Open government data, uncertainty and coronavirus: An infodemiological case study.

Governments around the world have made data on COVID-19 testing, case numbers, hospitalizations and deaths openly available, and a breadth of researchers, media sources and data scientists have curated and used these data to inform the public about the state of the coronavirus pandemic. However, it is unclear if all data being released convey anything useful beyond the reputational benefits of governments wishing to appear open and transparent. In this analysis we use Ontario, Canada as a case study to assess the value of publicly available SARS-CoV-2 positive case numbers. Using a combination of real data and simulations, we find that daily publicly available test results probably contain considerable error about individual risk (measured as proportion of tests that are positive, population based incidence and prevalence of active cases) and that short term variations are very unlikely to provide useful information for any plausible decision making on the part of individual citizens. Open government data can increase the transparency and accountability of government, however it is essential that all publication, use and re-use of these data highlight their weaknesses to ensure that the public is properly informed about the uncertainty associated with SARS-CoV-2 information.

Expressions of pro- and anti-vaccine sentiment on YouTube.

Billions of hours of YouTube content are viewed every day. Much of this content is aimed at entertainment, some of it is educational, and a considerable quantity is meant to influence or reinforce public opinion on a variety of matters, including health. Most of the content on YouTube is not created by professionals, public institutions or the traditional media, and instead is authored by private individual content creators. Given the potential impact of this medium for communicating health information, it is important for researchers and public health practitioners to understand the nature of health information as it is shared on YouTube. The primary objective of this research is to describe expressions of vaccine hesitancy content on YouTube, and specifically, compare the expression of pro- and anti-immunization sentiments. We do this by not only analyzing a systematic sample of influenza and measles immunization videos in terms of viewer analytics, but also by choice of language. We find that pro- and anti-immunization videos are common, but that videos with anti-immunization sentiment tend to be more 'liked'. We also find that a small number of words can be effectively used to identify anti-immunization content, an observation that could be useful for identifying trends in anti-immunization sentiment on social media. Our results suggest that public health experts may need to increase the profile of their content on YouTube, and that there may be some useful strategies for improving the public's disposition towards pro-immunization messaging.

Motor-vehicle collisions involving child pedestrians at intersection and mid-block locations.

We study motor-vehicle collisions involving child pedestrians walking to school in Hamilton, Ontario, Canada to understand and contrast collision risks at mid-block and intersection locations. We use a matched case-control study design and apply it to intersection and mid-block locations instead of people. Cases are intersections/mid-blocks where collisions occurred and controls are locations where collisions did not occur. We match cases to controls on geography, socio-economic status and year. We use conditional logistic regression to predict the log-odds of collision risk at intersections and mid-blocks as a function of various environmental measures while controlling for volume of child pedestrian activity. Our results suggest that child pedestrian injuries at intersections are associated with intersection control type, traffic volume, and land use characteristics. In contrast, mid-block child pedestrian collisions are not associated with small scale environmental features. The results of this study suggest that some factors associated with the risk of collision differ across location types. These findings may be useful in the planning of safer walking journeys to school.

Altering school attendance times to prevent child pedestrian injuries.

OBJECTIVES: The purpose of this research was to determine whether modifying school start time schedules can be used to reduce children's exposure to traffic on their morning walks to school. METHODS: We use models of pedestrian and motor vehicle commuting to estimate the frequency of encounters between child pedestrians and motor vehicles at intersections throughout the City of Hamilton, Ontario, Canada. We use a simple heuristic to identify the school-specific start times that would most reduce the local frequency of encounters between motor vehicles and pedestrians. RESULTS: Our analysis suggests that it may be possible to achieve an almost 15 percent reduction in the total number of encounters between child pedestrians and motor vehicles during the morning commute by staggering school start times such that the periods of high pedestrian activity are temporally staggered from periods of high motor vehicle activity. Our analysis suggests that small changes in school start times could be sufficient to see noteworthy reductions in pedestrian exposure to traffic. CONCLUSIONS: Changing school times may be an effective, inexpensive, and practical tool for reducing child pedestrian injuries in urban environments. Enhanced transportation models and community-based interventions are natural next steps for exploring the use of school-specific scheduling to reduce the risk of child pedestrian injury. Further research is required to validate our models before this analysis should be used by policy makers.

The effects of local and non-local traffic on child pedestrian safety: a spatial displacement of risk.

