How to conduct AMR data analysis

Matthijs S. Berends

02 June 2021

Note: values on this page will change with every website update since they are based on randomly created values and the page was written in R Markdown. However, the methodology remains unchanged. This page was generated on 02 June 2021.


Conducting AMR data analysis unfortunately requires in-depth knowledge from different scientific fields, which makes it hard to do right. At least, it requires:

Of course, we cannot instantly provide you with knowledge and experience. But with this AMR package, we aimed at providing (1) tools to simplify antimicrobial resistance data cleaning, transformation and analysis, (2) methods to easily incorporate international guidelines and (3) scientifically reliable reference data, including the requirements mentioned above.

The AMR package enables standardised and reproducible AMR data analysis, with the application of evidence-based rules, determination of first isolates, translation of various codes for microorganisms and antimicrobial agents, determination of (multi-drug) resistant microorganisms, and calculation of antimicrobial resistance, prevalence and future trends.


For this tutorial, we will create fake demonstration data to work with.

You can skip to Cleaning the data if you already have your own data ready. If you start your analysis, try to make the structure of your data generally look like this:

date patient_id mo AMX CIP
2021-06-02 abcd Escherichia coli S S
2021-06-02 abcd Escherichia coli S R
2021-06-02 efgh Escherichia coli R S

Needed R packages

As with many uses in R, we need some additional packages for AMR data analysis. Our package works closely together with the tidyverse packages dplyr and ggplot2 by RStudio. The tidyverse tremendously improves the way we conduct data science - it allows for a very natural way of writing syntaxes and creating beautiful plots in R.

We will also use the cleaner package, that can be used for cleaning data and creating frequency tables.


# (if not yet installed, install with:)
# install.packages(c("dplyr", "ggplot2", "AMR", "cleaner"))

Creation of data

We will create some fake example data to use for analysis. For AMR data analysis, we need at least: a patient ID, name or code of a microorganism, a date and antimicrobial results (an antibiogram). It could also include a specimen type (e.g. to filter on blood or urine), the ward type (e.g. to filter on ICUs).

With additional columns (like a hospital name, the patients gender of even [well-defined] clinical properties) you can do a comparative analysis, as this tutorial will demonstrate too.


To start with patients, we need a unique list of patients.

patients <- unlist(lapply(LETTERS, paste0, 1:10))

The LETTERS object is available in R - it’s a vector with 26 characters: A to Z. The patients object we just created is now a vector of length 260, with values (patient IDs) varying from A1 to Z10. Now we we also set the gender of our patients, by putting the ID and the gender in a table:

patients_table <- data.frame(patient_id = patients,
                             gender = c(rep("M", 135),
                                        rep("F", 125)))

The first 135 patient IDs are now male, the other 125 are female.


Let’s pretend that our data consists of blood cultures isolates from between 1 January 2010 and 1 January 2018.

dates <- seq(as.Date("2010-01-01"), as.Date("2018-01-01"), by = "day")

This dates object now contains all days in our date range.


For this tutorial, we will uses four different microorganisms: Escherichia coli, Staphylococcus aureus, Streptococcus pneumoniae, and Klebsiella pneumoniae:

bacteria <- c("Escherichia coli", "Staphylococcus aureus",
              "Streptococcus pneumoniae", "Klebsiella pneumoniae")

Put everything together

Using the sample() function, we can randomly select items from all objects we defined earlier. To let our fake data reflect reality a bit, we will also approximately define the probabilities of bacteria and the antibiotic results, using the random_rsi() function.

sample_size <- 20000
data <- data.frame(date = sample(dates, size = sample_size, replace = TRUE),
                   patient_id = sample(patients, size = sample_size, replace = TRUE),
                   hospital = sample(c("Hospital A",
                                       "Hospital B",
                                       "Hospital C",
                                       "Hospital D"),
                                     size = sample_size, replace = TRUE,
                                     prob = c(0.30, 0.35, 0.15, 0.20)),
                   bacteria = sample(bacteria, size = sample_size, replace = TRUE,
                                     prob = c(0.50, 0.25, 0.15, 0.10)),
                   AMX = random_rsi(sample_size, prob_RSI = c(0.35, 0.60, 0.05)),
                   AMC = random_rsi(sample_size, prob_RSI = c(0.15, 0.75, 0.10)),
                   CIP = random_rsi(sample_size, prob_RSI = c(0.20, 0.80, 0.00)),
                   GEN = random_rsi(sample_size, prob_RSI = c(0.08, 0.92, 0.00)))

