The prediction of hospital length of stay using unstructured data.
Ontology highlight
ABSTRACT: Objective
This study aimed to assess the performance improvement for machine learning-based hospital length of stay (LOS) predictions when clinical signs written in text are accounted for and compared to the traditional approach of solely considering structured information such as age, gender and major ICD diagnosis.Methods
This study was an observational retrospective cohort study and analyzed patient stays admitted between 1 January to 24 September 2019. For each stay, a patient was admitted through the Emergency Department (ED) and stayed for more than two days in the subsequent service. LOS was predicted using two random forest models. The first included unstructured text extracted from electronic health records (EHRs). A word-embedding algorithm based on UMLS terminology with exact matching restricted to patient-centric affirmation sentences was used to assess the EHR data. The second model was primarily based on structured data in the form of diagnoses coded from the International Classification of Disease 10th Edition (ICD-10) and triage codes (CCMU/GEMSA classifications). Variables common to both models were: age, gender, zip/postal code, LOS in the ED, recent visit flag, assigned patient ward after the ED stay and short-term ED activity. Models were trained on 80% of data and performance was evaluated by accuracy on the remaining 20% test data.Results
The model using unstructured data had a 75.0% accuracy compared to 74.1% for the model containing structured data. The two models produced a similar prediction in 86.6% of cases. In a secondary analysis restricted to intensive care patients, the accuracy of both models was also similar (76.3% vs 75.0%).Conclusions
LOS prediction using unstructured data had similar accuracy to using structured data and can be considered of use to accurately model LOS.
SUBMITTER: Chrusciel J
PROVIDER: S-EPMC8684269 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Hospital length of stay (LOS) after burn injury is commonly estimated as 1 day per percent burn, but LOS often exceeds that estimate. The purpose of this study is to develop a novel method for estimating burn hospital LOS at any time during hospitalization. The authors used the American Burn Association National Burn Repository from 2000 to 2009 to directly estimate the median residual LOS (MRLOS) of patients hospitalized for burn injuries and who survived to discharge. The MRLOS is the median of how many more days a burn patient will be hospitalized given that the person has been in hospital for a specified time period. The authors also estimated the 25th and 75th percentiles of residual life and quantified the relationship between MRLOS and LOS with ordinary least squares for all burn patients, by burn size and by presence of inhalation injury. MRLOS increased with increasing LOS, confirming that discharge estimates change over time. Patients with inhalation injury had longer MRLOS than patients without inhalation injury in the first 100 hospital days. Patients with large burns (>25%) had large MRLOS consistent with prolonged hospitalization, but patients with small burns (<25%) also had steadily increasing MRLOS during hospitalization (ie, the longer the patient was in the hospital, the longer the predicted LOS regardless of initial LOS estimate). Estimating remaining hospital LOS at any time during admission has been problematic; MRLOS can be used to provide an estimate of remaining hospital LOS and resource utilization to families, administrators, and other medical professionals.
| S-EPMC5014656 | biostudies-literature
Project description:ObjectiveTo develop a predictive model for real-time predictions of length of stay, mortality, and readmission for hospitalized patients using electronic health records (EHRs).Materials and methodsA Bayesian Network model was built to estimate the probability of a hospitalized patient being "at home," in the hospital, or dead for each of the next 7 days. The network utilizes patient-specific administrative and laboratory data and is updated each time a new pathology test result becomes available. Electronic health records from 32 634 patients admitted to a Sydney metropolitan hospital via the emergency department from July 2008 through December 2011 were used. The model was tested on 2011 data and trained on the data of earlier years.ResultsThe model achieved an average daily accuracy of 80% and area under the receiving operating characteristic curve (AUROC) of 0.82. The model's predictive ability was highest within 24 hours from prediction (AUROC = 0.83) and decreased slightly with time. Death was the most predictable outcome with a daily average accuracy of 93% and AUROC of 0.84.DiscussionWe developed the first non-disease-specific model that simultaneously predicts remaining days of hospitalization, death, and readmission as part of the same outcome. By providing a future daily probability for each outcome class, we enable the visualization of future patient trajectories. Among these, it is possible to identify trajectories indicating expected discharge, expected continuing hospitalization, expected death, and possible readmission.ConclusionsBayesian Networks can model EHRs to provide real-time forecasts for patient outcomes, which provide richer information than traditional independent point predictions of length of stay, death, or readmission, and can thus better support decision making.
