Nathaniel Bell

Deprivation Indices, Population Health and Geography: An Evaluation of the Spatial Effectiveness of Indices at Multiple Scales

Area-based deprivation indices (ABDIs) have become a common tool with which to investigate the patterns and magnitude of socioeconomic inequalities in health. ABDIs are also used as a proxy for individual socioeconomic status. Despite their widespread use, comparably less attention has been focused on their geographic variability and practical concerns surrounding the Modifiable Area Unit Problem (MAUP) than on the individual attributes that make up the indices. Although scale is increasingly recognized as an important factor in interpreting mapped results among population health researchers, less attention has been paid specifically to ABDI and scale. In this paper, we highlight the effect of scale on indices by mapping ABDIs at multiple census scales in an urban area. In addition, we compare self-rated health data from the Canadian Community Health Survey with ABDIs at two census scales. The results of our analysis confirm the influence of spatial extent and scale on mapping population health—with potential implications for health policy implementation and resource distribution.

Using GIS-based methods of multicriteria analysis to construct socio-economic deprivation indices

BackgroundOver the past several decades researchers have produced substantial evidence of a social gradient in a variety of health outcomes, rising from systematic differences in income, education, employment conditions, and family dynamics within the population. Social gradients in health are measured using deprivation indices, which are typically constructed from aggregated socio-economic data taken from the national census – a technique which dates back at least until the early 1970s. The primary method of index construction over the last decade has been a Principal Component Analysis. Seldom are the indices constructed from survey-based data sources due to the inherent difficulty in validating the subjectivity of the response scores. We argue that this very subjectivity can uncover spatial distributions of local health outcomes. Moreover, indication of neighbourhood socio-economic status may go underrepresented when weighted without expert opinion. In this paper we propose the use of geographic information science (GIS) for constructing the index. We employ a GIS-based Order Weighted Average (OWA) Multicriteria Analysis (MCA) as a technique to validate deprivation indices that are constructed using more qualitative data sources. Both OWA and traditional MCA are well known and used methodologies in spatial analysis but have had little application in social epidemiology.ResultsA survey of British Columbias Medical Health Officers (MHOs) was used to populate the MCA-based index. Seven variables were selected and weighted based on the survey results. OWA variable weights assign both local and global weights to the index variables using a sliding scale, producing a range of variable scenarios. The local weights also provide leverage for controlling the level of uncertainty in the MHO response scores. This is distinct from traditional deprivation indices in that the weighting is simultaneously dictated by the original respondent scores and the value of the variables in the dataset.ConclusionOWA-based MCA is a sensitive instrument that permits incorporation of expert opinion in quantifying socio-economic gradients in health status. OWA applies both subjective and objective weights to the index variables, thus providing a more rational means of incorporating survey results into spatial analysis.

Are injuries spatially related? Join–count spatial autocorrelation for small-area injury analysis

Objective: To present a geographic information systems (GIS) method for exploring the spatial pattern of injuries and to demonstrate the utility of using this method in conjunction with classic ecological models of injury patterns. Design: Profiles of patients’ socioeconomic status (SES) were constructed by linking their postal code of residence to the census dissemination area that encompassed its location. Data were then integrated into a GIS, enabling the analysis of neighborhood contiguity and SES on incidence of injury. Setting: Data for this analysis (2001–2006) were obtained from the British Columbia Trauma Registry. Neighborhood SES was calculated using the Vancouver Area Neighborhood Deprivation Index. Spatial analysis was conducted using a join–count spatial autocorrelation algorithm. Patients: Male and female patients over the age of 18 and hospitalized from severe injury (Injury Severity Score >12) resulting from an assault or intentional self-harm and included in the British Columbia Trauma Registry were analyzed. Results: Male patients injured by assault and who resided in adjoining census areas were observed 1.3 to 5 times more often than would be expected under a random spatial pattern. Adjoining neighborhood clustering was less visible for residential patterns of patients hospitalized with injuries sustained from self-harm. A social gradient in assault injury rates existed separately for men and neighborhood SES, but less than would be expected when stratified by age, gender, and neighborhood. No social gradient between intentional injury from self-harm and neighborhood SES was observed. Conclusions: This study demonstrates the added utility of integrating GIS technology into injury prevention research. Crucial information on the associated social and environmental influences of intentional injury patterns may be under-recognized if a spatial analysis is not also conducted. The join–count spatial autocorrelation is an ideal approach for investigating the interconnectedness of injury patterns that are rare and occur in only a small percentage of the population.

