Data Visualization Best Practices for Health Dashboards – Nic

Why Health Dashboards Matter

In healthcare, timely, clear information saves lives. Well-designed dashboards help:

✅ Clinicians - Monitor patient conditions in real-time ✅ Administrators - Track facility performance ✅ Public health officials - Detect disease outbreaks ✅ Researchers - Identify trends and patterns ✅ Policymakers - Make evidence-based decisions

Key Statistics: - Dashboards improve decision-making speed by 5x - Visual data is processed 60,000x faster than text - Good dashboards can reduce patient mortality by 15-20%

Dashboard Design Principles

1. The 5-Second Rule ⏱️

Your audience should understand the main message in 5 seconds.

Bad Example: - 20 different metrics crammed on one screen - No visual hierarchy - Inconsistent colors - Tiny text

Good Example: - 3-5 key metrics prominently displayed - Clear visual hierarchy - Consistent color scheme - Readable text (minimum 12pt)

2. Know Your Audience 👥

Audience	Needs	Dashboard Type
Hospital CEO	High-level KPIs, trends	Strategic
Ward Manager	Patient flow, staffing	Operational
Clinician	Patient vitals, alerts	Clinical
Data Analyst	Detailed data, filters	Analytical

3. Choose the Right Chart Type 📊

Common mistakes in health dashboards:

❌ Pie charts for > 3 categories ✅ Bar charts instead

❌ 3D charts (distort perception) ✅ 2D charts with clear labels

❌ Dual-axis charts (confusing) ✅ Separate charts or small multiples

Chart Selection Guide for Health Data

Time Series Data (Disease trends, admissions over time)

Best: Line Charts

# R example: COVID-19 cases over time
ggplot(covid_data, aes(x = date, y = cases)) +
  geom_line(color = "#00539B", size = 1.2) +
  geom_smooth(method = "loess", se = FALSE, 
              color = "#FFA500", linetype = "dashed") +
  labs(title = "Daily COVID-19 Cases",
       subtitle = "7-day moving average shown in orange",
       x = "Date", y = "Number of Cases") +
  theme_minimal() +
  theme(plot.title = element_text(size = 16, face = "bold"))

When to use: - Continuous time series - Showing trends and patterns - Comparing multiple time series (max 5 lines)

Comparisons (Between groups, facilities, treatments)

Best: Bar Charts (horizontal for long labels)

# Hospital comparison
ggplot(hospital_data, aes(x = reorder(hospital, mortality_rate), 
                           y = mortality_rate)) +
  geom_col(fill = "#00539B") +
  geom_text(aes(label = paste0(mortality_rate, "%")), 
            hjust = -0.2) +
  coord_flip() +
  labs(title = "Hospital Mortality Rates",
       x = NULL, y = "Mortality Rate (%)") +
  theme_minimal()

When to use: - Comparing categories - Ranking data - Showing discrete values

Part-to-Whole (Disease burden distribution)

Best: Stacked Bar Charts or Waffle Charts

# Disease burden by age group
ggplot(disease_data, aes(x = year, y = cases, fill = age_group)) +
  geom_col(position = "fill") +
  scale_y_continuous(labels = scales::percent) +
  scale_fill_brewer(palette = "Blues") +
  labs(title = "Malaria Cases by Age Group",
       y = "Proportion of Cases", 
       fill = "Age Group") +
  theme_minimal()

Avoid: Pie charts (except for 2-3 categories max)

Relationships (BMI vs. disease risk)

Best: Scatter Plots

# BMI vs Blood Pressure
ggplot(patient_data, aes(x = bmi, y = systolic_bp)) +
  geom_point(alpha = 0.5, color = "#00539B") +
  geom_smooth(method = "lm", color = "#FFA500") +
  labs(title = "Relationship Between BMI and Blood Pressure",
       x = "BMI (kg/m²)", 
       y = "Systolic BP (mmHg)") +
  theme_minimal()

Geographic Data (Disease outbreaks, facility locations)

Best: Choropleth Maps or Point Maps

# Disease prevalence map
library(sf)
library(viridis)

ggplot(kenya_counties) +
  geom_sf(aes(fill = malaria_prevalence)) +
  scale_fill_viridis(option = "plasma", 
                     name = "Prevalence (%)") +
  labs(title = "Malaria Prevalence by County") +
  theme_void()

Distributions (Patient age, wait times)

Best: Histograms or Box Plots

# Age distribution
ggplot(patient_data, aes(x = age)) +
  geom_histogram(binwidth = 5, fill = "#00539B", color = "white") +
  labs(title = "Patient Age Distribution",
       x = "Age (years)", y = "Count") +
  theme_minimal()

# Wait time by department
ggplot(patient_data, aes(x = department, y = wait_time)) +
  geom_boxplot(fill = "#00539B") +
  coord_flip() +
  labs(title = "Wait Times by Department",
       x = NULL, y = "Wait Time (minutes)") +
  theme_minimal()

Color Best Practices

1. Use Healthcare Color Psychology 🎨

Recommended colors: - Blue (#00539B) - Trust, calm, professional - Green (#009639) - Health, growth, positive outcomes - Orange (#FFA500) - Warning, attention needed - Red (#DC143C) - Critical, urgent, danger

Avoid: - Pure red/green combinations (colorblind accessibility) - Fluorescent colors - Too many colors (max 6-7)

