Geo-temporal patterns to design and cost-effectively assign and evaluate the temporary and differentiated allocation of scarce resources against non-zoonotic and zoonotic diseases

Abstract

A computer-implemented method for predicting a dissemination pattern of a disease is provided. The method includes obtaining data of infections caused by the disease ("infection data"), the infection data including at least geographical information of each infection, time information of each infection, and infection information for one or more species. The infection data is binned by species, predetermined time period, and geographic unit. One or more disease dissemination patterns is identified for each species of the one or more species using the binned infection data. A prediction of future disease spread is generated based on the one or more identified disease dissemination patterns.

Description

GEO-TEMPORAL PATTERNS TO DESIGN AND COST-EFFECTIVELY ASSIGN AND EVALUATE THE TEMPORARY AND DIFFERENTIATED ALLOCATION OF SCARCE RESOURCES AGAINST NON-ZOONOTIC AND ZOONOTIC DISEASESCross-Reference to Related Applications

[0001] This application claims priority to U.S. Provisional Application No. 63 / 608.670, filed on December 11, 2023, now pending, the disclosure of which is incorporated herein by reference.Field of the Disclosure

[0002] The present disclosure relates to epidemiology, and more particularly to biological, geo-referenced and temporal data analysis, cost-benefit oriented decision making, and economics.Background of the Disclosure

[0003] The COVID-19 pandemic brought many lessons and questions. One of them refers to whether epidemics should be countered with reactive or anticipatory approaches. In both cases, cost-benefit oriented studies are needed.Brief Summary of the Disclosure

[0004] Control of diseases, including, for example, zoonosis (diseases affecting human and non-human species) and non-zoonotic diseases, can benefit from geo-referenced procedures. Focusing on brucellosis, here the ability of two methods to distinguish disease dissemination patterns and promote cost-effective interventions was compared. Geographical data on bovine, ovine and human brucellosis reported in the country of Georgia between 2014 and 2019 were investigated with (i) the Hot Spot (HS) analysis and (ii) a bio-geographical (BG) alternative. The BG approach identified small areas with a twice higher case density than the HS method. The BG method also identified, in 2019, at least 5 times more cases in zoonotic (human and non- human) sites than sites that only reported human cases — a finding that, if corroborated, could support policies that prioritize interventions of zoonotic sites. This proof-of-concept provided a preliminary validation for a method for cost-effective interventions oriented to control zoonoses and non-zoonotic diseases.Description of the Drawings

[0005] For a fuller understanding of the nature and objects of the disclosure, reference should be made to the following detailed description taken in conjunction with the accompanying drawings, in which:Figure 1. A chart depicting a method according to the present disclosure.Figure 2. Brucellosis cases reported in cattle and sheep, in 2014-2015. A: cattle cases, 2014. B: sheep cases, 2014. C: cattle cases, 2015. D: sheep cases, 2015.Figure 3. Brucellosis cases reported in cattle and sheep, in 2016-2017. A: cattle cases, 2016. B: sheep cases, 2016. C: cattle cases. 2017. D: sheep cases, 2017.Figure 4. Brucellosis cases reported in cattle and sheep, in 2018-2019. A: cattle cases, 2018. B: sheep cases, 2018. C: cattle cases, 2019. D: sheep cases, 2019.Figure 5. Brucellosis cases reported in humans, in 2015-2018. A: cases reported in 2015. B: cases reported in 2016. C: cases reported in 2017. D: cases reported in 2018.Figure 6. Estimates of internal and external validity. Five variables (including two metrics — counts or percentage — and two species — cattle and sheep populations) supported the hypothesis that, within the timeframe investigated, brucellosis peaked in 2017 (blue rectangle).Figure 7. Estimates of external and statistical validity. Data from two ruminant species conveyed similar inferences: there is a linear and statistically significant relationship (p < 0.01) between case counts of (either species) and the case count of ruminant sites. A: Cattle data. B: Sheep data.Figure 8. Evaluation of construct validity’. To assess cost-effectiveness, the same data were analyzed by the bio-geographical method (BG) and the Hot Spot analysis (HS). Using ruminant data from 2018 as an example, it is shown that the BG method identified 139 25-sq km orange and red squares, which included 583 cases or 4. 19 cases / square (583 / 139, A). In contrast, the HS analysis found 521 cases in 194 squares or 2.68 cases / square (521 / 194, B). Consequently, the case density of the BG approach was 56.3% higher (4. 19 / 2.68) than that of the HS. This difference in potential cost-effectiveness was explained by two factors: (i) the HS missed large areas that included numerous cases (C) and, (ii) in particular, the HS analysis missed nine mini-areas with a very high case density (D). Such a difference was achieved while the BG analysis occupied an area 28.4% smaller (139 / 194 squares or 71.6%) than the area covered by the HS analysis. Ifcost effectiveness of interventions was measured as the ratio of benefits over costs (here expressed as cases captured / area unit), then the ratio of the BG method would be 2.18 (156.3 / 71.6), i.e., the BG method exhibited a benefit / cost ratio twice as large as the one shown by the HS analysis.Figure 9. A bio-geographical analysis may guide cost-benefit oriented, prioritized interventions. Using the BG method to explore 2019 data as an example, it is shown that zoonotic sites (geographical locations where human and non-human cases were detected) reported between 5.2 and 7.8 times more cases than sites that only included human cases (1018 / 188 and 1460 / 188, respectively, A). This metric — together with additional geobio-temporal information related to such sites — may provide decision-makers with a guide on where interventions may be prioritized. In this scenario, sites that report zoonotic cases could be the first priority, followed by sites where two species (not including humans) are infected, and finally, sites that only report human cases.Every thing else equal, such a policy could be at least 5 times more effective than policies that prioritize sites where only human cases are reported.Figure 10. Comparison of results — an embodiment of the presently-disclosed bio-geo- temporal (“BGT”) method (Figure 10(A)) was at least twice less costly / more beneficial than a classic spatial statistical test (Figure 10(B)-(D)).Detailed Description of the Disclosure

[0006] How, when, and where can implemented interventions lead to cost-benefit based results? To answer this composite question, the type of data analyzed is critical. The analysis of geo-referenced and temporal infectious disease-related data may determine whether intervening specific geographical sites induce cost-effective policies.

