A systematic approach to time series analysis of seasonal infections
Alarcon Falconi, Tania.
2018
-
Abstract:
Seasonal infections have well-known cyclic temporal patterns which may vary by pathogen,
population, and location. Epidemiologists study those temporal patterns using time
series analysis techniques, and by selecting models that best fit the underlying
statistical assumptions and distributions of the data. Even though changes to the
temporal unit of analysis and model selection have ... read morebeen shown to impact model results,
the relevant sensitivity analyses are rarely discussed in epidemiological literature.
The overarching objective of this research is to develop a systematic approach to time
series analysis of disease seasonality that includes a decision making process
considering data sources, temporal units of analysis, and model structure. To achieve
this goal, we used three case studies. In the first study, we assessed the seasonality
of legionellosis in the U.S. using two national databases and demonstrated a significant
increase in disease incidence starting in 2003, along with a shift in peak timing from
mid-September before 2003 to mid-August after 2003. We also highlighted discrepancies
between the databases. In the second study, we characterized seasonal patterns of
cholera hospitalizations in Vellore, India using harmonic regressions and Poisson,
quasi-Poisson, negative binomial, and logistic models. We found that cholera
hospitalization records are decreasing on average at a rate of 0.005 cases per month,
and have a peak between late June and mid-August based on both weekly and monthly
aggregation schemes and for all models, except logistic models. In the third study, we
focused on different aggregating schemes of longitudinally-observed episodes of diarrhea
and acute lower and upper respiratory infections (ALRI and AURI, respectively) in a
cohort of children in Quito, Ecuador. By aggregating episodes into daily, weekly and
monthly counts using a Gregorian calendar and a study "calendar", we demonstrated that
the aggregation with a Gregorian calendar structure creates irregularities affecting the
model results. The methods used in these studies and their results provide a road map to
systematically assess seasonal patterns of infections, paying close attention to the
decisions that researchers must make on data sources, temporal units of analysis, and
model structure.
Thesis (Ph.D.)--Tufts University, 2018.
Submitted to the Dept. of Civil Engineering.
Advisor: Elena Naumova.
Committee: Laurie Baise, Kenneth Chui, and Al Ozonoff.
Keywords: Environmental health, and Biostatistics.read less - ID:
- st74d344g
- Component ID:
- tufts:28586
- To Cite:
- TARC Citation Guide EndNote