← All projects
Healthcare Analytics & AI · Project Portfolio

🩺 Real health data, made explorable

Seven projects on public-health and hospital data from the CDC, NCHS, California's HCAI and SAS. Each starts from a raw dataset, is cleaned and analysed in Tableau, SAS Viya or Excel, checked number by number against the source, and ends in a model you can play with.

📥 Extract→🧹 Clean→ 📊 Analyse→✅ Verify→▶ Explore
Interactive models
7
Excel · Tableau · SAS · Python
Real datasets
6
CDC · NCHS · HCAI · GHI · SAS
Records analysed
~238K
largest file: 12,477 columns
Figures verified
43 / 44
recomputed from the source data
Preview of the Hospital Profit Simulator model

💰 Hospital Profit Simulator

An Excel decision-support model of hospital profit across five states. Ohio misses its target, so three plans are built to close the gap.

Ohio lifted from $479K to $959K net income; three different plans clear the $850K target
ExcelScenario modelling
▶ Interactive modelFindings ↓
Preview of the Causes of Death, 1999–2016 model

🩺 Causes of Death, 1999–2016

Ten leading causes re-rank year by year for the U.S. or any state, by age-adjusted rate or by deaths.

Heart disease −38%, Alzheimer's +84%; heart disease predicts overall mortality (r = 0.83)
TableauNCHS · 10,296 rows
▶ Interactive modelFindings ↓
Preview of the U.S. Adult Obesity model

⚖️ U.S. Adult Obesity

Hypothesis-driven analysis of a 110,880-row CDC survey: an animated state map, inactivity vs obesity, and the national trend.

Obesity 27.4% to 33.1%; the education gap (9.4 pts) is three times the income gap
TableauPythonCDC BRFSS
▶ Interactive modelFindings ↓
Preview of the Who Buys Organic? model

🛒 Who Buys Organic?

The real SAS decision tree on 111,115 loyalty customers, with shoppers flowing through it and a customer builder.

Best segment buys at 91.7%, the largest at just 11%; age and affluence drive it
SAS ViyaDecision treeClustering
▶ Interactive modelFindings ↓
Preview of the 444 Reports, 432 Hospitals model

🏥 444 Reports, 432 Hospitals

California hospital finances with a guided tour of executive questions and an animation that merges duplicate filings.

112 hospitals at a loss and 225 non-profits, once duplicate filings are caught
PythonExcelData integrity
▶ Interactive modelFindings ↓
Preview of the Heart Risk Tree model

❤️ Heart Risk Tree

Walk a patient down the risk tree fitted on 5,209 Framingham patients, plus three patient profiles from k-means.

Heart-disease share of deaths climbs 7% to 56% down the riskiest path; heavy smokers die about 10 years earlier
SAS Viyak-meansDecision tree
▶ Interactive modelFindings ↓
Preview of the Hunger Over Time model

🌍 Hunger Over Time

All 130 countries in the Global Hunger Index animated from 1990 to 2014, Gapminder-style.

Burundi's undernourishment rose from 44% to 67% while most of the world improved
Tableau130 countries
▶ Interactive modelFindings ↓

✅ Checked against the source

Every headline number was recomputed from the raw data files, not copied from the reports. A few examples:

ClaimSourceRecomputed
Heart disease vs all-cause mortality across statesNCHS, 2016r = 0.834 ✓
Alzheimer's age-adjusted death rate, 1999 → 2016NCHS+83.6% ✓
Physical inactivity vs obesity across 49 statesCDC BRFSS, 2023r = 0.70 ✓
Hospital reports → unique facilitiesHCAI disclosure file444 → 432 ✓
Best BigOrganics segment buy rateSAS decision-tree export2,390 / 2,605 = 91.7% ✓
Ohio Plan B net incomeExcel DSS model$959,145.04 ✓

Findings in depth

The charts and screenshots behind each project. Hover any chart for exact values.

💰 Hospital Profit Decision Support System

ExcelScenario modelling

An Excel model of hospital revenue, expenses and net income for five states. Ohio fell far short of an $850K net-income target, so three scenarios were built to close the gap. Every number is a live formula.

