Monitoring the Health Care Safety Net

Slide Presentation by John Billings

On September 25, 2003, John Billings made a presentation in the Web-Assisted Audioconference entitled Using Administrative Data To Monitor Access, Identify Disparities, and Assess Performance of the Safety Net.

This is the text version of Mr. Billings' slide presentation. Select to access the PowerPoint® slides (1.8 MB).

Using Administrative Data To Monitor Access, Identify Disparities, and Assess Performance of the Safety Net

John Billings
NYU Center for Health and Public Service Research
September, 2003

Slide 1

What Are "Administrative Data"

• Computerized records
• Gathered for some "administrative" purpose
  • Bill paying/reimbursement
  • Record keeping
• Typically containing information about individuals
  • Demographics
  • Utilization of services
  • Other events (birth, death, etc.)

Slide 2

Some Examples of "Administrative Data"

• o Birth/death records
• o Hospital admission/discharge abstracts
• o Emergency department billing records
• o Medicare and Medicaid claims files

Slide 3

Advantages of "Administrative Data"

• o They're already there
• o They're electronic [computerized]
• o They can be relatively inexpensive to analyze [sometimes]
• o They can tell you a lot about what is going on [sometimes]

Slide 4

Disadvantages of Administrative Data

• They can be "dirty" (caution is required)
  • Some data elements are a lot better than others
  • A good test is whether anyone will go to jail for bad data, or there is some other good reason to get it right
• They seldom tell the whole story (often raising more pesky questions)
• Not everyone is willing to share (which may be required)
• You're probably going to be dealing with bureaucrats not particularly interested in being helpful

Slide 5

Using Birth Records to Monitor Birth "Outcomes"

Slide 6

Percent Late/No Prenatal Care, New York City, 1997-8

Scatter plot depicting percentage of children receiving late or no prenatal care (Y-axis ranging from 0-18%) to the percent of the population below poverty (X-axis ranging from 0-35 %) Data points are zip code areas. Correlation is measured in r2 as .435. Source: NYU Center for Health and Public Service Research.

Slide 7

Map of New York City measuring the percentage of black mothers receiving late or no prenatal care in 1997 and 1998 on the same scale of scatter plot on slide 6.

Slide 8

Using Hospital Discharge Data

Preventable/Avoidable Hospitalizations
Ambulatory Care Sensitive (ACS) Conditions

ACS Conditions - Where timely and effective ambulatory care help prevent the need for hospitalization

• Chronic conditions - Effective care can prevent flare-ups (asthma, diabetes, congestive heart disease, etc.)
• Acute conditions - Early intervention can prevent more serious progression (ENT infections, cellulitis, pneumonia, etc.)
• Preventable conditions - Immunization preventable illness

Slide 9

ACS Admissions/1,000

By Zip Code Area Income
Baltimore - Age 0-17 - 1999

Scatter plot by zip code, Y-axis is numbers of admissions per 1000 people, X-axis is percent of households with incomes of less than $15,000.

R2 = .595
Low Inc/Hi Inc = 2.24
Mean Rate = 9.53

Source: NYU Center for Health and Public Service Research

Slide 10

ACS Admissions/1,000

By Zip Code Area Income
Baltimore - Age 40-64 - 1999

Scatter plot by zip code, Y-axis is numbers of admissions per 1000 people, X-axis is percent of households with incomes of less than $15,000.

R2 = .893
Low Inc/Hi Inc = 4.08
Mean Rate = 26.69

Source: NYU Center for Health and Public Service Research

Slide 11

Map of Atlanta Metro Area, including most of State of Georgia. Lines drawn by county boundary.

ACS Admissions/1,000
Age 40-64 - 1999

Source: AHRQ/HCUP - NYU Center for Health and Public Service Research

Slide 12

Map of Atlanta Metro Area in greater detail, this time showing smaller pockets of higher rates of ACS admissions within the city and outlying regions than the first map.

ACS Admissions/1,000
Age 40-64 - 1999

Source: AHRQ/HCUP - NYU Center for Health and Public Service Research

Slide 13

Scatter plot: ACS Admissions/1,000 By Zip Code Area Income
New York City - Age 18-64 - 2000

Arrows drawn to two data points with very different rates of ACS admission, but very similar rates of poverty.

Source: NYU Center for Health and Public Service Research

Slide 14

Using Emergency Department Data to Monitor the Safety Net

NYU ED Classification Algorithm

Flowchart: If not emergent, no decision to be made. If emergent, determine of ED care is required, or if primary care can treat the condition. If ED is required, then it should be assessed whether it was preventable or avoidable, or not.

Source: NYU Center for Health and Public Service Research

Slide 15

Preventable/Avoidable ED Use/1,000

By Zip Code Area Income
Baltimore - Age 18-64 - 2000

Scatter plot: R2 = .783
Low Inc/Hi Inc = 3.77
Mean Rate = 80

Source: AHRQ/HCUP - NYU Center for Health and Public Service Research

Slide 16

Preventable/Avoidable ED Use/1,000

By Zip Code Area Income
Map of Austin Metro Area - Age 0-17 - 2000.

Higher rates to the north and East of the city center.

Source: NYU Center for Health and Public Service Research

Slide 17

Some Cautions for Using Administrative Data

• The data can be "dirty" (see above)
• If a number is way high or way low, it's probably wrong
  • Unless it's not
  • (Some disparities are huge)
• Don't expect final answers
• Avoid the easy explanation - this stuff is complex

Current as of February 2004

Internet Citation:

Using Administrative Data To Monitor Access, Identify Disparities, and Assess Performance of the Safety Net. Text Version of a Slide Presentation at a Web-assisted Audioconference. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/news/ulp/safetynetaud/sess3/billingstxt.htm

Return to Audioconference
State & Local Policymakers
AHRQ Home Page