A two-stage cluster sampling method using gridded population data, a GIS, and Google Earth TM imagery in a population-based mortality survey in Iraq

BackgroundMortality estimates can measure and monitor the impacts of conflict on a population, guide humanitarian efforts, and help to better understand the public health impacts of conflict. Vital statistics registration and surveillance systems are rarely functional in conflict settings, posing a challenge of estimating mortality using retrospective population-based surveys.ResultsWe present a two-stage cluster sampling method for application in population-based mortality surveys. The sampling method utilizes gridded population data and a geographic information system (GIS) to select clusters in the first sampling stage and Google Earth TM imagery and sampling grids to select households in the second sampling stage. The sampling method is implemented in a household mortality study in Iraq in 2011. Factors affecting feasibility and methodological quality are described.ConclusionSampling is a challenge in retrospective population-based mortality studies and alternatives that improve on the conventional approaches are needed. The sampling strategy presented here was designed to generate a representative sample of the Iraqi population while reducing the potential for bias and considering the context specific challenges of the study setting. This sampling strategy, or variations on it, are adaptable and should be considered and tested in other conflict settings.

Modelling optimal location for pre-hospital helicopter emergency medical services

BackgroundIncreasing the range and scope of early activation/auto launch helicopter emergency medical services (HEMS) may alleviate unnecessary injury mortality that disproportionately affects rural populations. To date, attempts to develop a quantitative framework for the optimal location of HEMS facilities have been absent.MethodsOur analysis used five years of critical care data from tertiary health care facilities, spatial data on origin of transport and accurate road travel time catchments for tertiary centres. A location optimization model was developed to identify where the expansion of HEMS would cover the greatest population among those currently underserved. The protocol was developed using geographic information systems (GIS) to measure populations, distances and accessibility to services.ResultsOur model determined Royal Inland Hospital (RIH) was the optimal site for an expanded HEMS – based on denominator population, distance to services and historical usage patterns.ConclusionGIS based protocols for location of emergency medical resources can provide supportive evidence for allocation decisions – especially when resources are limited. In this study, we were able to demonstrate conclusively that a logical choice exists for location of additional HEMS. This protocol could be extended to location analysis for other emergency and health services.

GIS and Injury Prevention and Control: History, Challenges, and Opportunities

Intentional and unintentional injury is the leading cause of death and potential years of life lost in the first four decades of life in industrialized countries around the world. Despite surgical innovations and improved access to emergency care, research has shown that certain populations remain particularly vulnerable to the risks and consequences of injury. Recent evidence has shown that the analytical, data linkage, and mapping tools of geographic information systems (GIS) technology provide can further address these determinants and identify populations in need. This paper traces the history of injury prevention and discusses current and future challenges in furthering our understanding of the determinants of injury through the use of GIS.

A small-area population analysis of socioeconomic status and incidence of severe burn/fire-related injury in British Columbia, Canada.

Socioeconomic determinants of injury have been associated with risk of burn in the UK and USA, but the relative significance of this impact is largely unknown across Canadian populations. The purpose of this study is to determine the extent to which socioeconomic status (SES) is linked to risk of burn in the province of British Columbia (BC) and identify the extent to which these findings are generalizable across both urban and rural population groups. Measures of SES were based on province-wide comparisons using data obtained from the Canada Census using the Vancouver Area Neighbourhood Deprivation Index (VANDIX). Results illustrate that the effects of SES and increased injury risk are substantial, though the most pronounced variations were exhibited across each SES stratum for urban areas and with less demonstrable effect when itemized by injury type within rural areas. Although conservative, the results from this study illustrate that burns disproportionately affect populations of greater relative socioeconomic disadvantage and continued efforts to also address social inequities and their link to injury incidence is likely to be more effective than targeting individual behavior alone when trying to reduce and eliminate their occurrence.

