• active surveillance;
  • diagnostic code;
  • gastroenteritis;
  • hospitalisation;
  • rotavirus

Aim:  The number of rotavirus hospitalisations is usually estimated from assigned diagnosis codes for gastroenteritis despite lack of validation for these indirect methods. Reliable estimates before and after introduction of vaccines are needed to quantify the absolute impact of new immunisation programs.

Methods:  This 2-year study conducted at three hospitals prior to the licensure of the rotavirus vaccines in the USA compared two indirect methods for estimating hospitalisations for rotavirus gastroenteritis with estimates derived from prospective recruitment of children presenting with diarrhoea, vomiting or fever. For active surveillance, rotavirus gastroenteritis was confirmed by demonstration of stool antigen. The indirect residual and proportional methods assumed rotavirus to have caused a proportion of hospitalisations coded as acute gastroenteritis identified from computerised records.

Results:  There were 447 rotavirus hospitalisations among inpatients 31 days through 4 years of age admitted with vomiting and/or diarrhoea, compared with 306 and 228 hospitalisations identified by the two indirect methods. Only 52% of children hospitalised with gastroenteritis received a qualifying diagnosis code at discharge. Relative to active surveillance, the sensitivity and specificity (95% confidence interval (CI)) in identifying rotavirus-attributable hospitalisations was 45% (95% CI: 43–48%) and 89% (88–90%) for the residual method and 34% (30–39%) and 92% (90–94%) for the proportional method.

Conclusions:  Many children admitted to the hospital with diarrhoea, vomiting or fever were not assigned discharge codes for acute gastroenteritis. Consequently, standard indirect methods missed a substantial number of rotavirus-associated hospitalisations, thereby underestimating the absolute number of children who could potentially benefit from vaccination.