[0007] Because numerous (if not infinite) geo-temporal patterns may be found in disseminating infectious diseases, geo-referenced and temporal data may also inform on covariates, such as soil, elevation, meteorology, seasonality, and sociology. Because they can — visually — reveal interactions, geo-referenced data can inform more than tabular data. Because the geographical context surrounding diseases may be unique and it may influence (promoting or preventing) their dissemination, geo-temporal analysis of diseases can capture relationships that reductionist approaches may omit or not anticipate.

[0008] One question that decision-makers need to answer is where, exactly, interventions may lead to less costly, earlier and / or more beneficial results. To develop geo-referenced, decision-making oriented analyses, inter / transdisciplinary approaches have been recommended. Such approaches may consider bio-geographical and dynamical data that may feed models meant to interrupt disease transmission and / or be cost-effective.

[0009] Cost / benefit-oriented, inter / transdisciplinary approaches for prevention or control of diseases are highly needed — for example, as was revealed by the COVID-19 pandemic. In an aspect, the present disclosure provides methods and systems for bio-geo-temporal, cost / benefit analysis.

[0010] The present method provides a quadruple concept — a biological-geographical- temporal and cost-benefit oriented analysis. The method may use publicly available information, proprietary information, or combinations of both. In some embodiments, proprietary7methods (e.g., spatial tools) may be used as a part of the present method, and the present method provides further benefit in via bio-geo-temporal techniques. Results may be compared against alternative methods. Embodiments analyze variables that are investigated along (at least) four domains (biological, geographical, temporal, cost-benefit). Embodiments of the present disclosure include and exceed previous geographical methods or geo-temporal methods which involve epidemiology7, microbial identification, viral incubation periods, diagnostic tests, etc.

[0011] In an aspect, the present disclosure may be embodied as a computer-implemented method 100 of predicting a disease dissemination pattern. The method 100 includes obtaining 103, by a computer system, data of infections (‘"infection data”) caused by the disease for one or more species. In some embodiments wherein the obtained data includes at least two species, at least one of the species is human. The disease may' be a zoonotic disease or a non- zoonotic disease. For each infection represented in the infection data, the data includes geographical information (e.g.. spatial coordinates, subdivision type, subdivision identification, etc.), time information, and identification of affected species. The infection data is binned 106 by the computer system by species, predetermined time period, and geographic unit. The predetermined time period may be any time period relevant to the particular application. For example, the relevant time period may be calendar year, month, academic year, disease seasonality, etc. The geographic unit may be any relevant unit such as, for example, a regular rectangular grid (e.g., 5 km x 5 km grid), hexagonal grid, or any other geographic unit. In someembodiments, the geographic unit may be related to one or more preexisting constructs such as, for example, county or other political border, postal code, cellular tower grid, fire district, etc.

[0012] As further described below under the heading of Further Discussion, the method 100 includes identifying 109 a disease dissemination pattern for each species using the binned infection data. In embodiments where the obtained data includes two or more species, the method may include identifying geographic units where at least two species are affected during the same time period. In some embodiments, the method further includes identifying a hot spot (e.g., area of high infection) that recurs periodically. For example, the method may include identifying a geographic unit with periodic (e.g., annual) recurrence. Identify ing disease dissemination patterns may include identifying bins (e.g., distinct bio-geo-temporal units) having significantly different properties than neighboring bins. For example, a bin where the temporal variance is significantly different (test positivity increases 2 faster than test positivity of neighboring bins). Other properties may be used such as, for example, number of species affected, test positivity, population density, etc. or combinations of properties.

[0013] In some embodiments, identifying one or more disease dissemination patterns for each species using the binned infection data is further based on one or more feature selected from: (a) a percentage of ‘positive’ test results (individual diagnosed as affected, i.e., ‘cases’) over the total number of people tested (i.e., test positivity percentage); (b) a classification and / or subclassification of diagnostic tests (e.g. serological tests that indirectly measure previous exposure to a pathogen; molecular tests that directly measure presence of the pathogen, etc.); (c) epidemiological information classifying a diagnostic testing date into stages (sequences with a temporal order) (e.g., ‘very early (non-linear) epidemic phase’, ‘early (exponential growth) phase’; ‘late (post-peak) phase, etc.); (d) epidemiological information classifying the diagnostic testing date in reference to a pathogen (e.g., ‘strain X’, ‘strain Y’, ‘strain Z’, etc ); (e) a biological estimate of incubation time or transmission cycle of the pathogen (e.g., ‘estimated transmission cycle III’, ‘estimated transmission cycle VI’. etc ); (f) population-related information describing whether one or more species predominate (i.e., ‘residential’, ‘farm with sheep and cows’, etc.); (g) connectivity-related information (e.g., 'high (low) road density’; ‘with one or more highway intersections’, ‘railroad network’, ‘airport’, ‘harbor’; ‘major hub’ (road and railroad intersections, etc.); and (h) seasonality-related information (e.g, such as ‘tourist attractions’ and ‘stadium / concert hall’, etc.)

[0014] The method 100 includes generating 112 a prediction of future disease spread based on the one or more identified disease dissemination patterns. In embodiments where the obtained data includes two or more species, generating the prediction of future disease spread may be further based on the identified geographic units with at least two species affected during the same time period. In some embodiments, the method includes generating a map of the infection data. For example, the map may include the infection data binned by the geographic unit. Such a map may also include an overlay of the geographic units (e.g.. a grid overlay, shading, etc.) The map may include infection indicators which are differentiated by species. In various examples, the indicators may be symbols showing the location of each individual infection, shading or patterning of a geographic unit representing the number of infections in that unit, or other indicators. The indicators may be differentiated according to species by having different colors, patterns, symbols, shading, or other differences or combinations.