Using the left_join() function from the dplyr package, we can ‘map’ the gender to the patient ID using the patients_table object we created earlier:

data <- data %>% left_join(patients_table)

The resulting data set contains 20,000 blood culture isolates. With the head() function we can preview the first 6 rows of this data set:

date patient_id hospital bacteria AMX AMC CIP GEN gender
2017-08-29 U5 Hospital D Escherichia coli S I S S F
2012-10-19 U5 Hospital B Streptococcus pneumoniae S S S S F
2010-02-13 W10 Hospital A Escherichia coli R R S R F
2017-03-20 Y1 Hospital B Staphylococcus aureus R S S S F
2016-08-03 U6 Hospital B Staphylococcus aureus R S S S F
2015-06-16 X5 Hospital B Staphylococcus aureus S R R S F

Now, let’s start the cleaning and the analysis!

Cleaning the data

We also created a package dedicated to data cleaning and checking, called the cleaner package. It freq() function can be used to create frequency tables.

For example, for the gender variable:

data %>% freq(gender)

Frequency table

Class: character
Length: 20,000
Available: 20,000 (100.0%, NA: 0 = 0.0%)
Unique: 2

Shortest: 1
Longest: 1

Item Count Percent Cum. Count Cum. Percent
1 M 10,342 51.71% 10,342 51.71%
2 F 9,658 48.29% 20,000 100.00%

So, we can draw at least two conclusions immediately. From a data scientists perspective, the data looks clean: only values M and F. From a researchers perspective: there are slightly more men. Nothing we didn’t already know.

The data is already quite clean, but we still need to transform some variables. The bacteria column now consists of text, and we want to add more variables based on microbial IDs later on. So, we will transform this column to valid IDs. The mutate() function of the dplyr package makes this really easy:

data <- data %>%
  mutate(bacteria =

We also want to transform the antibiotics, because in real life data we don’t know if they are really clean. The as.rsi() function ensures reliability and reproducibility in these kind of variables. The is.rsi.eligible() can check which columns are probably columns with R/SI test results. Using mutate() and across(), we can apply the transformation to the formal <rsi> class:

# [1] "AMX" "AMC" "CIP" "GEN"

data <- data %>%
  mutate(across(where(is.rsi.eligible), as.rsi))

Finally, we will apply EUCAST rules on our antimicrobial results. In Europe, most medical microbiological laboratories already apply these rules. Our package features their latest insights on intrinsic resistance and exceptional phenotypes. Moreover, the eucast_rules() function can also apply additional rules, like forcing ampicillin = R when amoxicillin/clavulanic acid = R.

Because the amoxicillin (column AMX) and amoxicillin/clavulanic acid (column AMC) in our data were generated randomly, some rows will undoubtedly contain AMX = S and AMC = R, which is technically impossible. The eucast_rules() fixes this:

data <- eucast_rules(data, col_mo = "bacteria", rules = "all")

Adding new variables

Now that we have the microbial ID, we can add some taxonomic properties:

data <- data %>% 
  mutate(gramstain = mo_gramstain(bacteria),
         genus = mo_genus(bacteria),
         species = mo_species(bacteria))

First isolates

We also need to know which isolates we can actually use for analysis.

To conduct an analysis of antimicrobial resistance, you must only include the first isolate of every patient per episode (Hindler et al., Clin Infect Dis. 2007). If you would not do this, you could easily get an overestimate or underestimate of the resistance of an antibiotic. Imagine that a patient was admitted with an MRSA and that it was found in 5 different blood cultures the following weeks (yes, some countries like the Netherlands have these blood drawing policies). The resistance percentage of oxacillin of all isolates would be overestimated, because you included this MRSA more than once. It would clearly be selection bias.

