Slide Presentation for SHEA 18th Annual Scientific Meeting  
 
Slide 1

Enhanced Detection of Staphylococcus aureus-related Hospitalizations Using Administrative Databases, United States-1999-2005

Jhung MA, Banerjee SN, Fridkin S, Tenover FC, McDonald LC Division of Healthcare Quality Promotion Centers for Disease Control and Prevention

Presented at the 18th annual scientific meeting of the Society for Healthcare Epidemiology of America on April 6, 2008

Slide 2
Background

Staphylococcus aureus National Estimates

Data Source
Administrative Data Other Data Sources
ICD-9-CM discharge diagnoses Sentinel surveillance of HAIs (NHSN)
Nationwide Inpatient Sample (NIS) Population based surveillance (ABCs)
National Hospital Discharge Survey (NHDS) Prevalence survey (APIC)

NOTE: Recently, several studies have estimated the national burden of Staphylococcus aureus infections using different data sources.
Some have used administrative data, or ICD-9-CM codes, and examples of these are the Nationwide Inpatient Sample or NIS and the National Hospital Discharge Survey, or NHDS.
Others have used sentinel surveillance for healthcare associated infections, population-based surveillance, and prevalence surveys.

Slide 3
Background

Staphylococcus aureus National Estimates

Administrative Data

  • Relatively available, reliable and inexpensive
  • May be inconsistent and lack data elements
  • Variable sensitivity and positive predictive value

NOTE: Administrative databases are attractive for estimating the national burden of infections because they are relatively available, reliable, and inexpensive. However, their data may be inconsistent across providers, lack elements important for surveillance, and have highly variable sensitivity and positive predictive value.

Slide 4
Background

Staphylococcus aureus National Estimates

  • Common features of recently published estimates using administrative data
    • Varied in database, methodology, results Few ICD-9-CM codes
    • Single estimate for each infection type
    • Comparison to other estimates lacking
NOTE: Some commonalities of recently published estimates using administrative data are that they have varied in database, methodology, and results. They often use few ICD-9-CM codes, provide a single estimate for each infection type, and make no comparison to other estimates

Slide 5
Objectives

  • Determine national estimates of S. aureus-related hospitalizations
  • Pilot enhanced detection algorithm
  • Establish a range of estimates
  • Explore reasons for incidence trends
  • Compare to other analyses

NOTE: The objective of this study was to determine a national estimate for Staphylococcus aureus infections in the US, and in the process, attempt to identify an enhanced detection scheme that would define a range of estimates for each infection type, and provide enough granularity to explore reasons for any trends in incidence observed. We also sought to compare results from our study to other published analyses.

Slide 6
Methods

  • Nationwide Inpatient Sample (NIS) dataset
  • The Surveillance Network (TSN) for resistance
  • Categories of S. aureus infections
    • Bloodstream infections (septicemia)
    • Respiratory infections (pneumonia)
    • Skin infections
    • "All other"
  • Conservative, Moderate, Liberal estimates
  • "Case definitions" include healthcare and community associated disease

NOTE: We used the Nationwide Inpatient Sample (NIS) for ICD-9 data from 1999-2005 and obtained resistance data from published estimates from The Surveillance Network (TSN).

We defined 3 categories of Staphylococcus aureus infections
Bloodstream infections, which we labeled Septicemia,
Respiratory infections, which we labeled Pneumonia, and
Skin infections.

This resulted in a 4th category of remaining suspected Staphylococcus aureus infections, which we designated, an "all other" category
We then identified combinations of ICD-9-CM discharge codes to derive three levels of estimates: conservative, moderate, and liberal.

Note that the case definitions we used included both healthcare and community-associated disease.

Slide 7
Methods
ICD-9-CM Diagnosis Codes

S. aureus Specific Codes

041.11 Staphylococcus aureus
482.41 Staphylococcus aureus pneumonia
038.11 Staphylococcus aureus septicemia

Non-S. aureus Specific Codes

Respiratory Infection Codes
482.40 Staphylococcus pneumonia, unspecified
486 Pneumonia
510 Empyema
513 Abscess of lung and mediastinum

Bloodstream Infection Codes
038.1 Staphylococcus septicemia
038.9 Unspecified septicemia
790.7 Bacteremia

Skin Infection Codes
680 Carbuncle and furuncle
681 Cellulitis and abscess of finger and toe
682 Other cellulitis and abscess
684 Impetigo
686.9 Unspecified infection of skin
704.8 Other specified diseases of hair and hair follicles
707 Chronic ulcer of skin
998.3 Disruption of operation wound
998.5 Post-operative infection
998.83 Non-healing surgical wound

NOTE: The ICD-9-CM codes we used are shown here, separated into Staphylococcus aureus specific codes shown in the top box, and non-specific codes in the larger, bottom box.

