We also examined daily mortality for three different age
groups: below age 6, ages 6–49, and age 50 and above. In the
United States, about 70% of all deaths are in the group above
age 65. In Bangkok, about 65% of all deaths are individuals
age 50 and above. We therefore selected 50 and above as a
measure of mortality in older adults in Bangkok.
After totaling the daily mortality counts from the
individual death certificate data provided, it was clear that
recorded mortality counts for Bangkok were substantially
lower in 1994 and 1995 than in 1992 and 1993. To check
whether this might reflect a real drop in daily mortality
or a problem of missing data, we compared the total annual
mortality counts (calculated from the daily data) to
total annual mortality estimates for Bangkok from the
Ministry of Public Health which were based on monthly
counts of mortality. This comparison indicated that the
periods of low mortality counts were more likely caused
by missing data than by a real drop in mortality. We used
three alternative procedures in the statistical analysis (described
in greater detail in the Methods section) to address
the apparently varying degree of missing mortality
data during the study period: (1) excluding the suspect
months, (2) including binary variables in the regression
model to account for the period with the unexplained
drop in reported deaths, and (3) using locally weighted
smoothers to model the baseline mortality rate.