Home  |  About us  |  Editorial board  |  Ahead of print  | Current issue  |  Archives  |  Submit article  |  Instructions |  Search  |   Subscribe  |  Advertise  |  Contacts  |  Login 
  Users Online: 1103Home Print this page Email this page Small font sizeDefault font sizeIncrease font size  

 Table of Contents      
Year : 2022  |  Volume : 13  |  Issue : 1  |  Page : 12-16

Observational designs for real-world evidence studies

Medical Lead, Pfizer Ltd, Mumbai Maharashtra, India

Date of Submission23-Sep-2021
Date of Decision27-Sep-2021
Date of Acceptance27-Sep-2021
Date of Web Publication06-Jan-2022

Correspondence Address:
Dr. Santosh Ramesh Taur
Flat No. 1504, Building No. 68, Kurla Sahayog CHSL, MG Road, Nehru Nagar, Kurla (E), Mumbai - 400 024, Maharashtra
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/picr.picr_217_21

Rights and Permissions

In the era of evidence-based medicine, real-world evidence (RWE) studies have opened avenues to utilize real-world data (RWD) effectively for improving clinical decision-making. However, the transformation of RWD into a meaningful RWE can only be achieved when the researcher asks the right clinical question, selects the right RWD source for variables of interest, uses the right study design, and applies the right statistical analysis. The generated RWE needs to have internal as well as external validity to be actionable. The “fit-for-purpose” observational study designs include descriptive, case–control, cross-sectional, and cohort. This article focuses on the advantages and disadvantages including the inherent bias of each study design. The RWE study decision guide has also been provided to aid the selection of appropriate study designs.

Keywords: Bias, observational, real-world evidence, study designs

How to cite this article:
Taur SR. Observational designs for real-world evidence studies. Perspect Clin Res 2022;13:12-6

How to cite this URL:
Taur SR. Observational designs for real-world evidence studies. Perspect Clin Res [serial online] 2022 [cited 2023 Mar 30];13:12-6. Available from: http://www.picronline.org/text.asp?2022/13/1/12/335015

   Introduction Top

The evolution of real-world evidence (RWE) over the past years has improved our understanding of disease characteristics as well as the safety and effectiveness of treatment in clinical practice. Well-designed RWE may complement the randomized controlled trials (RCTs) in clinical decision-making. Digitalization of health-care systems advances the availability and utilization of real-world data (RWD), such as electronic medical records, registries, and wearable devices, by researchers. The generated RWE needs to be internally valid and generalizable to an identifiable target population to be actionable.[1] Turning RWD into RWE involves the interplay of a design layer – study design appropriate for the research question; a measurement layer – transforming the patient-level data into the variables of exposure and outcome; and an analysis layer – deriving the estimate of effect.[1] We will discuss RWE studies with focus on observational designs and the selection of appropriate designs that guide investigators and reviewers of RWE.

   Real-World Evidence Study Designs Top

RWE studies can be classified basis the assignment of intervention and requirement of the comparison group [Figure 1] and [Table 1].[2]
Figure 1: Classification of real-world evidence study designs; R – retrospective, P – prospective

Click here to view
Table 1: The design and analysis time frame (relative to the study start or index date) of different types of real-world evidence studies

Click here to view

In the observational study, the participants are not preassigned to exposure or an intervention. The choice of treatments is up to patients and their physicians.

In a retrospective observational study, both exposure and outcomes have already occurred. In subsequent sections, we shall discuss in detail about 3 retrospective observational study designs, namely cross-sectional, case-control, and cohort.

Cross-sectional studies

Cross-sectional studies involve the simultaneous assessment of exposure and outcome in a single group of patients at a specific point in time. They are typically used to assess the prevalence and infer the cause of conditions/outcomes.[2] The general design involves defining the target population, deriving a sample of that population, and defining the characteristics being studied. The definition of the condition and health characteristics under study should be standardized, reproducible, and feasible to apply on a large scale.[3]

Datta et al.[4] conducted a community-based descriptive cross-sectional study among individuals aged 30 years and above in urban field practice area of a tertiary care hospital. Pretested questionnaire was used to measure the prevalence of self-reported hypertension over the past 2 years.