[0015] In some embodiments, the map includes infection data during a subset of the predetermined time periods. For example, the map may show infection data for only a single time period (e.g, one year, etc.) In some embodiments, the map is interactive and a user is able to selectively modify the subset of predetermined time periods to be displayed. For example, the map may display a single time period and the user may use an input device to change the currently displayed time period.

[0016] The method may include prioritizing one or more interventions based on the identified geographic units where at least one species is affected during the same time period. In some embodiments, the method includes prioritizing one or more interventions based on the identified geographic unit.

[0017] In some embodiments, the present disclosure provides a procedure that (1) creates, integrates, structures, and orders, in a temporal sequence, a group of bio-geo-temporal variables that helps ameliorate and / or control rapidly disseminating emergencies of biological nature (those that may cover large territories and / or affect large numbers of people and / or nonhuman species, such as epidemics and epizootics), (2) produces quantitative information that is compared to a reference (an accepted practice), demonstrating the solution generated by the method is less costly and / or more beneficial than the one generated by the reference; and (3) identifies at least one specific area (defined by geographical coordinates) where application ofavailable resources is likely to be more effective in mitigating risks, hazards and / or resolve the emergency earlier.

[0018] To generate a potential control epidemiologic policy that applies to a specific territory and time, this information was or may be available at municipality (county) level: (i) area, (2) population, (3) time, (4) number of diagnostic tests conducted at a given time period, (5) cases (expressed as percentage of tests that resulted in ‘positive’ results or counts / sq km).

[0019] Integrating and applying combinatorial theory, network theory', geographic expertise, and biomedical expertise, the method described in #1-13 created numerous hypothetical and complex variables that supported the prioritized intervention of a specific small area, which was viewed as a major source of disease dispersal.

[0020] Embodiments of the present disclosure may use public information and / or proprietary information. For example, the following non-limiting example uses publicly available information on Puerto Rico.

[0021] Using public information, such as, for example, test positivity, date, specific country (area and population), etc. (e.g., https 7 / rconnect. df c-i . harvard, ed ty'cobttps: / / eo wikipedia.org / wiki / Miinicipalities_.Of _Puerto_.Rico. https:Z / www. census gov / data / tabies / time-senes / demo / popesl / 2020s-totai-puertG-ri co- municipios.html). an example, non-limiting two-component bio-geo-temporal method that analyzes highly connected, geographically specific areas affected by an emergency (such an epidemic) was generated. The method was able to achieve (a) a 12: 1 ratio of benefit / cost (i.e., with a 5% ‘cost’ (removing resources from non-priority counties), it was possible to increase the benefit to the priority' area(s) by 60%, which now receives more resources to control epidemics), and therefore, (b) it facilitates ‘differentiated resource allocation,’ for a period of time and without increasing the overall costs of a control campaign, this strategy could target small areas likely to promote disease dispersal (e.g.. those where a central county shows a test positivity percentage at least higher than the surrounding counties); and (c) the same procedure could later evaluate the success / failure of the decision and / or promote re-allocation of more resources to the same or other area(s). The method had objectives including:1 - Where:

[0022] (la) identifying a relatively small area where interventions against an emergency, if prioritized, are likely to be less costly and / or result in desirable outcomes earlier than alternatives;

[0023] (lb) identifying where resources can be temporarily removed from and reallocated to the prioritized area, and

[0024] (1c) frequently repeating the analysis so it is determined whether more (or less) resources are needed, when the original intervention may cease, and / or where and when resources should be re-allocated to prioritize other areas.2 - How much:

[0025] (2a) to estimate the optimal magnitude of resources to be re-allocated, such that, with the minimal magnitude to be removed from a non-prioritized area, a maximal magnitude of resources can be deployed at the prioritized area. For example, assuming that all 78 counties of a given territory have identical area, population, and receive the same resources against COVID. the re-deploying of 5% of resources assigned to 72 counties increases 60% the use of resources applied to a specific (6-county) priority area (72 x 5% = 3.6; 3.6 / 6 = 60%);

[0026] (2b) to consider whether areas of similar population, area, and / or biological features (e.g.. test positivity %) differ in geographical connectivity (e.g., the road density, the number of road or railroad intersections / square kilometer);

[0027] (2c) to compare the cost / benefit oriented solution generated by the method that considers bio-geo-temporal data on highly connected, small areas to the solution generated by established spatial statistical or alternative methods (see discussion of Figure 10 below, with explicit comparison against the Getis-Ord G test, which shows a twice smaller area is required using the present-disclosed method than the area to be intervened if an alternative method is used); and

[0028] (2d) to frequently repeat these analyses to determine if and when the intervention was successful and / or when and where a new re-allocation of resources is desirable and justified.

[0029] Examples of the present method are illustrated with data on the location of brucellosis cases reported in the “Further Discussion” section below. Figure 10 shows that the presently-disclosed bio-geo-temporal (“BGT”) method (Figure 10(A)) was at least twice less costly / more beneficial than a classic spatial statistical test (Figure 10(B)-(D)). To assess costeffectiveness, the same data were analyzed by the bio-geo-temporal graphical method (BGT) and the Hot Spot analysis (HS) as reported by the Getis-Ord G test. Using ruminant data from 2018 as an example, it was shown that the BGT method identified 139 25-sq km orange and red squares, which included 583 cases or 4.19 cases / square (583 / 139. Figure 10(A)). In contrast, the HS analysis found 521 cases in 194 squares or 2.68 cases / square (521 / 194, Figure 10(B)). Consequently, the case density7of the BGT approach was 56.3% higher (4. 19 / 2.68) than that of the HS. This difference in potential cost-effectiveness was explained by two factors: (i) the HS missed large areas that included numerous cases (Figure 10(C)) and, (ii) in particular, the HS analysis missed nine mini-areas with a very high case density (Figure 10(D)). Such a difference was achieved while the BGT analysis occupied an area 28.4% smaller (139 / 194 squares or 71.6%) than the area covered by the HS analysis. If cost effectiveness of interventions was measured as the ratio of benefits over costs (here expressed as cases captured / area unit), then the ratio of the BGT method would be 2.18 (156.3 / 71.6), i.e., the BGT method exhibited a benefit / cost ratio twice as large as the one shown by the HS analysis. The arrows of Figure 10(A) indicate a centrally-located area (bottom arrow ) and a non-centrally located area (top arrow ).

