Analysis of Emergency Deparment Visits and Admissions in New York City

Outline

• Data Overview
• Key Focus areas
• Exploration of Focus areas
• Future Scope
• Resources

Data Overview

The NYC dataset includes following variables:

• Extract Date: The date on which records are updated
• Date: The date when patient visited the emergency department
• Zip Code: Five digit ZCTA representing patient residence
• Total number of emergency department visits for any reason
• Total number of emergency department visits for influenza-like illness
• Total number of admissions for influenza-like illness

Note:
Influenza-like illness: i. Fever & cough ii. Fever & Sore throat iii. Mention of Influenza

Data Overview

• Time Period: 1st March-1st December 2020
• ~2M Emergency Department visits
• ~143K Emergency Department visits for Influenza like Illness
• ~47K, i.e. one-third of visits for Influenza like Illness resulted in the admission of the patients
  

Key Focus Areas

1. Which regions in New York city are experiencing high emergency department admissions?
2. What socio-economic factors could be responsible for the observed trends?
3. Can we help the hospitals to plan capacities as per future requirements?

1. Which regions in New York city are experiencing high emergency department admissions?

Admission Percentage for Influenza-like Illness across different zips

• Zip 11354 experienced highest proportion (66%) of visits resulting in admissions for influenza like illness
• Zip 12703 experienced lowest proportion (14%) of visits resulting in admissions for influenza like illness

Admission Percentage for Influenza-like Illness across different Boroughs

• On average, Bronx experienced lowest proportion (28%) of visits resulting in admissions for influenza like illness
• On average, Staten Island experienced highest proportion (35%) of visits resulting in admissions for influenza like illness

2. What socio-economic factors could be responsible for the observed trends?

I. Race and Median Household Income

NYC's population distribution by Race and Median Household Income

• Bronx includes ~56% of Hispanic population having a median household income of 30-40K and experienced lowest percentage of admissions for Influenza like illness
• Staten Island includes ~62% of White population having a median household income of 65K+ and experienced highest percentage of admissions for Influenza like illness

Note:

1. Size of bubble represents the percentage of one ethnic group's population present in a particular borough
2. The US uses the ethnonyms Hispanic or Latino to refer to a person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin regardless of race and states that Hispanics or Latinos can be of any race, any ancestry, any ethnicity

II. Age Distribution

NYC's population distribution by Age

• For Bronx and Staten Island, the median age of their population could affect their percentage of admissions for influenza like illness
• For Brooklyn and Queens, their similar proportion of population across different age groups could explain their similar admission percentage for influenza like illness

Note:

Size of bubble represents the percentage of one age group's population present in a particular borough

3. Can we help the hospitals to plan capacities as per future requirements?

Hospital Admission Counts from 1st Mar- 1st Dec'2020

Forecasted Admission count for influenza like illness in Bronx

• Actual count of admission in Bronx was 755 vs 718 predicted admissions by the model for 2nd December until 27th December
• Actual values differed from the predicted values by RMSE of 8.05

Forecasted Admission count for influenza like illness in Manhattan

• Actual count of admission in Manhattan was 669 vs 624 predicted admissions by the model for 2nd December until 27th December
• Actual values differed from the predicted values by RMSE of 7.04

Forecasted Admission count for influenza like illness in Staten Island

• Actual count of admission in Staten Island was 553 vs 647 predicted admissions by the model for 2nd December until 27th December
• Actual values differed from the predicted values by RMSE of 8.04

Forecasted Admission count for influenza like illness in Queens

• Actual count of admission in Queens was 1,361 vs 1,521 predicted admissions by the model for 2nd December until 27th December
• Actual values differed from the predicted values by RMSE of 10.14

Forecasted Admission count for influenza like illness in Brooklyn

• Actual count of admission in Brooklyn was 1,480 vs 1,197 predicted admissions by the model for 2nd December until 27th December
• Actual values differed from the predicted values by RMSE of 10.04

Future Scope

1. Research hospital bed availability in each borough
2. Improve predicted patient admission count using deep learning
3. Predict patient admission count at hospital level

Resources

1. For household income by race https://www.health.ny.gov/statistics/community/minority/county/
2. For Age groups in each borough https://censusreporter.org/
3. For New York ED data https://data.cityofnewyork.us/Health/Emergency-Department-Visits-and-Admissions-for-Inf/2nwg-uqyg

Thank you!

Questions?