The Clinical and Laboratory Standards Institute (CLSI) appoints this as follows:

(…) When preparing a cumulative antibiogram to guide clinical decisions about empirical antimicrobial therapy of initial infections, only the first isolate of a given species per patient, per analysis period (eg, one year) should be included, irrespective of body site, antimicrobial susceptibility profile, or other phenotypical characteristics (eg, biotype). The first isolate is easily identified, and cumulative antimicrobial susceptibility test data prepared using the first isolate are generally comparable to cumulative antimicrobial susceptibility test data calculated by other methods, providing duplicate isolates are excluded.
M39-A4 Analysis and Presentation of Cumulative Antimicrobial Susceptibility Test Data, 4th Edition. CLSI, 2014. Chapter 6.4

This AMR package includes this methodology with the first_isolate() function and is able to apply the four different methods as defined by Hindler et al. in 2007: phenotype-based, episode-based, patient-based, isolate-based. The right method depends on your goals and analysis, but the default phenotype-based method is in any case the method to properly correct for most duplicate isolates. This method also takes into account the antimicrobial susceptibility test results using all_microbials(). Read more about the methods on the first_isolate() page.

The outcome of the function can easily be added to our data:

data <- data %>% 
  mutate(first = first_isolate(info = TRUE))
# Determining first isolates using the 'phenotype-based' method and an
# episode length of 365 days
# ℹ Using column 'bacteria' as input for `col_mo`.
# ℹ Using column 'date' as input for `col_date`.
# ℹ Using column 'patient_id' as input for `col_patient_id`.
# Basing inclusion on all antimicrobial results, using a points threshold of
# 2
# => Found 10,622 first weighted isolates (phenotype-based, 53.1% of total
#    where a microbial ID was available)

So only 53.1% is suitable for resistance analysis! We can now filter on it with the filter() function, also from the dplyr package:

data_1st <- data %>% 
  filter(first == TRUE)

For future use, the above two syntaxes can be shortened:

data_1st <- data %>% 

So we end up with 10,622 isolates for analysis. Now our data looks like:

date patient_id hospital bacteria AMX AMC CIP GEN gender gramstain genus species first
2 2012-10-19 U5 Hospital B B_STRPT_PNMN S S S R F Gram-positive Streptococcus pneumoniae TRUE
3 2010-02-13 W10 Hospital A B_ESCHR_COLI R R S R F Gram-negative Escherichia coli TRUE
4 2017-03-20 Y1 Hospital B B_STPHY_AURS R S S S F Gram-positive Staphylococcus aureus TRUE
5 2016-08-03 U6 Hospital B B_STPHY_AURS R S S S F Gram-positive Staphylococcus aureus TRUE
6 2015-06-16 X5 Hospital B B_STPHY_AURS R R R S F Gram-positive Staphylococcus aureus TRUE
7 2010-03-21 Y9 Hospital B B_STRPT_PNMN R R R R F Gram-positive Streptococcus pneumoniae TRUE

Time for the analysis!

Analysing the data

You might want to start by getting an idea of how the data is distributed. It’s an important start, because it also decides how you will continue your analysis. Although this package contains a convenient function to make frequency tables, exploratory data analysis (EDA) is not the primary scope of this package. Use a package like DataExplorer for that, or read the free online book Exploratory Data Analysis with R by Roger D. Peng.

Dispersion of species

To just get an idea how the species are distributed, create a frequency table with our freq() function. We created the genus and species column earlier based on the microbial ID. With paste(), we can concatenate them together.

