We conducted a retrospective cohort analysis of trends in dengue diagnoses among hospitalized patients using the National Inpatient Sample, the largest all-payer database of hospital discharges in the United States. The database is maintained as part of the Healthcare Cost and Utilization Project by the Agency for Healthcare Research and Quality and consists of a 20% stratified sample of discharges from nonfederal acute care hospitals (11
). We first extracted all discharges from the National Inpatient Sample for hospital admission (i.e., the denominator of the incidence rate) from 2000 through 2007. Among this population, we then identified cases of dengue fever coded as either a primary or secondary diagnosis (i.e., code 061, dengue fever), according to the International Classification of Diseases, Ninth Revision, Clinical Modification.
For each yearly incidence rate, we calculated a 95% exact binomial confidence interval. To determine whether a significant trend in hospitalizations of patients with dengue fever occurred during the study period, we fit a logistic regression model using yearly incidence as the dependent variable and year as the independent variable. In addition, to accommodate the temporal association in the yearly incidence, we fit the model using generalized estimating equations, assuming an autoregressive correlation structure.
We also calculated the ratio of the yearly incidence rates at the beginning and the end of the study period (i.e., rates in 2000 and 2007). We tested whether this incidence ratio is significantly different from one using the Fisher exact test and computed a 95% exact confidence interval for the corresponding odds ratio using the hypergeometric distribution. (Because the incidence rates are fairly low, the odds ratio closely approximates the incidence ratio.)
Finally, we used the Monte Carlo variant of the Fisher exact test to investigate possible geographic variation in the incidence rate among the 4 US census regions. We tested for geographic homogeneity for every year in our 8-year sample. All statistical analyses were performed by using R version 2.10.1 (R Foundation for Statistical Computing; www.r-project.org
) and SAS version 9.2 (SAS Institute Inc., Cary, SC, USA.).
During 2000–2007, ≈1,250 patients were hospitalized for dengue fever. The mean age of this population was 38 years (range newborn–87 years). The length of stay for these patients ranged from 0 to 35 days (median 3 days). For the Monte Carlo variant of the Fisher exact test, we found the incidence rates for the 4 US Census regions were homogenous for all years, except for 2004 and 2007. In these 2 years, the Northeast Region had the highest incidence rate (p<0.0001 for each year). Over the study period, the estimated number of dengue cases more than tripled from 81 cases in 2000 to 299 cases in 2007. The trend in the incidence of patients hospitalized with dengue during the study period was upward and significant (trend estimate 0.1313, model-based SE 0.0258; p<0.0001) (). The increase from 2000 to 2007 was also significant (incidence ratio 3.5641, 95% confidence interval 2.0293–6.6232; p<0.0001).
Figure National estimates of dengue yearly incidence rates and 95% exact binomial confidence intervals (error bars), calculated by using data from the National Inpatient Sample, United States, 2000–2007. The trend (dotted line) is based on a logistic (more ...)