Slide 8
ICD-9-CM Diagnosis Codes
Conservative Estimate

S. aureus Specific Codes

041.11 Staphylococcus aureus*  
482.41 Staphylococcus aureus pneumonia*
038.11 Staphylococcus aureus septicemia*

Non-S. aureus Specific Codes

Respiratory Infection Codes
482.40 Staphylococcus pneumonia, unspecified
486 Pneumonia
510 Empyema
513 Abscess of lung and mediastinum

Bloodstream Infection Codes
038.1 Staphylococcus septicemia
038.9 Unspecified septicemia
790.7 Bacteremia

Skin Infection Codes
680 Carbuncle and furuncle*
681 Cellulitis and abscess of finger and toe*
682 Other cellulitis and abscess*
684 Impetigo*
686.9 Unspecified infection of skin*
704.8 Other specified diseases of hair and hair follicles*
707 Chronic ulcer of skin*
998.3 Disruption of operation wound*
998.5 Post-operative infection*
998.83 Non-healing surgical wound*

* Skin infection (SA code only)

NOTE: For our conservative estimate, we used only specific codes for Staphylococcus aureus septicemia and pneumonia.
The conservative estimate of skin infections consisted of discharges containing the organism code for Staphylococcus aureus (041.11) plus one of the skin infection codes listed here and no additional organism codes.
Once placed into an infection category, discharges were removed from the database and were not therefore able to be counted again.

Slide 9
ICD-9-CM Diagnosis Codes
Moderate Estimate

S. aureus Specific Codes

041.11 Staphylococcus aureus*  
482.41 Staphylococcus aureus pneumonia
038.11 Staphylococcus aureus septicemia

Non-S. aureus Specific Codes

Respiratory Infection Codes
482.40 Staphylococcus pneumonia, unspecified*
486 Pneumonia*
510 Empyema*
513 Abscess of lung and mediastinum*

Bloodstream Infection Codes
038.1 Staphylococcus septicemia*
038.9 Unspecified septicemia *
790.7 Bacteremia*

Skin Infection Codes
680 Carbuncle and furuncle*
681 Cellulitis and abscess of finger and toe*
682 Other cellulitis and abscess*
684 Impetigo*
686.9 Unspecified infection of skin*
704.8 Other specified diseases of hair and hair follicles*
707 Chronic ulcer of skin*
998.3 Disruption of operation wound*
998.5 Post-operative infection*
998.83 Non-healing surgical wound*

* Septicemia (SA code only), Pneumonia (SA code only), Skin infection (other organisms)

NOTE: The moderate estimates for septicemia and pneumonia were obtained by starting with the conservative estimates and adding to those discharges with the Staphylococcus aureus organism code and one of the non-specific bloodstream or respiratory infection codes and no other organism code.
The moderate skin infection estimate contained the Staphylococcus aureus organism code and one of the skin infection codes. Other organism codes were allowed in this estimate.

Slide 10
ICD-9-CM Diagnosis Codes
Liberal Estimate

S. aureus Specific Codes

041.11 Staphylococcus aureus
482.41 Staphylococcus aureus pneumonia
038.11 Staphylococcus aureus septicemia

Non-S. aureus Specific Codes

Respiratory Infection Codes
482.40 Staphylococcus pneumonia, unspecified*
486 Pneumonia*
510 Empyema*
513 Abscess of lung and mediastinum*

Bloodstream Infection Codes
038.1 Staphylococcus septicemia*
038.9 Unspecified septicemia *
790.7 Bacteremia*

Skin Infection Codes
680 Carbuncle and furuncle*
681 Cellulitis and abscess of finger and toe*
682 Other cellulitis and abscess*
684 Impetigo*
686.9 Unspecified infection of skin*
704.8 Other specified diseases of hair and hair follicles*
707 Chronic ulcer of skin*
998.3 Disruption of operation wound*
998.5 Post-operative infection*
998.83 Non-healing surgical wound*

* Septicemia x (.20), Pneumonia x (.15), Skin infection x (.58)

NOTE: The liberal estimates for septicemia, pneumonia and skin infections were obtained by starting with the moderate estimates and adding to those discharges containing non-specific respiratory, bloodstream, and skin infection codes with no organism code multiplied by a proportion, estimated from the literature, of each infection type that could be reasonably attributed to Staphylococcus aureus.