In most places, motor-vehicle traffic volume is associated with increased risk of child pedestrian injury; however, the burden of risk is geographically complex. In some neighbourhoods, proportionally fewer drivers may be local, meaning that the moral and practical responsibility of risk to children is displaced from one place (e.g., the suburbs) to another (e.g., downtown). Using the City of Toronto, Canada, as a case study, this research asks two related questions: 1) what is the variation in traffic volume by neighbourhood of origin and socioeconomic status and 2) what is the relationship between the geographical origin of traffic and the risk of collisions involving child pedestrians and motor-vehicles? We find that low-income downtown neighbourhoods have the highest proportion of non-local traffic. We also find that while higher local traffic activity is associated with lower risk of collision, higher flow-through traffic activity (excluding traffic from major thoroughfares) is associated with higher risk of collision. We interpret the former as very likely a proxy of parents' frequency of chauffeuring children to school, and the latter an illustration of the spatial displacement of risk between Toronto neighbourhoods. Our results suggest that more attention needs to be paid to account for the externalization of harm experienced by children, particularly in low-income downtown neighbourhoods.

Mapping commuter cycling risk in urban areas.

Cycling is becoming an increasingly important transportation option for commuters. Cycling offers exercise opportunities and reduces the burden of motor vehicle travel on society. Mapping the risk of collision between cyclists and motor vehicles in urban areas is important to understanding safe cyclist route opportunities, making informed transportation planning decisions, and exploring patterns of injury epidemiology. To date, many geographic analyses and representations of cyclist risk have not taken the concept of exposure into account. Instead, risk is either expressed as a rate per capita, or as a count of events. Using data associated with the City of Hamilton, Canada, we illustrate a method for mapping commuter cyclist collision risk per distance travelled. This measure can be used to more realistically represent the underlying geography of cycling risk, and provide more geographically and empirically meaningful information to those interested in understanding how cycling safety varies over space.

Spatial aberration vs. geographical substance: representing place in public health surveillance.

Public health surveillance involves the routine and ongoing collection, analysis and dissemination of health information for a variety of stakeholders-including both public health officials and the public. Much of the current focus of public health surveillance is on detecting aberrations in space-largely inspired by concerns about bioterrorism and newly emerging infectious diseases. We argue that the current focus on spatial aberrations has limited the development of public health surveillance by excluding a more explicit geographical understanding and representation of place. A more place-focused public health surveillance could represent geography in ways that are useful to a broader audience, provide information on the social and physical contexts related to health, facilitate a better understanding of health inequalities, and can benefit from local knowledge. Geographers can make important contributions to public health practice by contributing to more meaningful definitions of place in the design and operation of public health surveillance systems.

Child pedestrian injuries and urban change.

BACKGROUND: The health impacts of rapid changes in urban environments due to economic growth and/or retraction are not widely known. This study looks at the effects of urban change on the risk of child pedestrian injury in Edmonton, Alberta, a city that has experienced large economic and population growth following the expansion of the oil and gas industry in Canada. METHODS: A longitudinal ecological study design was used to model the relationships between several built and social environmental variables and the risk of child pedestrian injury and severe child pedestrian injury between 1996 and 2007. RESULTS: The incidence of child pedestrian injury was stable, but the incidence of severe injury increased over the study period. Areas with higher proportions of families on low incomes had higher injury incidence. While new residential development is associated with a lower incidence of injury in most areas, in poor areas, new residential development is associated with a higher incidence, even after controlling for urban planning features and traffic intensity. CONCLUSION: While suburban areas have a lower incidence of child pedestrian injury, residential development in poorer areas is associated with a higher child pedestrian injury risk. Child pedestrians may be less able to adapt to changes in the urban environment due to rapid growth and increasing income, and as a result, may be at greater risk of injury.

An integrated framework for the geographic surveillance of chronic disease.

BACKGROUND: Geographic public health surveillance is concerned with describing and disseminating geographic information about disease and other measures of health to policy makers and the public. While methodological developments in the geographical analysis of disease are numerous, few have been integrated into a framework that also considers the effects of case ascertainment bias on the effectiveness of chronic disease surveillance. RESULTS: We present a framework for the geographic surveillance of chronic disease that integrates methodological developments in the spatial statistical analysis and case ascertainment. The framework uses an hierarchical approach to organize and model health information derived from an administrative health data system, and importantly, supports the detection and analysis of case ascertainment bias in geographic data. We test the framework on asthmatic data from Alberta, Canada. We observe high prevalence in south-western Alberta, particularly among Aboriginal females. We also observe that persons likely mistaken for asthmatics tend to be distributed in a pattern similar to asthmatics, suggesting that there may be an underlying social vulnerability to a variety of respiratory illnesses, or the presence of a diagnostic practice style effect. Finally, we note that clustering of asthmatics tends to occur at small geographic scales, while clustering of persons mistaken for asthmatics tends to occur at larger geographic scales. CONCLUSION: Routine and ongoing geographic surveillance of chronic diseases is critical to developing an understanding of underlying epidemiology, and is critical to informing policy makers and the public about the health of the population.

