ABSTRACT
Survival data with a multi-state structure are frequently observed in follow-up studies. An analytic approach based on a multi-state model (MSM) should be used in longitudinal health studies in which a patient experiences a sequence of clinical progression events. One main objective in the MSM framework is variable selection, where attempts are made to identify the risk factors associated with the transition hazard rates or probabilities of disease progression. The usual variable selection methods, including stepwise and penalized methods, do not provide information about the importance of variables. In this context, we present a two-step algorithm to evaluate the importance of variables formulti-state data. Three differentmachine learning approaches (randomforest, gradient boosting, and neural network) as themost widely usedmethods are considered to estimate the variable importance in order to identify the factors affecting disease progression and rank these factors according to their importance. The performance of our proposed methods is validated by simulation and applied to the COVID-19 data set. The results revealed that the proposed two-stage method has promising performance for estimating variable importance.
ABSTRACT
Survival data with a multi-state structure are frequently observed in follow-up studies. An analytic approach based on a multi-state model (MSM) should be used in longitudinal health studies in which a patient experiences a sequence of clinical progression events. One main objective in the MSM framework is variable selection, where attempts are made to identify the risk factors associated with the transition hazard rates or probabilities of disease progression. The usual variable selection methods, including stepwise and penalized methods, do not provide information about the importance of variables. In this context, we present a two-step algorithm to evaluate the importance of variables for multi-state data. Three different machine learning approaches (random forest, gradient boosting, and neural network) as the most widely used methods are considered to estimate the variable importance in order to identify the factors affecting disease progression and rank these factors according to their importance. The performance of our proposed methods is validated by simulation and applied to the COVID-19 data set. The results revealed that the proposed two-stage method has promising performance for estimating variable importance. © 2023 Tech Science Press. All rights reserved.
ABSTRACT
Multi-state models are used to describe how individuals transition through different states over time. The distribution of the time spent in different states, referred to as 'length of stay', is often of interest. Methods for estimating expected length of stay in a given state are well established. The focus of this paper is on the distribution of the time spent in different states conditional on the complete pathway taken through the states, which we call 'conditional length of stay'. This work is motivated by questions about length of stay in hospital wards and intensive care units among patients hospitalised due to Covid-19. Conditional length of stay estimates are useful as a way of summarising individuals' transitions through the multi-state model, and also as inputs to mathematical models used in planning hospital capacity requirements. We describe non-parametric methods for estimating conditional length of stay distributions in a multi-state model in the presence of censoring, including conditional expected length of stay (CELOS). Methods are described for an illness-death model and then for the more complex motivating example. The methods are assessed using a simulation study and shown to give unbiased estimates of CELOS, whereas naive estimates of CELOS based on empirical averages are biased in the presence of censoring. The methods are applied to estimate conditional length of stay distributions for individuals hospitalised due to Covid-19 in the UK, using data on 42,980 individuals hospitalised from March to July 2020 from the COVID19 Clinical Information Network.
Subject(s)
COVID-19 , Humans , Models, Theoretical , Length of Stay , Computer Simulation , Intensive Care UnitsABSTRACT
BACKGROUND: The distribution of the duration that clinical cases of COVID-19 occupy hospital beds (the 'length of stay') is a key factor in determining how incident caseloads translate into health system burden. Robust estimation of length of stay in real-time requires the use of survival methods that can account for right-censoring induced by yet unobserved events in patient progression (e.g. discharge, death). In this study, we estimate in real-time the length of stay distributions of hospitalised COVID-19 cases in New South Wales, Australia, comparing estimates between a period where Delta was the dominant variant and a subsequent period where Omicron was dominant. METHODS: Using data on the hospital stays of 19,574 individuals who tested positive to COVID-19 prior to admission, we performed a competing-risk survival analysis of COVID-19 clinical progression. RESULTS: During the mixed Omicron-Delta epidemic, we found that the mean length of stay for individuals who were discharged directly from ward without an ICU stay was, for age groups 0-39, 40-69 and 70 +, respectively, 2.16 (95% CI: 2.12-2.21), 3.93 (95% CI: 3.78-4.07) and 7.61 days (95% CI: 7.31-8.01), compared to 3.60 (95% CI: 3.48-3.81), 5.78 (95% CI: 5.59-5.99) and 12.31 days (95% CI: 11.75-12.95) across the preceding Delta epidemic (1 July 2021-15 December 2021). We also considered data on the stays of individuals within the Hunter New England Local Health District, where it was reported that Omicron was the only circulating variant, and found mean ward-to-discharge length of stays of 2.05 (95% CI: 1.80-2.30), 2.92 (95% CI: 2.50-3.67) and 6.02 days (95% CI: 4.91-7.01) for the same age groups. CONCLUSIONS: Hospital length of stay was substantially reduced across all clinical pathways during a mixed Omicron-Delta epidemic compared to a prior Delta epidemic, contributing to a lessened health system burden despite a greatly increased infection burden. Our results demonstrate the utility of survival analysis in producing real-time estimates of hospital length of stay for assisting in situational assessment and planning of the COVID-19 response.
