Massively by HTML5 UP

Excel & Tableau

Preparing for Influenza Season

The project helped a medical staffing agency that provides temporary workers to clinics and hospitals on an as-needed basis. The analysis helped plan for influenza season, a time when additional staff are in high demand. The final results examined trends in influenza and how they can be used to proactively plan for staffing needs across the country. Here, I drawed statistical insights from joining two datasets with Excel and finalized an interim report containing the details of the analyses. In a second step, I visualized insights in a Tableau dashboard, which has been used for a final stakeholder presentation.



Objective:

Determine when to send staff, and how many, to each state to adequately treat vulnerable groups during influenza season.

Key Business Questions:

  • Staffing plan informing timing and spatial distribution of medical personnel throughout the US
  • Seasonality
  • Low-, medium-, or high-need states based on vulnerable population count

Find the Tableau dashboard here:

Tableau

The dataset

Population Data by Geography US Census Data: Population numbers per county (2009-2017), owned by US Census Bureau

Population Dataset

Influenza deaths provided by Centers for Disease Control and Prevention (CDC) (2009-2017)

Influenza Dataset

Procedures:

  • Excel:
  • Data preparation
  • Grouping, summarizing data
  • Descriptive analysis
  • Forecasting
  • Hypothesis testing
  • Tableau Dashboard

Analyses & Results:

Interim Report:

Interim_Report

Research Hypothesis:

If states have higher proportions of vulnerable groups (children under 5, people 65 years and above), then they have higher proportional flu death rates.

Hypothesis testing:

  • Median split for both age groups.
  • Group comparison: t-Test
    (two-sample assuming unequal variances, α = 0.05)

At an alpha of 0.05 or confidence level of 95 percent, proportionally higher flu death rates are due to states having higher proportions of inhabitants 65y and above.
At an alpha of 0.05 or confidence level of 95 percent, differences in flu death rate proportions are due to states having higher proportions of children < 5y.
Other than expected, states with higher proportions of children < 5y tend to have lower proportions of flu death rates.





Find the Tableau Dashboard visualization here:

Tableau Dashboard

Find the Vimeo video presentation here:

Vimeo Video Presentation

Future approach/ improvements with Python-coded advanced analytics algorithms:

Classification: Create various buckets for level of staffing required— low, medium, high. Predict which class each State/Province/County will fall into at a specific time.

Regression: Predict how many staff members are to be assigned to a State/Province/County at a specific time.

Clustering: Create entirely new groups from dataset based on distribution of data. Examples could include unvaccinated and dead from influenza, vaccinated and dead from influenza, well-staffed but major fatalities…etc.