key: cord-0907521-djpq7sgf authors: Romanyukha, A. A.; Sannikova, T. E.; Drynov, I. D. title: The origin of acute respiratory epidemics date: 2011-03-29 journal: Her Russ Acad Sci DOI: 10.1134/s1019331611010114 sha: 9075a27ffa39f18d8fd96ab732cf06102c127646 doc_id: 907521 cord_uid: djpq7sgf After studying the dependence of acute respiratory diseases of all etiologies on air temperature and population immunity dynamics, the authors proposed that the air temperature and the immunity level can affect disease resistance. Knowledge of the relationship between these factors must clarify the mechanisms that determine morbidity dynamics. Upper respiratory tract infections are the most widespread diseases in developed countries. An adult falls ill with acute respiratory diseases (ARDs) 2.5 times a year on average, and children under four, four to five times a year [1] . ARD induced mortality due to sequelae is 1-2%. Children, elderly people, and HIV patients are susceptible to the highest risk of mor bidity and sequelae [2] . Acute respiratory diseases lead to substantial economic loss because of temporary incapacity [3] . In recent years, specialists have largely focused on possible epidemics related to the new flu virus variants ("bird flu," "swine flu," etc.). The importance of the ARD and flu problem is widely acknowledged. However, despite considerable efforts to reduce morbidity, population data do not show a positive decrease in ARD frequency. In the search for the possible causes of the current situation, let us con sider factors that determine the dynamics of ARD morbidity. The traditional approach to reduce the frequency of infectious diseases is based on the identification of the causative agent, the development of a vaccine, and the immunization of the population, primarily the risk group. This approach to the suppression of a certain agent is efficient to control smallpox, hepatitides, roseola, and other viral infections. In the case of respi ratory diseases, this task is complicated by the variabil ity of viruses. The picture is distorted even more by the fact that ARDs include more than 200 etiologically independent diseases that are united in one group by three characters: the agent transfer mode (airborne, contact), the localization of the main pathological process (breathing passages), and the clinical charac ter. Some respiratory viruses can stay in latency or keep up an asymptomatic chronic infection. When immunity flags or external conditions change, chronic and latent infections can wake up and cause ARDs. All these infections localize in breathing passages; conse quently, resistance and immunity to them are based on the general local structures and resources of nonspe cific and immune protection. The described facts give reasons to consider all ARDs as a complex of related diseases. The main agents of diseases of respiratory passages are assumed to be the rhinovirus, coronavirus, group A and B viruses, parainfluenza virus, and adenoviruses [4, 5] . Different virus types have different seasonal patterns. Thus, the group A virus causes the strongest epidemics in the winter months. The beginning of the school year at educational establishments is annually accompanied by the outbreaks of respiratory infec tions caused by the type I and II parainfluenza viruses. The type III parainfluenza viruses cause ARDs in summer and late spring [5] . The ARD agents quickly lose infectivity in the environment; therefore, a high population density is necessary for agent transfer. The probability of conta gion increases in crowded places, like public transpor tation and cultural and educational establishments. Another condition for the development of an epi demic is a large quantity of the population. The point is that a patient radiates viruses for only a few days; then the patient's immune system eliminates the virus and forms immune memory for it. During the infectiv ity period, one patient must infect several people to create the right number of patients after several infec tivity cycles to call it an epidemic. If the population is partitioned, for example, includes several cities, the picture becomes distorted not only because of delays in agent transfer but also because of local differences, for example, in weather. Consequently, it is easier to study the development of an epidemic from the exam ple of one city. Regularities of epidemic development have been studied well [6] , but, in terms of general theory, this process needs an agent to which the majority of people are not immune. However, this mechanism does not explain regular rises in ARD morbidity during the From the Researcher's Notebook After studying the dependence of acute respiratory diseases of all etiologies on air temperature and population immunity dynamics, the authors proposed that the air temperature and the immunity level can affect disease resistance. Knowledge of the relationship between these factors must clarify the mechanisms that determine morbidity dynamics. winter season, which are not always related to the emergence of a new agent. A possible explanation for morbidity rises in winter is reduced to a head coldrelated factor of cold weather, which leads to reduced resistance of the immune system, and increased crowdedness, as people tend to spend more time indoors. At the same time, the scientific literature does not give us a qualitative assessment of the role of cli mate factors in the dynamics of ARD morbidity and the relationship between these factors and the level of specific and nonspecific human resistance. To investigate epidemic processes, we selected the Moscow population in the 1960s-1980s. The popula tion of more than eight million people was under the care of the state health system. Migration processes were rigidly regulated and had a limited influence on the population. ARD diagnostic standards practically did not change during the whole period under exami nation. Morbid events were registered during the issu ance of sick leave certificates, which stated the diagno sis. City data were summarized for the assessment of the epidemiological situation. This created a long term corpus of data on the morbidity with the ARD complex, including the flu, in a population with a known quantity. Note that Moscow territory has sev eral permanent weather stations, which make it possi ble to compare morbidity data with the continuous flow of weather reports. The time series under study is characterized by sea sonal fluctuations with a long term trend. The average annual morbidity grew at a rate of 0.1 events per 1000 people a year. Figure 1 shows that minimal mor bidity values grew in parallel with the average annual morbidity and maximal values (winter epidemic peaks) rather decreased. To detail these differences, we conducted analysis of the seasonal structure of the long term trend. To this end, we evaluated the coeffi cients of the trend curve slope for each week of the year and for each month (Fig. 2) . The positive values of the coefficient mean average morbidity growth during the corresponding week or month. We see that the average morbidity growth is