Introduction In 2016, in four municipalities of the Gao district in rural Mali, two similar nutrition-sensitive programmes were conducted simultaneously. Both programmes aimed to target vulnerable households (HH) but used two different methodologies: –a community-based method: the household economy approach (HEA); –a community-based method combined with a proxy-mean test (PMT). This offered a unique opportunity to compare the effectiveness of both methods. Methods A cross-sectional survey was conducted among randomly selected HH from the four municipalities: 545 HH targeted by the first method (group 1), 513 HH targeted by the second method (group 2), 526 HH that were not targeted by any of the two methods (group 3). Socioeconomic and demographic characteristics, expenditure, dietary diversity and food security data were collected through a standardized questionnaire. Characteristics of HH from groups 1 and 2 were compared using linear models. Performances of both targeting methods were assessed against a gold-standard measure of vulnerability, namely the household's monthly food expenditures (HMFE) per adult-equivalents living in the HH. Cut-off points for HMFE targeting were set at the level corresponding to the actual targeting rate observed in each municipality. The targeting effectiveness was assessed using the leakage rate and undercoverage rates. Sensitivity (Se) and specificity (Sp) were computed as well as the targeting differential [Se - (1-Sp)]. Results HH from groups 1 and 2 had the same mean size (7.0 and 7.2 - P = 0.32); same youth ratio (1.0 and 1.0 - P = 0.64) and same dependency ratio (1.6 and 1.5 - P = 0.47). They were mainly headed by a male (75.1% and 72.3% - P = 0.33). In both groups 27% of the heads of HH were literate (P = 0.87). Groups 1 and 2 had approximately the same HMFE (2810 CFA francs and 2316 CFA francs, per adult-equivalent - P = 0.22) and the same dietary diversity score (2.9 and 3.0 food groups, out of 12 - P = 0.37). However, group 1 had a lower proportion of food secure HH (6.4% vs. 7.5% - P = 0.02) according to the Food Insecurity Experience Scale. Forty-one percents of HH were miss-targeted by the first method and 43% by the second one. The leakage rates were 60% and 64% and the undercoverage rates 31% and 33% for methods one and two, respectively. Sensitivities were respectively 40% and 34% and specificities 69% and 68%, leading to targeting differentials equal to 8% and 3%. Conclusions Community-based method and community-based method combined with a PMT exhibited similar targeting performances, the former being slightly better.