  • Relatively quick and inexpensive compared with RCTs[2]
  • Simultaneous assessment of a wide variety of exposures and outcomes[3]
  • The most efficient way to determine the prevalence of a condition.[2]


  • As the data are collected at a single time point, a temporal relationship between the exposure and the outcome cannot be ascertained,[2] for example, if the cross-sectional study demonstrates an association between obesity and arthritis, for example, one cannot prove causal relationship whether arthritis led to obesity or the obesity caused arthritis or, whether some third factor caused them both.
  • Susceptible to selection bias, for example, patients who develop an outcome but die before the end of the study are not captured[2]
  • Unsuitable for studying rare diseases or for diseases of short duration due to low prevalence at a single point in time[3]
  • Often completed using questionnaires, which have inherent problems, including low response rates and susceptibility to various sources of bias.[2]

Case–control studies

Case–control studies are typically retrospective studies (”backward looking”) because the approach is to identify persons with the disease of interest and then look backward in time to identify factors that may have caused it [Figure 2].[3] Cases are the patients with the outcome of interest, and controls are matched groups of patients without this outcome derived from the same population. The exposure to potential causal variables such as risk factors and treatment are evaluated basis of the medical history to determine causality. Case–control studies have a longitudinal or temporal aspect to the data that cross-sectional studies do not.[3] To have the consistency of the associations and risk estimates between a case-control study and other types of studies, the three key assumptions need to be met:[3]
Figure 2: Design of case-control study

Click here to view

  1. cases are representative of all the cases,
  2. controls are representative of persons without the disease, and
  3. data are collected similarly in cases and controls.

The case–control studies help answer the research question leading to hypothesis generation.[2]

Swain et al.[5] conducted a population-based case–control study. Cases (disease) were confirmed dengue fever patients tested through IgM method and hospitalized to any public or private health facilities of the state, and controls (no disease) were from same living area of cases matched with sex and age (±2 years) with no history of dengue, chikungunya or acute febrile illness in the last 1 year. A structured validated questionnaire was used to collect information regarding exposures such as environmental practices and travel history for each consenting participant.


  • Suitable for rare outcomes or those with a long latency between exposure and disease[2]
  • Simultaneous assessment of multiple etiologic factors or potential predictors of the specific outcome[2]
  • The sample size requirement for the same effect size would be higher for a cohort study than that for a case–control study. Further, the lower prevalence of disease may result in much higher sample size of a cohort study but may not have significant impact on that for a case–control study.[3]


  • The study typically involves a single outcome[2]
  • No information about the base population from which the cases are drawn; so incidence and prevalence cannot be estimated[3]
  • Bias [Table 2] may skew the estimate of association.[3]
Table 2: Bias in case–control studies

Click here to view

Cohort studies

Cohort studies evaluate the association between a particular exposure or a risk factor and subsequent development of disease [Figure 3]. They are “prospective” (”forward looking”) in that exposure or risk factor information is collected about 2 cohorts (exposed and nonexposed), and then both cohorts are followed up to ascertain the development of the outcome. In concurrent cohort studies, people with or without exposures are identified at the initiation of the study and information is collected looking forward in time to identify disease outcomes. In nonconcurrent cohort studies (retrospective), exposure information has already been collected at some point in the past and participants are surveyed in the present to determine the presence or absence of disease. It has the advantage of providing long follow-ups without waiting for time to pass to obtain disease outcomes.
Figure 3: Concurrent and nonconcurrent cohort study design. Adapted from Johnson 2018[3]

Click here to view

Badyal et al.[6] conducted a prospective, observational, multicenter cohort study among health-care workers [HCW] (HCWs; likely to be exposed to COVID-19 cases) who were either taking or not taking prophylactic treatment (Exposure) for COVID-19 and assessed their COVID-19 positivity by reverse transcription-polymerase chain reaction (Disease).

Nested case–control study is a special study design that reduces most biases related to selection and data collection, which is typically seen in classic case–control studies, by selecting cases and controls from within a large-scale prospective cohort study. The biologic samples in the ongoing cohort study can be collected and stored until enough cases have accumulated to provide adequate study power. At that time, these baseline samples from the newly occurring cases can be thawed and measured, along with a comparison group of matched (or unmatched) controls, allowing a much more efficient approach to examining expensive or difficult-to-measure risk factors.[3]
Figure 4: Immortal time bias. Adapted from Franklin et al., 2017[7]

Click here to view

Nested case-cohort designs sample a priori so regardless of case status or time. This design is used for the analysis of “time to event/failure” in a large cohort where that event/failure is rare and enormous resources may be needed to ascertain covariate values. Nested case–cohort designs are more flexible than nested case–control designs when there are different choices for the time axis or several failure time outcomes.[3]


  • Find incident (new-onset) cases, rather than prevalent (existing) cases
  • Provide more information about the natural history of disease as well as direct estimates of incidence and relative risk
  • Firmly establish the temporal relationships between exposure and disease
  • Multiple disease outcomes can be studied in relationship to the exposure
  • Suitable to study a rare exposure, with exposed persons gathered at the beginning of the study.