The freq() function can be used like the base R language was intended:

freq(paste(data_1st$genus, data_1st$species))

Or can be used like the dplyr way, which is easier readable:

data_1st %>% freq(genus, species)

Frequency table

Class: character
Length: 10,622
Available: 10,622 (100.0%, NA: 0 = 0.0%)
Unique: 4

Shortest: 16
Longest: 24

Item Count Percent Cum. Count Cum. Percent
1 Escherichia coli 4,644 43.72% 4,644 43.72%
2 Staphylococcus aureus 2,644 24.89% 7,288 68.61%
3 Streptococcus pneumoniae 2,102 19.79% 9,390 88.40%
4 Klebsiella pneumoniae 1,232 11.60% 10,622 100.00%

Overview of different bug/drug combinations

Using Tidyverse selections, you can also select or filter columns based on the antibiotic class they are in:

data_1st %>% 
  filter(any(aminoglycosides() == "R"))
# ℹ For `aminoglycosides()` using column: 'GEN' (gentamicin)
date patient_id hospital bacteria AMX AMC CIP GEN gender gramstain genus species first
2012-10-19 U5 Hospital B B_STRPT_PNMN S S S R F Gram-positive Streptococcus pneumoniae TRUE
2010-02-13 W10 Hospital A B_ESCHR_COLI R R S R F Gram-negative Escherichia coli TRUE
2010-03-21 Y9 Hospital B B_STRPT_PNMN R R R R F Gram-positive Streptococcus pneumoniae TRUE
2012-11-04 X6 Hospital B B_ESCHR_COLI R S R R F Gram-negative Escherichia coli TRUE
2015-07-07 N5 Hospital C B_STRPT_PNMN S S R R M Gram-positive Streptococcus pneumoniae TRUE
2017-10-23 I1 Hospital B B_ESCHR_COLI S S S R M Gram-negative Escherichia coli TRUE

If you want to get a quick glance of the number of isolates in different bug/drug combinations, you can use the bug_drug_combinations() function:

data_1st %>% 
  bug_drug_combinations() %>% 
  head() # show first 6 rows
# ℹ Using column 'bacteria' as input for `col_mo`.
mo ab S I R total
E. coli AMX 2215 147 2282 4644
E. coli AMC 3410 185 1049 4644
E. coli CIP 3380 0 1264 4644
E. coli GEN 4031 0 613 4644
K. pneumoniae AMX 0 0 1232 1232
K. pneumoniae AMC 974 51 207 1232
data_1st %>% 
  select(bacteria, aminoglycosides()) %>% 
# ℹ For `aminoglycosides()` using column: 'GEN' (gentamicin)
# ℹ Using column 'bacteria' as input for `col_mo`.
mo ab S I R total
E. coli GEN 4031 0 613 4644
K. pneumoniae GEN 1103 0 129 1232
S. aureus GEN 2348 0 296 2644
S. pneumoniae GEN 0 0 2102 2102

This will only give you the crude numbers in the data. To calculate antimicrobial resistance in a more sensible way, also by correcting for too few results, we use the resistance() and susceptibility() functions.

Resistance percentages

The functions resistance() and susceptibility() can be used to calculate antimicrobial resistance or susceptibility. For more specific analyses, the functions proportion_S(), proportion_SI(), proportion_I(), proportion_IR() and proportion_R() can be used to determine the proportion of a specific antimicrobial outcome.

All these functions contain a minimum argument, denoting the minimum required number of test results for returning a value. These functions will otherwise return NA. The default is minimum = 30, following the CLSI M39-A4 guideline for applying microbial epidemiology.

As per the EUCAST guideline of 2019, we calculate resistance as the proportion of R (proportion_R(), equal to resistance()) and susceptibility as the proportion of S and I (proportion_SI(), equal to susceptibility()). These functions can be used on their own:

data_1st %>% resistance(AMX)
# [1] 0.540482

Or can be used in conjunction with group_by() and summarise(), both from the dplyr package:

data_1st %>% 
  group_by(hospital) %>% 
  summarise(amoxicillin = resistance(AMX))
hospital amoxicillin
Hospital A 0.5272898
Hospital B 0.5525126
Hospital C 0.5327044
Hospital D 0.5448872

Of course it would be very convenient to know the number of isolates responsible for the percentages. For that purpose the n_rsi() can be used, which works exactly like n_distinct() from the dplyr package. It counts all isolates available for every group (i.e. values S, I or R):

data_1st %>% 
  group_by(hospital) %>% 
  summarise(amoxicillin = resistance(AMX),
            available = n_rsi(AMX))
hospital amoxicillin available
Hospital A 0.5272898 3188
Hospital B 0.5525126 3761
Hospital C 0.5327044 1590
Hospital D 0.5448872 2083

