Page 60 - Contributed Paper Session (CPS) - Volume 3

CPS1944 Oyelola A.
Attributable risk measure
The attributable measures; attributable fraction (AF) and attributable
number (AN) are the most useful indicator of exposure-related health burdens
[18, 19]. We estimated the fraction of pneumonia cases attributed (AF) to
weekly mean temperature and total weekly rainfall, separately using the
optimum weather values as references.
Attributed fraction (AF) measure was derived from prediction of the overall
cumulative exposure-response relationship in the DLNM model. Using the
minimum incidence percentile, across the entire exposure spectrum as the
0
reference and cut-off for optimum temperature/rainfall value, we used a
backward perspective [18, 19], assuming that the risk at week t was
attributable to a series of exposure, x events in the past, − , … , − .
0
The attributable fraction ( − ) for a given exposure is derived as follows:
,

− , = 1 − − ∑ = 0 −
,

Where − , represented the risk associated (logRR) with lagged exposure, x
at time, − .

All statistical analyses were performed with R statistics software v3.4.0
[20], with the package “dlnm” to create the DLNM [13].

3. Result
The weekly time series distributions of cohort of pneumonia cases were
plotted in Figure 1. The time series decomposition shows increase trend and
seasonal patterns in cases of pneumonia over the years. Similarly, the pattern
of seasonality (alternating highs and low) of pneumonia cases is inversely
mirrored by mean weekly temperature and total weekly rainfall (not shown).
The summaries of cases and climate variables were presented in Table 1. There
were negative correlations between weekly pneumonia cases- and mean
temperature (r = -0.224, P<0.001), minimum temperature (r = -0.217,
P<0.001), maximum temperature (r = -0.218, P<0.001) and total rainfall
(r = -0.099, P=0.017).
Correlation among temperature variables were higher than 0.7 (not
shown), therefore to prevent issues with multi-collinearity, we based this study
on mean weekly temperature and total weekly rainfall.
The best DLNM model (out 128 candidates models) described the
weather-pneumonia association by lag up to 15 weeks and quadratic B-splines
function for temperature/rainfall-pneumonia relationship and linear function
for lag-pneumonia relationship with a total degrees of freedom of 6 (based
on smallest QAIC=3355.8). The model also include a natural cubic spline

49 | I S I W S C 2 0 1 9