Skip to main content
Public

Share visualization

Loading preview…

Data Analysis

vizhub.com

Shared links may look like this on X, LinkedIn, Slack, and Discord.

Choose how to deliver this visualization

LinkedIn caption

LinkedIn may not keep text supplied by the share sheet. Copy the caption and paste it into the post if needed. You can tag the VizHub company page when posting.

More options

Embed on a websiteCopy iframe embed code
Download & exportHTML, source ZIP

Viewing a saved revision (read-only)

cPBRTHE2mj_o9flz1_U0L

Data Analysis

0 stars1Public
Mmunashe gwenzi
Last edited Nov 11, 2026
Created on Nov 11, 2026

Created from react-d3-starter template

Diabetes Health Indicators Visualization

Interactive visualization exploring relationships between diabetes and various health/lifestyle factors.

Dataset Documentation

diabetes_health_indicators_vizhub_sample.csv

  • Rows: 9,000
  • Columns: 22
  • Subject: Survey responses related to diabetes and health indicators

Column Descriptions

Column Data Type Missing Values Cardinality Range/Mean Classification Description
Diabetes_012 Number 0 3 0-2, mean: 0.516 Ordinal Diabetes status (0=No, 1=Prediabetes, 2=Diabetes)
HighBP Number 0 2 0-1, mean: 0.498 Nominal High blood pressure (0=No, 1=Yes)
HighChol Number 0 2 0-1, mean: 0.476 Nominal High cholesterol (0=No, 1=Yes)
CholCheck Number 0 2 0-1, mean: 0.977 Nominal Cholesterol check in past 5 years (0=No, 1=Yes)
BMI Number 0 43 15-87, mean: 28.981 Quantitative Body Mass Index
Smoker Number 0 2 0-1, mean: 0.472 Nominal Smoked at least 100 cigarettes (0=No, 1=Yes)
Stroke Number 0 2 0-1, mean: 0.048 Nominal Ever had stroke (0=No, 1=Yes)
HeartDiseaseorAttack Number 0 2 0-1, mean: 0.114 Nominal Coronary heart disease/angina/heart attack
PhysActivity Number 0 2 0-1, mean: 0.716 Nominal Physical activity in past 30 days (0=No, 1=Yes)
Fruits Number 0 2 0-1, mean: 0.622 Nominal Consume fruit 1+ times per day (0=No, 1=Yes)
Veggies Number 0 2 0-1, mean: 0.8 Nominal Consume vegetables 1+ times per day (0=No, 1=Yes)
HvyAlcoholConsump Number 0 2 0-1, mean: 0.056 Nominal Heavy alcohol consumption (0=No, 1=Yes)
AnyHealthcare Number 0 2 0-1, mean: 0.952 Nominal Any kind of healthcare coverage (0=No, 1=Yes)
NoDocbcCost Number 0 2 0-1, mean: 0.079 Nominal Couldn't see doctor due to cost (0=No, 1=Yes)
GenHlth Number 0 5 1-5, mean: 2.636 Ordinal General health rating (1=Excellent, 5=Poor)
MentHlth Number 0 20 0-30, mean: 3.519 Quantitative Days of poor mental health in past 30 days
PhysHlth Number 0 23 0-30, mean: 4.66 Quantitative Days of poor physical health in past 30 days
DiffWalk Number 0 2 0-1, mean: 0.199 Nominal Difficulty walking (0=No, 1=Yes)
Sex Number 0 2 0-1, mean: 0.426 Nominal Gender (0=Female, 1=Male)
Age Number 0 13 1-13, mean: 8.333 Ordinal Age group (1=18-24, 13=75+)
Education Number 0 5 2-6, mean: 5.002 Ordinal Education level (1=Never attended, 6=College graduate)
Income Number 0 8 1-8, mean: 5.786 Ordinal Income scale (1=Less than $10k, 8=$75k+)

Data Notes

All values are numeric codes representing categorical or ordinal values. The dataset appears to be survey responses with binary indicators (0/1) for yes/no questions and ordinal scales for ratings.

Visualization Mapping

For our visualization:

  • Type: Scatterplot matrix with brushing
  • Fields:
    • X-axis: BMI (Quantitative)
    • Y-axis: Age (Ordinal)
    • Color: Diabetes_012 (Ordinal)
    • Size: PhysHlth (Quantitative)
  • Encodings:
    • Position (X/Y) for BMI and Age
    • Color for diabetes status
    • Size for physical health days
  • Interactions: Brushing to highlight subsets of data

Relationships & Join Keys

This project contains only one dataset, so there are no join relationships to consider.

