Preview

Journal of microbiology, epidemiology and immunobiology

Advanced search

Predicting incidence of Crimean-Congo hemorrhagic fever using satellite monitoring (remote sensing) data in the Stavropol Territory

https://doi.org/10.36233/0372-9311-213

Abstract

Introduction. With the epidemiological situation for Crimean-Congo hemorrhagic fever (CCHF) remaining tense in many countries worldwide, special attention should be focused on development and improvement of risk-based epidemiological prediction methods.

The aim of the study was to build a prediction model for CCHF incidence dynamics (based on the Stavropol Territory) using satellite monitoring (remote sensing) data.

Materials and methods. We analyzed the climate data obtained from the Space Research Institute of the Russian Academy of Sciences as well as the data of public statistics reports on CCHF incidence from 2005 to 2021. The prediction model incorporated the Bayes theorem and Wald sequential analysis. The information content of the factors was assessed using the Kullback method.

Results. Predictions for each of 26 districts were made stepwise (compared to threshold levels) to predict whether there will be at least one case of CCHF, whether the relative incidence per 100,000 population will exceed the median level (0.9 cases) or the average rate (3.5 cases) or the third quartile rate (4.7 cases). The highest values of information coefficients were obtained for soil temperature and moisture content (at depths of 10 and 40 cm), normalized relative vegetation index, relative humidity, maximum and average air temperature, relative air humidity. During the testing of the model in 2021, false-negative (erroneous) prediction was made for 2 districts.

Discussion. The model proved to be most efficient in prediction of occurrence or absence of cases. More accurate quantitative prediction may be difficult due to subjective factors (including misdiagnosing CCHF cases without hemorrhagic manifestations and administering treatment for other conditions with similar symptoms).

Conclusion. The tests of the model demonstrate its potential. The practical application of the prediction will make healthcare workers more alert when screening and detecting CCHF cases.

Introduction

Crimean-Congo hemorrhagic fever (CCHF) is an especially dangerous arboviral infection; the epidemiological situation has been tense for many years in countries of Central Asia and the Middle East as well as in the Russian Federation [1–12].

As there are currently no specific preventive vaccines against this dangerous infection, risk-based prediction of incidence in a population is given special attention to stabilize the epidemiological situation and to serve as a basis for scientifically grounded planning of non-specific preventive measures.

In Turkey, a multi-agent SIR model was offered for making epidemiological predictions [13]. In Russia, the "maximum stability" estimation method and regression analysis were used to build a short-term prediction model to estimate the expected number of CCHF cases (annual rates) in the current year by the incidence of the "key" month [14].

In the meantime, CCHF is a natural focal, transmissible infection; therefore, when selecting a prediction method, the priority should be given to the methods that address the effect of climatic factors on the population of specific vectors of its pathogen — ixodid ticks and, consequently, on the incidence rates, as is demonstrated by many studies in other countries.

In Iran, the researchers, using the time-series analysis and the seasonal auto-regression integrated moving average (SARIMA) model, found a strong correlation between the number of cases and the monthly average air temperature, monthly maximum relative humidity and accumulated precipitation, and using the Poisson regression analysis together with the McFadden pseudo R-squared – the correlation with the maximum temperature values of the previous month [15][16]. In Bulgaria, the one-way analysis of variance (ANOVA) showed that the increase in the average air temperature and the normalized difference vegetation index (NDVI) per unit resulted in a 5.5% increase in the CCHF incidence [17].

Earlier, using the correlation analysis, Bayes theorem and Wald sequential analysis, V.M. Dubyanskiy and D.A. Prislegina developed the method for quantitative risk-based prediction of CCHF incidence in each administrative district of the Stavropol Territory [18–20]. The calculations were based on monthly values of five factors of all seasons (air temperature, relative air humidity, precipitation amounts, snow depth and wind speed), which affected different stages in the life cycle of Hyalomma marginatum ticks, which are the main vectors and reservoirs of the CCHF virus in Russia [10][12][21–25]. The efficiency of the offered method was proved by the fact that the results obtained in 2018 and the actual data had a 87.4% match; the prediction results were used in planning preventive measures.