2. Accessible Color Palettes

# Colorblind-friendly palette
library(viridis)

# Sequential (for continuous data)
scale_fill_viridis_c(option = "viridis")  # Blue to yellow

# Diverging (for positive/negative)
scale_fill_gradient2(low = "#0571B0", mid = "white", high = "#CA0020",
                     midpoint = 0)

# Categorical (for groups)
library(RColorBrewer)
scale_fill_brewer(palette = "Set2")

Test your colors: - Color Oracle - Colorblind simulator - Viz Palette - Test combinations

3. Semantic Colors

Use colors consistently: - Targets met: Green - Warning: Orange/Yellow - Critical: Red - Neutral: Gray/Blue

Key Metrics for Health Dashboards

Hospital Operations Dashboard

Primary Metrics: 1. Bed Occupancy Rate (Target: 85-90%) 2. Average Length of Stay (Lower is often better) 3. Patient Wait Time (Emergency: <15 min) 4. Readmission Rate (Target: <8% within 30 days) 5. Staff-to-Patient Ratio

Visual example:

# KPI cards
library(flexdashboard)

# In your dashboard
valueBox(
  value = "87%",
  caption = "Bed Occupancy",
  icon = "fa-bed",
  color = ifelse(occupancy >= 85 && occupancy <= 90, "success", "warning")
)

Public Health Dashboard

Primary Metrics: 1. Disease Incidence Rate (per 100,000) 2. Vaccination Coverage (Target: >90%) 3. Outbreak Detection (Case counts, trend) 4. Geographic Hotspots 5. Healthcare Access (Distance to facility)

Clinical Dashboard

Primary Metrics: 1. Vital Signs (BP, HR, Temp, SpO2) 2. Lab Results (with reference ranges) 3. Medication Adherence 4. Risk Scores (Sepsis, Fall risk) 5. Alerts and Warnings

Dashboard Layout Best Practices

The F-Pattern Layout 👁️

Users read in an F-pattern: 1. Top left: Most important metric 2. Top row: Secondary metrics 3. Left column: Key visualizations 4. Center: Detailed charts 5. Bottom: Supplementary information

Example Layout Structure

+----------------------------------+
|  🏥 Hospital Name     📅 Date    |
+----------+----------+------------+
|  KPI 1   |  KPI 2   |   KPI 3    |
| (Large)  | (Medium) | (Medium)   |
+----------+----------+------------+
|            Main Chart             |
|         (Time Series)             |
+-------------------+---------------+
|  Detail Chart 1   | Detail Chart2 |
+-------------------+---------------+
|        Table of Recent Items      |
+----------------------------------+

Responsive Design

/* Mobile-first approach */
@media (max-width: 768px) {
  .kpi-card {
    width: 100%;
    margin-bottom: 10px;
  }
  
  .chart {
    height: 300px;
  }
}

@media (min-width: 769px) {
  .kpi-card {
    width: 30%;
    display: inline-block;
  }
  
  .chart {
    height: 500px;
  }
}

Tools for Building Health Dashboards

1. R Shiny (FREE, highly customizable) ⭐

Pros: - Complete control over design - Integration with R statistical packages - Can embed complex analyses - Free deployment options

Example:

library(shiny)
library(shinydashboard)
library(ggplot2)
library(dplyr)

ui <- dashboardPage(
  dashboardHeader(title = "Health Dashboard"),
  
  dashboardSidebar(
    sidebarMenu(
      menuItem("Overview", tabName = "overview"),
      menuItem("Patients", tabName = "patients"),
      menuItem("Analytics", tabName = "analytics")
    )
  ),
  
  dashboardBody(
    tabItems(
      tabItem(tabName = "overview",
              fluidRow(
                valueBoxOutput("total_patients"),
                valueBoxOutput("bed_occupancy"),
                valueBoxOutput("avg_wait_time")
              ),
              fluidRow(
                box(
                  title = "Daily Admissions",
                  plotOutput("admissions_plot"),
                  width = 12
                )
              )
      )
    )
  )
)

server <- function(input, output) {
  # Load data
  data <- reactive({
    read_csv("hospital_data.csv")
  })
  
  # KPI boxes
  output$total_patients <- renderValueBox({
    valueBox(
      value = nrow(data()),
      subtitle = "Total Patients",
      icon = icon("users"),
      color = "blue"
    )
  })
  
  # Plot
  output$admissions_plot <- renderPlot({
    data() %>%
      count(date) %>%
      ggplot(aes(x = date, y = n)) +
      geom_line(size = 1.2, color = "#00539B") +
      theme_minimal() +
      labs(title = "Daily Admissions", y = "Count")
  })
}

shinyApp(ui, server)

2. Tableau ($$, user-friendly)

Pros: - Drag-and-drop interface - Beautiful built-in templates - Strong community - Easy sharing

Best for: Non-programmers, quick prototypes

3. Power BI ($$, Microsoft ecosystem)

Pros: - Integration with Microsoft products - Good for enterprise - Mobile apps - Real-time data connections

Best for: Organizations using Microsoft infrastructure

4. Plotly Dash (Python, FREE)

Pros: - Python-based - Highly interactive - Good for ML integration - Deployment options

import dash
from dash import dcc, html
import plotly.express as px
import pandas as pd

# Load data
df = pd.read_csv('health_data.csv')

# Initialize app
app = dash.Dash(__name__)

app.layout = html.Div([
    html.H1("Health Dashboard"),
    
    html.Div([
        html.Div([
            html.H3("Total Patients"),
            html.H2(f"{len(df)}")
        ], className="kpi-card"),
        
        html.Div([
            html.H3("Bed Occupancy"),
            html.H2("87%")
        ], className="kpi-card")
    ]),
    
    dcc.Graph(
        figure=px.line(df, x='date', y='admissions',
                      title='Daily Admissions')
    )
])

if __name__ == '__main__':
    app.run_server(debug=True)

