The findings and conclusions in these presentations have not been formally disseminated by the Centers for Disease Control and Prevention and should not be construed to represent any agency determination or policy.
Wednesday, May 10, 2006 302
Time Series Analysis of Trends in Early Syphilis Morbidity in New York State (NYS)
Alison G. Muse1, Louise Anne McNutt2, Brian Close3, Igor Zurbenko3, and F. Bruce Coles1. (1) Bureau of STD Control, New York State Department of Health, 1168 Corning Tower, ESP, Albany, NY, USA, (2) School of Public Health, State University at Albany, 1 University Place, Rensselaer, NY, USA, (3) Department of Biometry and Statistics, University at Albany, 1 University Place, Rensselaer, NY, USA
Background: In response to an epidemic increase in reported cases of early syphilis in the late 1980's, NYS health officials implemented an intervention program in mid-1990 which expanded serologic screening in non-traditional sites. Following this intervention, early syphilis morbidity has declined to its lowest level since the 1950's.
Objective: To use the adaptive Kolmogorov-Zurbenko (KZA) algorithm, an iterative moving average filter for time series analysis, to assess the magnitude and time of the occurrence of shifts in syphilis surveillance data from 1975-1998 that may be attributable to the syphilis intervention.
Method: Early syphilis cases reported from 1975 to 1998 among residents of NYS (outside New York City) were included. Data were homogenized prior to transformation on a log-linear scale to stabilize the variance. The KZA algorithm estimated the average daily case rate and identified change points in the baseline data.
Result: A total of 17,618 early syphilis cases were reported during the study period. Use of the KZA algorithm identified three distinct time periods separated by abrupt changes. From 1975 – 1985, the annual rate remained constant. In December 1985, the annual rate increased 36% until January 1990, when the rate decreased 36%. The corresponding ratio of Primary and Secondary to Early syphilis was 1.28 (1975-85), 1.40 (1986-89), and 0.65 (1990-98).
Conclusion: The KZA algorithm identified two temporal shifts in the annual syphilis case rate. The first, an increase corresponding to the crack cocaine epidemic and the second, a decrease associated with the implementation of the screening intervention.
Implications: Researchers will become familiar with a time series model that identifies associations between temporal trends in STD surveillance data and program intervention while controlling for other data artifacts.