Net income by state

Base model, per 1,000 population

Getting Ohio over the line

Net income by scenario vs the $850K target
▶ Interactive model

🌍 Global Hunger Index Dashboard

Tableau2014 GHI

Four indicators behind the 2014 Global Hunger Index, compared across five countries. The same countries appear in the same order in every panel, so you can read across.

Tableau dashboard Hunger Index Undernourished Under-5 mortality
▶ Interactive model

🩺 Leading Causes of Death, 1999–2016

TableauNCHS · 10,296 rowsAge-adjusted rates

Five findings from the NCHS mortality file. All rates are age-adjusted deaths per 100,000, so states with older populations don't distort the comparison.

Opposite trends since 1999

U.S. age-adjusted death rate per 100,000

Leading causes, 2016

U.S. age-adjusted death rate per 100,000

Heart disease predicts overall mortality

Each dot is a state, 2016 ·

Suicide rate by state, 2016

Age-adjusted per 100,000 · top 10 and bottom 5 of 51 (all in the table)
Trend lines Suicide by state Scatter + trend line Cause share pie
▶ Interactive model

⚖️ U.S. Adult Obesity Analysis

TableauPython cleaningCDC BRFSS · 97,666 rows

Analysis driven by written hypotheses. A 110,880-row CDC survey file was cleaned down to 97,666 usable rows: suppressed values dropped, coordinates split, census regions added. Each hypothesis was then tested in Tableau.

Obesity up, overweight flat

U.S. adults, % · 2011–2024

Inactivity tracks obesity

Each dot is a state, 2023 ·

Education gap: 9.4 points

Adult obesity %, 2023, by education

Income gap: only 3.5 points

Adult obesity %, 2023, by household income
Summary dashboard State map Inactivity scatter By education
▶ Interactive model

🛒 BigOrganics Customer Analytics

SAS Visual Analytics111K customers

A full SAS Viya workflow: profile the customer base, impute missing values with documented rules, then segment customers and model who buys organic products.

1 · ProfileAge groups, bubble plot of loyalty × affluence, geo maps of frequency and spend
2 · Impute8 _IMP variables, e.g. Age → 53, Affluence → 8, Gender → U
3 · CompareBox plot of affluence and age by purchase target
4 · Cluster5 segments on affluence, age, tenure, purchase count, spend
5 · Decision tree10 predictors, leaf size 20, KS = 0.4707
Bubble plot Spend by region Box plot Clustering Decision tree
▶ Interactive model

🏥 California Hospital Financial Analysis

Python · pandasExcel live formulas444 reports × 12,477 columns

Eight executive questions answered from California's HCAI disclosure data. Answers were computed twice, in pandas and as live Excel formulas, and the two matched. The hard part was the data-integrity traps: 444 filed reports describe only 432 hospitals.

Highest debt ratio
6.05
Newport Bay Hospital
Hospitals at a loss
112
unique facilities (115 reports)
Top Medicare + Medi-Cal mix
99.76%
Newport Bay, near-tie with Laguna Honda
Watsonville charity care
2.62%
inside the 1–5% target band
Non-profit hospitals
225
not 233: that counts reports
Sutter Amador COVID stimulus
$13.3M
line P14_C1_L205
Facilities that filed twice
11
fiscal-year changes, not errors
Reports with no balance sheet
46
excluded, never treated as zero
New ratio column (formula) All answers as formulas Debt ratio ranking Repeat filers
▶ Interactive model

❤️ Framingham Heart Risk Analysis

SAS Visual Statisticsk-meansDecision tree5,209 patients

k-means clustering (k = 3) finds three cardiovascular risk profiles. A decision tree then traces how risk factors compound toward death from coronary heart disease (CHD).

Risk factors compound

Share of deaths from CHD along the riskiest branch of the tree

What the clusters say

  • Smoking costs the most life: heavy smokers (25 cig/day) die about 10.4 years earlier than the low-risk group.
  • High BP and cholesterol (183 mmHg, 259 mg/dL) cost about 1.6 years.
  • Blood pressure is the top predictor (importance 94.7). The tree splits on it first, and high BP alone more than doubles the CHD share.
  • The last leaf holds only 25 patients, so read its 56% as indicative, not precise.
Cluster output Decision tree Heavy-smoker cluster
▶ Interactive model