predetermined by increased mor bidity in spring, fall, and early winter. During the epi demic period (January, February), maximal morbidity decreased, and the length of its epidemic rise increased. Although the focus is on morbidity during the period of epidemic rise (week two through week nine in January and February), the morbidity contribution to total morbidity is 16% on average (from 2 to 33%), the bulk of ARD cases relating to the interepidemic period. Note that, even during an epidemic rise, the flu proper is 15-20% of the total ARD cases [2] . ARD morbidity also greatly changes during interepidemic periods (see Fig. 1 ). What factors affect these fluctua tions? Temperature changes, the "import" of new viral strains, the formation of immunity, or departure on vacations and to dachas? Do these factors act regularly or randomly? To assess the regularity of the observable ARD fluc tuations, let us calculate its autocorrelation function (ACF). This characterizes the dependence between the time series and the same series offset for a certain time period (lag). The ACF for weekly ARD morbidity data looks like an oscillating curve with a period of 52 weeks (Fig. 3) . The smoothness of the curve obtained shows regularity in annual morbidity cycles. It is interesting that the ACF has a local maximum equal to 0.52 with a lag of four years (207 weeks); this value is 15% higher than with a lag of one year and 8.9% higher with a lag of two years. This effect may be related to the fact that immunity to certain agents weakens in four years. The effect increases after deducting the linear dependence of morbidity on air temperature; the correlation value decreases; but the expression of growth increases under a lag divisible by four years. This shows the independence of the four year morbidity cycle from temperature cycles. In calculations, ACF data were averaged for the whole observation period. Therefore, only the most expressed and stable dependences were identified. At the same time, it is known that in different seasons res piratory infections are caused by agents with different epidemiological characteristics (the spread rate, the duration of latent and contagious periods, etc.). In addition, in winter the agent propagation process involves the majority of the population, and in sum mer morbidity is 20-30 times lower. Consequently, we may expect that correlation between the morbidity levels of neighboring weeks will differ. Figure 4 shows curves that describe the sequences of correlation coef ficients for the fifth and 22nd weeks. We see that the morbidity level of the 22nd week (end of May) corre lates with the morbidity levels of the following 15th-20th weeks. The morbidity of the fifth week (end of January) has a significant correlation only with the morbidity of the following five weeks. If we plot the dependence of the length of the cor relation period for morbidity levels on the week num ber, the graph will show a characteristic pattern: from week 20 through week 36 (mid May through early September), a monotonous decrease in the length of the correlation period occurs from 22 to four weeks (according to the data with a 95% confidence level) (Fig. 5) . The rest of the graph is characterized by irreg ular fluctuations of the correlation period length from four to eight weeks. Note also a less pronounced fea ture at the end of September (week 38 through week 41) with a correlation period of about 20 weeks. Although the confidence level here is slightly lower, this dependence is interesting because it implies a pos sible relation between morbidity levels at the end of September and during the winter epidemic rise between the end of January and the beginning of Feb ruary the next year. If this relation reflects a functional dependence, it is critical in terms of predicting and controlling the morbidity level during winter epidemi ological rises [2] . A lengthy correlation period of morbidity levels allows us to assume that, despite the diversity of agents, ARD morbidity is controlled by general fac tors. One of them is air temperature. The dependence of ARD morbidity on air temperature is shown in Fig. 6 . At temperatures above 0°C, morbidity grows linearly as temperature falls; at temperatures below zero, high morbidity values occur and break this linear dependence. These values relate to winter epidemics. To exclude them from analysis, we can use the notion of an epidemic threshold (a norm that medical admin istrations determine annually on the basis of the epi demic situation analysis). Assume that there is a flu and ARD epidemic if morbidity exceeds 15 per 1000 people during at least one week. Let us call this value the epidemic threshold. Analysis has shown that ARD morbidity during the interepidemic period grows linearly as air temperature drops. Let us consider the epidemic data in detail. From 1959 through 1988, ARD morbidity exceeded the epi demic threshold for 129 weeks, which was 8.6% of the total observation period and included 22 epidemics, 82% of which occurred from mid December through week number. Time intervals are presented during which the coefficient of correlation between morbidity in the first week and mor bidity in the following weeks is larger than zero (circles) and larger than zero with a 95% probability (crosses). The averaged morbidity dynamics during an epi demic is shown in Fig. 7a . To plot this graph, we syn chronized data about various epidemics by the time when the epidemic threshold was exceeded. Figure 7b also shows the dependence between the morbidity cor relation coefficients and air temperature for each epi demic week. We see that the morbidity level directly before the epidemic, as well as during the first three weeks, does not depend on air temperature. Starting with the fourth week, morbidity has a significant neg ative correlation with air temperature. Consequently, as temperature drops, epidemic duration increases. The results of processing a unique corpus of data on ARD morbidity in Moscow in show that, despite diverse agents, respiratory infections may be viewed as a complex of interrelated diseases, whose epidemiology depends on social and weather condi tions. Data analysis allows us to assume that winter flu epidemics are not isolated events caused by new viral strains. The amplitude of winter epidemic rises depends not only on virus characteristics but also on ARD morbidity levels in the previous year. Conse quently, flu epidemic controls should not be limited to immunization against certain strains of the flu virus but should be aimed to reduce the total infective load related to respiratory infections. Studies of the Community and Family: Acute Respiratory Illness and Infection The Basics of the Optimal Technology of Prevention of the Mass Spread of Acute Respiratory Diseases The Economic Burden of Non Influenza Related Viral Respiratory Tract Infection in the United States Viruses and Bacteria in the Etiology of the Common Cold Parainfluenza Viruses Infectious Diseases of Humans: Dynamics and Control This work was supported by the Russian Founda tion for Basic Research, project no. 09 01 00098 a.