[0030] In some embodiments, the present disclosure can be described as:

[0031] 1. Design a grid network (e.g., square cells, hexagons, governmental entities, etc.): A procedure to assess the classificational, spatial, temporal and variations of the raw data to determine the best spatial dissemination to provide the greatest flexibility for element aggregation. This involves the review of various short-term temporal-biological and long-term temporal-spatial data.

[0032] 2. Develop methodology for the identification of each grid entity (e.g, rows / columns, regional / municipal names, sequential numbers): A procedure that provides a classification schema to enable the identification of each spatial entity. The specific classification schema is dependent upon the anticipated use of the resulting analysis as well as the temporal / spatial resolution of the available data.

[0033] 3. Calculate temporal-geographic related impact factors for each spatial area (e.g., population, area, road density, percent land vs percent water, methods of spatial connectivity or separation): A procedure that assigns temporally fixed geographical feature data to each specific spatial entity. The assignment of geographical data enables the investigation of barriers or conductors potentially involved in a spatial shift of temporal-biological results.

[0034] 4. Import, clean and perform aggregation of temporal-biological data to be spatially linked with geographical data: A procedure to create an aggregated table of available temporal-biological data (e.g., human or animal test ty pes / results / dates and other individual factors) and in the process, filtering those tables to facilitate their utility the identification of geotemporal-biological changes.

[0035] 5. Link the temporal-biological data to the spatial entities: A procedure to update each spatial entity’s attribute data to include the temporally linked biological data. This enables the creation of new temporal-spatial-biological data (i.e., infection rate changes over time, percent population tested and / or positive, etc.)

[0036] 6. Once linked, many types of visual and tabular analysis can be performed: A procedure to enable the comparison of various factors to identify areas of increases of decreased infection rates as well as various spatial-biological-temporal variations.

[0037] In another aspect, the present disclosure may be embodied as a system for predicting a disease dissemination pattern. The system includes a processor programmed to perform the method of the methods disclosed herein. For example, the processor may be programmed to obtain data of infections caused by the disease ('‘infection data”), the infection data including at least geographical information of each infection, time information of each infection, and infection information for one or more species. Bin the infection data by species, predetermined time period, and geographic unit. Identify one or more disease dissemination patterns for each species of the one or more species using the binned infection data. Generate a prediction of future disease spread based on the one or more identified disease dissemination patterns.

[0038] In another aspect, the present disclosure may be embodied as a non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform any of the methods disclosed herein. For example, the instructions maycause the processor to obtain data of infections caused by the disease (“infection data”), the infection data including at least geographical information of each infection, time information of each infection, and infection information for one or more species; bin the infection data by species, predetermined time period, and geographic unit; identify one or more disease dissemination patterns for each species of the one or more species using the binned infection data; and generate a prediction of future disease spread based on the one or more identified disease dissemination patterns.Further Discussion

[0039] The following further discussion is related to a non-limiting study conducted to demonstrate the concept of the presently-disclosed biological-geographical (“BG”) method (sometimes referred to as the biological-geographical-temporal (“BGT”) method). The scope of the present disclosure is not limited to the parameters, materials, and methods used in the study.

[0040] An example embodiment is described for the use of a geo-temporal analysis of brucellosis to investigate and teach how zoonoses emerge and disseminate. Methods similar to the one reported may be considered in the design of geographical site-specific, time-sensitive, cost-effective interventions.

[0041] Geographically explicit, high-resolution, grid-based maps offer an actionable alternative to explore many sources of validity. Such maps have been used to investigate (non- infectious) interactions involving human and non-human species. These maps also circumvent the limitations of maps based on aggregate data, which miss local interactions among georeferenced variables. In contrast, grid-based maps can display a high level of granularity.Furthermore, bio-geographical methods do not assume space homogeneity — an assumption associated with classic spatial statistics, which also assume that neighbors are similar.

[0042] Brucellosis-related dissemination patterns can be explored in Georgia, the country located in the South Caucasus. With brucellosis being a substantial endemic problem, Georgia has a large geo-temporal dataset on cases affecting cattle, sheep, and humans.

[0043] Such a context is also adequate to explore One Health processes, in which the environment interacts with potential hosts and non-human and human species may infect oneanother. While numerous educational programs now focus on One Health, inter- / trans- disciplinary educational gaps have been reported in this field.

[0044] To evaluate cost-benefit oriented approaches, new concepts may be investigated. For example, the detection of small geographical sites where infections induced by the same bacterium affect two or more species (‘zoonotic disease clusters’) may be desirable.Data

[0045] Data on (ruminant) brucellosis were collected by the National Food Agency of the Ministry of Environmental Protection and Agriculture of Georgia between 2014 and 2019.Human data on brucellosis cases between 2015 and 2020 were provided by the Center for Disease Control and Public Health of Georgia.

[0046] The data were filtered to extract geographically referenced and time-stamped records (records with latitude, longitude, and date) for cattle, sheep, and humans. This filtering resulted in the identification of 7,643 records for the period of 2014-2019 (2,999 cattle and 4,644 sheep). Nine hundred and ninety-three human records were identified between 2015 and 2018.Geo-referenced method

[0047] These tabular data were brought into a geo-referenced platform and a geodatabase point feature was generated. Starting with this initial point feature information, yearly speciesspecific point features were generated for mapping purposes.

[0048] To identify areas of case concentrations, a 5km-by-5km country-wide grid (7,068 cells) was created, and scripts were developed to associate case data with the enclosing grid cell. Such size was selected as a compromise between larger areas (more likely to capture more cases but less precise in terms of specific case geo-location) and smaller areas (more precise in terms of specific case geo-location but more likely to miss cases).

