Using a proportional hazards model to determine the effect of covariates on pooled data — A case study

Pooled data, such as that found in shared databases increase aspects of data quality such as data scarcity but can also impact other aspects of data quality. One of these is due to pooled data not being sufficiently representative to the unique circumstances of that user, resulting in a loss of relevance to that user. In order to restore relevance of pooled data a proportional hazards model is used to model the effects of covariates. We include a case study that cleanses data from an organisation to be fit for purpose and fits a proportional hazards model to the cleansed data. A shared database for reliability data for the mining industry is being developed that is a future candidate for the use and extension of a proportional hazards model.