Summarising and presenting AMR data

Introduction

This module introduces common ways of summarising data visually, reviews the strengths and limitations of each approach, and discusses how you can use visual summaries to enhance the analysis of AMR-related data. The module explores the effectiveness of visual summaries in communicating important information to a wide range of people. You will have the opportunity to reflect on the most appropriate ways to visualise AMR-related data, to look for patterns and trends, and identify errors in this data.

This module is part of a series related to AMR data. It is recommended that you first complete the modules Fundamentals of data for AMR and Processing and analysing AMR data to understand AMR data and data analysis principles.

By the end of this module, you should be able to:

• describe the different ways to represent data visually
• explain why visual summaries of data are an essential part of AMR data analysis
• review the strengths and limitations of each visual presentation
• make a simple graph and map using AMR data.

1 Recap: AMR data and analysis

In the previous modules Fundamentals of data for AMR and Processing and analysing AMR data, you learnt how AMR data is transformed into information and the methods used to analyse data to answer important questions about AMR. This section briefly reviews key concepts from these modules to help you critically interpret visual data summaries and communicate key findings to different audiences.

1.1 Review of data types

Table 1 below reminds you of the definitions of data types previously described in the module Fundamentals of data for AMR.

You need to first understand the types of data you have (i.e. discrete, continuous, nominal or ordinal) before preparing to present it in visual formats. This is because the data type will determine the formats you can use to summarise and display findings.

To understand your dataset, first begin by examining individual records and summarising the data in tables. You may find that summary tables are sufficient for presenting the data, especially if the dataset is small. If the data are more complex, then graphs or maps can help highlight important findings, trends or errors that need to be corrected (such as data entry errors).

1.2 Review of different approaches to data analysis

Data analysis is about identifying and summarising meaningful patterns in data. In the module Processing and analysing AMR data, you were introduced to the basics of descriptive statistics and inferential statistics. It is recommended that you revisit the module to understand statistical concepts.

Here, we will briefly recap data analysis concepts relevant to visualising data.

1.2.1 Basic descriptive statistics

Descriptive analysis is the process of summarising data and generally involves characterising and visualising your dataset. Depending on the nature of the data, you may only need to undertake a descriptive analysis before reporting your findings. The approach to descriptive analysis varies depending on whether the data is categorical or numerical.

1.2.1.1 Categorical variables

Descriptive statistics for categorical variables include counts (frequencies) and proportions (relative frequencies). Percentages are obtained by multiplying proportions by 100.

AMR data is often presented as basic descriptive statistics in sentences in a report, in a frequency table, or in a visual format using a pie chart or a bar chart. An example of using text is in reporting the percentage of isolates that were resistant:

‘A large number of MRSA isolates showed resistance to levofloxacin (83.9%), ciprofloxacin (83%), erythromycin (77.7%), and clindamycin (72.3%).’

(Kot et al., 2020)

Activity 3: Approaches to presenting information

Timing: Allow about 10 minutes

Figure 1 and Table 3 displays data from the Third Australian Report on Antimicrobial Use and Resistance in Human Health (AURA), 2019. Table 3 is a frequency table showing the percentage of prescribed antibiotics, and Figure 1 shows the same data in a bar chart.

Table 3 The ten most commonly dispensed antibiotics under the Pharmaceutical Benefits Scheme/Repatriation Pharmaceutical Benefits Scheme, by percentage of all antibiotic prescriptions, 2017 (ACSQHC, 2019)

Antimicrobial class	Percentage (%)
Cefalexin	20.1
Amoxicillin	19.7
Amoxicillin-clavulanic acid	17.5
Doxycycline	8.0
Roxithromycin	5.1
Trimethoprim	3.4
Flucloxacillin	2.9
Clarithromycin	2.9
Metronidazole	2.5
Erythromycin	2.1
Total	100

Figure 1 The ten most commonly dispensed antibiotics under the Pharmaceutical Benefits Scheme/Repatriation Pharmaceutical Benefits Scheme, by percentage of all antibiotic prescriptions, 2017 (ACSQHC, 2019)

Show description|Hide description

An ordered list of the ten most commonly disposed antibiotics under the Pharmaceutical Benefits Scheme, by percentage, shown as a bar chart (RHS, B).

Figure 1 The ten most commonly dispensed antibiotics under the Pharmaceutical Benefits Scheme/Repatriation Pharmaceutical Benefits ...

Which type of data display (Table 3 or Figure 1) do you think most effectively conveys information about the most and least commonly dispensed antibiotics in Australia in 2017?

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Answer

While tables can summarise and display data to great effect, in this example, the bar chart (Figure 1) demonstrates a superior approach to presenting the data. When looking at the bar chart, it can be rapidly determined that there is a pronounced difference between the most commonly prescribed and least prescribed antimicrobials.

