Table 3.
No. | Author (Year), Country | Objective | Study Design | Sample | Statistical Test | Findings | Factors Associated | Determinants | Limitation |
---|---|---|---|---|---|---|---|---|---|
1 | Liu et al. (2016), United States | To determine the association between median household income and the risk of developing colorectal cancer in Texas | Retrospective cohort | 155,534 men and women with colorectal cancer in 1995–2011 | Getis-Ord G, Poisson regression model (ArcGIS v10.1) |
Higher median household income, measured from the third to the highest income quintile ranging between $38,040 and $80,876, was associated with decreased risk of colorectal cancer in Texas. The Hispanics showed higher incidence rate of CRC compared to the Non-Hispanic Whites and Blacks at all time period, with slight decrease trend across higher median household income. |
County median household income level (Median household incomes in 254 counties classified into five quintiles) Ethnicity (Non-Hispanic Whites, Non-Hispanic Blacks, Hispanics) |
Social Biology |
Income variable was the median household income measured at county level, thus potential ecological fallacy. Median household income alone may not be a good proxy for SES. Lack of accuracy in population estimates may led to biased calculation for CRC incidence |
2 | Mansori et al. (2018), Iran | To determine the factors associated with the spatial distribution of the CRC incidence in the neighborhoods of Tehran, Iran | Ecological | 2815 new cases of CRC from 2008 to 2011 | The Moran Index, BYM model (using OpenBUGS version 3.2.3, ArcGIS 10.3, GeoDa) |
There was spatial autocorrelation at the level of the neighborhoods. Significant association was observed between women head of household, living in a rental house, no daily milk consumption in the household and higher household health expenditures against higher risk of CRC respectively. | Socioeconomic variables (SES): unemployed people aged 15 years and above, educated women (university level of education) aged 17 years old and above, women head of household, households without a car, those living in a rental house, households with income below the poverty line, people without insurance coverage. Risk factors: Households without daily fruit consumption, households without daily milk consumption, overweight people aged 15 years and above, smoking households. Health costs: Household health expenditures, expenditures on diagnoses, expenditures on medicine, expenditures on hospitals, expenditures on medical visits | Social Biology |
Ecological fallacy. Edge effect, referring to border neighborhoods affected by size of adjacent regions. Misclassification in geocoding due to incomplete postal addresses. |
3 | Mansori et al. (2019), Iran | To determine effective factors on geographic distribution of the CRC incidence in Iran | Ecological | 2815 new cases of CRC from 2008 to 2011 | Geographically Weighted Poisson Regression Model (using GWR 4, Stata 14, ArcGIS 10.3) | The spatial variability was observed with more frequent utilization of health services as indicated by the household health expenditures (Median Incidence Risk Ratios (IRR): 1.39), the cost of diagnosis of the disease (Median IRR: 1.03), the cost of household medicine (Median IRR: 1.05), the cost of hospital admission (Median IRR: 1.09) and the cost of medical visits (Median IRR: 1.27) | Socioeconomic variables (SES) include employment status, education Daily fruit consumption, daily milk consumption, overweight, smoking Health costs variables |
Social Biology |
Ecological fallacy. Unclear addresses for geocoding. Covariates of age, older than 50 years and overweight |
4 | Pakzad et al. (2016), Iran | To investigate the spatial distribution of colorectal cancer in Iran | Ecological | 6210 cases of CRC in 2009 | Getis-Ord-Gi* (spatial statistics) | Higher incidence of CRC among men (11.31 per thousand people) than women (10.89 per thousand people) in the northern and central provinces of Iran. | Sex | Biology | Incomplete data registration. Lack of full report Poor data classification |
5 | Pourhoseingholi et al. (2020), Iran | To determine the distribution of CRC risk in a map with socioeconomic risk factors adjustment | Cross-sectional | 21,543 CRC cases between 2005 and 2008 | Generalized Linear Model (using WINBUGS program, ArcGIS v10) |
Hotspot areas of CRC cluster identified for men were in the North and Western regions (mean SIR 1.92) while the Central provinces reported higher rate for women (mean SIR 1.85). Unemployment rate and mean household income had minimal impacts on CRC cluster. | Unemployment rate (mean ± SD: 11.64% ± 3.18%), Mean household income (mean ± SD: 66.46 Rials ± 12.04 Rials) | Social | Incomplete and lack of up-to-date data on population. |
6 | Goungounga et al. (2016), France | To compare empirically different cluster detection methods to find spatial clusters of cancer cases | Cross-sectional | 3084 CRC cases between 1998 and 2007 | Moran’s I, the empirical Bayes index (EBI) and Potthoff-Whittinghill test (using SpODT, SaTScan an HBSM) |