  • Concurrent cohort design: Study duration may be exceedingly long, making difficult the maintenance of consistent study methods and enthusiasm of staff and participants
  • Follow-up of free-living populations may be expensive because people move or change their contact information and can be difficult to track
  • Requirement of large samples size and more so for rare diseases
  • Bias [Table 3] may skew the estimate of association[1],[7],[8]
Table 3: Bias in cohort studies

Click here to view

Selection of right real-world evidence study design

A well-conducted RWE has both successfully replicated and predicted findings of RCTs. Transparency in the conduct and presentation of RWE studies is critical to allow reviewers to evaluate the study's validity and have confidence in their decision-making.[1]

The decision guide for selection of RWE study design (modified from the RWE Framework flow diagram developed by Xia et al. 2019[9]) has been provided in [Table 4].
Table 4: Real-world evidence study decision guide

Click here to view

   Conclusion Top

The digitization of medical informatics has opened new avenues for the conduct of retrospective RWE studies. Observational RWE studies are less expensive and can be conducted quickly compared to the RCTs. However, observational RWE studies can be used only to find associations between risk factors and outcomes, but alone they cannot establish causation. In addition, there are inherent biases and issues of confounding associated with the observational studies impacting the “internal validity” of these studies. Well-designed and conducted observational RWE studies with careful analysis and interpretation will overcome most of these challenges and may help in hypothesis generation or complement the RCTs.

Financial support and sponsorship


Conflicts of interest

There are no conflicts of interest.

   References Top

Schneeweiss S, Patorno E. Conducting real-world evidence studies on the clinical outcomes of diabetes treatments. Endocr Rev 2021;42:1-33.  Back to cited text no. 1
Camm AJ, Fox KA. Strengths and weaknesses of 'real-world' studies involving non-vitamin K antagonist oral anticoagulants. Open Heart 2018;5:e000788.  Back to cited text no. 2
Johnson LL. Design of observational studies. In: Principles and Practice of Clinical Research. 4th ed. Academic Press: Elsevier; 2018. p. 231-48.  Back to cited text no. 3
Datta S, Sahu SK, Niranjjan R, Roy G. A community-based cross-sectional study on hypertension screening in Puducherry, India. Indian J Med Res 2019;150:199-202.  Back to cited text no. 4
[PUBMED]  [Full text]  
Swain S, Bhatt M, Biswal D, Pati S, Soares Magalhaes RJ. Risk factors for dengue outbreaks in Odisha, India: A case-control study. J Infect Public Health 2020;13:625-31.  Back to cited text no. 5
Badyal D, Chandy S, Chugh P, Faruqui A, Gupta YK, Hazra A, et al. Hydroxychloroquine for SARS CoV2 prophylaxis in healthcare workers – A multicentric cohort study assessing effectiveness and safety. J Assoc Physicians India 2021;69:11-2.  Back to cited text no. 6
Franklin JM, Schneeweiss S. When and how can real world data analyses substitute for randomized controlled trials? Clin Pharmacol Ther 2017;102:924-33.  Back to cited text no. 7
Lu CY. Observational studies: A review of study designs, challenges and strategies to reduce confounding. Int J Clin Pract 2009;63:691-7.  Back to cited text no. 8
Xia AD, Schaefer CP, Szende A, Jahn E, Hirst MJ. RWE framework: An interactive visual tool to support a real-world evidence study design. Drugs Real World Outcomes 2019;6:193-203.  Back to cited text no. 9


  [Figure 1], [Figure 2], [Figure 3], [Figure 4]

  [Table 1], [Table 2], [Table 3], [Table 4]


    Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
    Access Statistics
    Email Alert *
    Add to My List *
* Registration required (free)  

  In this article
    Real-World Evide...
    Article Figures
    Article Tables

 Article Access Statistics
    PDF Downloaded479    
    Comments [Add]    

Recommend this journal