Integrated spatiotemporal surveillance system: Data, Analysis and Visualization
Bibliographic record
Abstract
ObjectiveTo build an open source spatiotemporal system that integratesanalysis and visualization for disease surveillanceIntroductionMost surveillance methods in the literature focus on temporalaberration detections with data aggregated to certain geographicalboundaries. SaTScan has been widely used for spatiotemporalaberration detection due to its user friendly software interface.However, the software is limited to spatial scan statistics and suffersfrom location imprecision and heterogeneity of population. RSurveillance has a collection of spatiotemporal methods that focusmore on research instead of surveillanceMethodsBased in Ontario, Canada, we used postal codes for determiningthe location of cases of reportable infectious diseases. The variationin geographic sizes and shapes of the case and census geographiescreated challenges for developing a uniform temporal spatialsurveillance system, including:Linking case and population data due to misclassification errors,Distance based correlations due to irregularly shaped areas(e.g. FSA’s), andVisualization bias due to variation in population density, e.g. largearea with little population.To overcome these challenges, we developed the Ontario HybridInformation Map (OHIM) boundary, which is a combination ofPublic Health Unit boundaries (rural areas), census subdivisions(rural urban mixed) and regular grid cells (urban). The goal is tocapture population details in urban areas without losing informationin rural areas. OHIM has around 4600 geographies with more thanhalf located in urban centers. Population distribution by gender andage group was calculated for each OHIM geography. A lookup filewas also created to link all Ontario postal codes to OHIM geography.To create baselines, historical data for influenza A were used tomodel the seasonality and calculate expected case count for eachOHIM geography for each week. Standardized incident ratios (SIR)were calculated as exploratory statistics, and a spatiotemporal Besag-York-Mollie (BYM) model was used to calculate the probability thatthe risk is higher than a pre-specified threshold. Integrated NestedLaplace Approximation (R-INLA) was used in R to explore differenttypes of spatiotemporal interactions and for fast Bayesian inference.The ability to apply the models was verified by examining previousoutbreaks and seeking the opinion of staff that routinely performsurveillance on influenza.To ensure the visualization integrates with the analysis, R packageShiny was used to build an interactive spatiotemporal visualizationon OHIM boundary utilizing Open Street Map and html5. Theapplication not only allows users to pan and zoom in space and timeto explore the results and locate high risk areas, it also gives users theflexibility to change algorithm parameters for instant feedback. Figure1 demonstrates a zoomed-in OHIM boundary with pointers signalfor “high risk” area at user specified statistics exceeds a threshold(e.g., SIR > 2). Using the algorithms and visualization tools,surveillance experts pick the optimal time and place to be notifiedbased on historical data and therefore the optimal threshold, whichwill be verified by prospectively running the algorithms.ResultsThe OHIM boundaries build the foundation for efficient spatialmodelling and visualization for public health surveillance in Ontario.Together with the integrated modelling and visualization system,staff are able to interactively optimize the aberration thresholds andidentify potential outbreaks in real time. Staff reported preference ofSIR due to its faster computations and easier interpretation.One major challenge was scalability: the ability to handle highresolutions of spatiotemporal data. When the system was applied on4600 polygons by 200 weeks, significant delays were encountered inboth analysis and visualization. Difficulties in computational time,memory requirement and visualization interactivity created delaysand freezing, thereby limited user experience. This problem waspartially addressed by optimizing parameters for fast computationsConclusionsThis work shows the “proof of concept” for an open source,customizable spatiotemporal surveillance system that overcomesexisting data challenges in Ontario. However, more work is requiredto make this fully operational and efficient in production.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.002 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".