5. Flexdashboard (R Markdown, FREE) ⭐

Pros: - Static or dynamic dashboards - Easy to create from R Markdown - Beautiful layouts - Can host for free

Example:

---
title: "Hospital Dashboard"
output: 
  flexdashboard::flex_dashboard:
    orientation: rows
    vertical_layout: fill
---

```{r setup, include=FALSE}
library(flexdashboard)
library(tidyverse)
data <- read_csv("hospital_data.csv")
```

Row {data-height=150}
-----------------------------------------------------------------------

### Total Patients

```{r}
valueBox(
  value = nrow(data),
  icon = "fa-users",
  color = "primary"
)
```

### Bed Occupancy

```{r}
occupancy <- 87
valueBox(
  value = paste0(occupancy, "%"),
  icon = "fa-bed",
  color = ifelse(occupancy > 90, "danger", "success")
)
```

Row
-----------------------------------------------------------------------

### Daily Admissions

```{r}
ggplot(data, aes(x = date, y = admissions)) +
  geom_line(size = 1.2, color = "#00539B") +
  theme_minimal()
```

Real-World Dashboard Examples

Example 1: COVID-19 Monitoring Dashboard

Purpose: Track pandemic metrics for county health department

Key Components:

Hero Numbers (Top):
- Total Cases (with change from yesterday)
- Active Cases
- Total Deaths
- Vaccination Rate
Main Chart (Center):
- Daily new cases (7-day moving average)
- Hospitalization trend
Supporting Visuals:
- Cases by age group (bar chart)
- Geographic distribution (map)
- Testing positivity rate (gauge)
Table (Bottom):
- Recent cases by facility

Color Scheme: - Cases: Blue (#00539B) - Deaths: Dark gray (#4A4A4A) - Vaccinations: Green (#009639) - Warnings: Orange (#FFA500)

Example 2: Hospital Emergency Department Dashboard

Purpose: Real-time ED performance monitoring

Auto-refresh: Every 5 minutes

Key Metrics:

# Real-time ED dashboard metrics
metrics <- list(
  patients_waiting = sum(ed_data$status == "Waiting"),
  avg_wait_time = mean(ed_data$wait_time),
  patients_in_treatment = sum(ed_data$status == "In Treatment"),
  patients_admitted = sum(ed_data$disposition == "Admitted"),
  bed_availability = available_beds / total_beds * 100
)

# Color coding for wait times
wait_color <- case_when(
  metrics$avg_wait_time < 15 ~ "green",
  metrics$avg_wait_time < 30 ~ "orange",
  TRUE ~ "red"
)

Visualizations: 1. Patient flow (Sankey diagram) 2. Wait times by triage level (grouped bar) 3. Hourly arrival pattern (area chart) 4. Staff utilization (gauge charts)

Example 3: Public Health Surveillance Dashboard

Purpose: Disease outbreak detection

Update Frequency: Daily

Key Features:

# Outbreak detection algorithm
detect_outbreak <- function(current_cases, baseline) {
  threshold <- baseline + 2 * sd(baseline)
  alert <- current_cases > threshold
  return(alert)
}

# Visualize with threshold line
ggplot(disease_data, aes(x = date, y = cases)) +
  geom_line(size = 1.2) +
  geom_hline(yintercept = outbreak_threshold, 
             color = "red", linetype = "dashed") +
  geom_point(data = filter(disease_data, alert == TRUE),
             color = "red", size = 3) +
  annotate("text", x = max(disease_data$date), 
           y = outbreak_threshold,
           label = "Outbreak Threshold", 
           vjust = -0.5, color = "red")

Components: 1. Epidemic curve 2. Geographic hotspots (map) 3. Case demographics (population pyramid) 4. Alert system (conditional formatting)

Interactive Features to Include

1. Filters 🔍

# Shiny filters example
selectInput("facility", "Select Facility:",
            choices = unique(data$facility),
            multiple = TRUE)

dateRangeInput("dates", "Date Range:",
               start = Sys.Date() - 30,
               end = Sys.Date())

2. Drill-Down 📊

Allow users to click for details: - County → District → Facility → Ward

3. Tooltips 💬

# ggplot tooltips with plotly
library(plotly)

p <- ggplot(data, aes(x = date, y = cases, 
                      text = paste("Date:", date,
                                  "<br>Cases:", cases))) +
  geom_line()

ggplotly(p, tooltip = "text")

4. Export Options 📥

# Download button
downloadButton("download_data", "Download Data")

# Server
output$download_data <- downloadHandler(
  filename = function() {
    paste("health_data_", Sys.Date(), ".csv", sep = "")
  },
  content = function(file) {
    write_csv(filtered_data(), file)
  }
)

Performance Optimization

1. Data Aggregation

# Pre-aggregate data
daily_summary <- raw_data %>%
  group_by(date, facility) %>%
  summarize(
    total_patients = n(),
    avg_wait = mean(wait_time),
    max_wait = max(wait_time)
  )

# Use summary instead of raw data

2. Caching

# Cache expensive computations
predictions <- reactive({
  req(input$date_range)
  # Cache for 1 hour
  cached_result <- memoise::memoise(
    predict_admissions(data(), input$date_range),
    cache = cachem::cache_mem(max_age = 3600)
  )
})

