Browsing by Author "Mwambi, Henry"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Open Access
    A comparison of the conditional inference survival forest model to random survival forests based on a simulation study as well as on two applications with time-to-event data
    (2017) Nasejje, Justine B; Mwambi, Henry; Sabur, Natasha F; Lesosky, Maia
    Abstract Background Random survival forest (RSF) models have been identified as alternative methods to the Cox proportional hazards model in analysing time-to-event data. These methods, however, have been criticised for the bias that results from favouring covariates with many split-points and hence conditional inference forests for time-to-event data have been suggested. Conditional inference forests (CIF) are known to correct the bias in RSF models by separating the procedure for the best covariate to split on from that of the best split point search for the selected covariate. Methods In this study, we compare the random survival forest model to the conditional inference model (CIF) using twenty-two simulated time-to-event datasets. We also analysed two real time-to-event datasets. The first dataset is based on the survival of children under-five years of age in Uganda and it consists of categorical covariates with most of them having more than two levels (many split-points). The second dataset is based on the survival of patients with extremely drug resistant tuberculosis (XDR TB) which consists of mainly categorical covariates with two levels (few split-points). Results The study findings indicate that the conditional inference forest model is superior to random survival forest models in analysing time-to-event data that consists of covariates with many split-points based on the values of the bootstrap cross-validated estimates for integrated Brier scores. However, conditional inference forests perform comparably similar to random survival forests models in analysing time-to-event data consisting of covariates with fewer split-points. Conclusion Although survival forests are promising methods in analysing time-to-event data, it is important to identify the best forest model for analysis based on the nature of covariates of the dataset in question.
  • Thumbnail Image
    Item
    Open Access
    Diagnostics for joint models for longitudinal and survival data
    (2021) Singini, Isaac Luwinga; Gumedze, Freedom; Mwambi, Henry
    Joint models for longitudinal and survival data are a class of models that jointly analyse an outcome repeatedly observed over time such as a bio-marker and associated event times. These models are useful in two practical applications; firstly focusing on survival outcome whilst accounting for time varying covariates measured with error and secondly focusing on the longitudinal outcome while controlling for informative censoring. Interest on the estimation of these joint models has grown in the past two and half decades. However, minimal effort has been directed towards developing diagnostic assessment tools for these models. The available diagnostic tools have mainly been based on separate analysis of residuals for the longitudinal and survival sub-models which could be sub-optimal. In this thesis we make four contributions towards the body of knowledge. We first developed influence diagnostics for the shared parameter joint model for longitudinal and survival data based on Cook's statistics. We evaluated the performance of the diagnostics using simulation studies under different scenarios. We then illustrated these diagnostics using real data set from a multi-center clinical trial on TB pericarditis (IMPI). The second contribution was to implement a variance shift outlier model (VSOM) in the two-stage joint survival model. This was achieved by identifying outlying subjects in the longitudinal sub-model and down-weighting before the second stage of the joint model. The third contribution was to develop influence diagnostics for the multivariate joint model for longitudinal and survival data. In this setting we considered two longitudinal outcomes, square root CD4 cell count which was Gaussian in nature and antiretroviral therapy (ART) uptake which was binary. We achieved this by extending the univariate case i based on Cook's statistics for all parameters. The fourth contribution was to implement influence diagnostics in joint models for longitudinal and survival data with multiple failure types (competing risk). Using IMPI data set we considered two competing events in the joint model; death and constrictive pericarditis. Using simulation studies and IMPI dataset the developed diagnostics identified influential subjects as well as observations. The performance of the diagnostics was over 98% in simulation studies. We further conducted sensitivity analyses to check the impact of influential subjects and/or observations on parameter estimates by excluding them and re-fitting the joint model. We observed subtle differences, overall in the parameter estimates, which gives confidence that the initial inferences are credible and can be relied on. We illustrated case deletion diagnostics using the IMPI trial setting, these diagnostics can also be applied to clinical trials with similar settings. We therefore make a strong recommendation to analysts to conduct influence diagnostics in the joint model for longitudinal and survival data to ascertain the reliability of parameter estimates. We also recommend the implementation of VSOM in the longitudinal part of the two-stage joint model before the second stage.