Spatial Variability of Geriatric Depression Risk in a High-Density City

City: Hong Kong, Hong Kong, China
Organization: The Chinese University of Hong Kong,
Hong Kong Polytechnic University
Project Start Date: August 2001(start of data collection)
Project End Date: 30 August 2017 (article is accepted by the International Journal of Environmental Research and Public Health)
Reference: Ho, H. C. et al. (2017) ‘Spatial variability of geriatric depression risk in a high‐density city: A data-driven socio-environmental vulnerability mapping approach’, International Journal of Environmental Research and Public Health, 14(9). doi: 10.3390/ijerph14090994.
Problem: Problem:
Apply data-driven methods to predict the spatial variability of geriatric depression in a high-density city based on social vulnerability and environmental factors.

Context:
There is established literature that social vulnerability risk factors impact the risk of geriatric depression. Most of these studies do not include the influence of environmental factors (ie. the urban built environment). This is a transversal study which includes both social vulnerability and environmental factors for predicting geriatric depression risk in Hong Kong.
Technical Solution: Data preparation
  • Social vulnerability factors were gathered through a questionnaire and health records. It was then converted to usable social measures
  • Spatial data was prepared to “represent the conditions of the built environment within an approximately 400m of the subjects’ addresses” (Ho et al., 2017, p. 3) by “resample[ing] from their original spatial resolution into 1m resolution” (Ho et al., 2017, p. 3)
  • Out of a total of 4000 subjects, 3930 subjects were used in the training set. 364 subjects were identified as being geriatric depression cases and 3566 were identified as being controls
Modelling
  • Binomial logistic regression to estimate odds ratios (OR) using the 15 social vulnerability and environmental factors
  • 3 models were used to test for confounding factors
    • Model 1: only social vulnerability factors
    • Model 2: only environmental factors
    • Model 3: both social vulnerability and environmental factors
Evaluation
  • Statistically significant (p < 0.05) variables in Model 3 were used to construct a socio-environmental vulnerability index. Each variable in the index is weighted by its adjusted OR
  • Using this index and census information, a socio-environmental vulnerability heat map was generated and overlaid onto the “Tertiary Planning Unit” boundaries (Hong Kong’s most granular planning unit) for ease of interpretation
  • The training data was compared against the the vulnerability heat map. The comparison indicated a significant positive correlation between the two tools
Datasets Used: Dataset 1: age, gender, marital status, education level, living status, dementia, physical activity, alcohol consumption, smoking status
  • Title: social vulnerability factors
  • Source: questionnaire via recruitment notices in community centres for older adults and housing estates
  • Date: August 2001 to March 2003, 2002 to 2004
Dataset 2: cardiovascular-related diseases, respiratory-related diseases
  • Title: patient medical records
  • Source: unknown
  • Date: unknown
Dataset 3: % residential
  • Title: land utilization map with 10m spatial resolution
  • Source: Hong Kong Planning Department
  • Date: unknown
Dataset 4: % vegetation
  • Title: Normalized Difference Vegetation Index (NDVI) map at 15m spatial resolution
  • Source: IKONOS multispectral images
  • Date: unknown
Dataset 5: average building height, variation in building height
  • Title: building height and footprint information
  • Source: Hong Kong Planning Department
  • Date: unknown
Outcome: Pre-solution Performance:
No or limited prior prediction and mapping tools existed

Post-solution Performance:
A socio-environmental vulnerability index to predict geriatric depression risk was created. Using census data, this index was mapped onto the most granular planning unit in Hong Kong (“Tertiary Planning Unit”, or TPU) to create a planning and preventative health care tool for predicting geriatric depression risk.
Issues that arose: Problem understanding:
  • In accordance with research, the standard point ranges of the GDS-15 (“15 item geriatric depression scale") were adjusted to better suit an Asian population
Data preparation
  • Out of 4000 records, 98.6% was successfully geocoded and 98.3% of records had complete information
Evaluation
  • The study found no significant correlation between green space and geriatric depression, contrary to previous research and literature
  • The model may be biased toward high-density cities; that is, it may not perform well for medium- or low-density cities
Status: Operational

Further longitudinal studies may improve the flexibility of the socio-environmental vulnerability index through inclusion of multi-year data.

Further transversal studies may include temperature and air pollution environmental data as well as additional information on rural areas in Hong Kong.
Entered by: 28-Sep-2019: Daniel Tse, daniel.tse@mail.utoronto.ca


CEM1002,
Civil Engineering, University of Toronto
Contact: msf@eil.utoronto.ca