estimateCutoff: Estimate Epidemiological Cutoff Values

In sp2019-antibiotics/bacSplittR: Analysis of Antibiotic Resistance Data

Description

The function estimateCutoff contains an algorithm designed to estimate an Epidemiological Cutoff Value of Zone Data pairings of bacteria and antibiotics. The method was inspired by the paper "Statistical characterisation of bacterial wild-type MIC value distributions and the determination of epidemiological cut-off values".

Usage

estimateCutoff(obs, diam = 6:50, start = "mean", fit = "n_abs",
  q = 0.0075, plot = FALSE)

Arguments

obs	numeric vector giving the absolute frequency of observations per diameter
diam	numeric vector containing the diameter range (in mm), same length as obs, with default from 6:50
start	method for estimating mean start parameter; default: "mean", other option: "peak1"
fit	criterium for the fit; default: "n_abs", other options: "Pmean", "Psd", "Pn", "sigma"
q	numeric value for the quantile of the distribution; default: 0.0075
plot	logical; if FALSE (default) no plot is drawn

Details

Normal distributions are fitted to several subsets of the data, where (a part of) the most resistant data is removed. For this, Nonlinear Least Squares is used, which needs to be initzialized with starting values.

There are two options to estimate the start value of the parameter mean of the fitted normal distribution:

• "mean" = weighted mean

• "peak1" = the rightmost peak of the kernel density estimate, which surpasses a certain share (12.5 This can be especially useful, if the resistant observations also resemble a normal distribution.

To determine the subset which yields the best fitted normal distribution, one of the following criteria needs to be chosen:

• n_abs: the absolute deviation of the estimated number of observations in the subset to N

• Pmean: the p-value of the parameter mean in the NLS model

• Psd: the p-value of the parameter sd in the NLS model

• Pn: the p-value of the parameter n in the NLS model

• sigma: a parameter of the nls-function, which is the square root of the estimated variance of the random error

Here, the default value is "n_ab", which is also used in the original algorithm for MIC data.

Value

numeric value, estimate for the ECOFF

References

Turnidge, J., Kahlmeter, G., Kronvall, G. (2006) Statistical characterization of bacterial wild-type MIC value distributions and the determination of epidemiological cut-off values. Clin Microbial Infect 12: 418-425 doi: 10.1111/j.1469-0691.2006.01377.x

Examples

data("ZD", package = "EUCASTData") #load data
observations <- as.numeric(ZD[706,4:48])
estimateCutoff(observations, start = "peak1", fit = "n_abs", plot=TRUE)
example1 <- as.numeric(subset(ZD, Antimicrobial == "Ampicillin" & Bacterium == "Escherichia coli",
  grepl("^Z", colnames(ZD))))
estimateCutoff(example1, start = "peak1", fit = "n_abs", plot=TRUE)
example2 <- as.numeric(subset(ZD, Antimicrobial == "Piperacillin" & Bacterium == "Escherichia coli",
  grepl("^Z", colnames(ZD))))
estimateCutoff(example2, start = "peak1", fit = "n_abs", plot=TRUE)
example3 <- as.numeric(subset(ZD, Antimicrobial == "Mecillinam" & Bacterium == "Escherichia coli",
  grepl("^Z", colnames(ZD))))
estimateCutoff(example3, start = "peak1", fit = "n_abs", plot=TRUE)

sp2019-antibiotics/bacSplittR documentation built on Nov. 5, 2019, 9:14 a.m.