Relative moldiness index as predictor of childhood respiratory illness

Abstract

The results of a traditional visual mold inspection were compared to a mold evaluation based on the Relative Moldiness Index (RMI). The RMI is calculated from mold-specific quantitative PCR (MSQPCR) measurements of the concentration of 36 species of molds in floor dust samples. These two prospective mold evaluations were used to classify the mold condition in 271 homes of infants. Later, the development of respiratory illness was measured in the infants living in these homes and the predictive value of each classification system was evaluated. The binary classification of homes as either moldy or non-moldy by on-site visual home inspection was not predictive of the development of respiratory illness (wheeze and/or rhinitis) (P=0.27). Conversely, a method developed and validated in this paper, using the RMI index fit to a logistic function, can be used to predict the occurrence of illness in homes and allows stake-holders the choice among various levels of risk.

Figure 1

Notched box-plot of RMI values for all 271 study homes. The notch forms an approximate 95% confidence interval around the median. RMI values are shown along the left margin of the plot with corresponding percentiles of the empirical distribution shown along the right margin. This plot shows the overall range and distribution of the RMI values and illustrates the relationship between percentiles and corresponding quantiles.

Figure 2

(a) Plot showing odds ratios with 95% confidence interval spanning the quantiles corresponding to the middle 90% of the RMI distribution. (b) Prediction error rates from 10-fold cross-validation of logistic discriminant analysis using RMI values to predict illness (wheezing and/or rhinitis). The vertical line shows an example of an RMI threshold of −4.29 subsequently used for both predicting the incidence of illness (Table 2) and for dividing the homes into more moldy homes and less moldy homes (Table 3). False negatives are shown as minus signs and false positives as plus signs.