A Model for Identifying and Ranking Need for Trauma Service in Nonmetropolitan Regions Based on Injury Risk and Access to Services

BACKGROUND Timely access to definitive trauma care has been shown to improve survival rates after severe injury. Unfortunately, despite development of sophisticated trauma systems, prompt, definitive trauma care remains unavailable to over 50 million North Americans, particularly in rural areas. Measures to quantify social and geographic isolation may provide important insights for the development of health policy aimed at reducing the burden of injury and improving access to trauma care in presently under serviced populations. METHODS Indices of social deprivation based on census data, and spatial analyses of access to trauma centers based on street network files were combined into a single index, the Population Isolation Vulnerability Amplifier (PIVA) to characterize vulnerability to trauma in socioeconomically and geographically diverse rural and urban communities across British Columbia. Regions with a sufficient core population that are more than one hour travel time from existing services were ranked based on their level of socioeconomic vulnerability. RESULTS Ten regions throughout the province were identified as most in need of trauma services based on population, isolation and vulnerability. Likewise, 10 communities were classified as some of the least isolated areas and were simultaneously classified as least vulnerable populations in province. The model was verified using trauma services utilization data from the British Columbia Trauma Registry. These data indicate that including vulnerability in the model provided superior results to running the model based only on population and road travel time. CONCLUSIONS Using the PIVA model we have shown that across Census Urban Areas there are wide variations in population dependence on and distances to accredited tertiary/district trauma centers throughout British Columbia. Many of the factors that influence access to definitive trauma care can be combined into a single quantifiable model that researchers in the health sector can use to predict where to place new services. The model can also be used to locate optimal locations for any basket of health services.

Does direct transport to provincial burn centres improve outcomes? A spatial epidemiology of severe burn injury in British Columbia, 2001-2006.

BACKGROUND In Canada and the United States, research has shown that injured patients initially treated at smaller emergency departments before transfer to larger regional facilities are more likely to require longer stays in hospital or suffer greater mortality. It remains unknown whether transport status is an independent predictor of adverse health events among persons requiring care from provincial burn centres. METHODS We obtained case records from the British Columbia Trauma Registry for adult patients (age ≥ 18 yr) referred or transported directly to the Vancouver General Hospital and Royal Jubilee Hospital burn centres between Jan. 1, 2001, and Mar. 31, 2006. Prehospital and in-transit deaths and deaths in other facilities were identified using the provincial Coroner Service database. Place of injury was identified through data linkage with census records. We performed bivariate analysis for continuous and discrete variables. Relative risk (RR) of prehospital and in-hospital mortality and hospital stay by transport status were analyzed using a Poisson regression model. RESULTS After controlling for patient and injury characteristics, indirect referral did not influence RR of in-facility death (RR 1.32, 95% confidence interval [CI] 0.54- 3.22) or hospital stay (RR 0.96, 95% CI 0.65-1.42). Rural populations experienced an increased risk of total mortality (RR 1.22, 95% CI 1.00-1.48). CONCLUSION Transfer status is not a significant indicator of RR of death or hospital stay among patients who received care at primary care facilities before transport to regional burn centres. However, significant differences in prehospital mortality show that improvements in rural mortality can still be made.

Are We Failing Our Rural Communities? Motor Vehicle Injury in British Columbia, Canada, 2001–2007

In Canada, stratification by geographic area or socio-economic status remains relatively rare in national and provincial reporting and surveillance for injury prevention and trauma care. As injuries are known to affect some populations more than others, a more nuanced understanding of injury risk may in turn inform more effective prevention policy. In this study we assessed rates of hospitalization and death from motor vehicle collisions (MVC) in British Columbia (BC) by socio-economic status (SES) and by rural and urban status between 2001 and 2007. Excess risk in injury morbidity and mortality between different SES groups were assessed using a population attributable fraction (PAF). Over a six-year period rural populations in BC experienced a three-fold increase in relative risk of death and an average of 50% increase in relative risk of hospitalization due to injury. When assessed against SES, relative risk of MVC mortality increased from 2.36 (2.05-2.72) to 4.07 (3.35-4.95) in reference to the least deprived areas, with an estimated 40% of all MVC-related mortality attributable to the relative differences across SES classes. Results from this study challenge current provincial and national reporting practises and emphasize the utility of employing the PAF for assessing variations in injury morbidity and mortality.