3. Lazy Loading

# Load data only when tab is opened
output$detailed_table <- renderDT({
  req(input$tabs == "details")  # Only load if on this tab
  datatable(detailed_data())
})

Accessibility Checklist

✅ Color: - [ ] Not relying on color alone to convey information - [ ] Colorblind-friendly palette - [ ] Sufficient contrast (4.5:1 minimum)

✅ Text: - [ ] Minimum font size 12pt - [ ] Clear, readable fonts - [ ] Alternative text for images

✅ Navigation: - [ ] Keyboard accessible - [ ] Screen reader compatible - [ ] Clear focus indicators

✅ Content: - [ ] Meaningful titles and labels - [ ] Data tables have headers - [ ] Tooltips provide context

Common Mistakes to Avoid

❌ Mistake 1: Too Much Information

Problem: 20 charts crammed on one screen

Solution: Follow the “3-Click Rule” - Key metrics immediately visible - Details available in 1-2 clicks - Use tabs or pages for deep dives

❌ Mistake 2: Misleading Visualizations

Problems: - Y-axis doesn’t start at zero - Inconsistent scales - Cherry-picked date ranges

Solution: - Always start bar charts at zero - Use consistent scales for comparisons - Show full context

❌ Mistake 3: No Context

Problem: “87%” displayed with no interpretation

Solution: Add: - Target values or benchmarks - Trend indicators (↑↓) - Historical comparison - Reference ranges

valueBox(
  value = "87%",
  caption = "Bed Occupancy (Target: 85-90%)",
  icon = "fa-bed",
  color = "success"
)

❌ Mistake 4: Static Dashboard

Problem: Dashboard never updated or maintained

Solution: - Automated data refresh - Version control - Regular user feedback - Documented update schedule

Testing Your Dashboard

User Testing Checklist

Test with actual users:

Comprehension Test
- Can users identify the main message in 5 seconds?
- Do they understand what each metric means?
Task Completion
- Can they find specific information?
- Can they filter and interact effectively?
Decision Making
- Does the dashboard help them make decisions?
- What additional information do they need?
Performance
- Does it load quickly?
- Are interactions responsive?

A/B Testing

# Track which dashboard version performs better
log_interaction <- function(user_id, version, action, timestamp) {
  write_csv(
    data.frame(user_id, version, action, timestamp),
    "dashboard_analytics.csv",
    append = TRUE
  )
}

# Analyze results
analytics %>%
  group_by(version) %>%
  summarize(
    avg_time_on_page = mean(time_on_page),
    click_through_rate = sum(clicked) / n()
  )

Deployment Options

1. Shinyapps.io (Easiest for R)

# Install rsconnect
install.packages("rsconnect")

# Configure account
rsconnect::setAccountInfo(
  name = "your-account",
  token = "your-token",
  secret = "your-secret"
)

# Deploy
rsconnect::deployApp("path/to/dashboard")

Free tier: 5 apps, 25 active hours/month

2. RStudio Connect (Enterprise)

Internal hosting
Authentication
Scheduled updates
Email reports

3. Docker + Cloud (Most flexible)

# Dockerfile
FROM rocker/shiny:latest

# Install R packages
RUN R -e "install.packages(c('shiny', 'tidyverse', 'plotly'))"

# Copy app
COPY app.R /srv/shiny-server/

# Expose port
EXPOSE 3838

# Run
CMD ["/usr/bin/shiny-server"]

4. GitHub Pages (Static dashboards only)

# Build flexdashboard
Rscript -e "rmarkdown::render('dashboard.Rmd')"

# Push to gh-pages branch
git add dashboard.html
git commit -m "Update dashboard"
git push origin gh-pages

Maintenance and Updates

Regular Reviews

Monthly: - [ ] Check data accuracy - [ ] Review user feedback - [ ] Update metrics if needed - [ ] Test all interactive features

Quarterly: - [ ] Usability testing - [ ] Performance optimization - [ ] Design refresh if needed - [ ] Documentation update

Annually: - [ ] Complete redesign review - [ ] Technology stack evaluation - [ ] User needs assessment

Resources

Learning

Information is Beautiful - Inspiration
Flowing Data - Tutorials and examples
Data Visualization Catalogue - Chart selection
Chartio Data School - Best practices

Tools

Color Brewer - Color palettes
Coolors - Color scheme generator
FontPair - Font combinations
Figma - Dashboard prototyping

Communities

Conclusion

Creating effective health dashboards is both an art and a science. The best dashboards:

✅ Focus on what matters - Show key metrics prominently ✅ Tell a story - Guide users through data ✅ Enable action - Provide insights that drive decisions ✅ Are accessible - Work for all users ✅ Stay current - Update automatically

Remember: The goal is not to display all available data, but to provide actionable insights that improve health outcomes.

Start small, iterate based on feedback, and always keep your users’ needs first.

Tags: #DataVisualization #Dashboards #HealthAnalytics #RShiny #Tableau #BestPractices

What challenges have you faced creating health dashboards? Share your experiences below!

