S
everal algorithms are available in the literature
for selecting the appropriate study designs [1,2]. Conventional
algorithms focus more on what the various study designs are, rather than
facilitating how to approach a research question and arrive at the
appropriate design. To be able to use these algorithms, the user should
have a priori knowledge about the various study designs [3]. Most
of these approaches try to fit in all the types of study designs in a
single algorithm that makes it confusing for a novice researcher,
especially a postgraduate student [4].
A traditional study design algorithm lacks two key
elements; (a) the presence or absence of a comparative group, and
(b) the timeframe over which the data is to be collected. Without
explicitly stating the number of groups, single arm trials or one group
cohort studies may be missed [5]. The dimension of time is also
important in medical research for follow-up or repeated measurements.
For example, a study to determine the change in blood pressure will
require at least two readings at different time points. We propose a
simplified ‘3-Question (3Q) Approach’ that overcomes these limitations,
and is primarily targeted at the novice medical researcher.
The 3Q approach should be applied after the
researcher has framed the research question or the primary objective of
the study. A well-framed Research question (RQ) is an essential
pre-requisite for arriving at an appropriate study design. As the name
suggests, three questions are to be answered in a cascading manner; in
yes or no responses (Fig. 1).
The First Question
This aims to differentiate between an observational
and an interventional study.
Question 1: Are we trying to modify/ change the
outcome of interest, in the study?
Responses: No
®
Observational study;
Yes ®
Interventional study
The answer to the first question results in the
formation of two broad categories of study designs. An observational
study is defined as "a type of study in which individuals are observed
or certain outcomes are measured" [6]. The study does not intend to
change the patient outcome. The researcher may use certain tools for
observation such as microscopy for malaria parasite detection in a
malaria prevalence study. Carrying out procedures like microscopy or
endoscopy to find out the burden of disease or risk factors does not
make it an interventional study. Diagnostic studies to find out
sensitivity and/or specificity are also considered as observational
studies.
In the context of study design, intervention means
that the researcher is experimenting by changing or modifying some
existing variable and evaluating how it affects the outcome(s). This may
be a ‘novel drug’ for an existing condition, an ‘existing drug’ for a
‘novel indication,’ or a ‘novel diagnostic technique’ to be evaluated
for outcome [7]. The term interventional study is synonymous with
experimental study or trial [8].
The Second Question
The second question differs for an observational
study and for an interventional study, based on the response to the
first question.
For an observational study, its aim is to identify
the purpose of the study.
Question 2a: Do we have more than one group?
Responses: No
®
Descriptive study;
Yes ®
Comparative study
When the purpose of a study is to describe a single
population group, it is a descriptive study. These include studies that
compare population parameters such as rates, proportions, or means.
Summary statistics (such as mean, standard deviation, proportion) will
be applicable to descriptive studies. The statistical tests of
significance are applicable on comparative studies and not on
descriptive studies [9].
Studies of diagnostic accuracy wherein the objective is to assess
sensitivity, specificity etc. are also technically descriptive
studies as the focus is on single group, i.e. the ‘diseased.’
Traditional algorithms keep diagnostic accuracy studies as a separate
category. In the 3Q approach, these types of studies are integrated
within the whole framework.
In a case-control study, the diseased and
non-diseased groups are compared with respect to the presence or absence
of the risk factor. For a cohort study, a group with a risk factor under
evaluation and another without the risk factor are compared with respect
to the development of the disease.
For an interventional study, the aim is to identify
whether there is an intent to compare.
Question 2b: Do we have more than one group?
Responses: No
®
Single arm trial;
Yes ® Two-arm
trial
The possible responses to the second question
lead to two situations: if there is no comparison arm, it is called a
single arm trial or a before-after study [10, 11]; in two arm
trials, there is a treatment arm and a comparison arm.
The Third Question
This also differs according to whether it is an
observational or an interventional study.
For an observational study, the aim of this
question is to ascertain the time factor.
Question 3a: Do we have repeated measurements?