This study is a continuation of the previous research; its aim is to improve methods used for CCHF incidence prediction and to obtain new data on the relationship between the intensity of epidemic and epizootic processes of this especially dangerous infection and the combined effects of multiple, including previously unaddressed, climatic factors.

Materials and methods

The study incorporated combined methods of epidemiological analysis and mathematical statistics.

The materials included the hydro-meteorological data and NDVI received from the database of the SRI-Monitoring Center for Collective Use of Systems for Archiving, Processing and Analyzing Satellite Data of the RAS Space Research Institute as well as the data of public statistics reports for 2005-2021. To calculate the relative CCHF incidence (per 100,000 population) for administrative districts of the Stavropol Territory, we analyzed the epidemiological survey charts of the infectious disease cluster (Form 357/u), which were provided by the Department of the Federal Service for Surveillance on Consumer Rights Protection and Human Wellbeing (Rospotrebnadzor) for the Stavropol Territory, as well as the archival data from the Department of the Federal Statistics Service for the North Caucasian Federal District and the Federal Statistics Service for each year of the studied period.

The prediction model of the incidence dynamics was based on the Bayes theorem and Wald sequential analysis [26–28]. The Bayes theorem was intentionally used for building the model, as it helped incorporate divergent effects of multiple factors for each administrative district using data on the frequency of the respective values and occurrence/absence of disease cases as well as made it possible to obtain alternative results (one of the two possible prediction outcomes – absence/occurrence of the case, surpassing/non-surpassing of the selected thresholds by the incidence rates).

The threshold level of positive solution probability was set at 99% (the error probability of 1%). We used numeric values of 13 climatic factors (for each month of the studied period and annual means):

  • air temperature — average, maximum and minimum (°C);
  • soil temperature at depths of 10 and 40 cm (°C);
  • soil moisture content at depths of 10 and 40 cm (%);
  • snow depth (m);
  • area covered by snow (%);
  • pressure (Pa);
  • relative air humidity (%);
  • precipitation (kg/m2);
  • NDVI (relative units).

The information content of 169 parameters was estimated using the Kullback method [26–28]. The factor was deemed as informative, if it demonstrated a larger difference between distributions of two differentiated conditions of the object of study. For example, at the average air temperature of 28°C in June, the probability of at least one case of CCHF within the natural focus of infection is 2 times as high on average as the probability at the temperature of 24°C. This factor can be seen as informative, as the incidence prediction will be true at the ratio of 2:1 at the above temperature. However, the derivative of the information content – the diagnostic coefficient is more convenient to use. It is efficient for estimating the probability of the object being in one of the two studied conditions at certain values of several factors. For example, if the average air temperature in June is 28°C and the average barometric pressure in June is above 103,165.18 Pa, the probability of at least one case of CCHF within the natural focus of infection reaches 0.8.

In addition to using an increased number of factors, the offered model differs fundamentally from the former method of risk-based prediction [18–20] by organization of the data for identification of predictors. Considering a relatively short time series (15 years) and landscape heterogeneity of the Stavropol Territory, the data were organized as follows: The CCHF incidence data broken by administrative districts and years were combined into a cumulative series (a total of 364 values), while the weather and climate variables were distributed according to the incidence.

The incidence was measured in relative numbers of cases per 100,000 population.

Information and prediction coefficients were calculated automatically using the Microsoft Excel 2010-based program developed by the authors.

The step-by-step algorithm for making a prediction using the prediction model:

  1. Calculation of information and prediction coefficients of factors with the program.
  2. Making an optimized list of factors (with values of the information coefficient ≥5 in descending order).
  3. Predicting the occurrence (absence) of at least 1 case of disease per 100,000 population.
  4. Calculating the incidence per 100,000 population compared to the threshold median level (0.9).
  5. Calculating the incidence per 100,000 population compared to the average rate (3.5).
  6. Calculating the incidence per 100,000 population compared to the third quartile rate (4.7).

