The 2186 km China-Myanmar boundary consists of two sections: 189 km of the Tibet (China)-Myanmar and 1997 km of the Yunnan (China)—Myanmar border. Due to the high altitude and frigid climate, malaria cannot be transmitted across the Tibet-Myanmar border. There are 18 counties under six prefectures in Yunnan along the border between Yunnan and Myanmar. A population of 5.6 million is living in the Yunnan border area, where malaria-free status has been maintained since the last indigenous P. vivax malaria case was reported in April 2016 . The risk of malaria reintroduction is also very low in Nujiang Prefecture of the Yunnan due to its cold climate resulting from high altitude and seasonal snowfall on the mountains.
There are 23 townships under Kachin State and Shan State in the Myanmar side of the international border. A total 6321 malaria cases (5894 in Kachin, 286 in Northern Shan, and 141 in Eastern Shan) were reported from these townships with a population of 1.5 million in 2016 . Most parts of the Myanmar border area are managed by the five local ethnic minority governments: Kachin Special Region I (KR1), Kachin Special Region II (KR2), Kokang Autonomous Region (KAR), Shan Special Region II (SR2), and Eastern Shan Special Region IV (SR4). The Laiza City and the surrounding areas in the KR2 are an important malaria hot spot . Based on the epidemiology of the local malaria situation [17,18,19,20]this study was conducted in five prefectures in the Yunnan of China: Dehong, Baoshan, Lincang, Pu’er, and Xishuangbanna, as well as in the KR2 of Myanmar. The study sites in the Yunnan were originally hyperendemic areas with high receptivity and a high risk of malaria reintroduction . As one of the enrollment criteria, these participant health facilities must routinely perform microscopy for malaria diagnosis. At last, 37 health facilities in the five prefectures of the Yunnan and the Laiza City Hospital in the KR2 of Myanmar conducted this study (Fig. 1).
Study design and recruitment
In this cross-sectional study, the microscopy-confirmed malaria cases and the microscopy-excluded non-malaria febrile (NMF) patients who attended the same health facility were recruited (Fig. 2). Following one malaria case enrolled, two NMF patients were recruited within a week. This recruiting approach was repeated till to reaching the required simple size. The sample size was calculated using a 95% two-sided confidence level, 80% power, 20% of malaria cases with exposure to attending a health facility with laboratory testing for malaria within 48 h, and 10% of NMF patients with the same exposure in Epi Info 7.0 (Centers for Disease Control and Prevention, USA). The sample size obtained was at least a total of 158 malaria cases and 316 NMF patients. From May 2016 to October 2017, the research subjects were enrolled, and then an expert microscopist with the WHO Malaria Microscopy Level One Certificate re-read the blood slides to further confirm malaria cases and exclude malaria from the NMF patients.
Malaria diagnosis and treatment and data collection
Thin and thick smears for patients with documented fever (axillary temperature ≥ 37.5 ℃) or a history of fever in the previous 48 h were prepared and stained with Giemsa in the study health facilities. Laboratory technicians obtained informed consent, and then conducted microscopy for malaria parasites. Based on the results of microscopy, patients were assigned to either the confirmed malaria group or the NMF group. Patients were excluded from this study if they were in severe clinical conditions, as were those without informed consent, who were also excluded (Fig. 2). As a quality control, all blood slides of research subjects were re-read by an expert microscopist with the WHO Malaria Microscopy Capacity Level One Certificate for further confirmation, namely, to exclude subjects with false negativity and false positivity. Anti-malarial treatment is free of charge for all malaria cases, provided by the National Malaria Elimination Programme in China and the Malaria Project of the Global Fund to Fight AIDS, Tuberculosis, and Malaria in Myanmar.
A paper-based questionnaire was pre-tested for validation. Before the study, laboratory technicians who acted as interviewers underwent intensive training to ensure that the interviewers from the Laiza City Hospital, who were fluent in both Kachin and Chinese, could ask each question in Kachin and then fill out the questionnaire in Chinese. The questionnaire was administered to each subject in Mandarin in 37 health facilities in China, and in the Kachin ethnic language at the Laiza City Hospital in Myanmar. For the child participants, their parents or guardians helped to answer the questions. The questionnaire consisted of 36 questions on treatment-seeking behaviour for fever, socio-demographic characteristics, education, main sources of cash, places of residence, housing conditions, items of durable assets, malaria awareness and knowledge, and family decision-making. According to the principal household components, a quintile Family Wealth Index (FWI) was constructed in the questionnaire, namely, most poor, mid-low, middle, mid-high, and least poor (Additional file 1: Table S1) .
Data were entered and cleaned in Excel 2007 and then analysed in Epi Info 7.2. First, the proportions of TSB, including the time intervals between the onset of symptoms and attendance of health facilities with laboratory testing for malaria, self-medication, and other sites visited for treatment prior to attending the health facilities with laboratory testing, were calculated for the malaria group and the NMF group, respectively. Second, the proportions of TSB between the malaria group and the NMF group were compared using a two-tailed Fisher exact chi-squared test, with of P < 0.05 were considered significant. The same statistical analysis above was conducted for the data of the subsample collected in China and in Myanmar, respectively. Third, influence factors associated with appropriate TSB were analysed using unconditional multivariate logistic regression analysis to control for confounding factors (Fig. 2).
An ideal TSB for malaria case patients should be to visit a health facility within 24 h . However, the primary analysis results showed that the percentage of malaria case patients seeking treatment in a health facility with laboratory testing within 24 h was too low to be statistically significant. Most of the malaria cases were P. vivax along the China-Myanmar border. Taking into consideration of these two factors, the appropriate TSB was defined as seeking treatment in a health facility with laboratory testing (microscopy or rapid diagnostic test [RDT] or both) for malaria within 48 h. Otherwise, the TSB was deemed inappropriate. In logistic regression analysis (LRA), the outcome variable was the appropriate TSB. The independent variables were variables of socio-demographic characteristics, education, main cash sources, residing locations, housing condition, FWI, malaria awareness and knowledge, and family decision. Non-response answers were treated as missing values and, therefore, excluded from the analysis .