Visualization Ideas (20)

Important caveats

  • Association, not causation. Every view below shows co-occurrence in cross-sectional survey data. A chart showing higher diabetes share at higher BMI does not show that BMI causes diabetes. Reverse causation, confounding and measurement error are all possible. Use wording such as "is associated with" and "is more common among".
  • Adjusted sample. The 9,000-row file is a sample whose diabetes-category proportions were intentionally adjusted (the mean of Diabetes_012 is 0.516). Raw counts and the overall diabetes share are therefore not population prevalence. Compare within-group proportions or distributions, and label counts as "in this sample".
  • Coded values and outliers. Values are numeric codes (Age 1-13 are bands, Income 1-8 are bands). BMI has extreme values (up to 87), so offer a clipping or zoom control.
  • Accessibility defaults. Use a colorblind-safe palette for diabetes status (for example blue, orange and purple) with redundant shape or pattern cues. Provide keyboard-focusable marks, text summaries, and a data-table toggle.

Ideas

  1. Linked Status Dashboard

    • Purpose: the central hub for exploring the whole sample.
    • Chart: a BMI histogram, an Age-band bar chart, and an Income-band bar chart, all colored by Diabetes_012.
    • Interaction: brushing any chart filters the other two (cross-filtering). A reset button clears the filters.
    • Insight: shows which age, BMI and income segments have a higher share of prediabetes or diabetes in the sample.
  2. Age-by-BMI Heatmap of Diabetes Share

    • Purpose: replace the overplotted scatter with an aggregated view.
    • Chart: a heatmap with Age bands on one axis, BMI bins on the other, and color showing the % with Diabetes_012 = 2.
    • Interaction: hover shows the cell's n and percentage. A toggle switches between diabetes and prediabetes. Cells with small n are hatched.
    • Insight: the age and BMI combinations where diabetes is more common. The n warnings guard against over-reading thin cells.
  3. Blood Pressure Split Comparison

    • Purpose: compare diabetes status by HighBP.
    • Chart: side-by-side 100% stacked bars of Diabetes_012 for HighBP = 0 and HighBP = 1, with Age band as a facet.
    • Interaction: a dropdown to facet by Age band, Sex, or Income group.
    • Insight: whether the association between high blood pressure and diabetes persists within age bands (this hints at age as a confounder).
  4. Activity and Health Slope Chart

    • Purpose: compare general health between active and inactive respondents.
    • Chart: a slope chart of mean GenHlth for PhysActivity = 0 versus 1, one line per diabetes status.
    • Interaction: filter by Age band. Hover shows n and the mean.
    • Insight: whether the activity gap in self-rated health looks similar across diabetes categories. This is association only, since illness can also limit activity.
  5. Income Ladder Dot Plot

    • Purpose: show how diabetes status varies across Income bands.
    • Chart: a Cleveland dot plot with Income 1-8 on the y-axis, and the diabetes share (prediabetes and diabetes) on the x-axis. There is one dot per HighBP group.
    • Interaction: toggle the metric between the diabetes share and the mean BMI.
    • Insight: a possible gradient in the sample, which should not be read as population prevalence.
  6. GenHlth Sankey Flow

    • Purpose: show how respondents flow from Income to GenHlth to Diabetes_012.
    • Chart: a three-column Sankey diagram.
    • Interaction: clicking a node highlights all paths through it. A filter for age group is available.
    • Insight: which self-rated health levels are most common among low-income respondents with diabetes.
  7. Small-Multiple BMI Distributions

    • Purpose: compare BMI distributions across groups.
    • Chart: a grid of ridgeline or violin plots of BMI, one per Age band, colored by Diabetes_012.
    • Interaction: a switch for the facet variable (Age, Income, GenHlth) and a BMI range clip.
    • Insight: whether the BMI distributions of diabetes groups shift differently with age.
  8. Parallel Categories Explorer

    • Purpose: show combinations of risk factors.
    • Chart: parallel categories over HighBP, PhysActivity, GenHlth, Income group and Diabetes_012.
    • Interaction: drag to reorder axes and click a category to filter ribbons.
    • Insight: the most common profiles among respondents with diabetes, and the profiles that are rare.
  9. Odds-Style Comparison Panel

    • Purpose: summarize relative differences without modeling.
    • Chart: a forest-style dot plot of the ratio of diabetes shares (with and without each binary factor: HighBP, HighChol, PhysActivity, DiffWalk, Smoker) with bootstrap intervals.
    • Interaction: stratify by Age band or Sex, and select a reference group.
    • Insight: the unadjusted association size of each factor. A note states that these are not causal effects and are not adjusted for each other.
  10. Age Band Tile Cartogram

    • Purpose: give a compact overview of all 13 age bands.
    • Chart: a tile grid in which each tile is an Age band, colored by diabetes share, with a mini stacked bar inside it.
    • Interaction: clicking a tile filters every linked chart.
    • Insight: the age ranges where the sample's diabetes share rises.
  11. Sample Composition Reality Check