Calculations for item 2 were made against the selected minimum threshold value (the value of 0.000009, which is lower than the rate of 1 case per 100,000 population). The calculations for the prediction were made by summing up the values of prediction coefficients of the informative factors in accordance with their values for each administrative district until the numeric value of +20 or -20 was received to indicate the occurrence/absence of disease cases with a probability of 99%. Then, when the districts with positive results were identified, the prediction for them was made against other selected threshold levels of incidence.

Results

The prediction model was built for making prediction for CCHF epidemiological situation for each district using the climate data of the previous year. To date, there have been no similar studies in predicting CCHF incidence by using satellite monitoring (remote sensing) data and the identical algorithm.

The model is built for alternative-based predictions: Whether there will be or will not be at least 1 CCHF case in the district, whether the relative incidence (per 100,000 population) will exceed the median level (0.9 cases), the average rate (3.5), the third quartile rate (4.7). The informative predictors for each threshold are presented in Table 1.

Table 1. Climatic and environmental indicators used as predictors of the forecast model

A total of 57 predictors were used for the threshold "whether there will be or will not be at least one case of CCHF in the district"; 62 predictors were used for the threshold "whether the relative incidence will exceed the median level (0.9 cases per 100,000 population)"; 56 predictors were used for the threshold "whether the relative incidence will exceed the average rate (3.5 cases)"; 55 predictors were used for the threshold "whether the relative incidence will exceed the third quartile rate (4.7 cases)".

The assessment of the information content of factors showed the highest values of information coefficients for soil temperature and moisture content (at depths of 10 cm and 40 cm), NDVI, relative humidity, maximum and average air temperature (in June), relative air humidity and NDVI (in August). During the above months, these climatic factors (according to the literature data) have a strong impact on embryogenesis, survival rate and preimaginal phases and development of H. marginatum, the population of feeders of preimaginal phases, thus having a significant effect on the size of imago tick population in the next year (the year of prediction) [22][24][25].

The prediction model for CCHF incidence dynamics was checked using the historical data for 2018–2020 (Table 2).

Table 2. The results of checking the forecast model of the СCHF morbidity dynamics using retrospective data for  2018–2020

The model-based erroneous predictions were divided into 4 types:

  • false-positive — the predicted result is positive; however, no cases have been reported;
  • false-negative — the predicted result is negative; however, cases of disease have been detected;
  • overestimated — the actual incidence rate is lower than the predicted rate against the threshold level;
  • underestimated — the actual incidence rate is higher than the predicted rate against the threshold level.

In 2020, the epidemiological prediction for 2021 was prepared (Table 3).

Table 3. The results of approbation of the forecast model of the СCHF morbidity dynamics for 2021

Before we move on to the discussion of the efficiency of prediction, we would like to point out the following aspects. CCHF patients frequently do not develop hemorrhagic syndrome; they can have ARVI-like symptoms [8][9][19]; some patients are diagnosed with other conditions; some cases are not reported [8][9][19]. Therefore, the number of reported cases depends not only on objective environmental factors, but also on subjective factors, including the severity of the disease, the qualification of medical and laboratory personnel at some healthcare facilities [19]. The above subjective factors are not addressed during the prediction. Consequently, the occurrence or absence of disease cases in general can be predicted most efficiently. The more accurate data on incidence rates are more dependent on subjective factors and are less predictable.

The data presented in Table 3 prove the above. Out of 26 administrative districts of the Stavropol Territory, the erroneous prediction was made only for two of them: Grachevsky and Shpakovsky districts; thus, the prediction for the threshold "whether there will be or will not be at least one case of CCHF in the district" was erroneous by 7.7%. The above districts were expected not to have any cases, though, in fact, there were reported cases.

The overestimated number of cases was predicted for 11 districts. Here, it is unlikely that we deal with a serious mistake, as we cannot exclude poor diagnostics (misdiagnosis) or the probability that the case of infection was reported in another district (at the place of the patient’s residence).

