First, we divided Cambodia into two

First, we divided Cambodia into two geographic areas
based on the residence of the patient: (1) PP&K, the
intervention area and (2) elsewhere in Cambodia, the
control area and computed the reported number of dengue
hospitalizations and deaths in each area by year from 1995
to 2005. As a conservative approximation, we treated
PP&K as the only area where the large-scale intervention
was applied (despite the fact that 27.6% (i.e. 0.8 million ⁄2.9 million) of the population affected by the intervention lived outside these areas. This simplification was
necessary because epidemiologic data were not available
below the province level. Second, we divided the years
analysed into two periods: ‘before’ (1995–2000) and
‘during’ the intervention (2001–2005) and averaged the
number of hospitalized dengue cases and deaths for each
period. Third, we expressed the number of hospitalized
dengue cases in PP&K as a percentage of the number of
dengue hospitalizations elsewhere in Cambodia for each
year and for the annual average in the period. Fourth, we
assumed that if no intervention had occurred in PP&K, the
calculated annual average percentage from the ‘before’
period would have remained unchanged in the ‘during’
period. By multiplying the percentage in step three times
the reported dengue hospitalized cases elsewhere in Cambodia for each of the years 2001–2005, we obtained the
expected number of hospitalized dengue cases in PP&K
that would have occurred in the absence of intervention in
each of these years. Fifth, we calculated the number of
hospitalized dengue cases averted in PP&K in each
intervention year as the difference between the expected
and reported annual number of dengue hospitalizations in
PP&K. Sixth, we tested whether the mean number of cases
averted over the 5 years was significantly positive using a
one-tailed t-test with the level of significance of 0.05.