| S-EPMC7839923 | biostudies-literature
Project description:This work introduces a predictive Length of Stay (LOS) framework for lung cancer patients using machine learning (ML) models. The framework proposed to deal with imbalanced datasets for classification-based approaches using electronic healthcare records (EHR). We have utilized supervised ML methods to predict lung cancer inpatients LOS during ICU hospitalization using the MIMIC-III dataset. Random Forest (RF) Model outperformed other models and achieved predicted results during the three framework phases. With clinical significance features selection, over-sampling methods (SMOTE and ADASYN) achieved the highest AUC results (98% with CI 95%: 95.3-100%, and 100% respectively). The combination of Over-sampling and under-sampling achieved the second-highest AUC results (98%, with CI 95%: 95.3-100%, and 97%, CI 95%: 93.7-100% SMOTE-Tomek, and SMOTE-ENN respectively). Under-sampling methods reported the least important AUC results (50%, with CI 95%: 40.2-59.8%) for both (ENN and Tomek- Links). Using ML explainable technique called SHAP, we explained the outcome of the predictive model (RF) with SMOTE class balancing technique to understand the most significant clinical features that contributed to predicting lung cancer LOS with the RF model. Our promising framework allows us to employ ML techniques in-hospital clinical information systems to predict lung cancer admissions into ICU.
| S-EPMC8755804 | biostudies-literature
Project description:BackgroundLength of stay (LOS) is an important indicator of the optimization of health services and hospital financing efficiency in aneurysmal subarachnoid hemorrhage (aSAH) patients. The purpose of this study was to develop a scoring model to predict the LOS of patients with aSAH.MethodA clinical scoring was developed based on retrospectively collected data from the cerebral aneurysm registry of the National Brain Center Hospital, Jakarta, from January 2019 to June 2022. Multivariate logistic regression was used to determine the odds ratio for risk-adjusted prolonged LOS. LOS predictors were obtained based on the regression coefficients and converted into a point score model.ResultsOf the 209 aSAH patients observed, 117 patients had prolonged LOS (> 14 days of hospital stay). A clinical score was developed with a range of 0-7 points. Four variables were chosen as predictors of prolonged LOS: the presence of high-grade aSAH (1 point), aneurysm treatment (endovascular coiling: 1 point; surgical clipping: 2 points), cardiovascular comorbidities (1 point), and hospital-acquired pneumonia (3 points). The score showed good discrimination with an area under the receiving operating characteristics curve (AUC) of 0.8183 (SE 0.0278) and a p-value for the Hosmer-Lemeshow (HL) goodness-of-fit of 0.9322.ConclusionThis simple clinical score reliably predicted prolonged LOS in aneurysmal subarachnoid hemorrhage cases and may aid clinicians in improving patient outcomes and decreasing healthcare costs.
