Fig. 2
The effect of wall modifications and waning residual efficacy on the actual coverage of IRS campaigns. a Data from mud houses in South Africa (IRS with deltamethrin in 1999) and b from Koraput district in India (IRS with DDT 2004). The blue line represents actual (demographic or initial coverage as reported by programs and/or partners determined at the time of spraying), the yellow line represents observed wall modifications following real IRS-applications, the green line represents the residual bioefficacy on non-modified walls as monitored through WHO cone bioassays and the orange line represents the predicted impact of wall modifications combined with a waning residual efficacy on the effective IRS coverage. Logistic binomial models were fitted through actual residual bioefficacy data, obtained from (deltamethrin) [15] and (DDT) [43]. Bayesian models were fitted using Hamiltonian Monte Carlo sampling methods [51, 52]. Four chains were initialised to assess the convergence of 1000 iterations, the first 500 of each were discarded as burn in. The posterior distributions of parameters (4000 iterations) and 90% Bayesian credible intervals were estimated, posterior checks were performed using shinystan (version 1.0.0) [53] and visually confirmed to fit the data. The caveats for the South Africa data are: (1) repeated measures, as walls were washed or replastered on more than one occasion during the year, which leads to an overrepresentation of the extent of wall modifications, (2) initial coverage not being reported, necessitating us to use the self-reported 86.6% for IRS during the past 2 years, (3) no information on start date of IRS, so we assumed month 1 to be November, matching text with table. The caveat for the data from India includes initial coverage (reported for households and structures, we opted to show household-level data, as modification data were also reported at that level). For both studies it is not clear (probably not assessed) if all rooms and all walls in each room had been modified, or if only part of the house/structure was affected. Note that additional data (South Africa, concrete houses; India, Malkangiri district) are shown in Additional file 1: Fig. S1, and show similar patterns