Dataset Information

Predicting Prolonged Length of Hospital Stay for Peritoneal Dialysis-Treated Patients Using Stacked Generalization: Model Development and Validation Study.

ABSTRACT:

Background

The increasing number of patients treated with peritoneal dialysis (PD) and their consistently high rate of hospital admissions have placed a large burden on the health care system. Early clinical interventions and optimal management of patients at a high risk of prolonged length of stay (pLOS) may help improve the medical efficiency and prognosis of PD-treated patients. If timely clinical interventions are not provided, patients at a high risk of pLOS may face a poor prognosis and high medical expenses, which will also be a burden on hospitals. Therefore, physicians need an effective pLOS prediction model for PD-treated patients.

Objective

This study aimed to develop an optimal data-driven model for predicting the pLOS risk of PD-treated patients using basic admission data.

Methods

Patient data collected using the Hospital Quality Monitoring System (HQMS) in China were used to develop pLOS prediction models. A stacking model was constructed with support vector machine, random forest (RF), and K-nearest neighbor algorithms as its base models and traditional logistic regression (LR) as its meta-model. The meta-model used the outputs of all 3 base models as input and generated the output of the stacking model. Another LR-based pLOS prediction model was built as the benchmark model. The prediction performance of the stacking model was compared with that of its base models and the benchmark model. Five-fold cross-validation was employed to develop and validate the models. Performance measures included the Brier score, area under the receiver operating characteristic curve (AUROC), estimated calibration index (ECI), accuracy, sensitivity, specificity, and geometric mean (Gm). In addition, a calibration plot was employed to visually demonstrate the calibration power of each model.

Results

The final cohort extracted from the HQMS database consisted of 23,992 eligible PD-treated patients, among whom 30.3% had a pLOS (ie, longer than the average LOS, which was 16 days in our study). Among the models, the stacking model achieved the best calibration (ECI 8.691), balanced accuracy (Gm 0.690), accuracy (0.695), and specificity (0.701). Meanwhile, the stacking and RF models had the best overall performance (Brier score 0.174 for both) and discrimination (AUROC 0.757 for the stacking model and 0.756 for the RF model). Compared with the benchmark LR model, the stacking model was superior in all performance measures except sensitivity, but there was no significant difference in sensitivity between the 2 models. The 2-sided t tests revealed significant performance differences between the stacking and LR models in overall performance, discrimination, calibration, balanced accuracy, and accuracy.

Conclusions

This study is the first to develop data-driven pLOS prediction models for PD-treated patients using basic admission data from a national database. The results indicate the feasibility of utilizing a stacking-based pLOS prediction model for PD-treated patients. The pLOS prediction tools developed in this study have the potential to assist clinicians in identifying patients at a high risk of pLOS and to allocate resources optimally for PD-treated patients.

SUBMITTER: Kong G

PROVIDER: S-EPMC8173398 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundRadical prostatectomy (RP) is a leading treatment option for localised prostate cancer. Although hospital in-patient stays accounts for much of the costs of treatment, little is known about population-level trends in length-of-stay (LOS). We investigated factors predicting hospital LOS and readmissions in men who had RP following prostate cancer.MethodsIncident prostate cancers (ICD-O3: C61), diagnosed January 2002-December 2008 in men < 70 years, were identified from the Irish Cancer Registry, and linked to public hospital episodes. For those who had RP (ICD-9 CM procedure codes 60.3, 60.4, 60.5, 60.62) the associated hospital episode was identified. LOS was calculated as the number of days from date of admission to date of discharge. Patient-, tumour-, and health service-related factors predicting longer LOS (upper quartile, >9 days) were investigated using logistic regression. Patterns in day-case and in-patient readmissions within 28 days of discharge following RP were explored.ResultsOver the study period 9096 prostate cancers were diagnosed in men under 70, 26.5% of whom had RP by end of follow-up 31/12/2009. Two of eight public hospitals and eight of forty surgeons carried out 50% of all public-service RPs. Median LOS was 8 days (10th-90th percentile = 6-13 days) and fell significantly over time (2002, 9 days; 2008, 7 days; p < 0.001). In adjusted analyses men who were not married (OR = 1.71, 95% CI 1.25-2.34), had co-morbidities (OR = 1.64, 95% CI 1.25-2.16) or stage III-IV cancer (OR = 2.19, 95% CI 1.44-3.34) were significantly more likely to have prolonged LOS. Those treated in higher volume hospitals (annual median >49 RPs) or by higher volume surgeons (annual median >17 RPs) were significantly less likely to have prolonged LOS (OR = 0.34, 95% CI 0.26-0.45; OR = 0.55, 95% CI 0.42-0.71 respectively).ConclusionMedian LOS after RP decreased between 2002 and 2008 in Ireland but it remains higher than in both England and the US. Although volumes of RPs conducted in Ireland are low, there is considerable variation between hospitals and surgeons. Hospital and surgeon volume were strong predictors of shorter LOS, after adjusting for other variables. These factors point to a need for a comprehensive review of prostate cancer service provision.