The socioeconomic inequalities did not affect the spatial variations of CRC incidence | Socioeconomic disadvantage as proxy by the calculation of Townsend index of deprivation. The index considers the proportion of unemployed people in the workforce, proportion of households without car, proportion of households renting and the proportion of overcrowded households. Increase in the Townsend index indicates an increase in the deprivation level of the inhabitants. | Social | Power of spatial cluster detection methods increases with the event rate |
7 | Roquette et al. (2019), Portugal | To describe and discuss the geographical patterns of CRC incidence and mortality in Portugal municipalities | cross sectional | 37,543 CRC cases between 2007 and 2011 | Global Moran’s Index and Local Moran’s Index (LISA), geographically weighted regression (GWR) (using ArcGIS) |
The CRC incidence was relatively higher in the Norte region to women than men. Meanwhile, areas in the coastal Centro, the LVT and the Alentejo region showed markedly higher CRC cases among men than to women. | Sex | Biology | Limited data availability on other risk factors |
8 | Torres et al. (2018), United States | To evaluate the geographic distributions of colorectal cancer incidence among female residents in Baltimore City, Maryland and the neighborhood characteristics associated with those distributions | retrospective cohort | 1120 female CRC patients between 2000 and 2010 | Spatial clusters identified using Getis-Ord-Gi* statistic and local Moran’s I, global ordinary least square regression model (Using STATA, ArcGIS) |
Cluster spot for CRC was identified in two out of 55 Community Statistical Area (CSA) studied. The findings noted that every one percent increase in African-American residents resulted in CRC incidence increasing by 0.031 times per 1000 female residents. The CRC cluster spot experienced less crime with majority residents between the ages of 50 and 74 years old. | The 2012 Baltimore Neighborhood Indicators Alliance Vital Signs report was referred for the neighborhood characteristics. The indicators include females aged 50 to 74 years, percentage of African-American, female-headed households, percentage of households earning less than $25,000, percentage of vacant residential properties, housing violations, number of crime incidents per 1000 residents, count of emergency call for domestic violence, teen birth, employment rates, dirty streets, tree coverage and neighborhood associations. | Social Ecology |
Social determinants limited by the residential environment. No information on the length of residency on their addresses. |
9 | Halimi et al. (2019), Iran | To explore the spatial pattern of CRC incidence in Hamadan province, Iran | Cross-sectional | 805 CRC cases during 2007 to 2014 | Local Moran’s I (MS Excel, Arc GIS 10.5) |
High-high clusters of CRC incidence identified in Mohajeran and Lalejin areas. Majority of the CRC incidence were among men (54%) and those in the age group between 65 and 85 years. | Sex, Age | Biology | Lack of accuracy for registries. Census population data is used |
10 | Singh et al. (2017), Canada | To determine the variation of CRC incidence by average household income in area of residence | Cross-sectional | 19,484 CRC cases between 1985 and 2012 | Bayesian Poisson regression models (Using WinBUGS software, ArcGIS v10.3) |
There were few small geographic areas in the southwest rural Manitoba with persistent high CRC incidence | Sex, Age, Mean annual household income, proportion of recent immigrants (since 1961 to 2001), rate of visible minority status and unemployment status. | Biology Social |
Ecological analysis, hence result should be interpreted in the context of area of residence |
11 | Kuo et al. (2019), United States | To address spatial autocorrelation between CRC and county-level determinants | Cross sectional | 2003 to 2013 | Local Indicators of Spatial Association (LISA), Moran’s I (Using ArcMap, SAS, GeoDaSpace software) |
The location of high-high clusters identified in the north-eastern counties | Age, Adults with BMI ≥ 30 kg/m2 (Obesity), Current smoker adults who smoked at least 100 cigarettes in lifetime, Socioeconomic Status deprivation composed of education level, employment rate, income level, family and social support. Ethnicity studied include non-Hispanic Black, Hispanic, Native American and Asian. Percentage of population aged less than 65 years without health insurance. Urbanicity classified into urban, large town and rural counties. | Biology Social Ecology |
Result may not generalize to other geographic areas. Lack of person-level and tumor level data |
12 | Goshayeshi et al. (2019), Iran | To identify potential spatial factors contributing to its geographical distribution | Cross sectional | 1089 CRC cases between 2016 and 2017 | Local Moran’s I, Ordinary Least Square Regression (Using MS Excel, ArcGIS v10.6) |
CRC clusters identified in Rezashahr, Sarafrazan and Nofel-Loshato areas. The neighborhood of CRC hotspots areas was associated with high proportion of population with 50 years and above, obesity (Body Mass Index ≥30 kg/m2), daily fibre intake (≤25 g). | Age, Body Mass Index (BMI), daily consumption of red meat (gram), daily consumption of fibre (gram) | Biology | Study did not consider processed meats. Patients who had shifted were not included. Inability to geocode Persian addresses affect the accuracy |