Do you agree that your brain needs to work harder to get the same information from the table?

1.2.2 Plotting variability in data

Data obtained from a sample should be representative of the population of interest. Since it is rarely possible to include the entire population in a sample, there is always some degree of uncertainty or variability in how well the data represent the population (see the Sampling modules and Processing and analysing AMR data module for more information). It is good practice to indicate uncertainty in data either in text, tables or graphs. For graphs, uncertainty in the data is represented by error bars.

Error bars can be used to display the standard deviation, standard error, confidence intervals (e.g. 95% interval), or minimum and maximum values in a dataset. (To revise these terms, revisit the module Processing and analysing AMR data.) When using error bars on graphs, always state the measure used to represent uncertainty in the data as the measures are not the same.

Error bars are markers drawn over data points on a graph and either extend from the centre of a data point or the top edge of a column (such as a bar chart). You can get a sense of how precise the measurements are (which reflects the level of uncertainty in the data collected) by looking at the error bar's length: short error bars indicate less variability in the data, and in contrast, long error bars indicate more variability, that is, the values are more spread out.

You can also use error bars to gauge if there might be a significant difference between groups. Overlapping error bars may suggest there is not a significance difference between groups, while no overlap between error bars may suggest there may be a significance difference. You cannot conclude statistical significance by looking at the graphs alone, so always perform a statistical test when drawing conclusions. (For a recap on methods of testing statistical significance, revisit the module Processing and analysing AMR data.)

You often see error bars plotted on line graphs and bar charts, such as the one in Figure 2.

Activity 4: Plotting variability in data

Timing: Allow about 10 minutes

Figure 2 Time trend of vancomycin-resistant Enterococcus faecium from clinical isolates in German Hospitals. Time trend of vancomycin-resistant E. faecium as a proportion (%) of all E. faecium isolates with corresponding 95% confidence intervals (Markwart et al., 2019)

Show description|Hide description

Bar chart of the change in the proportion of vancomycin-resistant isolates of E. faecium in German hospitals from 2012 to 2017, with 95% confidence intervals shown as error bars.

Figure 2 Time trend of vancomycin-resistant Enterococcus faecium from clinical isolates in German Hospitals. Time trend of ...

Figure 2 shows confidence intervals for each year that the proportion of vancomycin-resistant Enterococcus faecium isolates were reported. What can you say about the confidence intervals?

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Answer

The confidence intervals appear to be wider in some years than others. For example, 2012, 2015 and 2016 appear to have wider confidence intervals than 2014. This may indicate that the data collected on clinical E. faecium in 2012, 2015 and 2016 was more variable than in 2014. Also, the confidence intervals for some years overlap, while for other years they do not. This may indicate a statistically significant increase in the proportion of vancomycin-resistant E. faecium between some years where the confidence interval bars do not overlap. This is most apparent between the years 2014 and 2017.

2 Summarising AMR data visually

Research has shown that our brains understand complex information more easily when it is presented in a visual format (Ware, 2012). That is why data are often best communicated visually, using graphs and maps rather than tables or spreadsheets. Visual summaries can be a great way of communicating your findings, particularly to people who may be unfamiliar with the data.

When choosing the most appropriate visual display, ask yourself the following:

• What is the primary aim of my visual analysis? For example, do I want to demonstrate a difference in prevalence of resistant pathogens by geographic area or between two groups? Do I want to highlight a change in antimicrobial use trends over time?
• What do I want to communicate, and with whom am I communicating? Graphs and maps can help you share public health information to people who may not be familiar with the topic.
• What type of data do I have? What is the data’s measurement scale (nominal, ordinal, discrete or continuous)? The data type will influence the kind of visual representation you can use.

Table 4 summarises the different analysis aims and gives examples of the best visual displays for presenting the analysis, described in more detail as we work through this module.

2.1 Graphs of ‘AMR data’

Graphs display data in a visual form to tell a story. They can be used to display patterns, trends, comparisons and differences in data that may not be otherwise evident. The goal of graphs should be to help people to understand patterns and relationships in data quickly.

2.1.1 Designing good graphs

Graphs are only as good as they are clear and understandable. A checklist of best practice to ensure a graph is conveying your message most effectively is given in Table 5 below:

In the following sections, you will learn about common graph types: histograms, boxplots, scatter plots, bar charts and line graphs. Not all of these graphs will be appropriate for the data and message that you want to convey. Understanding the strengths and limitations of each graph will help guide your decision on which graph to choose.

2.1.2 Histograms

Histograms show the distribution of values for a quantitative (continuous) variable. Histograms are useful when there are many observations and you want to understand the overall shape and spread of your data.

The x-axis is marked in the units of measurement for the independent variable (e.g. age, time, MIC, zone diameter). The y-axis is the scale that shows you the number of times (frequency, proportion) the value in an interval occurred.

