Study Design

Research methods need to ensure that valid and reliable answers are retrieved.

Different study designs are required to answer different research questions. Study designs may also reflect historical preferences for ways of doing research within different disciplines. The choice of study design is often affected by pragmatic issues such as funding or patient considerations (eg, pool of potential participants available).

Qualitative research is a naturalistic interpretive approach that seeks to describe and explain how and why people act and make decisions, and how they understand their situation. It may involve transcribing and analysing interviews, or participant observation methodologies.

Quantitative research seeks to identify factors or relationships in a sample that can be assumed to be true of the population from which the sample was drawn. Frequently a study will use mixed methods - that is, a combination of qualitative and quantitative research.

Within these two broad categories, there are many different types of studies. Each has advantages and disadvantages and has its own unique potential sources of bias. One study design is not in itself better than any other. It depends on the question being asked: some studies explore the natural history of a phenomenon or experience, or describe its impact from the perspective of patients or caregivers; others look for protective and risk factors, or causes; some test a new treatment or service to see if it improves the outcomes or experiences of patients; while others try to answer questions about patterns of practice, or of service use, or of adverse effects, and so on.

  • Descriptive studies are used to describe the frequency or possible determinants of a condition, or the experience of a condition and/or its natural history.
  • Cross-sectional studies are a snapshot at a particular time looking at the presence or absence of a disease or symptom or disease and the presence or absence of an exposure.
  • Cohort studies look forward to see if there is a correlation between exposure and an outcome.
  • Case control studies look backward to see if an outcome was preceded by a common event such as a disease, symptom or exposure to a material.
  • Non-randomised and randomised trials look forward but also compare the effects from a common starting point for two groups.
    • In a non-randomised group, investigators assign participants to a group whereas in a randomised study, participant allocation is by chance. Randomised allocation reduces the likelihood of a systematic variable affecting the outcome.
    • Randomised studies need to be carefully designed so that neither the researchers nor the participants know what treatment the patient is getting - this is called blinding (double blinding when neither knows).
    • Studies in which participants know which treatment they are receiving are called open label studies. Sometimes this is the only method possible, however the results can be very strongly affected by the “placebo effect” - in which patients respond positively to a treatment that they think will be of benefit to them.
    • Placebo controlled trials compare an intervention with a placebo (inactive comparator) so that it is possible to identify whether the intervention is responsible for the outcomes, or whether it is the placebo effect or the natural history of the condition being studied which produces the change.
      This is important when a condition fluctuates, for instance pain or delirium. When it is ethically feasible, a placebo controlled study of treatments or interventions produces the most reliable conclusions. However, at times it is more ethical and informative to patient care to compare a new treatment with current best practice rather than a placebo.
  • Other rigorous study designs that are particularly valuable for some situations in palliative care include n of 1 trials and crossover studies, where a patient tries one treatment and then another or a treatment and then a placebo or vice-versa.
  • Some questions relate not to individuals but to populations.
    • Research projects related to these questions may need to access routinely collected information such as cancer registries, census data, and other health related databases.
    • Some research programs might involve establishing a dataset to use for researching particular questions, for example longitudinal studies which follow a group of individuals over a long period of time, disease registries which collect all the cases of a particular disease, or repeated market research surveys which may be able to estimate the prevalence of a particular problem, need, or experience at a population level and over time.

To assess new therapeutic or pharmacological interventions, randomised controlled trials are commonly regarded as the gold standard, particularly where they measure against a control of current practice. The National Health and Medical Research Council (NHMRC) have developed a definition of levels of evidence that matches them to types of study design and research questions. [1]

Research in palliative care may also involve the study of complex interventions such as a new model of service delivery or the implementation of a new therapy in the community setting across multiple providers. Complex interventions are those in which there are multiple, interacting components and where context is likely to have a significant influence. The Medical Research Council (UK) has produced A Framework for development and evaluation of RCTs for Complex Interventions to Improve Health. (386kb pdf).

Last updated 20 January 2017