How to do (or not to do)…using causal loop diagrams for health system research in low and middle-income settings

Abstract

Causal loop diagrams (CLDs) are a systems thinking method that can be used to visualize and unpack complex health system behaviour. They can be employed prospectively or retrospectively to identify the mechanisms and consequences of policies or interventions designed to strengthen health systems and inform discussion with policymakers and stakeholders on actions that may alleviate sub-optimal outcomes. Whilst the use of CLDs in health systems research has generally increased, there is still limited use in low- and middle-income settings. In addition to their suitability for evaluating complex systems, CLDs can be developed where opportunities for primary data collection may be limited (such as in humanitarian or conflict settings) and instead be formulated using secondary data, published or grey literature, health surveys/reports and policy documents. The purpose of this paper is to provide a step-by-step guide for designing a health system research study that uses CLDs as their chosen research method, with particular attention to issues of relevance to research in low- and middle-income countries (LMICs). The guidance draws on examples from the LMIC literature and authors' own experience of using CLDs in this research area. This paper guides researchers in addressing the following four questions in the study design process; (1) What is the scope of this research? (2) What data do I need to collect or source? (3) What is my chosen method for CLD development? (4) How will I validate the CLD? In providing supporting information to readers on avenues for addressing these key design questions, authors hope to promote CLDs for wider use by health system researchers working in LMICs.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 65 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.