To create a histogram, you need to first group data from the independent variable into class intervals (bins) of equal width and then count the values in each interval (class frequency). Class frequencies are represented by bars on a histogram. The height of each bar corresponds to its class frequency. An example of how to construct a histogram from raw data is shown below:

Example

In this example, a series of observations have been recorded in a variable called ‘Age’. To construct the histogram, ‘Age’ is split into five class intervals. Each interval contains the count of occurrences.

Table 6 Variable name: Age

39	41	22	38	46	55	65	78	83	18
28	54	53	61	10	16	29	58	55	66

Table 7 Frequency of the variable: Age

Class interval	Class frequency	Observations
0–20	3	10, 16, 18
20.1–40	5	22, 28, 29, 38, 39
40.1–60	7	41, 46, 53, 54, 55, 55, 58
60.1–80	4	61, 65, 66, 78
80.1–100	1	83

The histogram is then constructed based on the number of class intervals which are plotted on the x-axis with the y-axis showing the frequency (number) of occurrences in each class interval.

Figure 3 Example histogram of the age groups of people

Show description|Hide description

A histogram showing the number of people in a sample in each of five equal age groups (from 0–20 to 80.1–100).

Figure 3 Example histogram of the age groups of people

Traditionally, a histogram is drawn with no space between classes to indicate all values of the variable are represented. (This is different from a bar chart, which has space between the classes.) However, sometimes histograms may be drawn with spaces between the classes for greater visual impact. You can see this in Figure 4.

Maximise

Figure 4 Histogram showing the distribution of zone sizes obtained for oxolinic acid against Aeromonas salmoncida (n=323) (grey bars) and the breakpoints currently being used (black bars) (Smith, 2008)

Show description|Hide description

Histogram showing a distribution of zone sizes obtained from testing oxolinic acid against A.salmoncida (grey bars, showing a bimodal distribution) with the breakpoints currently used superimposed as eight black bars. This is a typical distribution for AMR data.

Figure 4 Histogram showing the distribution of zone sizes obtained for oxolinic acid against Aeromonas salmoncida (n=323) (grey bars) and...

The purpose of the histogram (or, indeed, of any graph) is to help understand the data. When viewing a histogram, look for important features, including the shape and spread of the data and whether there are any deviations (outliers). Outliers are data points that lie a long way from the general pattern in the data.

A histogram can have different shapes: it can be unimodal (a single peak representing the interval with the most values e.g. a normal distribution), bimodal (two peaks) or multimodal (more than two peaks). A histogram can also be symmetrical (when the right and left sides of the midpoint are similar) or skewed, where the intervals are grouped to the right (positively skewed) or left side (negatively skewed).

AMR data often has a bimodal shape because there are often two separate populations of isolates – those that are susceptible and those resistant. Also, AMR data is often skewed. Depending on the population of interest, there might be more isolates that are susceptible than resistant, or vice versa. For example, in secondary care, there may be more isolates tested that are resistant to an antimicrobial agent than isolates that are susceptible because the people who are sampled are more likely to have been treated with antimicrobials and suffered treatment failure. Figure 4 demonstrates a bimodal distribution of zone diameter measurements obtained by testing the susceptibility of Aeromonas salmoncida to oxolinic acid. (Note, the purpose of including Figure 4 in this module is to demonstrate a typical distribution of AMR data. Therefore, it is not necessary to interpret the breakpoints depicted on the graph.)

The strengths and limitations of histograms are listed below:

2.1.3 Box-and-whisker plot

Another way of displaying information about the spread of data is the box-and-whisker plot (also referred to as the boxplot). Boxplots are graphs that summarise the five-number summary of continuous or discrete data. They consist of:

• a central box that spans the quartiles Q1 (lower quartile, 25th percentile) and Q3 (upper quartile, 75th percentile) – (note that the range from Q1 to Q3 is known as the interquartile range (IQR))
• a line in the box that marks the median (50th percentile)
• lines (whiskers) that extend from the box out to either the smallest (minimum) and largest (maximum) observations, excluding outliers, or 1.5 times the interquartile range on each side, also excluding outliers (if it is ambiguous which method is used, it is normally mentioned in the Figure’s legend)
• outliers, which are data values that are far away from other data values. On a boxplot, outliers are identified by a symbol such as a dot or an asterisk.

Boxplots are most beneficial when used for a side-by-side comparison of more than one distribution.

Activity 5: Understanding box-and-whisker plots

Timing: Allow about 15 minutes

Look at the box plot in Figure 5. Can you annotate one of the boxplots to show the five-number summary for a distribution?