Responses: No
®
Cross-sectional study;
Yes ®
Longitudinal study
Cross-sectional studies are those where measurement
is done only at one point of time. This measurement is either for the
risk factor or the outcome, or for both together. When repeated
measurements are done on the same individual, it is known as a
longitudinal study. Since interventional studies are always
longitudinal, this question is irrelevant for these studies.
For an interventional study, the aim of this
question is to determine the status of randomization in a two-arm trial.
Question 3b: Is randomization present?
Responses: No
® Non-Randomized
study;
Yes ®
Randomized study
For a two-arm trial, it is important to mention the
randomization status. It is well known that the process of
‘randomization’ increases the validity of the study. Randomized clinical
trials are true experimental studies. In a non-randomized trial, the
difference in the outcomes or the endpoint values in both the groups may
be due to the differences in baseline values. These are also known as
quasi-experimental studies.
Limitations of the 3 Q Approach
Validation studies have not been conducted for this
approach yet. However, the authors have used the 3Q approach in more
than 30 workshops for faculty and medical students on ‘selecting study
designs in medical research’ in different parts of the country till now,
and have found from the participants’ feedback that this approach
facilitated their understanding of study designs. We suggest that
studies to assess the validity of this approach be conducted.
To conclude, the 3Q approach is an easy-to-use
framework to decide study designs. However, it requires a well-framed
research objective. It gives the researcher insight into how the study
should be conducted, based on the responses that are obtained from these
three questions. It is easy to apply and can also be used to teach how
to choose study designs to novice researchers, including medical
students and younger faculty.
Acknowledgement: Dr Navjeevan Singh, former
Professor of Pathology, UCMS, and Coordinator, Medical Education Unit,
UCMS and GTB Hospital, Delhi, for his valuable contribution to this
article.
Contributors: AMK, PG: conceptualized, designed,
drafted, and critically revised the work; DM: contributed in drafting
the article and critically revising the work. All the authors approved
the final version to be published.
Funding: None; Competing interests:
None stated.
References
1. Oxford Centre for Evidence Based Medicine. Oxford:
Nuffield Department of Primary Health Care Sciences; c2018. Study
Designs. Available from: http://www.cebm. net/study-designs/.
Accessed May 08, 2018.
2. Evidence Analysis Library. Chicago: Academy of
Nutrition and Dietetics; c2018. Research Design Algorithm. Available
from:
https://www.andeal.org/vault/2440/web/files/Research_Design_Algorithm.pdf.
Accessed June 07, 2019.
3. Indrayan A, Satyanarayana L. Designs of medical
studies. Indian Pediatr. 1999;36:691-6.
4. Andrews J, Likis FE. Study design algorithm. J Low
Genit Tract Dis. 2015;19:364-8.
5. Evans SR. Clinical trial structures. J Exp Stroke
Transl Med. 2010;3:8-18.
6. PubMed Health. Bethesda (US): US National Library
of Medicine. Observational Study. Available from:
https://www.ncbi.nlm.nih.gov/pubmedhealth/PMHT0025839/. Accessed May
11, 2018.
7. Oprea TI, Bauman JE, Bologa CG, Buranda T, Chigaey
A. Edwards BS, et al. Drug repurposing from an academic
perspective. Drug Discov Today Ther Strateg. 2011;8: 61-9.
8. Thiese MS. Observational and interventional study
design types; an overview. Biochemia Medica. 2014;24:199-210.
9. Centers for Disease Control and Prevention.
National Centres for Health Statistics. Descriptive Statistics; updated
2013. Available from:
https://www.cdc.gov/nchs/tutorials/NHANES/NHANESAnalyses/descriptive
statistics/descriptive_statistics_intro.htm. Accessed May 11, 2018.
10. Sedgwick P. Before and after study designs. BMJ.
2014;349:g5074.
11. Goodacre S. Uncontrolled before-after studies: Discouraged by
Cochrane and the EMJ. Emerg Med J. 2015;32:507-8.