Conclusion

The application of the prediction model in 2021 demonstrated its potential. Based on the studied literature, no country has achieved better results in predicting CCHF incidence [24].

In the meantime, the model incorporates primarily weather variables serving as predictors. Although the information used for building the model is completely consistent with the literature data on indirect combined effects of climatic factors on the epidemiological situation of CCHF, the model should include biotic predictors [24][25][30].

The practical application of the prediction will increase the alertness of healthcare workers in the at-risk regions, will improve the detectability of CCHF cases and implementation of non-specific preventive measures.

Focusing on this area, we are planning to test the prediction model in other CCHF-endemic regions (the Rostov Region and the Republic of Kalmykia).

References

1. Shahhosseini N., Wong G., Babuadze G., Camp J.V., Ergonul O., Kobinger G.P., et al. Crimean-Congo hemorrhagic fever virus in Asia, Africa and Europe. Microorganisms. 2021; 9(9): 1907. https://doi.org/10.3390/microorganisms9091907

2. Obaidat M.M., Graziano J.C., Morales-Betoulle M., Brown S.M., Chiang C.F., Klena J.D. Rift Valley fever and Crimean-Congo hemorrhagic fever viruses in ruminants, Jordan. Emerg. Infect. Dis. 2021; 27(2): 653–55. https://doi.org/10.3201/eid2702.203713

3. Temur A.I., Kuhn J.H., Pecor D.B., Apanaskevich D.A., Keshtkar-Jahromi M. Epidemiology of Crimean-Congo hemorrhagic fever (CCHF) in Africa-underestimated for decades. Am. J. Trop. Med. Hyg. 2021; 104(6): 1978–90. https://doi.org/10.4269/ajtmh.20-1413

4. Dieng I., Barry M.A., Diagne M.M., Diop B., Ndiaye M., Faye M., et al. Detection of Crimean Congo haemorrhagic fever virus in North-eastern Senegal, Bokidiawé 2019. Emerg. Microbes Infect. 2020; 9(1): 2485–87. https://doi.org/10.1080/22221751.2020.1847605

5. Schulz A., Barry Y., Stoek F., Pickin M.J., Ba A., Chitimia-Dobler L., et al. Detection of Crimean-Congo hemorrhagic fever virus in blood-fed Hyalomma ticks collected from Mauritanian livestock. Parasit. Vectors. 2021; 14(1): 342. https://doi.org/10.1186/s13071-021-04819-x

6. Balinandi S., von Brömssen C., Tumusiime A., Kyondo J., Kwon H., Monteil V.M., et al. Serological and molecular study of Crimean-Congo hemorrhagic fever virus in cattle from selected districts in Uganda. J. Virol. Methods. 2021; 290: 114075. https://doi.org/10.1016/j.jviromet.2021.114075

7. Negredo A., Sánchez-Ledesma M., Llorente F., Pérez-Olmeda M., Belhassen-García M., González-Calle D., et al. Retrospective identification of early autochthonous case of Crimean-Congo hemorrhagic fever, Spain, 2013. Emerg. Infect. Dis. 2021; 27(6): 1754–6. https://doi.org/10.3201/eid2706.204643

8. Kulichenko A.N., Maletskaya O.V., Prislegina D.A., Vasilenko N.F., Semenko O.V., Gazieva A.Yu., et al. Epidemiological Situa tion on Natural Focal Infectious Diseases in the Southern and North Caucasian federal districts in 2019. Analytical Review [Epidemiologicheskaya obstanovka po prirodno-ochagovym infektsionnym boleznyam v Yuzhnom i Severo-Kavkazskom federal'nykh okrugakh v 2019 g]. Stavropol'; 2020. Available at: https://www.snipchi.ru/updoc/2020/An-Obzor%20po%20prirodno-oschagovim%202020.pdf (in Russian)

9. Maletskaya O.V., Taran T.V., Prislegina D.A., Dubyanskiy V.M., Volynkina A.S., Semenko O.V., et al. Natural-focal viral fevers in the South of the European part of Russia. Crimean-Congo hemorrhagic fever. Problemy osobo opasnykh infektsiy. 2020; (4): 75–80. https://doi.org/10.21055/0370-1069-2020-4-75-80 (in Russian)

