals were determined for which the rates of discharge or death were relatively constant. Poisson regression was used to develop a model for each type of terminal event, with risk factors on admission contributing proportionately to the subsequent rates for each outcome in each interval. Mean LOS and cumulative survival were calculated from a combination of the resulting piecewise exponential models. Age, Glasgow Coma Scale, Abbreviated Injury Scores, and specific mechanisms of injury were significant predictors of the rates of death and discharge, with effects that were variable in different time intervals. Predicted probability of survival and mean LOS from the model were similar to actual values for categorized patient groups. Piecewise exponential models may be useful in predicting LOS, especially if determinants of mortality are separated from determinants of discharge alive.