I ran into this 2008 paper doing hurricane work with Jesse. The results are not extremely surprising, but I really liked how they displayed their result. Many of us use high-dimensional data and multiple regression models to try and account for the many different processes that occur in social data, but it is often difficult to clearly display the effect of just one process while also being clear about all the other controls in the model. I like the approach of this team: they show predictions form the complex model (eg. with week and month fixed effects, socioeconomic controls, etc.) overlaid with the real data.
Factors determining vulnerability to diarrhoea during and after severe floods in Bangladesh
Masahiro Hashizume, Yukiko Wagatsuma, Abu S. G. Faruque, Taiichi Hayashi, Paul R. Hunter, Ben Armstrong and David A. Sack
Abstract: This paper identifies groups vulnerable to the effect of flooding on hospital visits due to diarrhoea during and after a flood event in 1998 in Dhaka, Bangladesh. The number of observed cases of cholera and non-cholera diarrhoea per week was compared to expected normal numbers during the flood and post-flood periods, obtained as the season-specific average over the two preceding and subsequent years using Poisson generalised linear models. The expected number of diarrhoea cases was estimated in separate models for each category of potential modifying factors: sex, age, socio-economic status and hygiene and sanitation practices. During the flood, the number of cholera and non-cholera diarrhoea cases was almost six and two times higher than expected, respectively. In the post-flood period, the risk of non-cholera diarrhoea was significantly higher for those with lower educational level, living in a household with a non- concrete roof, drinking tube-well water (vs. tap water), using a distant water source and unsanitary toilets. The risk for cholera was significantly higher for those drinking tube-well water and those using unsanitary toilets. This study confirms that low socio-economic groups and poor hygiene and sanitation groups were most vulnerable to flood-related diarrhoea.