Figure 5 Biofilm formation by Stenotrophomonas maltophilia isolated from patient samples according to patient and sample types. Biofilm biomass, assessed by spectrophotometric assay after crystal violet assay, was stratified according to patients with or without cystic fibrosis (CF, non-CF) and (B) sample type. (Significance level from Mann-Whitney test: * p

Show description|Hide description

Box-and-whisker plots showing the biomass of biofilm formed by S. maltophilia isolated from samples from patients with and without cystic fibrosis (CF). The left-hand plot (A) shows two plots obtained from CF and non-CF patients; the right-hand plot (B) shows three plots obtained from CF patients, respiratory samples from non-CF patients and blood samples.

Figure 5 Biofilm formation by Stenotrophomonas maltophilia isolated from patient samples according to patient and sample types. Biofilm ...

Answer

Figure 6 Annotated Figure 5

Show description|Hide description

Box-and-whisker plots showing the biomass of biofilm formed by S. maltophilia isolated from samples from patients with and without cystic fibrosis (CF). The left-hand plot (A) shows two plots obtained from CF and non-CF patients; the right-hand plot (B) shows three plots obtained from CF patients, respiratory samples from non-CF patients and blood samples. Labels have been added to the left-hand plot (A) to show where the Maximum, Median, Minimum, Q1 and Q3 are.

Figure 6 Annotated Figure 5

The box extends from the first (Q1) and third (Q3) quartiles. The line in the middle of the box is plotted at the median, while the ends of the whiskers represent the minima and the maxima of all the data. The lines extending from the interquartile range are called whiskers. The whiskers extend to the maximum and minimum values in the dataset, excluding outliers. Note, the asterisks in Figure 5 and Figure 6 are not outliers, rather they are indicating the statistical significance of the results.

The strengths and limitations of boxplots are listed below:

2.1.4 Scatter plots

Scatter plots are used when we are interested in the relationship between two different variables. Each point on the graph represents the values of a pair of variables. The value of one variable is plotted on the x-axis, and the value of the other variable is plotted on the y-axis. The variables generally have to be numeric and are commonly continuous, although they may also be discrete.

A scatter plot gives us a good idea of the correlation between the two variables and the nature of that correlation. You can plot several combinations to explore correlations and then investigate these further with more complex statistical analyses. As well as making it easy to identify any general pattern in the relationship between the variables, a scatter plot can help to identify outliers.

The scatter plot in Figure 7 shows the relationship between country-level antimicrobial consumption (x-axis) and resistance to aminopenicillins (y-axis) in European countries in 2015.

Figure 7 Relationship between penicillin-resistant S. pneumonia and total antimicrobial consumption by country in Europe, 2015. Note that DDD/1000 inhabitants per day is a technical unit of measure of antimicrobial consumption in non-hospital settings. The assumed average maintenance dose per day for a drug used for its main indication in adults

Show description|Hide description

Scatter plot showing the relationship between percentage aminopenicillin resistance (plotted on the y-axis) and antimicrobial consumption in non-hospital settings (x-axis) for every country in Europe in 2015. The distribution of dots shows a low positive correlation between the prevalence of resistance and antibiotic use.

Figure 7 Relationship between penicillin-resistant S. pneumonia and total antimicrobial consumption by country in Europe, 2015. Note that...

The relationship between two numeric variables is called correlation. There are three types of correlation (the strength and significance of which should be tested using formal statistical tests):

Positive correlation: as the values of one variable increases, so do the second variable's values. In Figure 7, as antimicrobial consumption increases at the national level, so does resistance to aminopenicillins. Note that this correlation is weak – although the trend is there, there are plenty of countries where it is not the case.

Negative correlation: as the values of one variable increases, the values of the other variable decreases. For example, as antimicrobial consumption increases, antimicrobial susceptibility (the opposite of resistance) decreases.

No correlation: there is no apparent relationship between the variables. For example, some studies have found that patient age is not correlated with AMR for specific antimicrobial classes and/or bacterial species.

The strengths and limitations of scatter plots are listed below:

In summary, scatter plots should be used when there are many different data points and you want to highlight similarities in the dataset. This is useful when looking for outliers or for understanding the distribution of your data.

2.1.5 Bar charts

Bar charts show relationships between different data series that are independent of each other. The most common data displayed in bar charts includes nominal, ordinal, discrete data or variables that can be treated as discrete.

The difference between histograms and bar charts is that histograms show the frequency of a continuous variable, whereas bar charts show the frequency of categorical or discrete data.

There are many bar charts: column (vertical), horizontal (left to right), grouped and stacked.

2.1.5.1 Horizontal and column bar charts

Horizontal and column bar charts are usually interchangeable. However, there are some reasons to choose one over the other. Horizontal charts work better when the labels are long and for data that needs to show ranking from largest to smallest. An example of a horizontal graph can be seen in Figure 8.

