Using hospitalization data to evaluate and improve invasive pneumococcal disease surveillance — New Mexico, 2007–2009

Background: Invasive pneumococcal disease (IPD) is vaccine-preventable and commonly manifests as septicemia or meningitis. IPD underreporting can influence policy decisions regarding vaccine use.

Objective: We evaluated reporting completeness and how hospitalization data can benefit surveillance.

Methods: Combined IPD data from passive surveillance and Emerging Infections Program Active Bacterial Core surveillance during 2007–2009 were linked with Hospital Inpatient Discharge Data (HIDD). Potential IPD cases in HIDD were identified by using two IPD-specific ICD-9 codes (meningitis or septicemia caused by Streptococcus pneumoniae) and three nonspecific codes. For potential cases identified only through HIDD, laboratory reports confirming S. pneumoniae isolation from normally sterile sites were reviewed for all those with IPD-specific codes and a systematic sample of 17% of those with nonspecific codes. Capture-recapture was used to estimate surveillance sensitivity.

Results: Surveillance detected 1,191 cases, 396 initially through active surveillance. HIDD identified 1,287 potential cases; 558 cases were identified by both HIDD and surveillance. Of 729 potential cases identified only by HIDD, 104 had IPD-specific and 625 had IPD-nonspecific codes; 32 were true IPD cases. We estimated 36 additional cases missed by both sources, yielding 95% overall system sensitivity. Predictive value positive was 78% for IPD-specific codes and 30% for nonspecific codes.

Conclusions: The high sensitivity of New Mexico IPD surveillance is likely because of combined active and passive surveillance. Hospitalization data yielded additional IPD cases not identified through surveillance. For states that rely on passive reporting, using IPD-specific ICD-9 codes from hospitalization data might be an efficient way to improve surveillance.