Table 8. Key IR study designs and overview

Descriptive
Cross-sectional (descriptive)	Provides insight into the presence, level and/or distribution of one or more variables of interest in a given population at a certain point in time. Descriptive cross-sectional studies can quantify the frequency or how much a certain variable or outcome of interest exists within a certain population.	Relatively simple and inexpensive, and useful for investigating contexts with many variables.	Requires appropriately sized sample, a scientific approach to sampling and a good response rate to generate valid results. Cannot be used to assess causal relationship between variables.	Cross-sectional studies can be used to assess reach and adoption. These are also good for analysis of factors associated with sub-optimal reach or adoption; for example, by looking at the relationship between individual or setting-related variables (such as socioeconomic status, location or number of staff). Cross-sectional studies done before and after an intervention can also be used to assess the effectiveness of an intervention.
Ecological	Ecological studies look at comparisons between groups, rather than individuals. Groups can be defined based on geography or time, and are compared based on outcomes of interest and relevant variables aggregated or averaged at the group-level.	Ecological studies typically utilize existing or routinely collected datasets so they can be used to quickly explore hypotheses and are inexpensive to conduct.	Analysis is based on aggregated data and extrapolating the findings to an individual may lead to "ecological fallacy", as any association observed between variables at the group level does not necessarily apply to each individual within the group.	Ecological studies could be used to assess effectiveness of an intervention at a community level.

Descriptive

Study design

Overview

Strengths

Limitations

Examples of use

Cross-sectional (descriptive)

Provides insight into the presence, level and/or distribution of one or more variables of interest in a given population at a certain point in time.

Descriptive cross-sectional studies can quantify the frequency or how much a certain variable or outcome of interest exists within a certain population.

Relatively simple and inexpensive, and useful for investigating contexts with many variables.

Requires appropriately sized sample, a scientific approach to sampling and a good response rate to generate valid results. Cannot be used to assess causal relationship between variables.

Cross-sectional studies can be used to assess reach and adoption. These are also good for analysis of factors associated with sub-optimal reach or adoption; for example, by looking at the relationship between individual or setting-related variables (such as socioeconomic status, location or number of staff).

Cross-sectional studies done before and after an intervention can also be used to assess the effectiveness of an intervention.

Ecological

Ecological studies look at comparisons between groups, rather than individuals. Groups can be defined based on geography or time, and are compared based on outcomes of interest and relevant variables aggregated or averaged at the group-level.

Ecological studies typically utilize existing or routinely collected datasets so they can be used to quickly explore hypotheses and are inexpensive to conduct.

Analysis is based on aggregated data and extrapolating the findings to an individual may lead to "ecological fallacy", as any association observed between variables at the group level does not necessarily apply to each individual within the group.

Ecological studies could be used to assess effectiveness of an intervention at a community level.

Analytical
Experimental/ quasi - experimental
Study design	Overview	Strengths	Limitations	Examples of use
Pragmatic randomized controlled trials/ cluster randomized controlled trials (CRT)	Pragmatic trials adopt a quasi-experimental approach and are designed to represent the setting in which an intervention will be used under real-life conditions. Pragmatic trials typically randomize at a cluster level (and are also referred to as pragmatic CRTs), such as hospital or clinic, and are used to evaluate an intervention against other, comparable interventions (established or not).	Pragmatic trials are suitable for IR as they are typically implemented within 'real-life' settings and as such, important, locally determined factors such as cost, feasibility and political concerns are considered within the context of the study.	May be expensive or difficult to implement given need for multiple clusters. Negative findings do not provide any insight into whether the intervention might be effective under optimal conditions (that is, is the intervention itself ineffective, or is the lack of impact due to the specific contextual factors the study was conducted in?). Additionally, because findings are specific to implementation (real life) contexts, the findings may not be widely generalizable or applicable to other settings.	Pragmatic trials are useful for identifying the effectiveness of an intervention at a group or community level. Can also be used to assess maintenance, implementation and adoption of the intervention over time.
Stepped-wedge trials	In quasi-experimental stepped-wedge trials, exposure to an intervention randomly occurs in waves or steps, so that all clusters/subjects will have spent time being both unexposed and exposed to an intervention. The outcome(s) of interest are measured among all subjects before and after exposure.	Commonly used in the evaluation of service delivery-type interventions. As all subjects are exposed to an intervention, they can act as their own control and therefore fewer units of investigation are needed.	As a relatively new study design, there is limited information available to guide the reporting and analysis of results.	Stepped-wedge trials are useful for identifying the effectiveness of an intervention. These designs can also explore maintenance by following the ongoing effect or implementation of an intervention overtime.
Pre/post with non-equivalent control group	In this quasi-experimental design, participants in an intervention group are tested before and after exposure to intervention to assess the effect on the research outcome(s) of interest. These results are compared to pre- and post- tests administered to a non-equivalent control group, which is a group selected on the basis of similar characteristics to the intervention group but without random allocation.	Less expensive and require fewer resources compared to pragmatic cluster randomized trials. Given these studies do not involve randomization, they are appropriate in situations where randomization would be considered unethical. Lastly, as these studies are pragmatic – that is, they evaluate the effectiveness of an intervention in real-world conditions – they are ideally suited for IR.	Because of the non-random allocation of intervention and control group participants, it is possible that observed differences between the two groups are due to selection differences rather than an underlying effect.	Useful for identifying the Effectiveness of an intervention at a group or community level. Can also be used to assess maintenance, implementation and adoption of the intervention over time.
Interrupted time series	In this design, the research outcome(s) of interest is measured at various time points before and after exposure to an intervention and changes observed throughout the measurements are used to infer intervention effect.	Useful for identifying an effect over time, including how long an effect may last for.	As only one group is observed, it is not possible to compare the results to a control group. As with other pragmatic designs, the unit of study are typically groups that already exist (i.e. participants are not randomized or allocated) and therefore there is a possibility of selection bias. Also, as multiple rounds of data collection are required, this may be a timely and/or costly study design.	Like stepped-wedge trials, interrupted time series are useful for identifying the effectiveness of an intervention. These designs can also explore maintenance by following the ongoing effect or implementation of an intervention over time.
Observational
Study design	Overview	Strengths	Limitations	Examples of use
Cohort	In a cohort study, a group of people are recruited and followed up over time to monitor the development of specific outcomes, typically health conditions or events. Cohort studies have a temporal element that can be used to produce high-quality data about the relationship between variables.	High-quality, individual-level data, enables researchers to examine if better implementation outcomes are associated with exposures at the individual level, including the timing and direction of any effects.	Can be difficult, expensive or timely to follow up participants over a long period of time. Cohort studies are vulnerable to loss to follow up or drop out of participants, which may weaken or undermine potential of study to identify an effect.	Cohort studies are useful to investigate effectiveness over time (e.g. looking at changes in treatment adherence throughout the course of an intervention). Cohort studies can also explore maintenance by following the ongoing implementation of an intervention by individuals or implementers.
Cross sectional (analytical)	Analytical cross-sectional studies can explore the relationship between the variable of interest and other factors	See above	See above	See above