Geographic hierarchies of diagnostic practice style in cerebrovascular disease.

Diagnostic practice style describes the ways in which physicians diagnose information about disease. Like practice style effects in general, diagnostic practice style effects may emerge as the result of training, inter-personal relationships between professionals, medical enthusiasm for particular diagnoses and patient-physician interactions. In this study we analyze the ways in which patterns of diagnostic practice style associated with cerebrovascular disease varies at different socio-geographical scales in the province of Alberta, Canada. We use hierarchical linear models to partition a measure of diagnostic practice style into four levels of observation: the physician level, the facility level, the municipality level and the regional (census division) level. We model a variety of fixed effects related to physician attributes, their practice, the facilities they work in and the municipalities within which their facilities operate. Our results suggest that attributes related to physicians and the facilities and municipalities in which they work all contribute to patterns of diagnostic practice style. Physicians working in rural and urban municipalities have different practice style patterns even after controlling for the types of facilities they work in, their professional medical specialization and their workload. Similar to other research, our results reveal that physicians have different diagnostic practice styles with members of the same sex than members of the opposite sex. Geographic variations in diagnostic practice style may obscure changes in the epidemiology of cerebrovascular disease in rural communities, and provide indirect evidence that the quality and/or timeliness of diagnosis may be worse in rural Alberta.

Parkinson's disease, multiple sclerosis and changes of residence in Alberta.

BACKGROUND: Our objective is to examine how persons diagnosed with Multiple Sclerosis (MS) and Parkinson's disease (PD) change residence following disease onset. We hypothesize that persons choose to change residence (locally or regionally) in different ways depending on whether or not they have been diagnosed with MS/PD. We also estimate the effects of residence change on measures of disease prevalence made at several different levels of geography. METHODS: Using fee-for service and hospitalization data, we identify cases of MS and PD between 1994 and 2004. Both of these case groups are matched to controls based on age, sex, socioeconomic status and municipality of residence. We tabulate and compare the changes of residence among persons in the case and control groups. We also use these data to estimate the effects that changes in residence have on disease prevalence at three different levels of geography. RESULTS: Both MS and PD patients were more likely to change residence following disease onset compared to groups of matched controls (p<=0.001). Most changes of residence occur within the same municipality. The total magnitude of these changes is small, however, and is unlikely to affect estimates of disease prevalence; over our study period, the largest change in geographical prevalence estimates due to individual changes in residence was about 1%. CONCLUSIONS: Persons diagnosed with MS and PD both have mobility characteristics that differ from those of their respective control groups, and in general, are more likely to move to or between Edmonton and Calgary, and less likely to move out of province. However, the balance of mobility characteristics of persons with PD and MS appear unlikely to greatly affect the patterns observed on maps of disease prevalence.

Adaptations for finding irregularly shaped disease clusters.

BACKGROUND: Recent adaptations of the spatial scan approach to detecting disease clusters have addressed the problem of finding clusters that occur in non-compact and non-circular shapes--such as along roads or river networks. Some of these approaches may have difficulty defining cluster boundaries precisely, and tend to over-fit data with very irregular (and implausible) clusters shapes. RESULTS & DISCUSSION: We describe two simple adaptations to these approaches that can be used to improve the effectiveness of irregular disease cluster detection. The first adaptation penalizes very irregular cluster shapes based on a measure of connectivity (non-connectivity penalty). The second adaptation prevents searches from combining smaller clusters into large super-clusters (depth limit). We conduct experiments with simulated data in order to observe the performance of these adaptations on a number of synthetic cluster shapes. CONCLUSION: Our results suggest that the combination of these two adaptations may increase the ability of a cluster detection method to find irregular shapes without affecting its ability to find more regular (i.e., compact) shapes. The depth limit in particular is effective when it is deemed important to distinguish nearby clusters from each other. We suggest that these adaptations of adjacency-constrained spatial scans are particularly well suited to chronic disease and injury surveillance.

West Nile virus: strategies for predicting municipal-level infection.