[0049] Each yearly species-specific grid cell polygon feature was added to a map and summarized by grid cell number. Each summarized table was exported to a text file for tabulation and analysis. Using the summarized data, a quantile (maximum of three) display of the grid cell’s total cases was generated, and a map layout was created.Construct, internal, external and statistical validity’

[0050] The degree to which the concept of interest was actually investigated by the operation implemented (construct validity) was estimated by comparing the bio-geographical model with an alternative — the Getis-Ord Gi* or Hot Spot analysis, a method performed using ArcGIS® Pro that assumes neighbors are bio-geographically similar. The method that yielded the highest case density’ (cases / square kilometer) was viewed as the most cost-effective.

[0051] Internal validity’ (lack of confounding) was assessed by testing several variables. Threats to internal validity were ruled out when two or more variables yielded similar results.

[0052] Standard statistical analysis was performed using commercial packages (Minitab 22 (Minitab Inc, State College, PA, USA)). Regression analysis explored relationships and proportions between the number of cases reported in a single host at specific types of sites and the number of cases found in multi-species sites (sites where either only ruminants or human and non-human species reported infections). By investigating two or more host species over two or more years, the external validity of the tool was also explored.

[0053] To investigate whether the bio-geographical tool could be used in different populations and / or different timeframes — i.e., external validity — cattle and sheep cases were plotted, side by’ side, annually (Figs. 1-3). It was observed that cattle cases (Figs. 1A, 2A, 3A) matched sheep cases (Figs. IB, 2B, 3C). However, time did not appear to be related with case location. For instance, earlier cases (Figs. 1 A, B) did not match later cases, even when a short temporal period (a year) was considered (Figs. 2A, B).

[0054] While cattle and sheep cases displayed noticeable geo-referenced changes over time, most human cases did not. Over four years, most human cases were reported in the same area — the eastern region of Georgia (Figs. 4A-D). Yet, a second pattern associated with human cases was also seen, which was heterogeneous and took place in the central municipalities of Georgia (Figs. 4 A-D).

[0055] Sites that reported cases affecting two species (cattle and sheep) seemed to differ from the remaining sites. When the counts or percentages of cases and sites were considered, ‘ruminant sites’ exhibited twice as many cases as uni-species sites (Table 1).

[0056] Findings supported the differentiation of brucellosis cases into three geo-temporal patterns: (i) one only observed in the eastern region, which included human cases; (ii) one also affecting humans, which took place outside the eastern region; and (iii) non-human cases, reported outside the eastern region. Within the last variety, two presentations were distinguished: (a) sites where only one ruminant species was infected, and (b) sites where both cattle and sheep cases were reported (ruminant sites).

[0057] While the historical nature of the data prevented inferences and predictions, possible educational applications included data-driven hypotheses on peak temporal patterns reaching in 2017 (Figure 6). The number of cases observed in either ruminant species predicted the total number of cases found in multi-species (ruminant) sites (p < 0.01, Figure 7). In one scenario under study, the bio-geographical procedure captured 78% more cases (or 927 versus 521) than the alternative (Figures 8 A, B). While the average case density was 0.107 cases per square kilometer in the Hot Spot analysis and, in addition, this method under-estimated large areas with numerous cases (Figures 8 B and C), the case density of a few (and small) disease clusters detected by the bio-geographical method was twice higher (Figure 8 D).

[0058] To explore possible applications in decision-making, the BG method was further applied to explore the annual number of cases according to the (zoonotic vs. non-zoonotic) content of the infected hosts. Using the data reported in 2019. the number of sites reporting zoonotic cases (including both human and cattle or human and sheep) was at least 5 times higher than the number of sites where only ruminants or only humans were infected (Figure 9).Therefore, if a cost-effective policy was designed to be applied in this scenario, the first priority' of interventions would focus on zoonotic cases, sites that only included ruminants would be the second priority and, as the last priority, sites only reporting infected humans would be intervened.

[0059] This study should not be construed to represent the current status of brucellosis- related conditions existing in Georgia but. instead, a realistic learning scenario that can support research and education on geo-epidemiology.

[0060] While zoonotic infections disseminate by only three (direct, indirect, or both direct and indirect) types of transmission, the expression of such transmissions may vary according to the local geography. Consequently, cost-effective interventions may be designed and selected according to specific geo-temporal expressions.Bio-geographical patterns of zoonotic disease transmission

[0061] The hypothesis of direct contacts between non-human infected species and susceptible humans was supported by the pattern observed in the eastern region of Georgia. Disease transmission, in this modality, is thought to be facilitated when migrant shepherds move their sheep, twice a year, between the southern and the northern borders of the country, along a path flanked by mountains — a geographical feature that determines a rather constant geographical pattern, detected regardless of time (Figs. 4A-D). In contrast, other human infections (reported outside the eastern region) are possibly facilitated by the consumption of contaminated milk and meat. The hypothesis of cattle-sheep contacts (apparent in several regions of Georgia) is facilitated by a common agricultural practice — also observed in many countries — in which cattle and sheep share summer pastures.

[0062] This study emphasized high-resolution, geo-referenced and cost-effective epidemic control measures. To avoid loss of resolution, this study was not centered on municipalities (which differ in area, population, connectivity and many other aspects) but on small areas of equal size (cells of 25 sq km). Such operation facilitated the detection of specific geographical sites where infections affecting two or more host species were found. While approaches that aggregate data and assume homogeneous data distributions over large spatial areas tend to result in large areas to be intervened (e.g, higher costs), the geo-biological method identified small areas with high case density, which are likely to induce more beneficial interventions (more cases to be covered) at lower costs (in smaller areas to be intervened), and — due to their smaller areas — may be completed earlier. Furthermore, small areas with high case density can be mapped over other layers and, consequently, inform on connectivity and many other variables, as described before.

