Observational studies using routinely collected health data can inform health technology assessments (HTA) and change clinical practice. However, when observational studies are used to investigate comparative treatment effects, a major concern is confounding bias. Careful study design is required to reduce this risk of bias and generate evidence that is useful for national and individual-level clinical decision-making. In this thesis, I aim to advance the use of routinely collected health data to assess comparative effectiveness. I use the target trial emulation (TTE) framework and instrumental variable analyses to reduce the risk of confounding in two case studies in England focused on the relative effectiveness of (1) sulfonylureas (SU), dipeptidyl peptidase-4 inhibitors (DPP4i), and sodium-glucose co-transporter 2 inhibitors (SGLT2i), all added to metformin, as second-line oral antidiabetic treatments for people with type 2 diabetes mellitus, and (2) invasive versus conservative cardiac management for acute myocardial infarction (AMI) among people with kidney impairment. In Case Study 1, I find evidence that SGLT2i are better than the alternatives at improving important clinical measures and reducing the hazards of cardiovascular and kidney outcomes. I exploit exogenous variation in second-line antidiabetic prescribing at the general practice group-level as a preference- based instrument. I also find evidence of treatment heterogeneity for DPP4i versus SU on mean change in haemoglobin A1c. In Case Study 2, I illustrate potential biases when using secondary care datasets to study AMI treatment in people with kidney impairment. I also demonstrate variation in alternative AMI treatment at the cardiology-centre level which could be used in future natural experiments. I conclude that careful study design and analyses helped reduce the risk of bias in these observational studies using routinely collected health data. I demonstrate challenges and opportunities for using these data to inform clinical practice and HTA, and to improve outcomes for patients.