Subject(s)
COVID-19 , SARS-CoV-2 , Humans , New South Wales/epidemiology , COVID-19/epidemiology , Australia , HospitalsABSTRACT
Using available phylogeographical data of 3585 SARS–CoV–2 genomes we attempt at providing a global picture of the virus's dynamics in terms of directly interpretable parameters. To this end we fit a hidden state multistate speciation and extinction model to a pre-estimated phylogenetic tree with information on the place of sampling of each strain. We find that even with such coarse–grained data the dominating transition rates exhibit weak similarities with the most popular, continent–level aggregated, airline passenger flight routes. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
ABSTRACT
BACKGROUND: Pulmonary embolism (PE) without overt deep vein thrombosis (DVT) was common in hospitalized coronavirus-induced disease (COVID)-19 patients and represented a diagnostic, prognostic, and therapeutic challenge. The aim of this study was to analyze the prognostic role of PE on mortality and the preventive effect of heparin on PE and mortality in unvaccinated COVID-19 patients without overt DVT. METHODS: Data from 401 unvaccinated patients (age 68 ± 13 years, 33% females) consecutively admitted to the intensive care unit or the medical ward were included in a retrospective longitudinal study. PE was documented by computed tomography scan and DVT by compressive venous ultrasound. The effect of PE diagnosis and any heparin use on in-hospital death (primary outcome) was analyzed by a classical survival model. The preventive effect of heparin on either PE diagnosis or in-hospital death (secondary outcome) was analyzed by a multi-state model after having reclassified patients who started heparin after PE diagnosis as not treated. RESULTS: Median follow-up time was 8 days (range 1-40 days). PE cumulative incidence and in-hospital mortality were 27% and 20%, respectively. PE was predicted by increased D-dimer levels and COVID-19 severity. Independent predictors of in-hospital death were age (hazards ratio (HR) 1.05, 95% confidence interval (CI) 1.03-1.08, p < 0.001), body mass index (HR 0.93, 95% CI 0.89-0.98, p = 0.004), COVID-19 severity (severe versus mild/moderate HR 3.67, 95% CI 1.30-10.4, p = 0.014, critical versus mild/moderate HR 12.1, 95% CI 4.57-32.2, p < 0.001), active neoplasia (HR 2.58, 95% CI 1.48-4.50, p < 0.001), chronic obstructive pulmonary disease (HR 2.47; 95% CI 1.15-5.27, p = 0.020), respiratory rate (HR 1.06, 95% CI 1.02-1.11, p = 0.008), heart rate (HR 1.03, 95% CI 1.01-1.04, p < 0.001), and any heparin treatment (HR 0.35, 95% CI 0.18-0.67, p = 0.001). In the multi-state model, preventive heparin at prophylactic or intermediate/therapeutic dose, compared with no treatment, reduced PE risk and in-hospital death, but it did not influence mortality of patients with a PE diagnosis. CONCLUSIONS: PE was common during the first waves pandemic in unvaccinated patients, but it was not a negative prognostic factor for in-hospital death. Heparin treatment at any dose prevented mortality independently of PE diagnosis, D-dimer levels, and disease severity.
ABSTRACT
Survival data with a multi-state structure are frequently observed in follow-up studies. An analytic approach based on a multi-state model (MSM) should be used in longitudinal health studies in which a patient experiences a sequence of clinical progression events. One main objective in the MSM framework is variable selection, where attempts are made to identify the risk factors associated with the transition hazard rates or probabilities of disease progression. The usual variable selection methods, including stepwise and penalized methods, do not provide information about the importance of variables. In this context, we present a two-step algorithm to evaluate the importance of variables for multi-state data. Three different machine learning approaches (random forest, gradient boosting, and neural network) as the most widely used methods are considered to estimate the variable importance in order to identify the factors affecting disease progression and rank these factors according to their importance. The performance of our proposed methods is validated by simulation and applied to the COVID-19 data set. The results revealed that the proposed two-stage method has promising performance for estimating variable importance.