The appearance and spread of West Nile virus (WNv) in North America represent a recent example of how mosquito-borne diseases can develop in new settings. Understanding the epidemiological, biological, and geographical aspects of WNv is critical to developing a greater understanding of how newly emerging, migrating, or evolving vector-borne infectious disease can develop globally. To aid in the allocation of resources that mitigate future outbreaks and to better understand the geographic nature of WNv in the North American prairies, we employ spatial and nonspatial modeling methods to predict municipal-level risk of human WNv infection rates. We use data based on a combination of routinely collected electronic data sources. Our findings suggest general agreement between spatial and nonspatial approaches, and results are consistent with seroprevalence-based estimates. We suggest that spatial models based on administrative data can offer estimates of relative risk in human populations at less cost, and in a timelier manner than estimates based on serology specimens.

Modelling geographic variations in West Nile virus.

BACKGROUND: This paper applies a method for modelling the spatial variation of West Nile virus (WNv) in humans using bird, environmental and human testing data. METHODS: We used data collected from 503 Alberta municipalities. In order to manage the effects of residual spatial autocorrelation, we used generalized linear mixed models (GLMM) to model the incidence of infection. RESULTS: There were 275 confirmed cases of WNv in the 2003 calendar year in Alberta. Our spatial model indicates that living in the grasslands natural region and levels of human testing are significant positive predictors of WNv; living in an urban area is a significant negative predictor. CONCLUSION: Infected bird data contribute little to our model. The variability of West Nile virus incidence in Alberta may be partly confounded by the variations in the rate of testing in different parts of the province. However, variation in infection is also associated with known environmental risk factors. Our findings are consistent with existing knowledge of WNv in North America.

Incident cerebrovascular disease in rural and urban Alberta.

STUDY OBJECTIVE: This study examines the pattern of incidence and health service utilisation of cerebrovascular disease cases in urban and rural areas within a publicly funded health care system. DESIGN: A population-based study covering a large geographic region, using population-wide administrative health data. Age- and sex-standardised incidence and mortality rates were calculated for rural and urban areas. Final status (discharge or death), place of service and place of residence were reported for all cases across several different subsets of cerebrovascular disease. SETTING: The province of Alberta, located in western Canada. PARTICIPANTS: Incident cases of cerebrovascular disease (stroke and transient ischaemic attack) and 4 different definitions of incident stroke were identified from data on emergency department admissions in the 1999/2000 fiscal year. MAIN RESULTS: The rate of cerebrovascular disease per 10,000 was similar between urban (13.24) and rural (13.82) areas. Rural residents frequently reported their incident episode to urban emergency departments. Although the mortality is similar between urban and rural residents, rural dwellers die more frequently in the emergency department setting than urban dwellers, who die more often as in-patients. CONCLUSIONS: Overall mortality is similar between urban and rural residents. A large proportion of rural residents receive diagnoses and treatment for cerebrovascular disease in urban areas. Location of service and location of death differs between rural and urban cases of cerebrovascular disease.

Zones of prevention: the geography of fall injuries in the elderly.

Our investigation of the geography of fall injuries considers the relationship between injury prevention and contextual approaches to health research. We use a geographic information system (GIS) to describe the pattern of emergency department reported falls of the elderly in the Capital Health Region, an administrative health area in Alberta, Canada. We used empirical Bayes estimates to obtain a geographic measure of fall incidence over the study area and a cluster detection statistic to measure the presence of a significant spatial cluster in the region. Inner-city Edmonton had the highest incidence of risk, suburban Edmonton the lowest, and surrounding rural regions and smaller communities had more moderate fall incidence. We argue that descriptive geography can enhance the effectiveness of injury prevention programs by identifying zones of high risk, even when the individual-level and contextual factors that explain the underlying patterns are unknown.

Regional comparisons of inpatient and outpatient patterns of cerebrovascular disease diagnosis in the province of Alberta.

The diagnosis of cerebrovascular disease (CBVD) from administrative data has been critically examined by epidemiologists in recent years. Much of the existing literature suggests that hospital discharge diagnoses based on ICD-9-CM codes are an unreliable source of information for determining a diagnosis of stroke, particularly when four- and five-digit codes are used. We examined how diagnoses for CBVD in hospital inpatient and outpatient facilities vary between rural and urban areas and among the 16 administrative health regions. Our analysis revealed differences in diagnostic patterns between the two sources of data, differences between rural and urban areas, and variation across most of the regions. Geographic variation in health service utilization, diagnostic practices, specialty of the physician making the diagnosis, and disease burden may explain our findings. Our results suggest that the diagnosis of patients attending rural facilities are either coded differently (and less precisely) than those of urban residents or are coded more precisely only after the patients attend urban facilities. Regional differences in coding practices show that any CBVD surveillance system based on administrative data requires a large-scale (in this case, province-wide) and person-oriented approach.