Clinical practice should be based on the best available evidence. Ideally such evidence is obtained through rigorously conducted, purpose-designed clinical studies such as randomised controlled trials and prospective cohort studies. However gathering information in this way requires a massive effort, can be prohibitively expensive, is time consuming, and may not always be ethical or practicable. When answers are needed urgently and purpose-designed prospective studies are not feasible, retrospective healthcare data may offer the best evidence there is. But can we rely on analysis with such data to give us meaningful answers? The current thesis studies this question through analysis with repeated psychological symptom screening data that were routinely collected from over 20,000 outpatients who attended selected oncology clinics in Scotland. Linked to patients’ oncology records these data offer a unique opportunity to study the progress of distress symptoms on an unprecedented scale in this population. However, the limitations to such routinely collected observational healthcare data are many. We approach the analysis within a missing data context and develop a Bayesian model in WinBUGS to estimate the posterior predictive distribution for the incomplete longitudinal response and covariate data under both Missing At Random and Missing Not At Random mechanisms and use this model to generate multiply imputed datasets for further frequentist analysis. Additional to the routinely collected screening data we also present a purpose-designed, prospective cohort study of distress symptoms in the same cancer outpatient population. This study collected distress outcome scores from enrolled patients at regular intervals and with very little missing data. Consequently it contained many of the features that were lacking in the routinely collected screening data and provided a useful contrast, offering an insight into how the screening data might have been were it not for the limitations. We evaluate the extent to which it was possible to reproduce the clinical study results with the analysis of the observational screening data. Lastly, using the modelling strategy previously developed we analyse the abundant screening data to estimate the prevalence of depression in a cancer outpatient population and the associations with demographic and clinical characteristics, thereby addressing important clinical research questions that have not been adequately studied elsewhere. The thesis concludes that analysis with observational healthcare data can potentially be advanced considerably with the use of flexible and innovative modelling techniques now made practicable with modern computing power.