| S-EPMC10273541 | biostudies-literature
Project description:BackgroundThe COVID-19 pandemic has placed an unprecedented strain on health systems, with rapidly increasing demand for healthcare in hospitals and intensive care units (ICUs) worldwide. As the pandemic escalates, determining the resulting needs for healthcare resources (beds, staff, equipment) has become a key priority for many countries. Projecting future demand requires estimates of how long patients with COVID-19 need different levels of hospital care.MethodsWe performed a systematic review of early evidence on length of stay (LoS) of patients with COVID-19 in hospital and in ICU. We subsequently developed a method to generate LoS distributions which combines summary statistics reported in multiple studies, accounting for differences in sample sizes. Applying this approach, we provide distributions for total hospital and ICU LoS from studies in China and elsewhere, for use by the community.ResultsWe identified 52 studies, the majority from China (46/52). Median hospital LoS ranged from 4 to 53 days within China, and 4 to 21 days outside of China, across 45 studies. ICU LoS was reported by eight studies-four each within and outside China-with median values ranging from 6 to 12 and 4 to 19 days, respectively. Our summary distributions have a median hospital LoS of 14 (IQR 10-19) days for China, compared with 5 (IQR 3-9) days outside of China. For ICU, the summary distributions are more similar (median (IQR) of 8 (5-13) days for China and 7 (4-11) days outside of China). There was a visible difference by discharge status, with patients who were discharged alive having longer LoS than those who died during their admission, but no trend associated with study date.ConclusionPatients with COVID-19 in China appeared to remain in hospital for longer than elsewhere. This may be explained by differences in criteria for admission and discharge between countries, and different timing within the pandemic. In the absence of local data, the combined summary LoS distributions provided here can be used to model bed demands for contingency planning and then updated, with the novel method presented here, as more studies with aggregated statistics emerge outside China.
| S-EPMC7467845 | biostudies-literature
Project description:BackgroundThe aim of this study was to improve understanding of hospital length of stay (LOS) in patients undergoing total joint arthroplasty (TJA) in a high-efficiency, hospital-based pathway.MethodsWe retrospectively reviewed 1401 consecutive primary and revision TJA patients across 67 patient and preoperative care characteristics from 2016 to 2019 from the institutional electronic health records. A machine learning approach, testing multiple models, was used to assess predictors of LOS.ResultsThe median LOS was 1 day; outpatients accounted for 16.5%, 1-day inpatient stays for 38.0%, 2-day stays for 26.4%, and 3-days or more for 19.1%. Patients characteristically fell into 1 of 3 broad categories that contained relatively similar characteristics: outpatient (0-day LOS), short stay (1- to 2-day LOS), and prolonged stay (3 days or greater). The random forest models suggested that a lower Risk Assessment and Prediction Tool score, unplanned admission or hospital transfer, and a medical history of cardiovascular disease were associated with an increased LOS. Documented narcotic use for surgery preparation prior to hospitalization and preoperative corticosteroid use were factors independently associated with a decreased LOS.ConclusionsAfter TJA, most patients have either an outpatient or short-stay hospital episode. Patients who stay 2 days do not differ substantially from patients who stay 1 day, while there is a distinct group that requires prolonged admission. Our machine learning models support a better understanding of the patient factors associated with different hospital LOS categories for TJA, demonstrating the potential for improved health policy decisions and risk stratification for centers caring for complex patients.
| S-EPMC10372176 | biostudies-literature
Project description:IntroductionEffective planning of elective surgical procedures requiring postoperative intensive care is important in preventing cancellations and empty intensive care unit (ICU) beds. To improve planning, we constructed, validated and tested three models designed to predict length of stay (LOS) in the ICU in individual patients.MethodsRetrospective data were collected from 518 consecutive patients who underwent oesophagectomy with reconstruction for carcinoma between January 1997 and April 2005. Three multivariable linear regression models for LOS, namely preoperative, postoperative and intra-ICU, were constructed using these data. Internal validation was assessed using bootstrap sampling in order to obtain validated estimates of the explained variance (r2). To determine the potential gain of the best performing model in day-to-day clinical practice, prospective data from a second cohort of 65 consecutive patients undergoing oesophagectomy between May 2005 and April 2006 were used in the model, and the predictive performance of the model was compared with prediction based on mean LOS.ResultsThe intra-ICU model had an r2 of 45% after internal validation. Important prognostic variables for LOS included greater patient age, comorbidity, type of surgical approach, intraoperative respiratory minute volume and complications occurring within 72 hours in the ICU. The potential gain of the best model in day-to-day clinical practice was determined relative to mean LOS. Use of the model reduced the deficit number (underestimation) of ICU days by 65 and increased the excess number (overestimation) of ICU days by 23 for the cohort of 65 patients. A conservative analysis conducted in the second, prospective cohort of patients revealed that 7% more oesophagectomies could have been accommodated, and 15% of cancelled procedures could have been prevented.ConclusionPatient characteristics can be used to create models that will help in predicting LOS in the ICU. This will result in more efficient use of ICU beds and fewer cancellations.