Project description:BackgroundThe epidemic of Coronavirus Disease 2019 (COVID-19) has become a global health emergency, but the clinical characteristics of COVID-19 are not fully described. We aimed to describe the clinical characteristics of COVID-19 outside of Wuhan city; and to develop a multivariate model to predict the risk of prolonged length of stay in hospital (ProLOS).MethodsThe study was conducted in a tertiary care hospital in Zhejiang province from January to February 20, 2020. Medical records of all confirmed cases of COVID-19 were retrospectively reviewed. Patients were categorized into the ProLOS and non-ProLOS groups by hospital length of stay greater and less than 14 days, respectively. Conventional descriptive statistics were applied. Multivariate regression model was built to predict the risk of ProLOS, with variables selected using stepwise approach.ResultsA total of 75 patients with confirmed COVID-19 were included for quantitative analysis, including 25 (33%) patients in the ProLOS group. ProLOS patients were more likely to have history of traveling to Wuhan (68% vs. 28%; P=0.002). Patients in the ProLOS group showed lower neutrophil counts [median (IQR): 2.50 (1.77–3.23) ×109/L vs. 2.90 (2.21–4.19) ×109/L; P=0.048], higher partial thrombin time (PT) (13.42±0.63 vs. 13.10±0.48 s; P=0.029), lower D-Dimer [0.26 (0.22–0.46) vs. 0.44 (0.32–0.84) mg/L; P=0.012]. There was no patient died and no severe case in our cohort. The overall LOS was 11 days (IQR, 5–15 days). The median cost for a hospital stay was 7,388.19 RMB (IQR, 5,085.39–11,145.44). The prediction model included five variables of procalcitonin, heart rate, epidemiological history, lymphocyte count and cough. The discrimination of the model was 84.8% (95% CI: 75.3% to 94.4%).ConclusionsOur study described clinical characteristics of COVID-19 outside of Wuhan city and found that the illness was less severe than that in the core epidemic region. A multivariate model was developed to predict ProLOS, which showed good discrimination.

Project description:BackgroundThe Blue Book (2005), recommended guidelines for patients care with fragility fractures. Together with introduction of a National Hip Fracture Database Audit and Best Practice Tariff model to financially incentivise hospitals by payment of a supplement for patients whose care satisfied six clinical standards), have improved hip fracture after-care. However, there is a lack of data-driven evidence to support its effectiveness. We aimed to verify the impact of an orthogeriatric service on hospital length of stay (LOS)-duration from admission to discharge.MethodsWe conducted a repeated cross-sectional study over a 10 year period of older individuals aged ≥ 60 years admitted with hip fractures to a hospital.ResultsAltogether 2798 patients, 741 men and 2057 women (respective mean ages; 80.5 ± 10.6 and 83.2 ± 8.9 years) were admitted from their own homes with a hip fracture and survived to discharge. Compared to 2009-2014, LOS during 2015-2019, when the orthogeriatric service was fully implemented, was shorter for all discharge destinations: 10.4 vs 17.5 days (P < 0.001). Each discharge destination showed reductions: back to own homes, 9.7 vs 17.7 days (P < 0.001); to rehabilitation units: 10.8 vs 13.1 days (P < 0.001); to residential care: 15.4 vs 26.2 days (P = 0.001); or nursing care, 24.4 vs 53.1 days (P < 0.001). During 2009-2014, the risk of staying > 3 weeks in hospital was greater by six-fold and pressure ulcers by three-fold. The number of bed days for every thousand patients per year was also shortened during 2015-2019 by: 1665 days for discharge back to own homes; 469 days with transfer to rehabilitation units; 1258 days for discharge to residential care, and 5465 days to nursing care. Estimated annual savings (2017 costs) per thousand patients after complete establishment of the service was about £2.7 m.ConclusionsImplementation of an orthogeriatric service generated significant reductions in hospital LOS for all patients, with associated cost-savings, especially for those discharged to nursing care.