10. Onishchenko G.G., Efremenko V.I., Beyer A.P. Crimean-Congo hemorrhagic fever [Krymskaya gemorragicheskaya likhoradka]. Moscow; 2005. (in Russian)

11. Platonov A.E., Avksent'ev N.A., Avksent'eva M.V., Derkach E.V., Platonova O.V., Titkov A.V., et al. Social and economic burden of five natural focal infections in the Russian Federation. Farmakoekonomika. Sovremennaya farmakoekonomika i farmakoepidemiologiya. 2015; 8(1): 47–56. https://doi. org/10.17749/2070-4909.2015.8.1.047-056 (in Russian)

12. Smirnova S.E. Crimean-Congo Hemorrhagic Fever (Etiology, Epi demiology, Laboratory Diagnostics) [Krymskaya-Kongo ge mor ragicheskaya likhoradka (etiologiya, epidemiologiya, laboratornaya diagnostika)]. Moscow: ATiSO; 2007. (in Russian)

13. Estrada-Peña A., Zatansever Z., Gargili A., Aktas M., Uzun R., Ergonul O., et al. Modeling the spatial distribution of Crimean-Congo hemorrhagic fever outbreaks in Turkey. Vector Borne Zoonotic Dis. 2007; 7(4): 667–78. https://doi.org/10.1089/vbz.2007.0134

14. Kharchenko T.V. Modern epidemiological features of the Crimean-Congo hemorrhagic fever in the Russian Federation: Diss. Stavropol'; 2011 (in Russian)

15. Ansari H., Shahbaz B., Izadi S., Zeinali M., Tabatabaee S.M., Mahmoodi M., et al. Crimean-Congo hemorrhagic fever and its relationship with climate factors in southeast Iran: a 13-year experience. J. Infect. Dev. Ctries. 2014; 8(6): 749–57. https://doi.org/10.3855/jidc.4020

16. Mostafavi E., Chinikar S., Bokaei S., Haghdoost A. Temporal modeling of Crimean-Congo hemorrhagic fever in eastern Iran. Int. J. Infect. Dis. 2013; 17(7): 524–8. https://doi.org/10.1016/j.ijid.2013.01.010

17. Vescio F.M., Busani L., Mughini-Gras L., Khoury C., Avellis L., Taseva E., et al. Environmental correlates of Crimean-Congo haemorrhagic fever incidence in Bulgaria. BMC Public Health. 2012; 12: 1116. https://doi.org/10.1186/1471-2458-12-1116

18. Dubyanskiy V.M., Prislegina D.A., Kulichenko A.N. Risk-oriented model for predicting epidemiological situation with Crimean-Congo hemorrhagic fever (on the example of Stavropol region). Analiz riska zdorov'yu. 2018; (1): 13–21. https://doi.org/10.21668/health.risk/2018.1.02 (in Russian)

19. Prislegina D.A., Dubyanskiy V.M., Maletskaya O.V., Kulichenko A.N., Vasilenko N.F., Manin E.A., et al. Crimean-Congo hemorrhagic fever in the Stavropol Region: contemporary clinical and epidemiological aspects and new approach to forecasting of morbidity. Infektsionnye bolezni: novosti, mneniya, obuchenie. 2018; 7(3): 49–56. https://doi.org/10.24411/2305-3496-2018-13007 (in Russian)

20. Prislegina D.A., Dubyanskiy V.M., Kulichenko A.N. Particular dangerous arbovirus fevers in the south of Russia: improvement of monitoring with modern information technology application. Meditsina truda i ekologiya cheloveka. 2019; (4): 50–8. https://doi.org/10.24411/2411-3794-2019-10047 in Russian)

21. Tokhov Yu.M., Degtyarev D.Yu., Dubyanskiy V.M. Ixodid Ticks (Morphology, Medical Significance, Population Regulation) [Iksodovye kleshchi (morfologiya, meditsinskoe znachenie, regulyatsiya chislennosti)]. Stavropol'; 2015. (in Russian)

