Anna B. Wheeler
1,
Susan Coffin1, Theoklis Zaoutis
1, Carolyn Bridges
2, Guillermo Herrera
3, Barbara Watson
4, and Ron Keren
5. (1) Division of Infectious Diseases, Children's Hospital of Philadelphia, Philadelphia, PA, USA, (2) Centers for Disease Control and Prevention, NIP/ESD, MS E-61, Atlanta, USA, (3) National Immunization Program, CDC, (4) Immunization Program, Philadelphia Department of Public Health, Philadelphia, PA, USA, (5) Division of General Pediatrics, Children's Hospital of Philadelphia, 3535 Market Street, 15th floor, Philadelphia, PA, USA
BACKGROUND:
Some population-based surveillance programs rely upon administrative data to asess the burden of vaccine-preventable infectious diseases and the impact of vaccine recommendations. However, little is known about the ability of administrative data to identify cases of pediatric laboratory-confirmed influenza (LCI) infection.
OBJECTIVE:
To examine the sensitivity and positive predictive value of administrative data used to identify cases of pediatric LCI.
METHOD:
We queried 2001-2004 inpatient billing records from a large urban children's hospital to identify patients discharged with an influenza-specific ICD-9 diagnostic code. The sensitivity and positive predictive value of influenza-specific ICD-9 codes to identify cases of LCI were assessed against the reference standard of laboratory test confirmation (routinely performed on all children admitted with flu-like symptoms during influenza season). Discrepancies between these two sources were resolved through detailed chart review.
RESULT:
We identified 726 cases of LCI from laboratory records, and found only 532 cases of influenza by ICD-9 code in billing records. Sixty-four cases identified by ICD-9 code were found to be false positives. In addition, 258 false negatives were identified (cases present in laboratory records but lacking an ICD-9 code for influenza). Thus ICD-9 codes displayed a sensitivity of 64% (468/726) and a positive predictive value of 88% (468/532) in identifying LCI hospitalizations.
Of the cases misidentified as influenza by ICD-9 code, miscoded cases of laboratory-confirmed parainfluenza (42) and Haemophilus influenzae (8) accounted for 78% of total errors. Other reasons for misclassification included mention of influenza-like-illness in the chart without laboratory confirmation (6), misinterpretation of abbreviations used in the medical record (2), and other (6).
Common ICD-9 codes for patients with LCI missed by administrative data included otitis media, hypovolemia, and asthma exacerbation NOS.
CONCLUSION:
Administrative data have a limited ability to identify LCI in hospitalized children.
LEARNING OBJECTIVES:
Supplemental case-finding strategies are needed to increase case capture and verify cases identified by review of administrative data.
See more of Epidemiology Track Workshop: Pediatric Influenza: A Serious and Often-Missed Vaccine-Preventable Illness
See more of The 39th National Immunization Conference (NIC)
