By the end of this chapter, you will be able to:
Mastering these skills will help you conduct effective and responsible nutrition research that can make a real difference in health and wellbeing.
Compiled by: Peter Kamau
Research design forms the foundation of scientific inquiry in nutrition and dietetics by providing a structured framework for collecting, analysing, and interpreting data. In the Kenyan context, where diverse populations and nutritional challenges exist, selecting an appropriate research design ensures that studies yield valid, reliable, and actionable findings. Nutrition professionals working in hospitals, community health programs, or food processing plants depend on robust research designs to evaluate interventions, assess dietary patterns, and influence policy. This chapter explores the concept of research design, the various types available, and critical considerations when choosing the best design for nutrition research.
A study design is the plan that decides how a research question will be answered: who will be studied, what will be measured, when and how often, and whether anything will be changed. A good design lets the researcher reach conclusions that can be trusted, within the time, money and ethical limits of the setting. A poor design cannot be rescued by careful analysis later, because the data collected will not answer the question.
A dietitian meets design choices often. A clinic manager may want to know how many mothers give fortified porridge, a ward team may want to know whether a feeding protocol reduces weight loss, and a community unit may ask whether a food voucher scheme improved children's growth. Each question calls for a different design. This topic defines research design, describes the main types, and sets out the factors that guide the choice.
Research design is the overall plan or blueprint for a study. It states the method of choosing participants, the variables to be measured, the tools and timing of data collection, the way the data will be analysed, and the steps taken to control bias and protect participants. It connects the research question to the evidence needed to answer it.
For example, a researcher who wants to know whether a daily serving of orange-fleshed sweet potato improves the vitamin A status of toddlers must decide whether to follow a group of toddlers over time, to compare two groups, or to look at clinic records that already exist. That decision is the design, and it determines what the study can and cannot claim.
Designs are grouped in several ways, and a single study can belong to more than one group. For example, a study may be descriptive, cross sectional and observational at the same time. The seven types below are used in nutrition work. Each is described by its purpose, how it works, and its strengths and limits.
A descriptive design describes the characteristics of a population, a condition or a practice as they are, without testing a cause. It answers questions such as how many, who, where and how. Examples are the proportion of adolescent girls in a district who skip breakfast, the foods eaten by older adults in a community, and the number of ward patients screened for malnutrition risk on admission.
Data may be numerical, such as weights and intakes, or in words, from interviews and focus group discussions. Descriptive studies may be qualitative or mixed methods, as defined under Types of research. For example, a researcher may hold group discussions with fathers and mothers to learn why households buy little fruit.
Strengths: it is relatively quick and cheap, it gives a picture of a problem on which to plan, and it can generate ideas for further study. Limits: it cannot show that one factor causes another.
A correlational design measures two or more variables in the same people and examines whether they change together. The researcher does not change anything. For example, the number of meals eaten in a day and the weight-for-age of 80 children might be measured to see whether children who eat more meals tend to have a higher weight-for-age.
The strength and direction of the link are given by a correlation coefficient, written r, which runs from -1 to +1. A value close to +1 means that the two variables rise together, a value close to -1 means that one rises as the other falls, and a value near 0 means there is no straight-line link. For example, r = +0.62 between meals a day and weight-for-age suggests that more meals go with a higher weight-for-age, but not that more meals caused the change.
Correlation does not prove causation. Another factor, such as household income, could raise both. The design is useful for finding links worth testing in an experiment.
An experimental design deliberately changes one factor, the intervention, and measures its effect on an outcome. Participants are divided into an intervention group, which receives the change, and a control group, which does not, or which receives the usual care. Because the researcher assigns the intervention, the design can show cause and effect more directly than the others.
The key features are:
Example: 100 undernourished children are assigned by chance, 50 to a supplement porridge and 50 to the usual porridge, and their weight gain over 12 weeks is compared.
Strengths: it shows the effect of the intervention directly. Limits: it is costly and slow, may not suit a food that cannot be hidden, and raises ethical questions when withholding a likely benefit.
A diagnostic design tests how well a method of assessment or screening identifies a condition when compared with an accepted reference method. In nutrition it is used to judge tools such as the mid-upper arm circumference tape for finding acute malnutrition, a haemoglobin meter against laboratory counts, or a screening questionnaire for the risk of malnutrition in hospital patients.
Each person is tested by both the new tool and the reference method, and the results are set out in a table:
| Condition present (reference) | Condition absent (reference) | Total | |
|---|---|---|---|
| New tool positive | 34 (true positive) | 9 (false positive) | 43 |
| New tool negative | 6 (false negative) | 51 (true negative) | 57 |
| Total | 40 | 60 | 100 |
The measures are then calculated:
A tool with high sensitivity misses few cases, which matters in screening. A tool with high specificity gives few false alarms, which matters when a positive result leads to costly or burdensome care. The numbers above are an invented illustration.