These functions can also be used to get the proportion of multiple antibiotics, to calculate empiric susceptibility of combination therapies very easily:

data_1st %>% 
  group_by(genus) %>% 
  summarise(amoxiclav = susceptibility(AMC),
            gentamicin = susceptibility(GEN),
            amoxiclav_genta = susceptibility(AMC, GEN))
genus amoxiclav gentamicin amoxiclav_genta
Escherichia 0.7741171 0.8680017 0.9760982
Klebsiella 0.8319805 0.8952922 0.9829545
Staphylococcus 0.8033283 0.8880484 0.9799546
Streptococcus 0.5413892 0.0000000 0.5413892

To make a transition to the next part, let’s see how this difference could be plotted:

data_1st %>% 
  group_by(genus) %>% 
  summarise("1. Amoxi/clav" = susceptibility(AMC),
            "2. Gentamicin" = susceptibility(GEN),
            "3. Amoxi/clav + genta" = susceptibility(AMC, GEN)) %>% 
  # pivot_longer() from the tidyr package "lengthens" data:
  tidyr::pivot_longer(-genus, names_to = "antibiotic") %>% 
  ggplot(aes(x = genus,
             y = value,
             fill = antibiotic)) +
  geom_col(position = "dodge2")


To show results in plots, most R users would nowadays use the ggplot2 package. This package lets you create plots in layers. You can read more about it on their website. A quick example would look like these syntaxes:

ggplot(data = a_data_set,
       mapping = aes(x = year,
                     y = value)) +
  geom_col() +
  labs(title = "A title",
       subtitle = "A subtitle",
       x = "My X axis",
       y = "My Y axis")

# or as short as:
ggplot(a_data_set) +

The AMR package contains functions to extend this ggplot2 package, for example geom_rsi(). It automatically transforms data with count_df() or proportion_df() and show results in stacked bars. Its simplest and shortest example:

ggplot(data_1st) +
  geom_rsi(translate_ab = FALSE)

Omit the translate_ab = FALSE to have the antibiotic codes (AMX, AMC, CIP, GEN) translated to official WHO names (amoxicillin, amoxicillin/clavulanic acid, ciprofloxacin, gentamicin).

If we group on e.g. the genus column and add some additional functions from our package, we can create this:

# group the data on `genus`
ggplot(data_1st %>% group_by(genus)) + 
  # create bars with genus on x axis
  # it looks for variables with class `rsi`,
  # of which we have 4 (earlier created with `as.rsi`)
  geom_rsi(x = "genus") + 
  # split plots on antibiotic
  facet_rsi(facet = "antibiotic") +
  # set colours to the R/SI interpretations (colour-blind friendly)
  scale_rsi_colours() +
  # show percentages on y axis
  scale_y_percent(breaks = 0:4 * 25) +
  # turn 90 degrees, to make it bars instead of columns
  coord_flip() +
  # add labels
  labs(title = "Resistance per genus and antibiotic", 
       subtitle = "(this is fake data)") +
  # and print genus in italic to follow our convention
  # (is now y axis because we turned the plot)
  theme(axis.text.y = element_text(face = "italic"))

To simplify this, we also created the ggplot_rsi() function, which combines almost all above functions:

data_1st %>% 
  group_by(genus) %>%
  ggplot_rsi(x = "genus",
             facet = "antibiotic",
             breaks = 0:4 * 25,
             datalabels = FALSE) +

Plotting MIC and disk diffusion values

The AMR package also extends the plot() and ggplot() functions for plotting minimum inhibitory concentrations (MIC, created with as.mic()) and disk diffusion diameters (created with as.disk()).