--- title: "Data Visualization Best Practices for Health Dashboards" subtitle: "Design Dashboards That Inform, Engage, and Drive Action in Healthcare Settings" author: "Nichodemus Amollo" date: "2025-10-26" categories: [Data Visualization, Dashboards, Health Analytics, Design, Best Practices] --- ## Why Health Dashboards Matter In healthcare, **timely, clear information saves lives**. Well-designed dashboards help: ✅ **Clinicians** - Monitor patient conditions in real-time ✅ **Administrators** - Track facility performance ✅ **Public health officials** - Detect disease outbreaks ✅ **Researchers** - Identify trends and patterns ✅ **Policymakers** - Make evidence-based decisions **Key Statistics:** - Dashboards improve decision-making speed by **5x** - Visual data is processed **60,000x faster** than text - Good dashboards can reduce patient mortality by **15-20%** --- ## Dashboard Design Principles ### 1. **The 5-Second Rule** ⏱️ **Your audience should understand the main message in 5 seconds.** **Bad Example:** - 20 different metrics crammed on one screen - No visual hierarchy - Inconsistent colors - Tiny text **Good Example:** - 3-5 key metrics prominently displayed - Clear visual hierarchy - Consistent color scheme - Readable text (minimum 12pt) ### 2. **Know Your Audience** 👥 | Audience | Needs | Dashboard Type | |----------|-------|----------------| | Hospital CEO | High-level KPIs, trends | Strategic | | Ward Manager | Patient flow, staffing | Operational | | Clinician | Patient vitals, alerts | Clinical | | Data Analyst | Detailed data, filters | Analytical | ### 3. **Choose the Right Chart Type** 📊 **Common mistakes in health dashboards:** ❌ **Pie charts for > 3 categories** ✅ **Bar charts** instead ❌ **3D charts** (distort perception) ✅ **2D charts** with clear labels ❌ **Dual-axis charts** (confusing) ✅ **Separate charts** or **small multiples** --- ## Chart Selection Guide for Health Data ### **Time Series Data** (Disease trends, admissions over time) **Best: Line Charts** ```r # R example: COVID-19 cases over time ggplot(covid_data, aes(x = date, y = cases)) + geom_line(color = "#00539B", size = 1.2) + geom_smooth(method = "loess", se = FALSE, color = "#FFA500", linetype = "dashed") + labs(title = "Daily COVID-19 Cases", subtitle = "7-day moving average shown in orange", x = "Date", y = "Number of Cases") + theme_minimal() + theme(plot.title = element_text(size = 16, face = "bold")) ``` **When to use:** - Continuous time series - Showing trends and patterns - Comparing multiple time series (max 5 lines) ### **Comparisons** (Between groups, facilities, treatments) **Best: Bar Charts (horizontal for long labels)** ```r # Hospital comparison ggplot(hospital_data, aes(x = reorder(hospital, mortality_rate), y = mortality_rate)) + geom_col(fill = "#00539B") + geom_text(aes(label = paste0(mortality_rate, "%")), hjust = -0.2) + coord_flip() + labs(title = "Hospital Mortality Rates", x = NULL, y = "Mortality Rate (%)") + theme_minimal() ``` **When to use:** - Comparing categories - Ranking data - Showing discrete values ### **Part-to-Whole** (Disease burden distribution) **Best: Stacked Bar Charts or Waffle Charts** ```r # Disease burden by age group ggplot(disease_data, aes(x = year, y = cases, fill = age_group)) + geom_col(position = "fill") + scale_y_continuous(labels = scales::percent) + scale_fill_brewer(palette = "Blues") + labs(title = "Malaria Cases by Age Group", y = "Proportion of Cases", fill = "Age Group") + theme_minimal() ``` **Avoid: Pie charts** (except for 2-3 categories max) ### **Relationships** (BMI vs. disease risk) **Best: Scatter Plots** ```r # BMI vs Blood Pressure ggplot(patient_data, aes(x = bmi, y = systolic_bp)) + geom_point(alpha = 0.5, color = "#00539B") + geom_smooth(method = "lm", color = "#FFA500") + labs(title = "Relationship Between BMI and Blood Pressure", x = "BMI (kg/m²)", y = "Systolic BP (mmHg)") + theme_minimal() ``` ### **Geographic Data** (Disease outbreaks, facility locations) **Best: Choropleth Maps or Point Maps** ```r # Disease prevalence map library(sf) library(viridis) ggplot(kenya_counties) + geom_sf(aes(fill = malaria_prevalence)) + scale_fill_viridis(option = "plasma", name = "Prevalence (%)") + labs(title = "Malaria Prevalence by County") + theme_void() ``` ### **Distributions** (Patient age, wait times) **Best: Histograms or Box Plots** ```r # Age distribution ggplot(patient_data, aes(x = age)) + geom_histogram(binwidth = 5, fill = "#00539B", color = "white") + labs(title = "Patient Age Distribution", x = "Age (years)", y = "Count") + theme_minimal() # Wait time by department ggplot(patient_data, aes(x = department, y = wait_time)) + geom_boxplot(fill = "#00539B") + coord_flip() + labs(title = "Wait Times by Department", x = NULL, y = "Wait Time (minutes)") + theme_minimal() ``` --- ## Color Best Practices ### 1. **Use Healthcare Color Psychology** 🎨 **Recommended colors:** - **Blue (#00539B)** - Trust, calm, professional - **Green (#009639)** - Health, growth, positive outcomes - **Orange (#FFA500)** - Warning, attention needed - **Red (#DC143C)** - Critical, urgent, danger **Avoid:** - Pure red/green combinations (colorblind accessibility) - Fluorescent colors - Too many