VectorWatch 🦟📡

VectorWatch is an R Shiny malaria surveillance dashboard that visualizes county prevalence, detects LISA hotspot clusters, and shows trend + 3D profiles from weekly/monthly data extracts.

Live app: https://markchweya.shinyapps.io/VectorWatch/

---

Demo logins

• user / user123 (viewer)
• admin / admin123 (admin)

Note: these credentials are hardcoded for demo purposes.

---

What it does

• Secure login with roles (viewer/admin)
• Dataset folder picker (local device) + manual path input
• Coverage switching: high, medium, low
• Auto-detects extract frequency
  • Weekly files (W01..W52) OR monthly files (01..12)
• Interactive tabs
  • Map: County prevalence heat map (Leaflet)
  • Hotzones: LISA clusters (High-High, Low-Low, High-Low, Low-High)
  • Trends: Positivity over time (Plotly)
  • 3D: County profile scatter (Mean Age vs Mean Hb vs Prevalence)
• Export
  • Download hotspots/cluster results as CSV

---

Dataset requirements (important)

Folder structure

VectorWatch expects a base dataset folder containing these subfolders:

<DATASET_FOLDER>/
  high/
  medium/
  low/

File naming conventions

Inside each coverage folder, files must match one of these patterns:

Weekly

vectorwatch_malaria_2024_W01_<coverage>.csv
vectorwatch_malaria_2024_W02_<coverage>.csv
...
vectorwatch_malaria_2024_W52_<coverage>.csv

Monthly

vectorwatch_malaria_2024_01_<coverage>.csv
vectorwatch_malaria_2024_02_<coverage>.csv
...
vectorwatch_malaria_2024_12_<coverage>.csv

Where <coverage> is one of: high, medium, low.

---

Expected columns (CSV schema)

Your CSVs should include (or be mappable to) the following fields:

Required

• county
• date (convertible to Date)
• parasite_detected (1 = positive, 0 = negative)
  • The app also tries to map common alternatives like positive, pos, result, malaria_positive.

Optional (enables extra insights)

• age (years)
• hemoglobin_g_dl (or hb, hgb, hemoglobin)
• wbc_cells_ul (or wbc, wbc_ul, white_blood_cells)

The app standardizes column names and attempts to auto-map common variations.

---

Spatial layer & hotspot analytics

• Kenya county boundaries are fetched via: geodata::gadm("KEN", level = 1)
• Data is joined to polygons using a normalized county key (handles punctuation/spacing differences).
• Hotzones are computed with:
  • Global Moran’s I (spatial autocorrelation)
  • Local Moran’s I (LISA) to label clusters:
    • High-High, Low-Low, High-Low, Low-High, Not significant

---

How it calculates key metrics

• Tests = total rows in selected files
• Cases = sum of parasite_detected == 1
• Prevalence (%) = 100 * cases / tests
• County summaries also include:
  • mean age, mean Hb, mean WBC (if available)
• Filters:
  • Minimum tests per county (min_n)
  • Cluster category filter (All or specific LISA cluster types)

---

Tech stack

Core packages used:

• shiny, bslib, htmltools, shinyjs
• dplyr, janitor, stringr, lubridate
• leaflet, plotly
• sf, spdep, geodata
• digest (password hashing)
• shinyFiles (folder picker)

---

Run locally

1) Install dependencies

install.packages(c(
  "shiny","bslib","htmltools","dplyr","janitor","stringr","lubridate",
  "digest","leaflet","plotly","shinyFiles","shinyjs",
  "sf","spdep","geodata"
), dependencies = TRUE)

2) Run the app

From the folder containing your app file:

shiny::runApp()

Tip: If you run into sf / spatial library install issues, install system dependencies for your OS first (GDAL/GEOS/PROJ), then reinstall sf.

---

Security note (important)

This project includes demo credentials for convenience. If you publish the source code:

• Remove hardcoded users/passwords
• Use environment variables or an external authentication provider
• Never commit real credentials to GitHub

---

Deployment

Deployed on shinyapps.io: https://markchweya.shinyapps.io/VectorWatch/

---

Author

Mark Chweya