ABSTRACT
This article discusses the spread of infectious diseases using the multi-state SVIRS model with the assumption that a discrete-time Markov chain (DTMC) occurs in a closed population that is regularly examined. This article aims to generate transition probabilities, which are then used to predict the number of confirmed cases in the next period. The multi-state SVIRS model uses four states, namely susceptible, vaccinated, infected, and recovered, followed by calculating the probabilities of each transition between states that are different from the compartment model. The model was applied to the COVID-19 data in Indonesia, which was analyzed using the statistical software R. The result showed that the transition probability of a person being infected according to the multistate model with the assumption of DTMC SVIRS on the COVID-19 data was around 25.38% including those with and without vaccination. In comparison, the probability of being recovered was about 92.34%. Then this transition probability was used to predict the confirmed cases of COVID-19 in the next few days. The prediction results were highly accurate with a MAPE value less than 10%. The main contribution of this research is the use of the DTMC assumption, which is a stochastic model in determining the parameters of the differential equation formed by the compartment model and adding the vaccinated state in the model. The vaccinated cases in this article used the proportion of the efficacy of each vaccine used by several susceptible individuals, which, according to WHO recommendations, should be given in two doses. The multi-state model with the assumption of DTMC can model chronic diseases and infectious diseases. This can be seen from the results of the analysis of the COVID-19 data in Indonesia, in which the short-term prediction results had a high level of accuracy. [ FROM AUTHOR] Copyright of Engineering Letters is the property of Newswood Limited and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full . (Copyright applies to all s.)
ABSTRACT
Combating SARS-CoV-2 is the first concern and goal of the whole world faced with the global health crisis. Since 2019, the SARS-CoV-2 infection (COVID-19) and even mutated infection cases have been increasing rapidly. From 2019 through 27 August 2021, a total of 214,468,601 individuals were confirmed cases of SARS-CoV-2, including 4,470,969 death toll. Some of these individuals were able to access treatment and some could not, but for a while there was complete uncertainty. It was not known whether those who accessed treatment were lucky, but treatment was based on trial and error because of this uncertainty around the world until data was collected. Therefore, the aim of this study was to model SARS-CoV-2 infectious disease progression from the date of polymerase chain reaction (PCR) test to the date of negative outcome via Bayesian multi-state model approaches considering risk factors such as gender, age, and antiviral treatment. Data from 746 inpatients were collected from August 1st until the December 1st, 2020. For the multi-state model, five various discrete states were selected according to the Republic of Turkey Ministery of Health treatment algorithm. The results showed that Bayesian multi-state models with the Weibull distributed baseline hazard function were more appropriate models in the presence of risk factors and antiviral treatment.
ABSTRACT
BACKGROUND: Understanding the risk factors associated with hospital burden of COVID-19 is crucial for healthcare planning for any future waves of infection. METHODS: An observational cohort study is performed, using data on all PCR-confirmed cases of COVID-19 in Regione Lombardia, Italy, during the first wave of infection from February-June 2020. A multi-state modelling approach is used to simultaneously estimate risks of progression through hospital to final outcomes of either death or discharge, by pathway (via critical care or not) and the times to final events (lengths of stay). Logistic and time-to-event regressions are used to quantify the association of patient and population characteristics with the risks of hospital outcomes and lengths of stay respectively. RESULTS: Risks of severe outcomes such as ICU admission and mortality have decreased with month of admission (for example, the odds ratio of ICU admission in June vs March is 0.247 [0.120-0.508]) and increased with age (odds ratio of ICU admission in 45-65 vs 65 + age group is 0.286 [0.201-0.406]). Care home residents aged 65 + are associated with increased risk of hospital mortality and decreased risk of ICU admission. Being a healthcare worker appears to have a protective association with mortality risk (odds ratio of ICU mortality is 0.254 [0.143-0.453] relative to non-healthcare workers) and length of stay. Lengths of stay decrease with month of admission for survivors, but do not appear to vary with month for non-survivors. CONCLUSIONS: Improvements in clinical knowledge, treatment, patient and hospital management and public health surveillance, together with the waning of the first wave after the first lockdown, are hypothesised to have contributed to the reduced risks and lengths of stay over time.