In an observational design the researcher observes and records what happens naturally, without giving or withholding any intervention. It is used when an experiment is not possible or not ethical, for example when studying whether a poor diet leads to disease. The main types are:
A longitudinal design suits the study of change over weeks, months or years, such as growth in the first two years of life. Strengths of observational designs: they study real-life exposures and can be done when intervention is not possible. Limits: confounding factors may affect the result, and cause cannot be shown as clearly as in an experiment.
A cross-sectional study measures exposure and outcome in a population at one point in time, like a snapshot. For example, a survey of 400 adults at a health centre measures weight, height and blood pressure on a single visit to find how many are overweight and how many have raised blood pressure.
It is suited to finding how common a condition is and to comparing groups. It is quick and relatively cheap, and it can study many factors at once. Its limits are that it cannot show which came first, the exposure or the outcome, and it tells nothing about change over time. It is therefore descriptive, or at most a start for studies that look at causes.
An intervention study tests the effect of a planned action on health or nutrition. The action may be a supplement, a food, a counselling programme, a feeding policy or a change in how a service is delivered. Intervention studies may be experimental, with randomised groups, or quasi-experimental, where groups are formed without random assignment, as when one community receives the programme and a similar one does not.
A simple pattern is to measure the outcome before the intervention (baseline), apply the intervention, and measure again after it (endline). Comparison with a group not receiving the intervention shows whether the change is due to the intervention and not to time or other events.
Example: a baseline survey in two communities finds that 28 per cent of children aged 6 to 23 months in each are anaemic. One community receives a programme that includes iron-rich foods and counselling for 6 months. At endline the proportions are 18 per cent in the programme community and 27 per cent in the other. The difference of 9 percentage points suggests that the programme may have contributed, although a statistical test is needed before the difference can be trusted. The figures are invented.
Strengths: it shows whether a programme works in practice. Limits: it needs resources, careful follow-up and ethical approval, and results from one setting may not apply to another.
No single design suits every study. The choice depends on several factors, which are weighed together.
The design must fit the question. To describe how many children are stunted, a cross-sectional survey is enough. To test whether a supplement improves growth, an experimental or intervention design is needed. To find reasons behind feeding practices, a qualitative design is suitable.
Time, funds, staff and equipment limit the options. A long cohort study or a large trial may be out of reach for a small team, while a cross-sectional survey or a qualitative study may be possible.
The size of the population, how easy it is to reach, and how varied it is affect the choice. Where people move often, a long follow-up will lose participants. Groups at risk, such as pregnant women and infants, call for designs that keep risk to a minimum.
Informed consent, confidentiality and fairness apply to all designs. Experimental designs need particular care, since withholding a treatment that is likely to help may not be acceptable, and an ethics committee must approve the plan.
The design must match the tools available and the skills to analyse the data. Biochemical tests, for instance, need a laboratory, and a study with many variables needs help with statistics.
Designs differ in how well they control bias and confounding. Randomisation and blinding in experiments reduce these. In observational designs, the researcher must reduce them by choosing the sample carefully and by adjusting for other factors in the analysis.
A short time frame suits a cross-sectional survey. A study of outcomes that take years to appear, such as the effect of childhood diet on adult disease, needs a longitudinal design, or a case-control study that looks back.
Create a free account to open more of this chapter.
Free: practical guides, quick cards, workplace scenarios and more.
Create a free accountThis chapter explored the essential components involved in determining a nutrition research design, beginning with a clear understanding of what research design entails and its various types, including descriptive, correlational, experimental, diagnostic, observational, cross-sectional, and intervention studies. It outlined the importance of carefully selecting a study design based on specific research objectives and contextual factors. The concept of the study population was defined along with practical steps to establish an appropriate population for the research. Sampling techniques were examined, highlighting their advantages and disadvantages, as well as criteria for selecting the most suitable method and calculating an adequate sample size to ensure representativeness. Different data types and collection methods were discussed, covering surveys, interviews, observations, focus groups, experiments, and secondary data analysis. The chapter also emphasized the necessity of budgeting and resource planning to support the research process effectively. Finally, it addressed the critical ethical considerations related to the treatment of participants, responsibilities towards sponsors, and the conduct expected of researchers and their teams during the study.
At the start of this chapter we promised you would be able to:
Tick each one you can genuinely do.
So, are you there yet?
You're competent when you can confidently do 50% or more of what this chapter promised.
Sign in to record how you're doing.