A novel nonparametric time-dependent precision-recall curve estimator for right-censored survival data.

In order to assess prognostic risk for individuals in precision health research, risk prediction models are increasingly used, in which statistical models are used to estimate the risk of future outcomes based on clinical and nonclinical characteristics. The predictive accuracy of a risk score must be assessed before it can be used in routine clinical decision making, where the receiver operator characteristic curves, precision-recall curves, and their corresponding area under the curves are commonly used metrics to evaluate the discriminatory ability of a continuous risk score. Among these the precision-recall curves have been shown to be more informative when dealing with unbalanced biomarker distribution between classes, which is common in rare event, even though except one, all existing methods are proposed for classic uncensored data. This paper is therefore to propose a novel nonparametric estimation approach for the time-dependent precision-recall curve and its associated area under the curve for right-censored data. A simulation is conducted to show the better finite sample property of the proposed estimator over the existing method and a real-world data from primary biliary cirrhosis trial is used to demonstrate the practical applicability of the proposed estimator.

Joint Bayesian modeling of time to malaria and mosquito abundance in Ethiopia.

BACKGROUND: This paper studies the effect of mosquito abundance and malaria incidence in the last 3 weeks, and their interaction, on the hazard of time to malaria in a previously studied cohort of children in Ethiopia. METHODS: We model the mosquito abundance and time to malaria data jointly in a Bayesian framework. RESULTS: We found that the interaction of mosquito abundance and incidence plays a prominent role on malaria risk. We quantify and compare relative risks of various factors, and determine the predominant role of the interaction between incidence and mosquito abundance in describing malaria risk. Seasonal rain patterns, distance to a water source of the households, temperature and relative humidity are all significant in explaining mosquito abundance, and through this affect malaria risk. CONCLUSION: Analyzing jointly the time to malaria data and the mosquito abundance allows a precise comparison of factors affecting the spread of malaria. The effect of the interaction between mosquito abundances and local presence of malaria parasites has an important effect on the hazard of time to malaria, beyond abundance alone. Each additional one km away from the dam gives an average reduction of malaria relative risk of 5.7%. The importance of the interaction between abundance and incidence leads to the hypothesis that preventive intervention could advantageously target the infectious population, in addition to mosquito control, which is the typical intervention today.

Risk Factors for Mortality among Adult HIV/AIDS Patients Following Antiretroviral Therapy in Southwestern Ethiopia: An Assessment through Survival Models.

Introduction: Efforts have been made to reduce HIV/AIDS-related mortality by delivering antiretroviral therapy (ART) treatment. However, HIV patients in resource-poor settings are still dying, even if they are on ART treatment. This study aimed to explore the factors associated with HIV/AIDS-related mortality in Southwestern Ethiopia. Method: A non-concurrent retrospective cohort study which collected data from the clinical records of adult HIV/AIDS patients, who initiated ART treatment and were followed between January 2006 and December 2010, was conducted, to explore the factors associated with HIV/AIDS-related mortality at Jimma University Specialized Hospital (JUSH). Survival times (i.e., the time from the onset of ART treatment to the death or censoring) and different characteristics of patients were retrospectively examined. A best-fit model was chosen for the survival data, after the comparison between native semi-parametric Cox regression and parametric survival models (i.e., exponential, Weibull, and log-logistic). Result: A total of 456 HIV patients were included in the study, mostly females (312, 68.4%), with a median age of 30 years (inter-quartile range (IQR): 23-37 years). Estimated follow-up until December 2010 accounted for 1245 person-years at risk (PYAR) and resulted in 66 (14.5%) deaths and 390 censored individuals, representing a median survival time of 34.0 months ( IQR: 22.8-42.0 months). The overall mortality rate was 5.3/100 PYAR: 6.5/100 PYAR for males and 4.8/100 PYAR for females. The Weibull survival model was the best model for fitting the data (lowest AIC). The main factors associated with mortality were: baseline age (>35 years old, AHR = 3.8, 95% CI: 1.6-9.1), baseline weight (AHR = 0.93, 95% CI: 0.90-0.97), baseline WHO stage IV (AHR = 6.2, 95% CI: 2.2-14.2), and low adherence to ART treatment (AHR = 4.2, 95% CI: 2.5-7.1). Conclusion: An effective reduction in HIV/AIDS mortality could be achieved through timely ART treatment onset and maintaining high levels of treatment adherence.

Identification of different malaria patterns due to Plasmodium falciparum and Plasmodium vivax in Ethiopian children: a prospective cohort study.

BACKGROUND: The identification of epidemiological pattern of infection with Plasmodium falciparum and Plasmodium vivax in malaria-endemic area, where multiple episodes are common, is important for intervention programmes. METHODS: A longitudinal cohort study based on weekly house-to-house visits was conducted between July 2008 and June 2010 in 2040 children less than 10 years of age, living nearby the Gilgel-Gibe hydroelectric power dam reservoir in order to determine factors associated with increased P. vivax and P. falciparum incidence. Two types of multivariate frailty models were applied (using time-to-first malaria episode data and time-to-recurrent malaria episode data), allowing the estimation of adjusted hazard ratios (AHR) of potential risk factors (gender, age, proximity to the dam reservoir, and season) for species-specific malaria incidence. RESULTS: Of 2040 children in 96 weeks of follow up, 864 children experienced at least one malaria episode: 685 due to P. falciparum in 548 children, and 385 due to P. vivax in 316 children. Plasmodium vivax and P. falciparum malaria incidence rates were 8.2 (95 % CI: 7.3-9.1) and 14.6 (95 % CI: 13.4-15.6) per 1000 children per month, respectively. According to the time-to-recurrent event models, children aged ≥7 years had a lower risk of presenting P. vivax episodes (AHR = 0.6; 95 % CI: 0.4-0.9), but a higher risk of P. falciparum episodes, when compared with children under ≤3 years (AHR = 1.2; 95 % CI: 1.1-1.6). In addition, P. vivax (AHR = 2.7; 95 % CI: 2.2-3.5) and P. falciparum (AHR = 16.9; 95 % CI: 14.3-20.2) episodes were respectively 2.7 and 16.9 times more frequent in the dry season than in the long rainy season. CONCLUSIONS: The analysis of all malaria episodes (first and recurrent episodes) in the malaria cohort suggests different species-specific patterns of malaria disease in children, with mild seasonality in the incidence of P. vivax episodes mostly observed in younger age groups, and with marked seasonality in the incidence of P. falciparum episodes mainly seen in older children.