Identification of Chinese Plague Foci from Long-Term Epidemiological Data
Total Page:16
File Type:pdf, Size:1020Kb
Identification of Chinese plague foci from long-term epidemiological data Tamara Ben-Aria,b,1, Simon Neerinckxa, Lydiane Agiera, Bernard Cazellesb,c, Lei Xud, Zhibin Zhangd, Xiye Fange, Shuchun Wange, Qiyong Liue,2, and Nils C. Stensetha,2 aCentre for Ecological and Evolutionary Synthesis, Department of Biology, University of Oslo, N-0316 Oslo, Norway; bCentre National de la Recherche Scientifique, Ecole Normale Supérieure, Unité Mixte de Recherche 7625, Université Pierre et Marie Curie, 75230 Paris Cedex 05, France; cUnité de Modélisation Mathématique et Informatique des Systèmes Complexes, UMI 209, Institut de Recherche pour le Développement et Université Pierre et Marie Curie, 93142 Bondy Cedex, France; dState Key Laboratory of Integrated Management on Pest Insects and Rodents, Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China; and eState Key Laboratory for Infectious Diseases Prevention and Control, National Institute for Communicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing 102206, China Edited* by Hans R. Herren, Millennium Institute, Arlington, VA, and approved April 2, 2012 (received for review July 1, 2011) Carrying out statistical analysis over an extensive dataset of the administrative constraint on the data and strive to objectively human plague reports in Chinese villages from 1772 to 1964, we establish the contours of plague foci. identified plague endemic territories in China (i.e., plague foci). Our dataset covers the period 1772–1964 over all of China. It Analyses rely on (i) a clustering method that groups time series is potentially hindered by some shortcomings, which we discuss based on their time-frequency resemblances and (ii) an ecological in detail. The time coverage includes epidemic (respectively niche model that helps identify plague suitable territories charac- endemic) periods, that is, periods when above (respectively. terized by value ranges for a set of predefined environmental below) number of villages reported plague. Endemic periods, variables. Results from both statistical tools indicate the existence when outbreaks are of smaller amplitude and localized in space, are of great interest to us because they provide the opportunity of two disconnected plague territories corresponding to Northern fi to identify territories where plague is maintained without regular and Southern China. Altogether, at least four well de ned inde- external input. Indeed, the main focus of this paper is to identify pendent foci are identified. Their contours compare favorably with fi plague foci with a typical resolution of hundreds of kilometers. eld observations. Potential and limitations of inferring plague We thus go beyond simple data aggregation (Preamble on the foci and dynamics using epidemiological data is discussed. Data) and to define foci by a combination of similarity in the local time series (TS) (Clustering) and similarity of environments Yersinia pestis | Ecological Niche Modeling | wavelet analysis | [Ecological Niche Modeling (ENM)]. plague dynamics First, local TS are clustered on the basis of a decomposition using wavelet decompositions. Subsequently, local time series lague is an ancient disease caused by the bacterium Yersinia are classified according to their degree of wavelet spectrum re- Ppestis. It is hosted by small mammals (mainly rodents) and semblance (14). This allows us to identify common patterns vectored through bites from infected fleas (1). In the natural without any a priori assumption on time series stationarity or on fi cycle of plague, humans are secondary hosts (i.e., they do not the consistency of plague dynamics within a de ned geographical participate in plague’s maintenance cycle). Human plague unit. We show that local TS tend to display spatial coherency. reports are nonetheless the main available source of worldwide Three regions robustly emerge from our analysis, namely a southwest, a northwest, and an eastern area. data on plague (2), with the notable exception in Asia of – Kazakhstan and other previous Soviet Union countries (3). It is Second, we use ecological niche modeling (15 18) [see Eco- therefore important that we find ways to efficiently use human logical Niche Modeling (ENM)] to characterize and segregate different plague niches in China. Results from the heuristic, data to infer where and how plague is endemically maintained. spectral, and ecological approaches are reconciled in Discussion Plague has a long epidemic history in China (4). It has been to propose a map of plague enzootic foci in China. claimed (5, 6) that the first two plague pandemics both originated in China. To our knowledge, the oldest recorded evidence of Preamble on the Data plague can be found around 1353 in the northeastern Hebei The dataset is a compilation of human plague cases recorded in province (4) (Fig. S1 provides cited provinces). The third pan- local gazetteers (or difangzhi) (19). We calculated a yearly demic most likely started in China is responsible for