[0063] Such a geographical approach can be complemented with a biological emphasis. When the data are divided into classes (e.g, zoonotic or human and non-human, only ruminants, and only humans), the number of cases may be much higher in zoonotic sites and, consequently, interventions that prioritize such sites may be less costly and / or more beneficial. (Figure 9).Validation

[0064] Several estimates of construct, internal, external and / or statistical validity were facilitated by the method under study. Although the historical nature of the data is prone toseveral threats to validity, the fact that three host species were investigated (and tested for several years) demonstrates that this tool may possess external validity. Because the analysis of five variables yielded similar patterns (and, therefore, confounding was ruled out) and revealed statistically significant associations, internal and statistical validities were supported (Figure 6). Because the bio-geographical approach captured more cases per sq km than the alternative method, construct validity7was not ruled out (Figures 8 and 9).

[0065] However, the previous comparison did not consider all possible scenarios (which may include additional bio-geo-temporal variables). To better estimate cost-effective interventions, the benefits associated with interrupting disease transmission (short- and longterm, multi-species-related bio-geographical interactions) need to be considered.Further applications

[0066] These geo-referenced findings facilitate the development of several new applications. One refers to prioritizing interventions in locations that report multi-species (e.g.. both cattle and sheep) cases. Regardless of disease prevalence, the detection of small areas where two species are affected suggests a site-specific interaction that promotes disease dissemination.

[0067] Such an interaction suggests zoonotic contacts when, in addition to ruminant cases, human cases are also found in a small area. Because ‘zoonotic disease’ sites can induce secondary' infections along ruminant and non-ruminant hosts, they may be prioritized when costbenefit oriented interventions are planned. Support for ranks that prioritize where interventions should be implemented may be facilitated by a cost-benefit oriented analysis that considers these findings. For example, when there is evidence that multi-species and zoonotic mini-sites capture more cases than sites presenting with uni-species cases (as shown in Table 1 and Figure 9), sites that show the highest case density and are potentially zoonotic could be prioritized, followed by sites that do not show zoonotic cases, and, finally, sites that display the lowest case density and are not zoonotic. Such priority ranks could be expanded with time-related information and a calculation on pair of cases located on an explicit road network. For example, everything else equal, older (earlier) cases could receive the first priority to be intervened when they also display the shortest inter-case along-road distance.

[0068] The second potential application refers to migrant shepherds. Because their cases are consistently reported, every year, at the same places, it appears that some behaviors arerepeated over time, which occur at specific places where interventions are likely to be beneficial. Because available information suggests there may be inadequate brucellosis-related, educational campaigns in the Tusheti and Kakheti regions, further applications may focus on sociological- educational variables at specific geographical areas, such as associations between educational5 packages received on the role of uncooked meat and / or unpasteurized milk consumption and disease occurrence.

[0069] Furthermore, additional applications based on these or similar geo-referenced procedures could explore alternative methods that estimate the costs induced by bacterial zoonoses and the potential benefits of preventive campaigns. While similar applications have10 been conducted for viral zoonoses (1), no cost-benefit estimates on bacterial zoonoses-related decision-making have been emphasized.Table 1: Brucellosis cases reported in ruminants between 2014 and 2019.Average number 6.476.689.34 in ruminant sites(2421 / 374) (1670 / 250) (2102 / 225)Across time, between 27 and 57% of all ruminant cases were found in ruminant sites. Across time, the average number of cases was at least twice as large in ruminant sites than in sites where only one ruminant species reported infected cases. Specifically, that number was: (i) 2.2-2.45 times larger in 2014 (2.74 / 1. 16, 2.74 / 1.25); (ii) 5.3-6.6 times larger in 2015 (8.41 / 1.27, 8.41 / 1.58); (hi) 5.7-6.4 times larger in 2016 (8.2 / 1.29, 8.2 / 1.45); 3.4-4.1 larger in 2017 (6.47 / 1.51; 6.47 / 1.9); (iv) 4.1 times larger in 2018 (6.68 / 1.59, 6.68 / 1.61); and (v) 4.9-5.5 times larger in 2019 (9.34 / 1.89, 9.34 / 1.71).Additional Example Embodiments:

[0070] Example 1. A computer-implemented method for predicting a dissemination pattern of a disease, the method including: obtaining, by a computer system, data of infections caused by the disease ("infection data’7), the infection data including at least geographical information of each infection, time information of each infection, and infection information for one or more species; binning, by the computer system, the infection data by species, predetermined time period, and geographic unit; identifying one or more disease dissemination patterns for each species of the one or more species using the binned infection data; and generating, by the computer system, a prediction of future disease spread based on the one or more identified disease dissemination patterns.

[0071] Example 2. The method of example 1, wherein the obtained data includes infection information for two or more species, and further including: identifying geographic units where at least two species are affected during the same time period; and wherein generating, by the computer system, a prediction of future disease spread is further based on the identified geographic units with at least two species affected during the same time period.

[0072] Example 3. The method of example 2, further including: ranking, by the computer 5 system, geographic units for intervention based on case density, prediction of future disease spread, and, optionally, one or more biological, geographical, and temporal variables; generating, by the computer system, a visual output including a map of the geographical region displaying predicted disease dissemination patterns, ranked intervention zones, and potential transmission pathways.

[0073] Example 4. The method of example 3, wherein the map includes the infection data binned by the geographic unit.

[0074] Example 5. The method of example 4, wherein the map includes an overlay of the geographic units.

[0075] Example 6. The method of example 4, wherein the map includes infection indicators which are differentiated by species.

[0076] Example 7. The method of example 4, wherein the map includes infection data during a subset of the predetermined time periods.

[0077] Example 8. The method of example 7, wherein the map is interactive and a user is able to selectively modify the subset of predetermined time periods to be displayed.

[0078] Example 9. The method of example 1, further including identifying one or more geographic units for resource prioritization based on the predicted future disease spread.

[0079] Example 10. The method of example 1, further including: obtaining updated infection data: updating the binned infection data with the updated infection data; updating the identified one or more disease dissemination patterns for each species of the one or more species using the updated binned infection data; and updating the prediction of future disease spread based on the one or more updated identified disease dissemination patterns.

