Data analysis
All interviews and observations were transcribed from the local language into Mandarin and translated into English. All transcripts were imported into MAXQDA release 12 statistical software (VERBI Software, Berlin, Germany) for data management and analysis. The analysis process incorporated both deductive and inductive approaches, and followed a process of initial coding, identification of new themes, primary coding and identification and analysis of emerging themes.25 An initial codebook associated with the five core themes of the interview guide was developed a priori. After a close reading of the transcripts, two authors used the initial codebook to independently code two transcripts in their entirety, making notes on emerging themes and specified subthemes. Subsequently, the two authors adapted the codebook and used the modified codebook to code all of the transcripts. During the coding process, the two authors met when any major new themes or concepts emerged to decide on any necessary revisions to the codebook, until no new themes emerged and no new information was obtained from the coding. The final codebook was restructured with five sections: (1) demographics, (2) biosecurity in human environments, (3) human–animal contact, (4) illness, treatment and death and (5) animal taxa; subthemes were defined under each section.
After completion of the coding, a code report was generated from MAXQDA. Internal reliability was assessed by comparing the coded segments from two authors on the same two transcripts to reach a minimum code interaction rate of 80%.26 A saturation grid was built using the ‘Segment Retrieval’ function in MAXQDA to ensure saturation was reached.27 Coded segments were categorized into protective factors and risk factors based on their known associations with disease transmission, and the analysis was stratified at the individual, community and policy or regulation level.28 At the individual level, both risk and protective factors were analysed in terms of the individuals' knowledge, attitudes and practices to better understand the risk factors for identifying context-based strategies.29 (Figure 1).
Table 1  Demographic characteristics of the study participants
Characteristic  Participants (n=88)
Frequency, n  %
Gender
Male  58  66
Female  30  34
Age (y)
18–30  8  9
31–50  55  63
>50  25  28
Province
Yunnan  36  41
Guangxi  25  28
Guangdong  27  31
Source of livelihood
Government employee  10  11
Private company employee  7  8
School teacher  5  6
Cash crop production (e.g. fruit tree, bamboo)  23  26
Grain crop production (e.g. corn, rice)  32  36
Household animal raising for sale  13  15
Domestic animal farmer  1  1
Wild animal farmer  2  2
Health worker  2  2
Construction worker  10  11
Nature reserve worker  8  9
Small business (e.g. restaurant, grocery store owner)  16  18
Student  1  1
Mineworker  1  1
Other casual or out-migrating work (non-specific)  30  34
Has worked or work on multiple jobs to make a living  35  40