1. Choose an appropriate study design

Research can be conducted using numerous types of study designs. The study design that you choose to conduct your research will depend on a variety of considerations: your research question, study hypothesis, the outcome / disease and exposures / risk factors that you are investigating, and the time and financial resources available.

Generally, studies fall into two types; Experimental or Observational
Experimental (intervention) Study: Experimental Studies involve an active attempt to change a disease determinant – such as an exposure or behaviour – or the progress of a disease through treatment.[4] An example of an experimental study is a randomised controlled trial (RCT). It is the gold standard of intervention studies.

In an RCT, a sample is chosen from the population, baseline variables are measured, and participants are randomly offered or not offered the study intervention. The outcome factor is then measured and compared. Eg. Overweight and obese children (and their families) attending paediatricians in the Adelaide region are randomised to receive either standard treatment or additional services in the form of 6 education sessions on healthy eating and living.

It is fundamental to a RCT that the randomization of participants to the active treatment or control group is done indiscriminately. When sample numbers are large, simple randomization will balance the groups for patient characteristics and other factors which may bias outcomes. For example, parent BMI in a child obesity intervention study. When numbers are small, further precautions must be taken to ensure patient baseline characteristics are approximately evenly distributed across the groups to avoid an imbalance of gender, age, disease progression etc.

There are various techniques to ensure rigorous randomization. For example, block and stratified randomization further ensure that participant characteristics are evenly distributed among the treatment groups. What method you use to physically randomise your participants will require careful consideration. Computer-generated random numbers or random-number tables are available, but you may benefit from expert help or consider using a randomisation service affiliated with a university or hospital. Further information on randomisation services can be found at the Wombat collaboration and IMPACT: Interdisciplinary Maternal Perinatal Australasian Collaborative Trials Network.

Furthermore, to eliminate or reduce selection bias in the randomisation process allocation concealment should be used and where practicable blinding / masking.

Observational Study: Observational Studies are studies in which “nature is allowed to take its course”. Changes or differences in a variable (eg. whether or not people receive a specific treatment) are studied in relation to an outcome (eg. a change or difference), without the intervention of the investigator. Examples of observational studies are as follows:

Case-control studies
Retrospectively investigates the level or prevalence of exposure to a factor comparing two groups. At the start of the study one group has the outcome/disease (cases) and one group does not have the outcome/disease (controls) (eg. Vitamin D exposure in pregnancy comparing children with and without birth defects).

Cohort studies(also known as longitudinal or follow-up studies)
Prospectively investigates the incidence of outcome/disease over time in one group compared to another. At the start of the study one group has been exposed to a potential risk factor and the other group has not. The incidence of asthma in children who are exposed to passive smoking in the first 12 months of life compared with those who are not.

Cross-sectional studies
Measures the prevalence of a disease. Investigates potential or suggested associations between outcomes and exposures. Causation can not be established. The study collects data at one point in time only. Surveys are a popular type of cross-sectional study eg A survey examining the age of diagnosis of autism and family socioeconomic status

Ecological (correlation) studies
This study uses populations or groups for units of analysis. Individuals are not used. Generally the number of people exposed and the number of people with the outcome are known eg Average sales of prophylactic antibiotics and rates of urinary tract infections in Australia or Trends in childhood obesity rates in Australia and New Zealand from 1950-2010.

2. Calculating your sample size

The number of participants that your recruit for your study (the size of your sample) is a critical factor to consider when designing your study. The number of participants required will depend on the expected size of the treatment effect (eg a 1% reduction in HbA1c) and the degree of variability in the measured effect (eg the variation in HbA1c reduction experienced by participants). On the one hand if there are too few participants in your study (sample) your findings will not have statistically sufficient power. On the other hand recruiting more people than what you require may place an unnecessary burden on your time, resources and finances.

Whilst there are websites which provides free sample size calculators, it is recommended that you consult with someone with expertise in statistics to correctly determine an adequate sample size, guided by your clinical input. Contact the hospital or university you’re submitting your project through for ethics approval for advice about statistical consulting services.

A useful article as an estimate on Sample size is:
Lehr R. Sixteen S-squared over D-squared: A relation for crude sample size estimates. Statistics in Medicine. Volume 11, Issue 8, 1992, Pages: 1099–1102

3. Design your questionnaire

Surveys can be paper based or online. The following websites may be able to assist you develop a questionnaire.

Epi Info is a free software package for developing a questionnaire or form, developing a data entry process, and entering and analysing data.

Survey monkey is a free (& fee for service) internet based survey tool; you can create surveys (the free service is limited to 10 questions), collect and analyse data online.

REDCap is a secure, web-based application designed exclusively to support data capture for research studies and may be of interest for collaborative multi-site research. This application is only available to institutional partners therefore you will need to have affiliations with the relevant institutions. The Murdoch Childrens Research Institute in Melbourne is a partner.

When designing a questionnaire it is highly recommended that you consult with a statistician to ensure that future data analysis will be straightforward and feasible.

[4] Bonita, R, Beaglehole, R, Kjellstrom, T. 2006, Basic epidemiology, 2nd Ed, WHO, Switzerland.