[0080] Example 11. The method of example 10, wherein the updated infection data is obtained periodically (e.g, daily, weekly, etc.)

[0081] Example 12. The method of example 1, wherein the pre-determined time period is a calendar year

[0082] Example 13. The method of example 1, where the geographic units have a same area.

[0083] Example 14. The method of example 1, wherein the geographic unit is a polygon, (e.g, a grid).

[0084] Example 15. The method of example 1, wherein the geographic unit is a preexisting construct (e.g., county, postal code, fire district, etc.)

[0085] Example 16. The method of example 1, wherein the geographic information comprises spatial coordinates.

[0086] Example 17. The method of example 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data includes identifying case clusters based on case density and species-species interactions within each geographic unit.

[0087] Example 18. The method of example 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data incorporates at least one environmental covariate selected from elevation, soil type, meteorological data, and humananimal interaction pathways.

[0088] Example 19. The method of example 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on one or more feature selected from: (a) a percentage of ‘positive’ test results (individual diagnosed as affected, i.e., ‘cases’) over the total number of people tested (i.e., test positivity percentage); (b) a classification and / or subclassification of diagnostic tests (e.g.. serological tests that indirectly measure previous exposure to a pathogen; molecular tests that directly measure presence of the pathogen, etc.); (c) epidemiological information classifying a diagnostic testing date into stages (sequences with a temporal order) (e.g., ‘very early (non-linear) epidemic phase’, ‘early (exponential growth) phase’; Tate (post-peak) phase, etc.); (d) epidemiological information classifying the diagnostic testing date in reference to a pathogen (e.g., ‘strain X’, 'strain Y’, ‘strain Z’, etc.); (e) a biological estimate of incubation time or transmission cycle of the pathogen (e.g., ‘estimated transmission cycle III’, ‘estimated transmission cycle VI’, etc.); (!) population- related information describing whether one or more species predominate (i.e.. 'residential’, ‘farm with sheep and cows’, etc.); (g) connectivity-related information (e.g., ‘high (low) road densify’; ‘with one or more highway intersections’, ‘railroad network’, ‘airport’, ‘harbor’; ‘major hub’ (road and railroad intersections, etc.); and (h) seasonality-related information (e.g., such as ‘tourist attractions' and ‘ stadium / concert hall’, etc.)

[0089] Example 20. The method of example 19, further including identifying a feature of a geographic unit that has a value which differs from the same feature in one or more neighboring geographic unit by at least a factor of 2.

[0090] Example 21. The method of example 19, further including identifying clusters of contiguous geographic units having a similar value of one or more features (e.g., a value within ±20%, ±10%, ±5%).

[0091] Example 22. The method of example 1, further including identifying one or more subdivisions of a geographic unit which have higher cases than the corresponding geographic unit.

[0092] Example 23. The method of example 1, further including identifying at least one central geographic unit within the geographic units of the identified one or more disease dissemination patterns.

[0093] Example 24. The method of example 1, further including: obtaining updated infection data; updating the binned infection data with the updated infection data; updating the identified one or more disease dissemination patterns for each species of the one or more species using the updated binned infection data; and updating the prediction of future disease spread based on the one or more updated identified disease dissemination patterns.

[0094] Example 25. The method of example 24, wherein the updated infection data is obtained periodically (e.g.. daily, weekly, etc.)

[0095] Example 26. The method of example 1, further including performing, using the computer system, a cost-benefit analysis configured to evaluate a ratio of decision-related costs to costs resulting from an alternative prediction method.

[0096] Example 27. The method of example 1, further including performing, using the computer system, a cost-benefit analysis to evaluate a ratio of expected disease cases mitigated to the estimated cost of intervention.

[0097] Example 28. The method of example 27, wherein the cost-benefit analysis includes minimizing a geographical area in which to provide interventions.

[0098] Example 29. The method of example 28, wherein the cost-benefit analysis includes identifying geographic units to contribute additional resources.

[0099] Example 30. The method of example 1, wherein one of the one or more species is human.

[0100] Example 31. The method of example 1, further including prioritizing one or more interventions based on the identified geographic units where at least two species are affected during the same time period.

[0101] Example 32. The method of example 1, wherein the identifying the disease dissemination pattem(s) further comprises using both case densify (number of cases per unit area) and infections affecting both human and non-human species.

[0102] Example 33. The method of example 1, further including: determining an alternative disease dissemination pattern using one or more biological, geographical, and temporal variables; and selecting an advantageous disease intervention based on a cost-benefit analysis of the disease dissemination pattern and the alternative disease dissemination pattern.

[0103] Example 34. The method of example 33, wherein the cost-benefit analysis includes determining greater benefit, lower cost, or both greater benefit and lower cost.

[0104] Example 35. The method of example 1, wherein the disease is a zoonotic disease.

[0105] Example 36. A system for predicting a disease dissemination pattern, including a processor programmed to perform the method of any one of examples 1-35.

[0106] Example 37. A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform the method of any one of examples 1-35.

[0107] Although the present disclosure has been described with respect to one or more particular embodiments, it will be understood that other embodiments of the present disclosure may be made without departing from the spirit and scope of the present disclosure.

Claims

We claim:

1. A computer-implemented method for predicting a dissemination pattern of a disease, the method comprising: obtaining, by a computer system, data of infections caused by the disease (“infection data”), the infection data including at least geographical information of each infection, time information of each infection, and infection information for one or more species; binning, by the computer system, the infection data by species, predetermined time period, and geographic unit; identifying one or more disease dissemination patterns for each species of the one or more species using the binned infection data; and generating, by the computer system, a prediction of future disease spread based on the one or more identified disease dissemination patterns.

2. The method of claim 1, wherein the obtained data includes infection information for two or more species, and further comprising: identifying geographic units where at least two species are affected during the same time period; and wherein generating, by the computer system, a prediction of future disease spread is further based on the identified geographic units with at least two species affected during the same time period.