With the random_mic() and random_disk() functions, we can generate sampled values for the new data types (S3 classes) <mic> and <disk>:

mic_values <- random_mic(size = 100)
# Class <mic>
#   [1] 0.25     8        0.25     2        4        2        1        0.5     
#   [9] 8        16       0.25     >=128    8        2        <=0.0625 64      
#  [17] <=0.0625 64       32       32       2        1        64       64      
#  [25] 0.125    0.25     2        <=0.0625 32       16       8        2       
#  [33] <=0.0625 4        1        4        0.25     <=0.0625 32       >=128   
#  [41] 0.125    >=128    0.25     0.25     <=0.0625 64       2        16      
#  [49] 16       0.5      0.5      8        >=128    64       >=128    0.125   
#  [57] 64       4        0.5      1        8        16       4        <=0.0625
#  [65] 64       <=0.0625 4        2        64       16       32       16      
#  [73] 2        0.5      4        0.25     1        2        0.25     0.5     
#  [81] 0.25     2        4        1        0.125    0.25     8        0.125   
#  [89] 16       4        <=0.0625 4        1        1        <=0.0625 0.125   
#  [97] >=128    >=128    <=0.0625 1
# base R:

# ggplot2:

But we could also be more specific, by generating MICs that are likely to be found in E. coli for ciprofloxacin:

mic_values <- random_mic(size = 100, mo = "E. coli", ab = "cipro")

For the plot() and ggplot() function, we can define the microorganism and an antimicrobial agent the same way. This will add the interpretation of those values according to a chosen guidelines (defaults to the latest EUCAST guideline).

Default colours are colour-blind friendly, while maintaining the convention that e.g. ‘susceptible’ should be green and ‘resistant’ should be red:

# base R:
plot(mic_values, mo = "E. coli", ab = "cipro")

# ggplot2:
ggplot(mic_values, mo = "E. coli", ab = "cipro")

For disk diffusion values, there is not much of a difference in plotting:

disk_values <- random_disk(size = 100, mo = "E. coli", ab = "cipro")
# ℹ Translation is uncertain of one microorganism. Use `mo_uncertainties()`
#   to review it.
# Class <disk>
#   [1] 23 17 25 30 29 25 17 26 27 21 20 25 17 29 20 18 20 17 18 22 27 31 30 17 19
#  [26] 28 26 20 20 18 30 30 28 24 25 19 27 31 31 24 29 20 28 20 29 22 29 22 22 26
#  [51] 31 28 21 31 20 24 27 24 19 25 31 29 18 17 29 31 20 17 23 17 25 18 27 28 31
#  [76] 30 24 26 29 25 19 30 25 27 27 23 21 28 29 26 31 31 27 27 17 28 24 29 18 17
# base R:
plot(disk_values, mo = "E. coli", ab = "cipro")

And when using the ggplot2 package, but now choosing the latest implemented CLSI guideline (notice that the EUCAST-specific term “Incr. exposure” has changed to “Intermediate”):

       mo = "E. coli",
       ab = "cipro",
       guideline = "CLSI")

Independence test

The next example uses the example_isolates data set. This is a data set included with this package and contains 2,000 microbial isolates with their full antibiograms. It reflects reality and can be used to practice AMR data analysis.

We will compare the resistance to fosfomycin (column FOS) in hospital A and D. The input for the fisher.test() can be retrieved with a transformation like this:

# use package 'tidyr' to pivot data:

check_FOS <- example_isolates %>%
  filter(hospital_id %in% c("A", "D")) %>% # filter on only hospitals A and D
  select(hospital_id, FOS) %>%             # select the hospitals and fosfomycin
  group_by(hospital_id) %>%                # group on the hospitals
  count_df(combine_SI = TRUE) %>%          # count all isolates per group (hospital_id)
  pivot_wider(names_from = hospital_id,    # transform output so A and D are columns
              values_from = value) %>%     
  select(A, D) %>%                         # and only select these columns
  as.matrix()                              # transform to a good old matrix for fisher.test()

#       A  D
# [1,] 25 77
# [2,] 24 33

We can apply the test now with:

# do Fisher's Exact Test
#   Fisher's Exact Test for Count Data
# data:  check_FOS
# p-value = 0.03104
# alternative hypothesis: true odds ratio is not equal to 1
# 95 percent confidence interval:
#  0.2111489 0.9485124
# sample estimates:
# odds ratio 
#  0.4488318

As can be seen, the p value is 0.031, which means that the fosfomycin resistance found in isolates from patients in hospital A and D are really different.