colors (max 6-7) ### 2. **Accessible Color Palettes** ```r # Colorblind-friendly palette library(viridis) # Sequential (for continuous data) scale_fill_viridis_c(option = "viridis") # Blue to yellow # Diverging (for positive/negative) scale_fill_gradient2(low = "#0571B0", mid = "white", high = "#CA0020", midpoint = 0) # Categorical (for groups) library(RColorBrewer) scale_fill_brewer(palette = "Set2") ``` **Test your colors:** - [Color Oracle](https://colororacle.org/) - Colorblind simulator - [Viz Palette](https://projects.susielu.com/viz-palette) - Test combinations ### 3. **Semantic Colors** Use colors consistently: - **Targets met:** Green - **Warning:** Orange/Yellow - **Critical:** Red - **Neutral:** Gray/Blue --- ## Key Metrics for Health Dashboards ### Hospital Operations Dashboard **Primary Metrics:** 1. **Bed Occupancy Rate** (Target: 85-90%) 2. **Average Length of Stay** (Lower is often better) 3. **Patient Wait Time** (Emergency: <15 min) 4. **Readmission Rate** (Target: <8% within 30 days) 5. **Staff-to-Patient Ratio** **Visual example:** ```r # KPI cards library(flexdashboard) # In your dashboard valueBox( value = "87%", caption = "Bed Occupancy", icon = "fa-bed", color = ifelse(occupancy >= 85 && occupancy <= 90, "success", "warning") ) ``` ### Public Health Dashboard **Primary Metrics:** 1. **Disease Incidence Rate** (per 100,000) 2. **Vaccination Coverage** (Target: >90%) 3. **Outbreak Detection** (Case counts, trend) 4. **Geographic Hotspots** 5. **Healthcare Access** (Distance to facility) ### Clinical Dashboard **Primary Metrics:** 1. **Vital Signs** (BP, HR, Temp, SpO2) 2. **Lab Results** (with reference ranges) 3. **Medication Adherence** 4. **Risk Scores** (Sepsis, Fall risk) 5. **Alerts and Warnings** --- ## Dashboard Layout Best Practices ### The F-Pattern Layout 👁️ Users read in an F-pattern: 1. **Top left:** Most important metric 2. **Top row:** Secondary metrics 3. **Left column:** Key visualizations 4. **Center:** Detailed charts 5. **Bottom:** Supplementary information ### Example Layout Structure ``` +----------------------------------+ | 🏥 Hospital Name 📅 Date | +----------+----------+------------+ | KPI 1 | KPI 2 | KPI 3 | | (Large) | (Medium) | (Medium) | +----------+----------+------------+ | Main Chart | | (Time Series) | +-------------------+---------------+ | Detail Chart 1 | Detail Chart2 | +-------------------+---------------+ | Table of Recent Items | +----------------------------------+ ``` ### Responsive Design ```css /* Mobile-first approach */ @media (max-width: 768px) { .kpi-card { width: 100%; margin-bottom: 10px; } .chart { height: 300px; } } @media (min-width: 769px) { .kpi-card { width: 30%; display: inline-block; } .chart { height: 500px; } } ``` --- ## Tools for Building Health Dashboards ### 1. **R Shiny** (FREE, highly customizable) ⭐ **Pros:** - Complete control over design - Integration with R statistical packages - Can embed complex analyses - Free deployment options **Example:** ```r library(shiny) library(shinydashboard) library(ggplot2) library(dplyr) ui <- dashboardPage( dashboardHeader(title = "Health Dashboard"), dashboardSidebar( sidebarMenu( menuItem("Overview", tabName = "overview"), menuItem("Patients", tabName = "patients"), menuItem("Analytics", tabName = "analytics") ) ), dashboardBody( tabItems( tabItem(tabName = "overview", fluidRow( valueBoxOutput("total_patients"), valueBoxOutput("bed_occupancy"), valueBoxOutput("avg_wait_time") ), fluidRow( box( title = "Daily Admissions", plotOutput("admissions_plot"), width = 12 ) ) ) ) ) ) server <- function(input, output) { # Load data data <- reactive({ read_csv("hospital_data.csv") }) # KPI boxes output$total_patients <- renderValueBox({ valueBox( value = nrow(data()), subtitle = "Total Patients", icon = icon("users"), color = "blue" ) }) # Plot output$admissions_plot <- renderPlot({ data() %>% count(date) %>% ggplot(aes(x = date, y = n)) + geom_line(size = 1.2, color = "#00539B") + theme_minimal() + labs(title = "Daily Admissions", y = "Count") }) } shinyApp(ui, server) ``` ### 2. **Tableau** ($$, user-friendly) **Pros:** - Drag-and-drop interface - Beautiful built-in templates - Strong community - Easy sharing **Best for:** Non-programmers, quick prototypes ### 3. **Power BI** ($$, Microsoft ecosystem) **Pros:** - Integration with Microsoft products - Good for enterprise - Mobile apps - Real-time data connections **Best for:** Organizations using Microsoft infrastructure ### 4. **Plotly Dash** (Python, FREE) **Pros:** - Python-based - Highly interactive - Good for ML integration - Deployment options ```python import dash from dash import dcc, html import plotly.express as px import pandas as pd # Load data df = pd.read_csv('health_data.csv') # Initialize app app = dash.Dash(__name__) app.layout = html.Div([ html.H1("Health Dashboard"), html.Div([ html.Div([ html.H3("Total Patients"), html.H2(f"{len(df)}") ], className="kpi-card"), html.Div([ html.H3("Bed Occupancy"), html.H2("87%") ], className="kpi-card") ]), dcc.Graph( figure=px.line(df, x='date', y='admissions', title='Daily Admissions') ) ]) if __name__ == '__main__': app.run_server(debug=True) ``` ### 5. **Flexdashboard** (R Markdown, FREE) ⭐ **Pros:** - Static