3. The method of claim 2, further comprising: ranking, by the computer system, geographic units for intervention based on case density, prediction of future disease spread, and, optionally, one or more biological, geographical, and temporal variables; and generating, by the computer system, a visual output comprising a map of the geographical region displaying predicted disease dissemination patterns, ranked intervention zones, and potential transmission pathways.

4. The method of claim 3, wherein the map includes the infection data binned by the geographic unit.

5. The method of claim 4, wherein the map includes an overlay of the geographic units.

6. The method of claim 4, wherein the map includes infection indicators which are differentiated by species.

7. The method of claim 4, wherein the map includes infection data during a subset of the predetermined time periods.

8. The method of claim 7, wherein the map is interactive and a user is able to selectively modify the subset of predetermined time periods to be displayed.

9. The method of claim 1, further comprising identifying one or more geographic units for resource prioritization based on the predicted future disease spread.

10. The method of claim 1. further comprising: obtaining updated infection data: updating the binned infection data with the updated infection data; updating the identified one or more disease dissemination patterns for each species of the one or more species using the updated binned infection data; and updating the prediction of future disease spread based on the one or more updated identified disease dissemination patterns.

11. The method of claim 10, wherein the updated infection data is obtained periodically.

12. The method of claim 1, wherein the pre-determined time period is a calendar year13. The method of claim 1, where the geographic units have a same area.

14. The method of claim 1, wherein the geographic unit has a polygon shape.

15. The method of claim 1, wherein the geographic unit is a preexisting construct.

16. The method of claim 1, wherein the geographic information comprises spatial coordinates.

17. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data includes identifying case clusters based on case density and species-species interactions within each geographic unit.

18. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data incorporates at least one environmental covariate selected from elevation, soil type, meteorological data, and human-animal interaction pathways.

19. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on one or more feature selected from: a percentage of ‘positive’ test results (individual diagnosed as affected, i.e., ‘cases’) over the total number of people tested (i.e., test positivity percentage); a classification and / or subclassification of diagnostic tests (e.g., serological tests that indirectly measure previous exposure to a pathogen; molecular tests that directly measure presence of the pathogen, etc.): epidemiological information classifying a diagnostic testing date into stages (sequences with a temporal order) (e.g., ‘very early (non-linear) epidemic phase’, ‘early (exponential growth) phase’; ‘late (post-peak) phase, etc.); epidemiological information classifying the diagnostic testing date in reference to a pathogen (e.g. 'strain X’, ‘strain Y’, ‘strain Z’, etc.); a biological estimate of incubation time or transmission cycle of the pathogen (e.g., ‘estimated transmission cycle III’, ‘estimated transmission cycle VI’, etc.); population-related information describing whether one or more species predominate (i.e.. 'residential’, 'farm with sheep and cows’, etc.); connectivity-related information (e.g., ‘high (low) road density’; ‘with one or more highway intersections’, ‘railroad network’, ‘airport’, ‘harbor’; ‘major hub’ (road and railroad intersections, etc.); and seasonality-related information (e.g.. such as ‘tourist attractions’ and ‘stadium / concert hall’, etc.)20. The method of claim 19, further comprising identifying a feature of a geographic unit that has a value which differs from the same feature in one or more neighboring geographic unit by at least a factor of 2.

21. The method of claim 19, further comprising identifying clusters of contiguous geographic units having a similar value of one or more features (e.g, a value within ±20%. ±10%, ±5%).

22. The method of claim 1, further comprising identifying one or more subdivisions of a geographic unit which have higher cases than the corresponding geographic unit.

23. The method of claim 1, further comprising identifying at least one central geographic unit within the geographic units of the identified one or more disease dissemination patterns.

24. The method of claim I . further comprising: obtaining updated infection data; updating the binned infection data with the updated infection data; updating the identified one or more disease dissemination patterns for each species of the one or more species using the updated binned infection data; and updating the prediction of future disease spread based on the one or more updated identified disease dissemination patterns.

25. The method of claim 24, wherein the updated infection data is obtained periodically (e.g., daily, weekly, etc.)26. The method of claim 1 , further comprising performing, using the computer system, a costbenefit analysis configured to evaluate a ratio of decision-related costs to costs resulting from an alternative prediction method.

27. The method of claim 1, further comprising performing, using the computer system, a costbenefit analysis to evaluate a ratio of expected disease cases mitigated to the estimated cost of intervention.

28. The method of claim 27, wherein the cost-benefit analysis includes minimizing a geographical area in which to provide interventions.

29. The method of claim 28, wherein the cost-benefit analysis includes identifying geographic units to contribute additional resources.

30. The method of claim 1. wherein one of the one or more species is human.

31. The method of claim 1, further comprising prioritizing one or more interventions based on the identified geographic units where at least two species are affected during the same time period.

32. The method of claim 1, wherein the identifying the disease dissemination pattem(s) further comprises using both case density (number of cases per unit area) and infections affecting both human and non-human species.

33. The method of claim 1. further comprising: determining an alternative disease dissemination pattern using one or more biological, geographical, and temporal variables; and selecting an advantageous disease intervention based on a cost-benefit analysis of the disease dissemination pattern and the alternative disease dissemination pattern.

34. The method of claim 33, wherein the cost-benefit analysis includes determining greater benefit, lower cost, or both greater benefit and lower cost.

35. The method of claim 1, wherein the disease is a zoonotic disease.

36. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on a percentage of ‘positive’ test results divided by the total number of people tested.

37. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on a classification and / or subclassification of diagnostic tests.38 The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on epidemiological information classifying a diagnostic testing date into stages (sequences with a temporal order).39 The method of claim 1. wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on epidemiological information classifying the diagnostic testing date in reference to a pathogen.

40. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on a biological estimate of incubation time or transmission cycle of the pathogen.

41. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on population-related information describing whether one or more species predominate.

42. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on connectivity -related information.

45. The method of claim 1, wherein identifying one or more disease dissemination patterns for each species using the binned infection data is further based on seasonality-related information.

46. A system for predicting a disease dissemination pattern, comprising a processor programmed to perform the method of any one of claims 1-45.

47. A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform the method of any one of claims 1-45.

Images