GoKawiilData
Compiling the database
We compiled our database by identifying all sources providing data on wild meat consumption in seven Central African countries between 2000 and 2022: Cameroon, Central African Republic, Democratic Republic of the Congo, Equatorial Guinea, Gabon and Republic of the Congo. We also included data from the Cross River Forest landscape in southeast Nigeria, as forests there are contiguous with protected areas in Cameroon. We considered peer-reviewed articles, technical reports, PhD and master’s dissertations, online data repositories and unpublished data, adopting a snowball sampling approach70 to search reference lists and online libraries. We used wildmeat, wild meat, bushmeat, bush meat and viande de brousse as main keywords, and consumption, nutrition and food as secondary keywords. We defined a study as a set of data collected using a single methodology in a specific study area over a determined timeframe. In this way, each data source could provide more than one study. For example, large projects that monitored multiple regions in different countries were split so that each study area represented a single study. For consistency, we restricted our research to studies investigating wild meat consumption at the household level, discarding those monitoring consumption of individual consumers who could not be aggregated to households, for example by enquiring people randomly met in the streets, a methodology mostly used in cities, where household surveys are difficult to implement. This is also the reason for the limited number of cities included in our database. Most recent studies used KoboToolBox (https://www.kobotoolbox.org/) and different versions of the KoboCollect App (v.2.020.40 and subsequent releases). Older studies recorded data with pen and paper. When possible, we downloaded the raw data from online resources (such as publicly available databases). Alternatively, we contacted the authors to request the raw data.
Data preparation
Individual studies underwent a preliminary phase of data cleaning and standardization to conform with the format required by the database https://www.wildmeat.org/. The resulting database included studies providing three different datatypes: (1) consumption/non-consumption; (2) frequency of consumption; (3) quantity consumed, each requiring specific formatting of the raw data.
Consumption/non-consumption data were provided by all studies and were therefore available for all recall events in our database (that is, at the recall level). If a household declared to have consumed wild meat during the specific recall, we recorded a ‘consumption event’ and coded the recall as 1. By contrast, if no wild meat consumption was reported during the recall period, we coded the recall as 0. Here, we considered a minimum recall period of 24 h (see the ‘Statistical analyses’ section for a description of how the number of monitored days was accounted for in the analyses). Therefore, if wild meat consumption was available for multiple meals within 24 h, we aggregated the information available for single days. In other terms, if wild meat was consumed twice (for example, in the morning and afternoon) within the same 24 h, we coded the 24 h recall as 1.
Frequency of consumption—defined as the number of consumption events recorded over the number of days a household’s consumption was monitored—was provided by 24 out of the 30 studies included in our analysis, representing 11,582 households and 107,896 recall events. For studies that recorded the frequency of consumption in categories such as daily, weekly and monthly, we calculated the frequency on a scale from 0 to 1. For example, if a household reported consuming wild meat monthly, it was assigned a frequency of consumption of 0.033 (12/365 days). However, for studies that recorded several recall events from the same household, we calculated the frequency of consumption by dividing the number of consumption events, by the total number of recalled days. For example, if a household was interviewed for 6 days about wild meat consumption over the previous 24 h and consumption occurred on two occasions, we calculated frequency as 2/6 days = 0.33. In this way, the frequency of consumption referred to individual households, rather than to single recall events. Each household had to be monitored for a minimum of 2 days to be considered (see the ‘Statistical analyses’ section for a description of how the number of monitored days was accounted for in the analyses). In other words, studies that recorded wild meat consumption over 24 h, in a single occasion for each household, were considered as consumption/non-consumption data.
Finally, 19 studies provided the quantity consumed (in g, kg, or local units such as leg, piece or entire animal) by the households over a recall period, as well as information on the wild animal species consumed. These studies included 9,189 households and 105,503 recall events. Data were available at the recall level, and we standardized the data as the quantity (in kg) consumed per household per day. Therefore, if the recorded quantities represented the cumulative consumption over a recall period of >24 h, we divided the reported quantities by the duration of the recall (in days). So, if a household reported having consumed 12 kg of undressed meat over a 72 h recall (that is, 3 days), we considered the quantity consumed by the household in a day to be 4 kg. Conversely, if quantities consumed were recorded for multiple meals within 24 h, we summed the quantity of wild meat reported for a single day. Thus, if a household reported to have consumed 0.5 kg of wild meat in the morning and 1 kg in the evening, the quantity of wild meat consumed by the household in that day would be 1.5 kg. Finally, when consumed quantities were reported in local units of measure (Extended Data Table 1), we estimated consumed kilograms following procedures specific for each unit. If the consumed units were reported in local units (such as entire, half, quarter or gigot), we assigned the species-specific average mass value using data available from the literature71 or empirical data obtained in Gabon by the authors of this study (L.C., Dibouka, 2001–2010; D.F. and D.C., Lastourville area, 2021). For all other units, including piece, pile and plate, we used estimated conversion factors, based on empirical observation collected by various authors of this study (L.C., K.A., F.S., D.D.). Because in Central Africa wild meat is generally sold and cooked along with bones and sometimes skin, we considered the quantities in our database as the quantity of undressed meat (in kg) consumed per household per day.
Ethics statement
The procedure used to compile the database was approved by the ethics review committee of CIFOR/ICRAF (SLF6430000-UFW044-AI2; 13/12/2021) and included the anonymization of all sensitive data. All included studies obtained (1) an ethical review of data collection protocols, (2) the agreement of the local communities (focal groups/authorization of the communities’ representatives) before data collection; (3) prior informed consent from all respondents (Extended Data Table 1).
... continue reading