Page 1 of 10
Article DOI: http://dx.doi.org/10.3201/eid2103.141845
Regional Spread of Ebola Virus, 2014
This Technical Appendix provides further details on the methods used as well as additional
Data and Definitions
Case Count Data: We obtained the cumulative number of confirmed, probable and suspect case
counts for each of the 63 districts in Liberia, Sierra Leone, and Guinea from WHO Situation Reports
posted weekly on the WHOâ€™s website (1). We defined a district as having become affected if it had least
one suspect, probable, or confirmed Ebola case in the WHO reports. We considered the week a district
first reported a case as the week it became â€œaffectedâ€. We also used the case counts data from the WHO
Situation Reports in our calculations of the weighted sum of inverse distances (see Calculations section
below). We first identified the number of â€œnew casesâ€ in a single given week by subtracting the
cumulative case count for a district in a given week from the cumulative case count reported for the
week prior. We then summed the new cases values for every three week period in our outbreak dataset
to obtain the number of new cases over the prior three weeks.
For some districts, defining the week a district became affected and calculating new cases was
complicated by reductions in the cumulative case count from week to week or gaps in reporting.
Technical Appendix Table 1 describes how case counts data were used to define the week of first report
(i.e. affected) and for case count weighting in these special circumstances.
Other studies (2â€“4) examining the role of distance as a predictor of disease spread used
confirmed cases only to determine when a new area had became affected. We did not rely on confirmed
cases alone due to heterogeneity in the reporting delay of confirmed cases reported by country (5). For
example, in Guinea (using data through week 33), the reporting of confirmed cases across all affected
districts occurred on just 4 differen