Hospital Length of Stay (ALOS) Analysis

Improving Healthcare Outcomes with SEPTEE and Donabedian's Model

Goal

Inefficiencies and inequalities in patient care cost more than just billions of dollar annually, but lives. This is why hospitals and companies seek to improve their operations, and why this projects aims to address costly problems

Process

A key step was understanding healthcare quality, which guided the data used.
PostgreSQL queries group and analyze data, while Tableau visualizes the results

Insights

1. Hospice and Skilled Nursing Facility (SNF) encounters have highest ALOS
2. Regional differences in healthcare access and quality
3. Racial disparities in ALOS are inconclusive

Project Overview:

This project analyzes the Average Length of Stay (ALOS) by race, zip code, and encounter class in a hospital setting. The goal is to use SEPTEE (Safety, Efficiency, Patient-centeredness, Timeliness, Equity, Effectiveness) and Donabedian's model (structure, process, outcome) to identify areas for improvement in patient care and hospital efficiency. By visualizing this data through Tableau, actionable insights can be derived to address disparities and enhance patient care outcomes across different demographics and encounter classes.


Data Sources:

Synthea Artifical Healthcare Data prepared by Josh Matlock of Data Wizardry

Tools

  • PostgreSQL (data cleaning, entity resolution)
  • Tableau (dashboard)

Data Cleaning and Preparation

  • Checked for relevant NULLs: None
  • Learned about several metrics important in adminstrative healthcare data

Exploratory Data Analysis

  1. What is the Average Length of Stay (ALOS)?
  2. Are there disparities in ALOS by Race?
  3. What regional differences exist based on Zip Codes?
  4. How does Encounter Class (emergency vs. scheduled) affect ALOS?

Data Analysis


Under the hood of visualization is an associated query in which we're essentially grouping dependent variables and ordering ALOS from greatest to least


Sample SQL Queries:


-- ALOS by Encounter Class
SELECT encounterclass
,AVG(EXTRACT(EPOCH FROM (stop - start)) / 86400) AS alos
FROM public.encounters
GROUP BY encounterclass
ORDER BY alos DESC;
                                    


-- ALOS by Race
SELECT p.race
,AVG(EXTRACT(EPOCH FROM (stop - start)) / 86400) AS alos
FROM public.encounters e
JOIN public.patients p
ON e.patient = p.id
GROUP BY race
ORDER BY alos DESC;

Results

Average Length of Stay is roughly 4 Hours and 12 minutes.


The analysis identified regional disparities in ALOS, while racial divides were inconclusive.


Emergency encounters, as expected, had a higher ALOS compared to scheduled admissions, with Hospice and SNF care having the highest ALOS of nearly 22 and 20 days, respectively.

Recommendations

  1. Investigate possible racial and regional disparities (SEPTEE focus on equity).
  2. Reduce ALOS by improving care coordination and discharge planning, improving Process Measures.
  3. Increase efficiency and focus on patient-centered care by improving Structure Measures (staffing, training).

Limitations

Synthetic Data:
While the Synthea dataset is modelled from existing relationships, it is synthetic data

References

Tableau Dashboard with all of the visualizations

Here is an overview of Donabedian's Model of Healthcare Quality Measures

Six Domains of Healthcare Quality, originally developed in Crossing the Quality Chasm: A New Health System for the 21st Century from The Institute of Medicine (IOM)