the death of number of plague locations within a regular grid. Fig. S2 presents 12 million people in China and India alone (7). It appeared in the location of these in 0.5*0.5° grid cells separated in three time Hong Kong in 1894 and spread to major world ports through windows corresponding to below (period 1: 1772–1893 and pe- merchant ship transports (8). The plague source for the third riod 3: 1953–1964) or above (period 2: 1894–1952) average pandemic may have been northwestern Yunnan province, where number of village reports for all of China. Separation years fi fi it remained con ned since the earliest con rmed records in the (vertical lines in Fig. S2 B–D), 1894 and 1952, correspond to city of Dali (in 1772 or earlier according to different sources) (9). some outstanding dates for plague. The former is the start of the Expansion of plague to the southeast occurred during the first third pandemic and the latter the end of the Chinese civil war and part of the twentieth century, while human plague incidence also increased in northern China (10). In modern times, plague remains an important public health concern to Chinese authori- Author contributions: T.B.-A., S.N., and N.C.S. designed research; T.B.-A., S.N., and L.A. ties as exemplified by the 2009 pulmonary outbreak in the city of performed research; T.B.-A., S.N., L.A., B.C., L.X., Z.Z., X.F., and S.W. analyzed data; and Ziketan, Qinghai province (11). T.B.-A. wrote the paper. There have been few studies focusing on occurrences of hu- The authors declare no conflict of interest. man plague in China. The most comprehensive study (12) *This Direct Submission article had a prearranged editor. revealed that plague intensity shows a positive correlation with 1Present address: Institut National de la Recherche Agronomique (INRA), 147 rue de a wetness index up to a threshold level dependent upon location l’université 75007, Paris, France. in northern or southern China. Another study also linked plague 2To whom correspondence may be addressed. E-mail: [email protected] or n.c.stenseth@bio. intensity in China to climate at the province scale (13): Their uio.no. results emphasized the existence of different, separated plague This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. environments within these predetermined regions. Here we relax 1073/pnas.1110585109/-/DCSupplemental. 8196–8201 | PNAS | May 22, 2012 | vol. 109 | no. 21 www.pnas.org/cgi/doi/10.1073/pnas.1110585109 Downloaded by guest on September 27, 2021 the start of the plague control program around 1950 (20). We will group 2, and the eastern part of NE plus SE for group 3 TS. henceforth name periods 1 and 3 (Fig. S2 B and D) endemic as Groups 1 and 3 thus cover the four suspected nonepidemic opposed to the epidemic period 2 (Fig. S2C). During the first plague territories revealed by Fig. S2, whereas group 2 TS are period, plague only occurs in southern China, mainly Yunnan and associated with the 1894–1952 epidemic period. The robustness to a lesser extent in the Guangdong and Fujian provinces; there of the aggregation is demonstrated by the comparison between are only sporadic reports of plague in the northern provinces of grouping results for n = 3 and n = 6. Adding more groups only Hebei and Inner Mongolia. The beginning of the third plague removes a very limited number of outliers from each group (not pandemic in 1894 led to high plague incidence over the most shown for clarity). Other parameters than N could also influence extended area. Finally, the short period 3 extends from 1952 to our results. To check their robustness, we carried out a series of 1964 and corresponds to a contraction of the plague area to the sensitivity tests. A small number of TS (<10%) may belong to western part of Qinghai province and the eastern part of the one group or the other depending on changes in the parameters Inner Mongolia province. Noticeably, this period is characterized (e.g., N), which does not invalidate the main results above or the by a drastic reduction of plague reports in the entire south. interpretations made in the discussion. Overall, endemic periods suggest the existence of five separate TS for each of the three groups are presented in Fig. 1B and portions of territory subsequently called northwest (NW, a few reflect the importance of time patterns in the grouping. Main dispersed occurrences), north central (NC), northeast (NE), characteristics of each group TS are as follows: reports distributed southwest (SW), and southeast (SE), in which there are more throughout the whole time period with a long period of above- than one report per grid cell. Note that grid cells with rarer normal (epidemic) conditions for group 1, a vast majority of pla- plague occurrences (1 or 2 to up to 10 villages) tend to surround gue reports between 1937 and 1952 for group 2 and finally, two to the region with the highest plague levels, suggesting that these three bursts in 1917–1918, 1928, and 1932 with recurrent low in- plague territories are composed of “endemic cores.” A visual cidence reports after 1900 (no cases before) for group 3.