Figure 8 Percentage of specimen type sent for bacterial culture, Health Facility X, Quarter 3, 2015

Show description|Hide description

Horizontal bar chart (essentially, a typical bar chart rotated by 90^o) showing the percentage of patients with different types of specimens sent for bacterial culture.

Figure 8 Percentage of specimen type sent for bacterial culture, Health Facility X, Quarter 3, 2015

Column charts work better for grouped data and time-series data. Simple column charts compare categories of one variable, while grouped (or clustered) column charts compare categories of two or more variables as shown in Figure 9. The bars within a group adjoin, but there is a space between groups. It is important that the bars are distinctive and described in the legend.

Figure 9 Percentage of female and male patients infected by type of bacteria, Health Facility X, Quarter 3, 2015

Show description|Hide description

Grouped (or clustered) conventional bar chart showing the percentages of female patients (LH bar of each pair, blue) and male patients (RH bar, green) infected with different bacteria.

Figure 9 Percentage of female and male patients infected by type of bacteria, Health Facility X, Quarter 3, 2015

Activity 6: Interpreting bar charts

Timing: Allow about 20 minutes

Review Figure 10 and describe what you see, what you think are the key findings and how might you improve the graph.

Figure 10 Number of countries reporting to GLASS by economic status in 2017–2018 (WHO, 2019)

Show description|Hide description

Grouped bar chart showing the numbers of countries reporting both information on surveillance systems (LH bar of each pair, blue) and AMR rates (RH bar, green) to GLASS by both year (2017, 2018) and economic status.

Figure 10 Number of countries reporting to GLASS by economic status in 2017–2018 (WHO, 2019)

To use this interactive functionality a free OU account is required. Sign in or register.

Interactive feature not available in single page view (see it in standard view).

Answer

In Figure 10, groups of countries are categorised by economic status (low, low-middle, upper-middle, high – this wasn’t made clear in the caption) and year (2017, 2018) on the x-axis, and the number of countries in each category participating in GLASS on the y-axis. Within each group, there are two bars. One bar represents the frequency of countries that report information on their surveillance system, and the other bar represents the frequency of countries that report information on AMR rates in bacteria of interest to GLASS.

The graph shows that, overall, more high-income countries participate in GLASS compared to lower-income countries. Reporting to GLASS increased from 2017 to 2018 for all types of countries. When looking at the bar lengths, it appears there is an approximate doubling in reporting on surveillance systems from 2017 to 2018 for most categories. While the number of countries reporting AMR rates also increased, the increase was not as large when compared to countries reporting on their surveillance systems, other than in high-income countries.

The graph could be improved by reporting the exact frequencies in each bar so that the actual change in reporting rates in each category can be determined. The graph could also be improved by defining the country group acronyms (LIC, LMC, UMC, HIC). Another way to present the data is demonstrate proportional representation in each economic category which can be done by plotting percent on the y-axis rather than the count.

2.1.5.2 Stacked bar chart

In a stacked bar chart, each bar represents a total amount broken down into subsections. The same colour is used for the equivalent subsection in each bar. This allows you to compare both the whole picture and the components of each bar. A 100% component bar chart is a variation of the stacked bar chart, and shows all of the bars at the same height (100%) and the components as percentages of the total. This type of chart is useful for comparing the contribution made by different subgroups within the categories of the main variable.

Figure 11 consolidates data into one stacked bar chart making it much clearer to compare the distribution of resistance of Campylobacter jejuni found in chickens, cattle and people in Denmark between 2016 and 2019. The subsections have been ordered so that the largest changes are first and last. This choice of order highlights changes that may be of specific interest when interpreting the graph. The most noticeable feature to compare between different hosts’ susceptibility is the proportion of C. jejuni isolates with resistance to both ciprofloxacin and tetracycline from people returning from travel overseas compared to domestic cases in people, chickens and cattle. Also, unlike what was observed in cattle, resistance to ciprofloxacin and tetracycline in C. jejuni isolated from poultry and from local cases of campylobacteriosis in people appears to have risen over time. However, statistical significance should be tested before making any firm conclusions from this graph.

Maximise

Figure 11 Distribution (%) of AMR profiles in Campylobacter jejuni from broilers, cattle and human cases in Denmark (Korsgaard et al., 2020). Note: the number of isolates included each year is shown in parentheses, where broilers include isolates from Danish broiler meat. CIP: all isolates with ciprofloxacin resistance but not tetracycline resistance, TET: all isolates with tetracycline resistance but not ciprofloxacin resistance, CIP TET: all isolates with both ciprofloxacin and tetracycline resistance, Other resistance: all isolates without both ciprofloxacin and tetracycline resistance. CIP TET, CIP and TET isolates may be resistant to erythromycin, nalidixic acid or streptomycin

Show description|Hide description

Stacked bar chart, showing the distribution of AMR profiles in C. jejuni samples extracted from broiler chickens, cattle, people with no recent travel history and people returning from overseas travel from 2015 to 2019.

