Brief profile of study area
We used the typical sampling method to select 6 provinces where malaria was most prevalent in Eastern, Central and Western China as study sites, which were, Zhejiang and Jiangsu provinces from Eastern China; Henan and Anhui provinces from Central China; Yunnan and Sichuan provinces from Western China, as shown in Figure 1[11, 27]. The population of Zhejiang stands at 57 million, and Jiangsu is the fifth most populous and the most densely populated nationally during the 2019 census. Both of them have a humid subtropical climate. Anhui and Henan are landlocked province in Central China, with the population of 63 million and 96 million in 2019, respectively. Most of their cities have a continental monsoon humid climate. Sichuan adjoins the Tibetan Plateau in the west, with the third largest population in China. Yunnan, a mountain and plateau region on the southwestern border of China, with a subtropical climate. Both Jiangsu and Zhejiang ranked top fifth in Gross Domestic Product (GDP), followed by Henan, Sichuan, and Anhui, while Yunnan's economy ranked behind in 2019. Compared with the provinces of Western China, the provinces of Eastern and Central China have well-developed transportation[28]. During 2014-2016, the average annual malaria incidence rates in Zhejiang, Jiangsu, Anhui, Henan, Sichuan, and Yunnan were 3.1, 3.0, 1.9, 1.3, 2.5, and 7.5 per million population, respectively.
Data source
Data on the numbers of malaria cases reported by each health facility in sample provinces during the period from January 2014 to December 2016 was obtained from the Infectious Diseases Information Reporting Management System (IDIRMS) with the help of the National Health Commission of the People’s Republic of China. These data was used to analyze the distribution of malaria patients in different types of health facilities. Due to the data obtained from IDIRMS lacked patients residential address information, it could not be used directly for the analysis of the proportions of patients seeking care in hospitals within county and prefectural administrative area. So we used the stratified sampling method to select two provincial hospitals, five prefecture-level hospitals, ten county-level hospitals with most malaria cases in each province. Then we used the clustering sampling method to obtain medical record data in hospitals. Variables collected from the medical record included patient demographics, patient residential address, hospital course. According to our sampling expectation, a total of 1868 cases from 102 hospitals should be investigated. In fact, we sampled 1,633 cases from 61 hospitals. The actual number of samples accounted for 87.4% of the total expected number of samples. The actual number of samples accounted for 54.2% of the total number of hospital cases (3011 cases). Among them, Anhui Province has the most samples by default.
Variables and definition
The basic characteristics of hospitals (including address, hospital tier) were obtained from the National Medical Institution Inquiry System[29, 30]. The patient's address was derived from the medical records. Each patient had two residences in the medical record, one is permanent home residence and the other is current residence. The current residence was used in our study, which usually refers to the place where the patients who had been living and working in the past three months, including the patients who were long-term migrants living and working in one location. The address used for analysis only needs to be accurate to the county level, so for patients who lacked detailed address information, we used the county government address of the county where he currently lived to replace his current residence. Then we geocoded the address using Baidu Map Application Programming Interface (API).
Within county patients mean that patients sought care in the hospitals within the county area. Within city patients mean that patients sought care in the hospitals within the prefecture-level area where they currently lived. A province could be divided into several prefecture-level areas, and each prefecture-level areas could be divided into one city and several counties in China. A more detailed terms description of within county and within city patients was shown in Figure 2. The data we collected was based on cases reported by health facilities in sample province. There may be some patients in the sample province who went to another province to seek care. The study assumes that the number of cases flowing in and out was equal. Cases reported at the health facilities in the sample province but whose current residence was not the sample province are counted in the sample province. And defined as (a) Proportion of between-county cases in A province = Number of inter-county cases / Total cases reported in A province. And the principle of calculating the proportion of inter-city cases, the proportion of inter-province cases is the same as above.
China’s hospital system was administrated by governments at the provincial, prefecture-level and county-level, so hospitals can be classified into provincial hospitals, prefecture-level hospitals and county-level hospitals. Hospitals of higher administrative level tend to have better medical resources. In a county level region, township hospitals and county-level hospitals belong to grassroots medical institutions which provide basic medical services. Hospital levels are rated by hospital size and treatment level. Generally, the treatment level of the tertiary hospital is higher than that of the secondary hospital. Almost county-level hospitals belong to secondary hospitals, and most provincial-level hospitals and municipal hospitals belong to tertiary hospitals in China. Usually provincial hospitals are located in the provincial capital city and prefecture-level hospitals are located in the city.
Data analysis and mapping
Paper medical records were double entered into the Microsoft Excel 2016 (Microsoft Corporation, USA) and then were merged to electronic record data, and a descriptive analysis was conducted. Point maps were created to visualize the spatial distribution of hospitals reporting malaria cases using the ArcGIS software version 10.7 (ESRI, USA)[31]. Flow maps were created to illustrate the patient spatial flow using the “XY To Line” analyst tool in ArcGIS.
Ethical approval
The study was approved by of the Ethics Committee of the Tongji Medical College of Huazhong University of Science and Technology (IORG0003571). Permission was granted by the National Health Commission of People’s Republic China and the manager of each hospital. Patient information was anonymised and deidentified before analysis.