Validation Assessment of the RTI Micropem V3.2 as Part of Preliminary Research for the Ghana Randomised Air Pollution and Health Study (GRAPHS)

Abstract

Background: Improved exposure assessment methods are needed in biomass studies to better define dose-response relationships and inform how clean improved stoves must be to improve health. The recently-developed RTI MicroPEM monitor appears to be a major step forward in monitoring personal exposure to fine particulate matter (PM2.5), but independent validation results in rural settings have not been reported. Objectives: To validate the ability of the RTI MicroPEM to provide precise and accurate estimates of personal exposure to PM2.5 Methods: Traditional personal samplers (BGI 4S pumps and KTL 4 LPM cyclones (Waltham, MA)) and duplicate MicroPEM monitors were worn by 3 rotating cooks during 15 controlled cooking sessions at the Kintampo Health Research Center. As a control, co-located monitors at 2 fixed sites recorded ambient PM2.5 concentrations. Accuracy was assessed by comparing filters (target = 8 sets), which integrated over multiple cooking sessions, from both types of monitors. Measurements of duplicate microPEM monitors were compared evaluate precision (target= 60). Impact of pump duty cycle was investigated by comparing distribution of data when the sampler was constantly on to distributions for different data subsampling rates as a proxy for duty cycles. Data completeness was assessed during thirty-one 72 hour deployments of MicroPEMs (50% duty cycle) worn by subjects in communities of rural Ghana. Results: Regression analysis finds a highly significant relationship between the filters from the BGI and microPEM monitors (N= 7, slope = 1.09 +/- .18, r2 = 0.97, p<0.001), with the data distributed tightly about the identity line. Likewise,a highly significant relationship is found between session averages of original and duplicate MicroPEM monitors (N=56); (Slope =1.00+/-.01, r2 = 0.98, p=0, intercept near zero). A 50% duty cycle did not impact the distribution of the PM2.5 data. The mean data capture rate for the 72-hour deployments (N=31) was 97.9%.