Figure 11 Distribution (%) of AMR profiles in Campylobacter jejuni from broilers, cattle and human cases in Denmark (Korsgaard et al., ...

The strengths and limitations of bar charts are listed below:

Table 11 Strengths and limitations of bar charts

Strengths	Limitations
Represent three or more variables on a single chart (unlike a histogram or scatterplot)	May require an additional explanation in text if many groups are compared or using stacked graphs
Understood by a wide audience, commonly used, and the scales and bars are easily read	Data can be manipulated or misrepresented – care must be taken when presenting data. For example, if the scale is exaggerated or minimised it can change the impression of the magnitude of differences between categories
Useful for showing time trends	Doesn’t show relationships between variables

2.1.6 Line graphs

Line graphs show trends in data and help explore relationships with time-related variables (such as age or calendar year). They help to determine the relationship between two sets of values, with one variable dependent on the other. The independent time-related variable is plotted on the x-axis and the dependent variable on the y-axis. Several dependent variables can be plotted against one independent variable on the same graph if they have the same units and range.

Line graphs are particularly useful when there are many data points and when you want to compare several trend lines on a single graph.

Figure 12 shows data from Australia which reports on trends in resistance and multi-drug resistance patterns to four first-line antimicrobials in Neisseria gonorrhoeae between 2000 and 2017. The graph shows that resistance rates to these antimicrobials have changed in different ways over this time period. Resistance to most antimicrobials was lower in 2017 compared to previous years, except for resistance to azithromycin, which has increased since 2015. This graph raises several questions, one of which is what explains the decrease in benzylpenicillin and ciprofloxacin resistance levels to around 30% after peaking at 45–50% in 2008? Time series data presented in line graphs prompts important questions about the data that allow you to undertake exploratory analyses to look for associations in the data.

Figure 12 Trends in resistance and multidrug-resistance patterns, and decreased susceptibility to ceftriaxone, in Neisseria gonorrhoeae

Show description|Hide description

Line graph showing trends in the percentage resistance to four antibiotics in N. gonorrhoeae from 2000 to 2017.

Figure 12 Trends in resistance and multidrug-resistance patterns, and decreased susceptibility to ceftriaxone, in Neisseria gonorrhoeae

The strengths and limitations of line graphs are listed below:

2.2 Mapping AMR data

Geospatial data surround us. Every day many of us turn to maps to help navigate our way or look at different countries in the world. Maps are a great way to analyse geospatial data, which is data with a geographic value such as Global Positioning System (GPS) coordinates, Geographic Information System (GIS) data, address or postcode. With the increasing use of maps (particularly electronic ones) in daily life, they can offer an intuitive data presentation for non-specialist data users.

Maps are used extensively in disease surveillance because they can be used to display information on the spread of diseases across different geographic regions. For example, Figure 13 is a map where different colours are used to represent different categories of the indicator of interest (in this case, the number of countries enrolled in GLASS).

Maximise

Figure 13 Number of countries enrolled in GLASS in 2017–2018 (WHO, 2019)

Show description|Hide description

World map showing countries enrolled in GLASS in 2017–18 highlighted in pink.

Figure 13 Number of countries enrolled in GLASS in 2017–2018 (WHO, 2019)

Because maps are so effective at telling a story, they are used by governments, NGOs, non-profits, public health departments, the media and at local levels to understand the situation in different areas.

Maps answer spatial questions about your data and can help you understand trends or patterns. It is a good idea to ask yourself whether a map is the best way to display data. If the answer is yes, then you need to ensure your map represents your data accurately, in sufficient detail and attractively.

The strengths and limitations of maps are listed below:

3 Creating data visualisations using common software

3.1 Basic software tools

Many statistical software packages are available where you can quickly and easily construct different types of graphs to explore the data and determine which type best conveys your key messages. These software packages are especially helpful when larger data sets are involved or complex analyses are required.

The most basic and widely used programme for summarising and presenting data is Microsoft Excel. Advantages of MS Excel include:

• flexible storage, modelling and manipulation of datasets
• a range of automated functions, including data analysis and chart functions
• users can select from a range of automated charts to display the data graphically. This process requires minimal amount of user input.

While MS Excel is familiar to many and easy to use, the programme has several limitations, including:

• fewer functions for data analysis and graphic displays than dedicated statistical programmes
• a limited number of total cells compared to these programmes.

While MS Excel is convenient for data entry, it is generally recommended that MS Excel spreadsheets be exported to a dedicated statistical software programme for more than basic analysis.

Google Maps is an open-source service that provides information about geographical sites around the world. It offers conventional road maps, aerial and satellite views of many places. It can be customised to visualise geographic statistical data, such as economic and population statistics, and can be used to create disease prevalence maps. Google Maps is an obvious tool to use to visualise geospatial data as many people are familiar with the platform.

