- Home/
- Our Work/
- Health Emergencies/
- Research Methods/
- Sections and chapters/
- Section 4: Study design/
- Chapter 4.14 Natural experiments in a hazard context
Section 4: Study design
Chapter 4.14 Natural experiments in a hazard context
Research Methods for Health EDRM
Section navigation
- Section 4: Study design
- Chapter 4.14 Natural experiments in a hazard context
- Chapter 4.1 Basic principles in designing studies to assess the effects of interventions
- Chapter 4.2 Measuring the problem: Basic statistics
- Chapter 4.3 Cluster randomized trials
- Chapter 4.4 Collection and management of good quality data
- Chapter 4.5 Advanced statistical techniques
- Chapter 4.6 Health-related risk modelling
- Chapter 4.7 Evaluating economic impacts in health emergency and disaster risk management
- Chapter 4.8 Geographic information systems
- Chapter 4.9 Real-time syndromic surveillance
- Chapter 4.10 Using logic models in research and evaluation of Health EDRM interventions
- Chapter 4.11 Researching communication and communicating research in Health EDRM
- Chapter 4.12 Qualitative research
- Chapter 4.13 Addressing complexity through mixed methods
- Chapter 4.15 Monitoring and evaluation
Authors: Kim HM, Stewart AG, Schluter PJ.
Chapter 4.14 describes the potential utility of natural experiments in health emergency and disaster risk management (Health EDRM), including:
- Process of conducting a natural experiment in a disaster context.
- Framework for, and outcomes of, natural experiments.
- Important strengths and limitations of natural experiments.
What is this chapter about?
Implementing experimental designs, such as randomized trials, to study cause-effect relationships might not be feasible or ethical in some natural or human-instigated hazard contexts. In these circumstances, natural experiments provide researchers with alternative ways to investigate topics of relevance to Health EDRM that are not amenable to those experimental designs.
This chapter discusses natural experiments as an alternative method for studying causal associations. It briefly describes the key components of a causal framework for natural experiments and looks at how natural experiments can be used in a hazard or disaster context. The chapter describes strengths and limitations associated with using natural experiments in Health EDRM and uses three case studies of natural experiments to illustrate these points.
Case studies presented in the chapter:
- Children’s vulnerability to weather shocks: A natural experiment from the October 1998 Hurricane Mitch in Nicaragua.
- Residential relocation and obesity after a disaster: A natural experiment from the 2011 Great East Japan Earthquake and tsunami.
- Differences in endemic goitre prevalence in the Karakoram mountains, north Pakistan: a natural experiment suggesting an unrecognized cause.
What are the key messages of this chapter?
- In natural or human-instigated hazard contexts, implementing the traditional experimental design to study cause-effect relationship can be unfeasible or unethical.
- When people are assigned into exposure/treatment and control groups by chance, but in a way that resembles true randomization, natural experiments can be used to infer relationships between exposures and outcomes, just as in a traditional experiment.
- The credibility and validity of natural experiments depend on the persuasiveness of the ‘as if’ random assignment argument. The randomization ensures that the exposed and control groups are similar in their pre-exposure characteristics and hence mitigates the effects of observed and unobserved confounders.
- Quantitative analyses of pre-exposure characteristics and qualitative evidence around context and processes are useful for establishing the credibility of natural experiment design.
- If the assumption of random, or ‘as if’ random, assignment is persuasive, then the estimation of causal (or treatment) effect is as simple as taking the difference between the means of outcome from the treatment and control groups.