    • Purpose: make the sample adjustment transparent.
    • Chart: a waffle chart of the 9,000 respondents by Diabetes_012, with a short note about the adjusted proportions.
    • Interaction: toggle between "counts in sample" and "share within each group" to switch views.
    • Insight: teaches viewers why within-group rates matter more than raw counts here.
  12. Physical Health Days Beeswarm

    • Purpose: show the distribution of PhysHlth, which is heavily skewed (many 0s and 30s).
    • Chart: a beeswarm or strip plot of PhysHlth by Diabetes_012, with the median marked.
    • Interaction: filter by PhysActivity and DiffWalk. Hover on a dot to see the respondent's profile.
    • Insight: whether poor-health days are bimodal and whether the mix of 0 days and 30 days differs by status.
  13. Mental versus Physical Health Density

    • Purpose: relate MentHlth and PhysHlth.
    • Chart: a 2D binned density plot with a Diabetes_012 selector.
    • Interaction: brush a region to see the profile of the selected respondents (mean BMI, HighBP share, income).
    • Insight: clusters of respondents with both poor mental and poor physical health days.
  14. Healthcare Cost Barrier Matrix

    • Purpose: examine access issues by income.
    • Chart: a heatmap with Income band on one axis and Diabetes_012 on the other, colored by the share with NoDocbcCost = 1.
    • Interaction: toggle the metric to AnyHealthcare or CholCheck.
    • Insight: where cost barriers co-occur with a diabetes diagnosis in this sample.
  15. Risk Profile Builder (Story Mode)

    • Purpose: let viewers build a profile and see what the data says.
    • Chart: toggles for Age band, HighBP, PhysActivity and Income group drive a bar of Diabetes_012 shares among matching respondents.
    • Interaction: toggle controls, and a visible "n matching" counter with a warning below a minimum n.
    • Insight: how the mix changes as filters are added. It is framed as "people like this in the sample", not as personal risk prediction.
  16. Scrollytelling Journey Through One Cohort

    • Purpose: provide an engaging narrative for healthcare storytelling.
    • Chart: a scroll-driven sequence that starts with all respondents as dots, then splits them by Age, then by HighBP, then by GenHlth, recoloring by status at each step.
    • Interaction: scroll or arrow keys advance the steps, and the dots animate between layouts with a reduced-motion option.
    • Insight: shows step by step how groups differ and where the associations weaken once you compare within groups.
  17. Simpson's Paradox Spotlight

    • Purpose: demonstrate confounding.
    • Chart: a paired view, with the overall PhysActivity versus diabetes share on the left and the same comparison split by Age band on the right.
    • Interaction: a slider or toggle to "adjust" for Age, Income or Sex, with the bars animating between pooled and stratified views.
    • Insight: how much an apparent association changes once Age is held fixed, and why causal claims need more than this data.
  18. Respondent Constellation (Small-Multiple Similarity Map)

    • Purpose: find groups of similar respondents.
    • Chart: a 2D projection (for example PCA computed in-browser on the standardized numeric columns) with points colored by Diabetes_012.
    • Interaction: lasso selection shows a profile panel with mean BMI, HighBP share and GenHlth. A color-by selector switches the variable.
    • Insight: whether the diabetes categories separate or overlap. Overlap is a reminder that no single indicator is decisive.
  19. Difficulty Walking Pathway Radial

    • Purpose: link mobility with other indicators.
    • Chart: a radial bar chart of the share of respondents with DiffWalk = 1 for each Age band, with a ring segment per diabetes status.
    • Interaction: hover for values, and click a band to update a linked GenHlth bar chart.
    • Insight: how mobility difficulty rises with age and whether it differs by status and general health.
  20. Accessible Data Table and Sonification Companion

    • Purpose: make the analyses usable for everyone.
    • Chart: a sortable summary table (n, mean BMI, % HighBP, % PhysActivity, mean GenHlth) by Diabetes_012 and Age band, with inline sparkbars. An optional audio mode plays pitch for the diabetes share across Age bands.
    • Interaction: sort and filter the table. Selecting a row filters the linked charts, and keyboard navigation and a screen-reader summary are included.
    • Insight: lets people check precise figures behind the visual patterns and keeps the dashboard inclusive.

Most innovative picks for healthcare storytelling

Ideas 11 (sample reality check), 15 (profile builder), 16 (scrollytelling), 17 (Simpson's paradox spotlight), 18 (respondent constellation) and 20 (accessible table and sonification) are the strongest fits for an engaging narrative project. They combine transparency about the adjusted sample, careful association-versus-causation messaging, and accessibility.

Implementation notes

  • Load the CSV at runtime with d3.csv and parse every needed column with +d[column], using the exact column names from the dataset.
  • Apply one shared filter state in React so that all linked charts update together.
  • Show n for each aggregated mark, and suppress or flag small groups. EOF_MARKER