RESUMO
Support vector regression (SVR) is particularly beneficial when the outcome and predictors are nonlinearly related. However, when many covariates are available, the method's flexibility can lead to overfitting and an overall loss in predictive accuracy. To overcome this drawback, we develop a feature selection method for SVR based on a genetic algorithm that iteratively searches across potential subsets of covariates to find those that yield the best performance according to a user-defined fitness function. We evaluate the performance of our feature selection method for SVR, comparing it to alternate methods including LASSO and random forest, in a simulation study. We find that our method yields higher predictive accuracy than SVR without feature selection. Our method outperforms LASSO when the relationship between covariates and outcome is nonlinear. Random forest performs equivalently to our method in some scenarios, but more poorly when covariates are correlated. We apply our method to predict donor kidney function 1 year after transplant using data from the United Network for Organ Sharing national registry.
Assuntos
Algoritmos , Análise de Regressão , Humanos , Máquina de Vetores de SuporteRESUMO
The net reclassification improvement (NRI) is a widely used metric used to assess the relative ability of 2 risk models to distinguish between low- and high-risk individuals. However, the validity and usefulness of the NRI have been questioned. Criticism of the NRI focuses on its use comparing nested risk models, whereas in practice it is often used to compare nonnested risk models derived from distinct data sources. In this study, we evaluated the performance of the NRI in a nonnested context by using it to compare competing cardiovascular risk-prediction models. We explored the NRI's sensitivity to variations in risk categories and to the calibration of the compared models. We found that the NRI was very sensitive to changes in the definition of risk categories, especially when at least 1 model was miscalibrated. To address these shortcomings, we describe a novel alternative to the usual NRI that uses percentiles of risk instead of cutoffs based on absolute risk. This percentile-based NRI demonstrates the relative ability of 2 models to rank patient risk. It displays more stable behavior, and we recommend its use when there are no established risk categories or when models are miscalibrated.
Assuntos
Medição de Risco/métodos , Registros Eletrônicos de Saúde , Métodos EpidemiológicosRESUMO
Background. To support proactive decision making during the COVID-19 pandemic, mathematical models have been leveraged to identify surveillance indicator thresholds at which strengthening nonpharmaceutical interventions (NPIs) is necessary to protect health care capacity. Understanding tradeoffs between different adaptive COVID-19 response components is important when designing strategies that balance public preference and public health goals. Methods. We considered 3 components of an adaptive COVID-19 response: 1) the threshold at which to implement the NPI, 2) the time needed to implement the NPI, and 3) the effectiveness of the NPI. Using a compartmental model of SARS-CoV-2 transmission calibrated to Minnesota state data, we evaluated different adaptive policies in terms of the peak number of hospitalizations and the time spent with the NPI in force. Scenarios were compared with a reference strategy, in which an NPI with an 80% contact reduction was triggered when new weekly hospitalizations surpassed 8 per 100,000 population, with a 7-day implementation period. Assumptions were varied in sensitivity analysis. Results. All adaptive response scenarios substantially reduced peak hospitalizations relative to no response. Among adaptive response scenarios, slower NPI implementation resulted in somewhat higher peak hospitalization and a longer time spent under the NPIs than the reference scenario. A stronger NPI response resulted in slightly less time with the NPIs in place and smaller hospitalization peak. A higher trigger threshold resulted in greater peak hospitalizations with little reduction in the length of time under the NPIs. Conclusions. An adaptive NPI response can substantially reduce infection circulation and prevent health care capacity from being exceeded. However, population preferences as well as the feasibility and timeliness of compliance with reenacting NPIs should inform response design. Highlights: This study uses a mathematical model to compare different adaptive nonpharmaceutical intervention (NPI) strategies for COVID-19 management across 3 dimensions: threshold when the NPI should be implemented, time it takes to implement the NPI, and the effectiveness of the NPI.All adaptive NPI response scenarios considered substantially reduced peak hospitalizations compared with no response.Slower NPI implementation results in a somewhat higher peak hospitalization and longer time spent with the NPI in place but may make an adaptive strategy more feasible by allowing the population sufficient time to prepare for changing restrictions.A stronger, more effective NPI response results in a modest reduction in the time spent under the NPIs and slightly lower peak hospitalizations.A higher threshold for triggering the NPI delays the time at which the NPI starts but results in a higher peak hospitalization and does not substantially reduce the time the NPI remains in force.
RESUMO
Objective: To assess the frequency of preventative COVID-19 behaviors and vaccination willingness among United States (US) college and university students during the COVID-19 pandemic. Participants: Participants (N = 653) were ≥18 years old and students at institutions for higher education in the US in March 2020. Methods: Students self-reported preventative behaviors, willingness to be vaccinated, and social contact patterns during four waves of online surveys from May-August 2020. Results: Student engagement in preventative behaviors was generally high. The majority of students intended to be vaccinated (81.5%). Overall, there were no significant differences in the proportion adopting preventative behaviors or in willingness to be vaccinated by sex or geographic location. The most common reason for willingness to get vaccinated was wanting to contribute to ending COVID-19 outbreaks (44.7%). Conclusions: Early in the pandemic, college students primarily reported willingness to vaccinate and adherence to preventative behaviors. Outreach strategies are needed to continue this momentum.
RESUMO
After an outbreak of meningococcal B (MenB) disease at a university, we surveyed students regarding their vaccination status 2 months and 20 months after campus-led vaccination campaigns and compared students' self-report to vaccination records. Nearly all participants accurately reported the number of vaccine doses at both visits. Among those who received two doses of the vaccine, accurate recall of the timing of MenB vaccination was 85.7% (95% CI: 82.7-88.6) in the short term and 62.9% (95% CI: 56.0-69.8) in the long term. After the outbreak, only one-third reported feeling 'very confident' in their MenB disease and vaccine knowledge. Our findings suggest that the validity of self-reported vaccination status among university students in an outbreak setting is high, but that if the duration of protection is unknown and additional doses of vaccine may be needed, documented vaccination records may be preferred over self-report to assess timing of vaccine receipt.