or dynamic dashboards - Easy to create from R Markdown - Beautiful layouts - Can host for free **Example:** ````markdown --- title: "Hospital Dashboard" output: flexdashboard::flex_dashboard: orientation: rows vertical_layout: fill --- ```{r setup, include=FALSE} library(flexdashboard) library(tidyverse) data <- read_csv("hospital_data.csv") ``` Row {data-height=150} ----------------------------------------------------------------------- ### Total Patients ```{r} valueBox( value = nrow(data), icon = "fa-users", color = "primary" ) ``` ### Bed Occupancy ```{r} occupancy <- 87 valueBox( value = paste0(occupancy, "%"), icon = "fa-bed", color = ifelse(occupancy > 90, "danger", "success") ) ``` Row ----------------------------------------------------------------------- ### Daily Admissions ```{r} ggplot(data, aes(x = date, y = admissions)) + geom_line(size = 1.2, color = "#00539B") + theme_minimal() ``` ```` --- ## Real-World Dashboard Examples ### Example 1: COVID-19 Monitoring Dashboard **Purpose:** Track pandemic metrics for county health department **Key Components:** 1. **Hero Numbers (Top):** - Total Cases (with change from yesterday) - Active Cases - Total Deaths - Vaccination Rate 2. **Main Chart (Center):** - Daily new cases (7-day moving average) - Hospitalization trend 3. **Supporting Visuals:** - Cases by age group (bar chart) - Geographic distribution (map) - Testing positivity rate (gauge) 4. **Table (Bottom):** - Recent cases by facility **Color Scheme:** - Cases: Blue (#00539B) - Deaths: Dark gray (#4A4A4A) - Vaccinations: Green (#009639) - Warnings: Orange (#FFA500) ### Example 2: Hospital Emergency Department Dashboard **Purpose:** Real-time ED performance monitoring **Auto-refresh:** Every 5 minutes **Key Metrics:** ```r # Real-time ED dashboard metrics metrics <- list( patients_waiting = sum(ed_data$status == "Waiting"), avg_wait_time = mean(ed_data$wait_time), patients_in_treatment = sum(ed_data$status == "In Treatment"), patients_admitted = sum(ed_data$disposition == "Admitted"), bed_availability = available_beds / total_beds * 100 ) # Color coding for wait times wait_color <- case_when( metrics$avg_wait_time < 15 ~ "green", metrics$avg_wait_time < 30 ~ "orange", TRUE ~ "red" ) ``` **Visualizations:** 1. Patient flow (Sankey diagram) 2. Wait times by triage level (grouped bar) 3. Hourly arrival pattern (area chart) 4. Staff utilization (gauge charts) ### Example 3: Public Health Surveillance Dashboard **Purpose:** Disease outbreak detection **Update Frequency:** Daily **Key Features:** ```r # Outbreak detection algorithm detect_outbreak <- function(current_cases, baseline) { threshold <- baseline + 2 * sd(baseline) alert <- current_cases > threshold return(alert) } # Visualize with threshold line ggplot(disease_data, aes(x = date, y = cases)) + geom_line(size = 1.2) + geom_hline(yintercept = outbreak_threshold, color = "red", linetype = "dashed") + geom_point(data = filter(disease_data, alert == TRUE), color = "red", size = 3) + annotate("text", x = max(disease_data$date), y = outbreak_threshold, label = "Outbreak Threshold", vjust = -0.5, color = "red") ``` **Components:** 1. Epidemic curve 2. Geographic hotspots (map) 3. Case demographics (population pyramid) 4. Alert system (conditional formatting) --- ## Interactive Features to Include ### 1. **Filters** 🔍 ```r # Shiny filters example selectInput("facility", "Select Facility:", choices = unique(data$facility), multiple = TRUE) dateRangeInput("dates", "Date Range:", start = Sys.Date() - 30, end = Sys.Date()) ``` ### 2. **Drill-Down** 📊 Allow users to click for details: - County → District → Facility → Ward ### 3. **Tooltips** 💬 ```r # ggplot tooltips with plotly library(plotly) p <- ggplot(data, aes(x = date, y = cases, text = paste("Date:", date, "<br>Cases:", cases))) + geom_line() ggplotly(p, tooltip = "text") ``` ### 4. **Export Options** 📥 ```r # Download button downloadButton("download_data", "Download Data") # Server output$download_data <- downloadHandler( filename = function() { paste("health_data_", Sys.Date(), ".csv", sep = "") }, content = function(file) { write_csv(filtered_data(), file) } ) ``` --- ## Performance Optimization ### 1. **Data Aggregation** ```r # Pre-aggregate data daily_summary <- raw_data %>% group_by(date, facility) %>% summarize( total_patients = n(), avg_wait = mean(wait_time), max_wait = max(wait_time) ) # Use summary instead of raw data ``` ### 2. **Caching** ```r # Cache expensive computations predictions <- reactive({ req(input$date_range) # Cache for 1 hour cached_result <- memoise::memoise( predict_admissions(data(), input$date_range), cache = cachem::cache_mem(max_age = 3600) ) }) ``` ### 3. **Lazy Loading** ```r # Load data only when tab is opened output$detailed_table <- renderDT({ req(input$tabs == "details") # Only load if on this tab datatable(detailed_data()) }) ``` --- ## Accessibility Checklist ✅ **Color:** - [ ] Not relying on color alone to convey information - [ ] Colorblind-friendly palette - [ ] Sufficient contrast (4.5:1 minimum) ✅ **Text:** - [ ] Minimum font size 12pt - [ ] Clear, readable fonts - [ ] Alternative text for images ✅ **Navigation:** - [ ] Keyboard accessible - [ ] Screen reader compatible - [ ] Clear focus indicators ✅ **Content:** - [ ] Meaningful titles and labels - [ ] Data tables have headers - [ ] Tooltips provide context --- ## Common Mistakes to Avoid ### ❌ **Mistake 1: Too Much Information** **Problem:** 20 charts crammed on one screen **Solution:** Follow the "3-Click Rule" - Key metrics immediately visible - Details available in 1-2 clicks - Use tabs or pages for deep dives ### ❌ **Mistake 2: Misleading Visualizations** **Problems:** - Y-axis doesn't start at zero - Inconsistent scales - Cherry-picked date ranges **Solution:** - Always start bar charts at zero - Use consistent scales for comparisons - Show full context ### ❌ **Mistake 3: No Context** **Problem:** "87%" displayed with no interpretation **Solution:** Add: - Target values or benchmarks - Trend indicators (↑↓) - Historical comparison - Reference ranges ```r valueBox( value = "87%", caption = "Bed Occupancy (Target: 85-90%)", icon = "fa-bed", color = "success" ) ``` ### ❌ **Mistake 4: Static Dashboard** **Problem:** Dashboard never updated or maintained **Solution:** - Automated data refresh - Version control - Regular user feedback - Documented update schedule --- ## Testing Your Dashboard ### User Testing Checklist **Test with actual users:** 1. **Comprehension Test** - Can users identify the main message in 5 seconds? - Do they understand what each metric means? 2. **Task Completion** - Can they find specific information? - Can they filter and interact effectively? 3. **Decision Making** - Does the dashboard help them make decisions? - What additional information do they need? 4. **Performance** - Does it load quickly? - Are interactions responsive? ### A/B Testing ```r # Track which dashboard version performs better log_interaction <- function(user_id, version, action, timestamp) { write_csv( data.frame(user_id, version, action, timestamp), "dashboard_analytics.csv", append = TRUE ) } # Analyze results analytics %>% group_by(version) %>% summarize( avg_time_on_page = mean(time_on_page), click_through_rate = sum(clicked) / n() ) ``` --- ## Deployment Options ### 1. **Shinyapps.io** (Easiest for R) ```r # Install rsconnect install.packages("rsconnect") # Configure account rsconnect::setAccountInfo( name = "your-account", token = "your-token", secret = "your-secret" ) # Deploy rsconnect::deployApp("path/to/dashboard") ``` **Free tier:** 5 apps, 25 active hours/month ### 2. **RStudio Connect** (Enterprise) - Internal hosting - Authentication - Scheduled updates - Email reports ### 3. **Docker + Cloud** (Most flexible) ```dockerfile # Dockerfile FROM rocker/shiny:latest # Install R packages RUN R -e "install.packages(c('shiny', 'tidyverse', 'plotly'))" # Copy app COPY app.R /srv/shiny-server/ # Expose port EXPOSE 3838 # Run CMD ["/usr/bin/shiny-server"] ``` ### 4. **GitHub Pages** (Static dashboards only) ```bash # Build flexdashboard Rscript -e "rmarkdown::render('dashboard.Rmd')" # Push to gh-pages branch git add dashboard.html git commit -m "Update dashboard" git push origin gh-pages ``` --- ## Maintenance and Updates ### Regular Reviews **Monthly:** - [ ] Check data accuracy - [ ] Review user feedback - [ ] Update metrics if needed - [ ] Test all interactive features **Quarterly:** - [ ] Usability testing - [ ] Performance optimization - [ ] Design refresh if needed - [ ] Documentation update **Annually:** - [ ] Complete redesign review - [ ] Technology stack evaluation - [ ] User needs assessment --- ## Resources ### Learning 1. **[Information is Beautiful](https://informationisbeautiful.net/)** - Inspiration 2. **[Flowing Data](https://flowingdata.com/)** - Tutorials and examples 3. **[Data Visualization Catalogue](https://datavizcatalogue.com/)** - Chart selection 4. **[Chartio Data School](https://dataschool.com/)** - Best practices ### Tools 1. **[Color Brewer](https://colorbrewer2.org/)** - Color palettes 2. **[Coolors](https://coolors.co/)** - Color scheme generator 3. **[FontPair](https://fontpair.co/)** - Font combinations 4. **[Figma](https://figma.com/)** - Dashboard prototyping ### Communities 1. **[RStudio Community](https://community.rstudio.com/)** 2. **[Data Visualization Society](https://www.datavisualizationsociety.org/)** 3. **[Tableau Community](https://community.tableau.com/)** --- ## Conclusion Creating effective health dashboards is both an art and a science. The best dashboards: ✅ **Focus on what matters** - Show key metrics prominently ✅ **Tell a story** - Guide users through data ✅ **Enable action** - Provide insights that drive decisions ✅ **Are accessible** - Work for all users ✅ **Stay current** - Update automatically **Remember:** The goal is not to display all available data, but to provide **actionable insights** that improve health outcomes. **Start small, iterate based on feedback, and always keep your users' needs first.** --- **Related Posts:** - [Why Reproducible Research Matters in Public Health](#) - [A Beginner's Guide to R for Health Researchers](#) - [Dashboard Design Principles](../19-dashboard-design-principles/) **Tags:** #DataVisualization #Dashboards #HealthAnalytics #RShiny #Tableau #BestPractices --- *What challenges have you faced creating health dashboards? Share your experiences below!*