Epidemiology of Coronavirus Disease in Gansu Province, China, 2020

Jingchun Fan; Xiaodong Liu; Weimin Pan; Mark W. Douglas; Shisan Bao


Emerging Infectious Diseases. 2020;26(6):1257-1265. 

In This Article

Materials and Methods

We split the 12-day study period in half. The early period began January 23, the date of the first confirmed case of COVID-19 reported in Gansu Province, and continued through January 28, the date when the Gansu government decreed it mandatory to wear face masks in public places. The late period, the subsequent 6 days, extended from January 29 through February 3. To analyze the epidemiology of the COVID-19 outbreak in Gansu Province, we compared cases diagnosed during the early and late periods. Our aim was to compare groups to determine if the restriction order was effective for controlling transmission. The definition of primary versus secondary cases refers to whether persons traveled from Wuhan (primary) or never left Gansu Province (secondary). The aim of this distinction was to explore potential transmission.

The first wave of infection was seen in persons who arrived from Wuhan, whereas later infections resulted from virus transmission from the first group of persons. To compare the demographic and clinical characteristics of persons with primary and secondary infection, we collected data about transmission and family clusters, including the date of return from epidemic areas, first day of close contact, date of symptom onset, date of first medical visit and hospitalization, and relationships between patients with primary and secondary cases.


Gansu Province (32°31′N–42°57′N, 92°13′E–108°46′E) is located in northwestern China (Figure 1). Gansu is the seventh largest province in China, comprising 12 prefecture-level cities and 2 autonomous prefectures (86 counties and districts), with a total land area of 454,000 km2 and a population of 26,257,100 in 2019.[11] It is a long, handle-shaped province, and Lanzhou is located on the Yellow River. The complex landforms of Gansu Province include mountainous regions, plateaus, plains, river valleys, and desert. With a population of 3.8 million and 13,100 km2, the population density of Lanzhou is 287 persons/km2 (Figure 2), although Lanzhou is classified as a third-tier city in China.[10]

Figure 1.

Location of Gansu Province and Wuhan, China. Circles indicate capital cities.

Figure 2.

Population density of Gansu Province, China, in 2018.


COVID-19 diagnoses in Gansu Province from January 23 through February 3, 2020, were confirmed in the laboratory of Gansu Provincial Centre for Disease Control and Prevention.[12] Suspected cases of COVID-19 infection were identified in hospitals and confirmed in the same laboratory by specific nucleic acids. We collected demographic data, including patient sex, age, occupation, place of residence, and exposure history, from the official website of the Gansu Provincial Center for Disease Control and Prevention (http://gscdc.net).

Within each prefecture or prefecture-level city in Gansu Province are districts, counties or autonomous counties, and county-level cities. For this study, we classified all counties and county-level cities as counties for simplicity and for data analysis. To conduct a geographic information system (GIS)–based analysis on the spatial distribution of COVID-19 cases, we applied the county-level polygon map (county layers) at 1:250,000 scale from Data Sharing Infrastructure of Earth System Science (http://www.geodata.cn), on which we generated a county-level layer containing information regarding latitudes and longitudes of cases in county layers (central points) of each county.

Statistical Analyses

The 54 patients were assigned numbers from 1 to 54 according to time of diagnosis. The statistical descriptions included demographic characteristics, exposure history, whether the cases were primary or secondary, and potential spread of disease. Because the ages of case-patients were not normally distributed, we performed nonparametric Brown-Mood tests to compare medians between early and late cases. For expected cell sizes <5, we used the Fisher exact test to compare the frequency between or among groups; otherwise, we used the χ2 test. We estimated days from illness onset to first medical visit and days from illness onset to hospitalization by fitting a Weibull distribution to the dates of illness onset, first medical visit, and hospital admission.[13]

GIS Mapping and Spatial Analyses

We geocoded all COVID-19 cases and matched them to the county-level layers of polygon and point by administrative codes by using ArcGIS 10.2.2 software (https://www.arcgis.com). To explore the spatial distribution pattern of COVID-19 cases on the county level during the study periods, we applied local indicators of spatial association (LISA[14]). Using LISA, we could identify the type and degree of spatial clustering, including significant hot spots (high-high), cold spots (low-low), and spatial outliers (high-low and low-high) between a given location and surrounding spatial units by calculating the local Moran's I.[14,15] We used the Z statistic to determine the significance of clustering based on a significance level of 0.05. A significant positive Z indicates high-value regions surrounded by high-value regions (high-high) or low value regions surrounded by low-value regions (low-low). A significant negative Z indicates high-value regions surrounded by low-value regions (high-low) or low-value regions surrounded by high-value regions (low-high).[16]

Ethics Approval

Our study was approved by the institutional review board, Gansu University of Chinese Medicine. We collected data from the official website of Gansu Provincial Center for Disease Control and Prevention, which was considered exempt from approval.