Figures in Medical Writing: A Review

by Ramonaon Posted on

The second half of Chapter 4 of the AMA Manual of Style deals with Figures. A figure is any graphical display of data, including graphs, diagrams, maps, illustrations, and images. In medical writing, figures play a crucial role in presenting complex data in an easily digestible format. Whether used to illustrate trends, clarify methodologies, or provide visual evidence, figures should be accurate, clear, and concise. This week, we review the different types of figures and best practices for their creation.

Graphs

Line Graphs

  • Show associations between two or more variables.
  • Typically have dependent variables on the y-axis and independent variables on the x-axis.
  • Axes should be continuous, with tick marks for reference points.
  • Data points should align with tick marks rather than fall between them.

Survival Plots

  • Used for time-to-event outcomes (e.g., Kaplan-Meier survival analysis).
  • The y-axis usually starts at 100% and steps downward over time.
  • Step-wise curves should not be smoothed.
  • Numbers for each step should be displayed below the x-axis.

Scatterplots

  • Show relationships between independent and dependent variables.
  • Include a mathematically generated trend line or curve for the ‘center’ of the data.
  • Should include correlation or regression coefficients, sample size, and p-values in the legend.

Histograms and Frequency Polygons

  • Represent frequency or incidence on the y-axis against a continuous x-axis.
  • The y-axis should always start at zero, and must not be broken.
  • If data is represented by bars, the bars should be equal widths.
  • If data is represented by frequency polygons, they can be connected with a curve.

Bar Graphs

  • Display categorical frequencies on the x-axis, and can also show up to 3 subcategories per group.
  • Bars should be equally wide and separated by spaces.
  • Stacked bar graphs are used to show proportions of a total.
  • If error bars are used, they must be bidirectional.
  • Less ideal for showing trends, bar graphs are not appropriate for representing summary statistics.

Dot Graphs

  • Represent quantitative data with discrete markers.
  • Preferably include error bars.

Box-and-Whisker Plots

  • Useful for displaying quartiles and error bars.
  • Boxes represent the interquartile range, and whiskers extend to the upper and lower adjacent values.
  • Horizontal line within the box represents the median, and outliers are shown as dots or open circles outside the whiskers.

Individual-Value Plots

  • Show full distribution of findings.
  • Mean or median is indicated with a horizontal line, +/- error bars.

Paired Data and Spaghetti Plots

  • Track individual trends over time in longitudinal studies.

Forest Plots

  • Used in meta-analyses and systematic reviews to summarize data from multiple studies.
  • Plotted on a log scale with a center at 1.
  • Display hazard ratios with confidence intervals.
  • Generally presented with adjacent columns providing further information, such as the source and its discrete data.

Funnel Plots

  • Scatterplots of effect size used for data in meta-analyses to detect bias.
  • If no bias is present, it should resemble an inverted funnel; if bias is present, it will be skewed and asymmetrical. 

Hybrid Graphs

  • Combine multiple graphical elements to improve data visualization.

Pie Charts

  • Should be avoided in scientific publications

Diagrams

Flowcharts

  • Show sequences of events and used to display study protocols, participant recruitment, or inclusion/exclusion criteria.
  • Randomization points are oval-shaped, while all other entries are rectangular.

Decision Trees

  • Used for clinical decision-making or cost-effectiveness analysis.
  • Decision nodes are squares, while probability nodes are circles.

Algorithms

  • Branched pathways for clinical diagnoses or treatment decisions.
  • Decision points are diamonds/hexagons, and action steps are rectangles/squares.

Pedigrees

  • Illustrate familial relationships and can show inherited conditions.
  • Use standardized symbols for gender, status, and condition presence.

Venn Diagrams

  • Show overlapping relationships but are not recommended for scientific publications.

Maps

  • Used to demonstrate location and distance relationships, such as disease spread.
  • Choropleth maps: Use color gradients to indicate frequency.
  • Heat maps: Show structure within a data matrix, and commonly used for gene expression.
  • Network maps: Represent connectivity among data points.

Illustrations and Imaging

  • Used to explain physiology, techniques, or procedures.
  • Can be photographs of patients, biopsy specimens, or medical imaging.
  • If a patient can be identified within the image, the author needs written permission from the patient to publish the image

Best Practices for Creating Figures

  • Scales and Axes: Axes should be continuous, labeled with data type and units, and include clearly placed tick marks.
  • Data Representation: Data lines should be thicker than scale lines, with labels using sentence-style capitalization.
  • Titles: Titles should be concise (10-15 words) and should not start with a figure type descriptor (example: “Bar graph of…”).
  • Legends: Legends should be in sentence format, no longer than 40 words.
  • Color and Symbols: Ensure all elements are clearly distinguishable and explained in a legend.
  • Error Bars: Must be bidirectional and properly defined.
  • Reproduced Figures: Must include acknowledgments in the legend.

Bonus: Non-Tabular Material

  • Text Boxes: Highlight key information.
  • Sidebars: Provide supplementary references or reading suggestions.

Reference:

AMA Manual of Style Committee. AMA Manual of Style: A Guide for Authors and Editors. 11th ed. Oxford University Press; 2020