Slide 11
S. aureus Infections Increase over Time
Line chart
S. aureus Infections by Type and Year (Conservative Estimate)

All infections increased over time period, Poisson regression p < .01

This slide shows rates of Staphylococcus aureus related discharges from 1999-2005.
Total rates, and rates for each infection type increased significantly over the time period.
2005 discharge counts for each category, show that by 2005, skin infections contributed substantially more to the total count than did septicemia or pneumonia.

Slide 12
S. aureus Infections Increase over Time
Line chart
S. aureus Infections by Type and Year (Conservative Estimate)

Skin infections made up nearly half of all Staphylococcus aureus related discharges in 2005.
Since they comprised such a large portion of all infections, we took a closer look to see what these skin infections were.

Slide 13
3 Major Skin Infection Classes

2004 Skin Infection Distribution (pie chart)

  • 2%
    • Carbuncle and Furuncle
    • Impetigo
    • Diseases of hair and follicles
    • Other local skin & subcutaneous infections
  • 26% Surgical Site Infection
  • 8% Chronic Ulcer
  • 64% Cellulitis and Abscess

NOTE: For each year in the time frame we studied, three major classes of infections comprised the vast majority of all discharges falling into the skin infection category.
Shown here are data from 2004, when chronic ulcers, surgical site infections, and cellulitis and abscesses made up 98% of all skin infections identified.

Slide 14
Cellulitis and Abscess Increasingly Important

S aureus skin infections by type and year (line chart)

Cochran-Armitage test for trend p < .01 for all infection types

Looking at these three classes of skin infections over time, we found significant changes in the proportion each contributed to the overall skin infection category.

This graph shows the proportion of all Staphylococcus aureus skin infections attributed to the three main classes identified on the previous slide.

Note that the proportion of chronic ulcers and SSIs decreased, and that of the cellulitis and abscess class increased significantly over time.

By 2005, 73% of all skin infections fell into the cellulitis and abscess class.

Slide 15
Cellulitis and abscess in Younger Age Groups

Proportion of S. aureus Cellulitis and Abscess by Age Group

Cochran-Armitage test for trend p < .01

Taking a closer look at cellulitis and abscess infections, we also found significant changes across age groups.

This graph contains information on the cellulitis and abscess class only and shows the proportion contributed to this class by each of 4 age groups.

We found that while the percent contribution by the two youngest age groups increased significantly, proportions for the two oldest age groups either stayed constant or decreased over time.

Slide 16
S. aureus Septicemia Increases from Conservative to Liberal Estimates

2005 S. aureus Septicemia Estimates (bar chart)

  Conservative 109,921
  Moderate 132,756
  Liberal 169,481 x % MRSA (TSN) = 98,299
                                                           ~ 94,360 ABCs estimate invasive MRSA

Klevens, et al. JAMA, October 2007.

We turn now to our estimates by Staphylococcus aureus infection category. As expected, liberal estimates were larger than moderate which were larger than conservative.

An example of this is shown here in a graph depicting these 3 levels for 2005 Staphylococcus aureus septicemia discharges. For septicemia, the moderate estimate was 21% greater than the conservative, and the liberal estimate was 54% greater than the conservative.

Notice that the liberal estimate of nearly 170,000, when multiplied by the overall percent of methicillin resistance obtained from TSN, corresponds well to the national estimate from ABC surveillance for invasive MRSA infections (Klevens, et al. JAMA, October 2007).

Slide 17
NIS Estimate Larger than NHDS Estimate

S. aureus Infections by Type and Year Conservative Estimates (line chart)

Klein, et al Emerg Infect Dis. December 2007.

The last result slide demonstrates a potential sensitivity benefit that using the NIS database might have.

This graph shows total Staphylococcus aureus related discharge numbers for both the Nationwide Inpatient Sample and the National Hospital Discharge Survey, each year from 1999 to 2005, the Nationwide Inpatient Sample estimate is substantially larger than the National Hospital Discharge Survey estimate.

By 2005, there is a 25% difference between the Nationwide Inpatient Sample and the National Hospital Discharge Survey estimates.

This may be due in part to the fact that the Nationwide Inpatient Sample database contains up to 15 discharge diagnosis codes while the National Hospital Discharge Survey is limited to seven.

