Background The number of people living with dementia and mild cognitive impairment (MCI) is increasing. A supportive urban environment can prevent or delay the progress of cognitive decline. There is evidence for the existence of causal mechanisms between the urban environment and cognitive decline, but the interrelations between these mechanisms are unclear. In this study, we aimed (1) to map the causal mechanisms by which urban environment factors impact cognitive decline in older adults creating a causal loop diagram (CLD), and (2) to translate a CLD into a series of directed acyclic graphs (DAGs).

Methods A 2-day workshop with 12 researchers was conducted using the group model building methodology. The workshop aimed to create a CLD that identifies established and potential urban environment, lifestyle, health, and physiological factors determining dementia and MCI, and the dynamic interrelationships between these factors. After the workshop, the CLD was reviewed to ensure that main hypothesised causal mechanisms were captured. From the CLD we focused on the relationship between ‘green space’ and ‘dementia’ and all the indirect connections between both. We used the focused CLD to develop a DAG. The use of CLDs and DAGs has been suggested to be incompatible due to DAGs being apparently unsuitable for modelling systems containing feedback loops. Because of that, when a feedback loop was identified, one arrow was considered as the main one based in the predominant relationship in the urban health literature. Once the main DAG was drawn, all the variables were identified as measured or unmeasured based in the available data.

Results The final CLD contains 34 factors and 109 connections. All factors were classified in nine main themes: urban design, social environment, travel behaviours, by-products, lifestyle, mental health conditions, disease/physiology, brain physiology, and dementia and MCI. Thousands of feedback loops were identified and five were selected to illustrate the main dynamics detected in the system. After focusing on green space and dementia, the resulted DAG presented a total of 19 variables being: one exposure (green space), one outcome (dementia), four confounders (material deprivation, education, sensory impairment, private vehicle use), and 14 mediators.

Conclusion The CLD detailed numerous plausible causal mechanisms and feedback loops between the urban environment and cognitive decline. The CLD has important implications to inform future analyses and identification of possible systems-based interventions to prevent cognitive decline. Making CLD and DAGs methodologies compatible can bring a broader scope to health and population research.