3.2 Specialist software tools

There are now several freely downloadable statistical programs for conducting more complex analyses. Even for simple analyses, there are many advantages to using statistical software, including the ability to reproduce graphs and maps if data is updated, or if another user needs to conduct analyses. Some software programs have steep learning curves but are very powerful once mastered.

R is an example of a free statistical software programme that can be used to summarise and present data in interesting ways, including producing a range of graphs and maps. You can download R by visiting the R project website.

QGIS is an open-source Geographical Information System (GIS) platform that can be run on all operating systems and can support a number of database formats and functionalities. It allows users to analyse and edit spatial information as well as creating and exporting graphical maps.

Box 1: Steps for generating a graph from the NARMS interactive website

Step 1: From the NARMS website, click on the icon ‘Antimicrobial Resistance by Year’ (there is also a tab at the top of the interactive display).

Figure 18 Screenshot of the NARMS ‘Antimicrobial Resistance by Year’ option

Show description|Hide description

Screenshot of the ‘Antimicrobial Resistance by Year’ option on the NARMS home page.

Figure 18 Screenshot of the NARMS ‘Antimicrobial Resistance by Year’ option

Step 2: On the ‘Antimicrobial Resistance by Year’ page, make the following selections:

1. Select a Bacterium – select ‘Salmonella’.
2. Select a host species – select the icon of a person.
3. Select a source and antimicrobial agent(s) – from the drop-down menu select ‘Ceftriaxone’ and click on ‘Apply’.

Figure 19 Screenshot of part of the ‘Antimicrobial Resistance by Year’ page in NARMS

Show description|Hide description

Screenshot of part of the ‘Antimicrobial Resistance by Year’ page in NARMS, highlighting options to select bacterium, host, source and antimicrobial agents.

Figure 19 Screenshot of part of the ‘Antimicrobial Resistance by Year’ page in NARMS

Box 2: Steps for generating a map from the NARMS interactive website

Step 1: From the NARMS website, click on the icon ‘Map of Resistance’ (there is also a tab at the top of the interactive display).

Figure 20 Screenshot of the NARMS ‘Map of Resistance’ option

Show description|Hide description

Screenshot of the ‘Map of Resistance’ option on the NARMS home page.

Figure 20 Screenshot of the NARMS ‘Map of Resistance’ option

Step 2: On the ‘Map of resistance’ page, make the following selections:

1. Select a Bacterium – select ‘Salmonella’.
2. Select a source – select ‘humans’.
3. Select an antimicrobial agent – from the drop-down menu select ‘Ceftriaxone’ and click on ‘Apply’.
4. Use the slider to select the year – ‘2018’.

Figure 21 Screenshot showing the options available on the NARMS ‘Map of Resistance’ page

Show description|Hide description

Screenshot showing the options available on the ‘Map of Resistance’ page in NARMS. Options highlighted are: bacterium, source, antimicrobial agent and year.

Figure 21 Screenshot showing the options available on the NARMS ‘Map of Resistance’ page

4 End-of-module quiz

Well done – you have reached the end of this module and can now do the quiz to test your learning.

This quiz is an opportunity for you to reflect on what you have learned rather than a test, and you can revisit it as many times as you like.

• End-of-module quiz

Open the quiz in a new tab or window by holding down ‘Ctrl’ (or ‘Cmd’ on a Mac) when you click on the link.

5 Summary

In this module, you have learnt about the most common visual ways to display data. You reviewed the strengths and limitations of each and looked at examples of when a particular visual display is better than another. If you are trying to decide whether to use a table or a graph (and the type of graph), ask yourself how the data will be used, consider your target audience and decide the best way to map out your information. Think about the utility of your visual and let that help drive your decision-making.

You should now be able to:

• describe the different ways to visually represent data
• explain why visual summaries of data are an important part of AMR data analysis
• review the strengths and limitations of each visual presentation
• make a simple graph and map using AMR data.

Now that you have completed this module, consider the following questions:

• What is the single most important lesson that you have taken away from this module?
• How relevant is it to your work?
• Can you suggest ways in which this new knowledge can benefit your practice?

When you have reflected on these, go to your reflective blog  and note down your thoughts.

When you have reflected on your answers and your progress on this module, go to your reflective blog and note down your thoughts.

6 Your experience of this module

Now that you have completed this module, take a few moments to reflect on your experience of working through it. Please complete a survey to tell us about your reflections. Your responses will allow us to gauge how useful you have found this module and how effectively you have engaged with the content. We will also use your feedback on this pathway to better inform the design of future online experiences for our learners.

Many thanks for your help.

Now go to the survey.