22. Trukhachev V.I., Tokhov Yu.M., Lutsuk S.N., Dylev A.A., Tolokonnikov V.P., D'yachenko Yu.V. Distribution and ecological characteristics of Hyalomma ixodid ticks in the ecosystems of the Stavropol Region. Yug Rossii: ekologiya, razvitie. 2016; 11(2): 59–69. https://doi.org/10.18470/1992-1098-2016-2-59-69 (in Russian)

23. Tokhov Yu.M., Chumakova I.V., Lutsuk S.N., D'yachenko Yu.V., Kotenev E.S., Zaytsev A.A. Ticks as the reservoir of contagious diseases in the Stavropol territory. Vestnik veterinarii. 2013; (2): 19–21. (in Russian)

24. Prislegina D.A., Dubyanskiy V.M., Platonov A.E., Maletskaya O.V. Effect of the natural and climatic factors on epidemiological situation related to natural focal infections. Infektsiya i immunitet. 2021; 11(5): 820–36. https://doi.org/10.15789/2220-7619-EOT-1631 (in Russian)

25. Kulichenko A.N., Prislegina D.A. Climatic prerequisites for changing activity in the natural Crimean-Congo hemorrhagic fever focus in the South of the Russian Federation. Infektsiya i immunitet. 2019; 9(1): 162–72. https://doi.org/10.15789/22207619-2019-1-162-172 (in Russian)

26. Gubler E.V. Computational Methods for Analysis and Identification of Pathological Processes [Vychislitel'nye metody analiza i raspoznavaniya patologicheskikh protsessov]. Leningrad: Meditsina; 1978. (in Russian)

27. Gubler E.V., Genkin A.A. Application of Statistical Non-Parametric Criteria in Medico-Biological Investigations [Primeneniye neparametricheskikh kriteriyev statistiki v mediko-biologicheskikh issledovaniyakh]. Leningrad: Meditsina; 1973. (in Russian)

28. Dubyanskiy M.A., Kenzhebaev A., Stepanov V.M., Asenov G.A., Dubyanskaya L.D. Prognostication of Plague Epizootic Activity in Sub-Aral and Kyzylkum Areas [Prognozirovanie epizooticheskoy aktivnosti chumy v Priaral'e i Kyzylkumakh]. Nukus: Karakalpakstan; 1992. (in Russian)

29. Yigit G.K. An example of tick-Crimean Congo hemorrhagic fever (CCHF) in Eflani district, Karabuk, Turkey. Sci. Res. Essays. 2011; 6(11): 2395–402. https://doi.org/10.5897/SRE11.574

30. Cherkasskiy B.L. Risk in Epidemiology [Risk v epidemiologii]. Moscow: Prakticheskaya meditsina; 2007. (in Russian)


About the Authors

V. M. Dubyanskiy
Stavropol Plague Control Research Institute; Central Research Institute of Epidemiology
Russian Federation

Vladimir M. Dubyanskiy — D. Sci. (Biol.), Head, Department of epizoo tological monitorning and prognostication, Stavropol Plague Control Research Institute; member of the temporary research team for the implementation of the grant of the Russian Science Foundation, Central Research Institute of Epidemiology.

Stavropol, Moscow



D. A. Prislegina
Stavropol Plague Control Research Institute; Central Research Institute of Epidemiology
Russian Federation

Daria A. Prislegina — Cand. Sci. (Med.), senior researcher, Laboratory of epidemiology, Stavropol Plague Control Research Institute; member of the temporary research team for the implementation of the grant of the Russian Science Foundation, Central Research Institute of Epidemiology,

Stavropol, Moscow



A. E. Platonov
Central Research Institute of Epidemiology
Russian Federation

Alexander E. Platonov — D. Sci (Biol.), Prof., chief researcher, Laboratory of zoonoses.

Moscow



Review

Views: 144


Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.


ISSN 0372-9311 (Print)
ISSN 2686-7613 (Online)