This project was completed as part of the AWS AI & ML Scholars Program using AWS PartyRock (Whiskers AI).
The objective was to explore a Multiple Cause of Death (MCOD) dataset from 2022 and use generative AI to identify mortality trends, leading causes of death, and the relative impact of disease-related and injury-related mortality.
The project demonstrates how Large Language Models (LLMs) can assist with exploratory data analysis by generating insights from structured datasets through natural-language prompts.
- AWS PartyRock (Whiskers AI)
- Generative AI Prompting
- GitHub
- CSV Dataset Analysis
Future enhancements include:
- Python
- Pandas
- Matplotlib
- Jupyter Notebooks
Dataset: Fact_MCOD2022_Causes_Info_Horizontal.csv
The dataset contains over 500,000 mortality records and includes:
- Underlying Cause of Death
- Contributing Causes
- Injury-related Causes
- ICD-10 Cause Groupings
The full dataset is not included in this repository due to file size limitations.
The following prompts were used in AWS PartyRock:
- What are the top 10 causes of death?
- What are the least common causes of death?
- How important are injuries as a cause of mortality compared with diseases?
- List 3 to 5 insights discovered from the data.
| ICD-10 Code | Death Count |
|---|---|
| R99 | 67,581 |
| E14 | 29,880 |
| I64 | 24,855 |
| X59 | 24,162 |
| I10 | 21,892 |
| J18 | 19,338 |
| I50 | 15,762 |
| A16 | 15,681 |
| B33 | 13,394 |
| B20 | 12,042 |
| Category | Deaths | Percentage |
|---|---|---|
| Disease / Natural Causes | 449,021 | 87.16% |
| Injury / External Causes | 66,156 | 12.84% |
R99 (Other ill-defined and unspecified causes of mortality) was the leading cause of death.
Diabetes, hypertension, stroke, and heart failure were among the leading causes of death.
Tuberculosis, HIV, and pneumonia remained prominent contributors to mortality.
Nearly 87% of deaths were attributed to disease-related causes compared to approximately 13% caused by injuries.
Several causes appeared only once in the dataset.
- No age information available.
- No sex or gender information available.
- No geographic information available.
- High prevalence of R99 classifications may affect interpretation.
- Findings are based on AI-assisted exploratory analysis and should be independently validated.
AWS PartyRock enabled rapid exploration of a large mortality dataset through natural-language prompts. The exercise demonstrated how generative AI can assist analysts in identifying trends and generating hypotheses for further investigation.
Future work will focus on validating these findings using Python-based data analysis techniques.
Completed as part of the AWS AI & ML Scholars Program using AWS PartyRock.