References

Acknowledgements

This free course was collaboratively written by Melanie Bannister-Tyrrell, Skye Badger and Clare Sansom, and reviewed by Siddharth Mookerjee, Claire Gordon, Natalie Moyen, Peter Taylor and Hilary MacQueen.

Except for third party materials and otherwise stated (see terms and conditions), this content is made available under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 Licence.

The material acknowledged below is Proprietary and used under licence (not subject to Creative Commons Licence). Grateful acknowledgement is made to the following sources for permission to reproduce material in this free course:

Images

Course Image: NosorogUA/Shutterstock

Figure 1: Australian Commission on Safety and Quality in Health Care (ACSQHC). AURA 2019: third Australian report on antimicrobial use and resistance in human health. Sydney: ACSQHC; 2019.

Figure 2: Markwart, R., Willrich, N., Haller, S. et al. (2019) The rise in vancomycin-resistant Enterococcus faecium in Germany: data from the German Antimicrobial Resistance Surveillance (ARS). Antimicrob Resist Infect Control 8, 147 (2019) https://doi.org/ 10.1186/ s13756-019-0594-3. This file is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/ licenses/ by/ 4.0/

Figure 3: Ausvet Pty Ltd

Figure 4: Smith, P. (2008) Antimicrobial resistance in aquaculture, Revue Scientifique et Technique, 2008, 27 (1), 243-264

Figure 5: Pompilio, A., Savini, V., Fiscarelli, E., Gherardi, G., Di Bonaventura, G. (2020) Clonal Diversity, Biofilm Formation, and Antimicrobial Resistance among Stenotrophomonas maltophilia Strains from Cystic Fibrosis and Non-Cystic Fibrosis Patients. Antibiotics 2020, 9, 15. https://doi.org/ 10.3390/ antibiotics9010015 . An open access article distributed under the terms of the Creative Commons Attribution (CC BY) license, http://creativecommons.org/ licenses/ by/ 4.0/

Figure 6: Pompilio, A., Savini, V., Fiscarelli, E., Gherardi, G., Di Bonaventura, G. (2020) Clonal Diversity, Biofilm Formation, and Antimicrobial Resistance among Stenotrophomonas maltophilia Strains from Cystic Fibrosis and Non-Cystic Fibrosis Patients. Antibiotics 2020, 9, 15. https://doi.org/ 10.3390/ antibiotics9010015 . An open access article distributed under the terms of the Creative Commons Attribution (CC BY) license, http://creativecommons.org/ licenses/ by/ 4.0/

Figure 10: Global antimicrobial resistance surveillance system (GLASS) report: early implementation 2017-2018. Geneva: World Health Organization; 2018. This file is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 IGO (CC BY-NC-SA 3.0 IGO) license, https://creativecommons.org/ licenses/ by-nc-sa/ 3.0/ igo/

Figure 11: DANMAP 2019 - Use of antimicrobial agents and occurrence of antimicrobial resistance in bacteria from food animals, food and humans in Denmark. ISSN 1600-2032

Figure 12: Ausvet Pty Ltd

Figure 13: Global antimicrobial resistance surveillance system (GLASS) report: early implementation 2017-2018. Geneva: World Health Organization; 2018. This file is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 IGO (CC BY-NC-SA 3.0 IGO) license, https://creativecommons.org/ licenses/ by-nc-sa/ 3.0/ igo/

Figures 14–16: Food and Drug Administration (FDA). NARMS Now. Rockville, MD: U.S. Department of Health and Human Services. Available from URL: https://www.fda.gov/ animal-veterinary/ national-antimicrobial-resistance-monitoring-system/ narms-now-integrated-data. Accessed 05/11/2021.

Figure 17: © 2021 Mapbox © OpenStreetMap

Figures 18–21: Food and Drug Administration (FDA). NARMS Now. Rockville, MD: U.S. Department of Health and Human Services. Available from URL: https://www.fda.gov/ animal-veterinary/ national-antimicrobial-resistance-monitoring-system/ narms-now-integrated-data. Accessed 05/11/2021.

Quiz Question 4: Food and Drug Administration (FDA). NARMS Now. Rockville, MD: U.S. Department of Health and Human Services. Available from URL: https://www.fda.gov/ animal-veterinary/ national-antimicrobial-resistance-monitoring-system/ narms-now-integrated-data. Accessed 05/19/2021.

Tables

Table 1: Ausvet Pty Ltd

Table 2: Ausvet Pty Ltd

Table 3: Data source: Gadzhanova S, Roughead E. Analysis of 2013–2017 Pharmaceutical Benefits Scheme (PBS) data for the National Report on Antimicrobial Use and Resistance

Tables 5–13: Ausvet Pty Ltd

Every effort has been made to contact copyright owners. If any have been inadvertently overlooked, the publishers will be pleased to make the necessary arrangements at the first opportunity.