BACKGROUND: From 2005 through 2015, a total of 881 (confirmed and probable) pertussis cases in Houston jurisdiction were reported to Texas NEDSS Based System (NBS), and the annual average was roughly 80 cases. Respectively in 2009 and 2013, two peaks of pertussis cases (147 and 196) were observed. Among the 881 cases, number of female patients was higher than males, but the differences were statistically insignificant. Most of the patients were under 5 and the two peaks of cases were observed in almost all age groups. Moreover, a statistically significant trend was found. The main objective is to apply scan statistics to the pertussis surveillance data and determine geographic clusters.
METHODS: The case data were extracted from NBS, and the census tract data were downloaded from the US Census Bureau's website. Eighteen pertussis cases with invalid/incomplete addresses were excluded. The annual number of pertussis cases is considered to follow a Poisson distribution, and the census tract is used as the basic geographic unit. The estimated number of pertussis cases adjusted for age in each tract is obtained by direct standardization. The observed and expected number of cases in each tract is also stratified by reporting year. In the analysis, first, taking the 863 cases as a whole and not adjusting for time, a pure spatial analysis is performed. Then, same analysis is repeated on cases reported in 2013. In the next step, pure temporal and space-time analysis are conducted on the date stratified by year. The SaTScan v9.4.2 is used for the analysis.
RESULTS: Not adjusting for time, three significant clusters of pertussis cases are detected. The pure spatial analysis on the 2013 cases doesn’t generate any significant clusters, but two clusters from the 2013 data overlap with the significant clusters from the overall pure spatial analysis not adjusted for time. Though only one statistically significant cluster is generated, the clusters from stratified data overlap well with clusters from non-stratified data. The space-time analysis doesn’t generate any statistically insignificant cluster.
CONCLUSIONS: After adjusting for time and population age, and based on patient addresses, the scan statistics does not render any statistically significant clusters of pertussis cases. We could either conclude that pertussis infections in Houston are geographically independent, or that the analysis based on available sample doesn’t have enough power to detect statistically significant clusters.