| S-EPMC2206463 | biostudies-literature
Project description:Machine learning can predict outcomes and determine variables contributing to precise prediction, and can thus classify patients with different risk factors of outcomes. This study aimed to investigate the predictive accuracy for mortality and length of stay in intensive care unit (ICU) patients using machine learning, and to identify the variables contributing to the precise prediction or classification of patients. Patients (n = 12,747) admitted to the ICU at Chiba University Hospital were randomly assigned to the training and test cohorts. After learning using the variables on admission in the training cohort, the area under the curve (AUC) was analyzed in the test cohort to evaluate the predictive accuracy of the supervised machine learning classifiers, including random forest (RF) for outcomes (primary outcome, mortality; secondary outcome, length of ICU stay). The rank of the variables that contributed to the machine learning prediction was confirmed, and cluster analysis of the patients with risk factors of mortality was performed to identify the important variables associated with patient outcomes. Machine learning using RF revealed a high predictive value for mortality, with an AUC of 0.945 (95% confidence interval [CI] 0.922-0.977). In addition, RF showed high predictive value for short and long ICU stays, with AUCs of 0.881 (95% CI 0.876-0.908) and 0.889 (95% CI 0.849-0.936), respectively. Lactate dehydrogenase (LDH) was identified as a variable contributing to the precise prediction in machine learning for both mortality and length of ICU stay. LDH was also identified as a contributing variable to classify patients into sub-populations based on different risk factors of mortality. The machine learning algorithm could predict mortality and length of stay in ICU patients with high accuracy. LDH was identified as a contributing variable in mortality and length of ICU stay prediction and could be used to classify patients based on mortality risk.
| S-EPMC9334583 | biostudies-literature
Project description: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.
| S-EPMC8295642 | biostudies-literature
Project description:BackgroundPostoperative length of stay is a key indicator in the management of medical resources and an indirect predictor of the incidence of surgical complications and the degree of recovery of the patient after cancer surgery. Recently, machine learning has been used to predict complex medical outcomes, such as prolonged length of hospital stay, using extensive medical information.ObjectiveThe objective of this study was to develop a prediction model for prolonged length of stay after cancer surgery using a machine learning approach.MethodsIn our retrospective study, electronic health records (EHRs) from 42,751 patients who underwent primary surgery for 17 types of cancer between January 1, 2000, and December 31, 2017, were sourced from a single cancer center. The EHRs included numerous variables such as surgical factors, cancer factors, underlying diseases, functional laboratory assessments, general assessments, medications, and social factors. To predict prolonged length of stay after cancer surgery, we employed extreme gradient boosting classifier, multilayer perceptron, and logistic regression models. Prolonged postoperative length of stay for cancer was defined as bed-days of the group of patients who accounted for the top 50% of the distribution of bed-days by cancer type.ResultsIn the prediction of prolonged length of stay after cancer surgery, extreme gradient boosting classifier models demonstrated excellent performance for kidney and bladder cancer surgeries (area under the receiver operating characteristic curve [AUC] >0.85). A moderate performance (AUC 0.70-0.85) was observed for stomach, breast, colon, thyroid, prostate, cervix uteri, corpus uteri, and oral cancers. For stomach, breast, colon, thyroid, and lung cancers, with more than 4000 cases each, the extreme gradient boosting classifier model showed slightly better performance than the logistic regression model, although the logistic regression model also performed adequately. We identified risk variables for the prediction of prolonged postoperative length of stay for each type of cancer, and the importance of the variables differed depending on the cancer type. After we added operative time to the models trained on preoperative factors, the models generally outperformed the corresponding models using only preoperative variables.ConclusionsA machine learning approach using EHRs may improve the prediction of prolonged length of hospital stay after primary cancer surgery. This algorithm may help to provide a more effective allocation of medical resources in cancer surgery.
| S-EPMC7939945 | biostudies-literature