Slide 18
Summary

  • Staphylococcus aureus-related discharges have increased significantly from 1999-2005
  • Estimates based on this algorithm are higher than other published estimates
  • Septicemia liberal estimates may be more accurate than conservative
  • Majority of increase in Staphylococcus aureus-related discharges is due to skin infections (cellulitis and abscess) in patients < 45 years of age (Community associated disease?)

In summary, we found that Staphylococcus aureus -related discharges increased significantly from 1999-2005 across all infection types.

Even our conservative estimates were higher than other published values that used administrative data sources.

Using ABC surveillance as a de facto gold standard, we found evidence that our liberal estimate, at least for septicemia, may be more accurate than the conservative value.

Finally, the bulk of the increase in Staphylococcus aureus -related discharges over the six year period appeared to be due to increases in skin infections, particularly cellulitis and abscesses, among patients less than 45 years of age.

This raises the possibility that the majority of the overall increase in Staphylococcus aureus -related discharges is to due to community-associated disease.

Slide 19
Limitations

  • Administrative data
    • Not primarily intended for surveillance
    • Unable to distinguish community from healthcare onset
    • Unit of analysis is discharge not patient
  • Analysis
    • Codes may not represent S. aureus

This study is subject to the following limitations:
First, administrative data are primarily intended to assist with hospital billing, not disease surveillance
Second, we were unable to determine community vs. healthcare acquisition of Staphylococcus aureus infections using these data
Third, the unit of analysis for these databases is a hospital discharge, not a patient and it is unclear how many unique patients are represented in these discharges

Finally, since Staphylococcus aureus organism codes are not linked directly to infection codes, the discharges identified by our algorithm may not represent true Staphylococcus aureus infections

Slide 20
Conclusions

Surveillance Using Administrative Data

  • Multiple codes and a range of estimates may yield the most useful results
  • A database with a large number of diagnosis fields may increase sensitivity
  • Further study needed to determine appropriate estimate for pneumonia and skin infections

Nonetheless, we may still be able to conclude the following about using administrative data to derive national estimates of Staphylococcus aureus infections

First an approach employing multiple codes to identify a range of estimates may yield the most accurate and useful results

Second, a database with a large number of diagnosis fields, perhaps greater than 15, may increase sensitivity

Finally, further study is needed to determine which estimate from the range we identified is the best measure of the burden of Staphylococcus aureus pneumonia and skin infections.

Slide 21
Acknowledgments

Division of Healthcare Quality Promotion
Centers for Disease Control and Prevention
Katherine Ellingson
Rachel Gorwitz
Jeffrey Hageman
John Jernigan
Melissa Schaefer

The findings and conclusions in this presentation are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention

Agency for Healthcare Quality and Research
Anne Elixhauser

Slide 22
Citations for Published Estimates

Elixhauser A. Infections with Methicillin-Resistant Stapyhlococcus aureus (MRSA) in US Hospitals, 1993-2005. In: AHRQ/HCUP Statistical Brief #35; 2007:1-10
Jarvis WR. National prevalence of methicillin-resistant Staphylococcus aureus in inpatients at US health care facilities, 2006. Am Jour of Infect Cont. 2007 Dec;35(10):631-7.
Klein E. Hospitalizations and deaths caused by methicillin-resistant Staphylococcus aureus, United States, 1999-2005. Emerg Infect Dis. 2007 Dec;13(12):1840-6.
Klevens RM. Invasive methicillin-resistant Staphylococcus aureus infections in the United States. JAMA. 2007 Oct 17;298(15):1763-71.
Kuehnert MJ. Methicillin-resistant-Staphylococcus aureus hospitalizations, United States. Emerg Infect Dis. 2005 Jun;11(6):868-72.
Noskin GAl. The burden of Staphylococcus aureus infections on hospitals in the United States: an analysis of the 2000 and 2001 Nationwide Inpatient Sample Database. Arch Intern Med. 2005 Aug 8-22;165(15):1756-61.
Noskin GA. National trends in Staphylococcus aureus infection rates: impact on economic burden and mortality over a 6-year period (1998-2003). Clin Infect Dis. 2007 Nov 1;45(9):1132-40.
Styers D. Laboratory-based surveillance of current antimicrobial resistance patterns and trends among Staphylococcus aureus: 2005 status in the United States. Ann of Clin Micro and Antimicrobials. 2006;5:2

 

 

Date last modified: May 1, 2008
Division of Healthcare Quality Promotion (DHQP)
National Center for Preparedness, Detection, and Control of Infectious Diseases