11.522: UIS Research Seminar (Fall 2014) - Discussion notes
Over the last few discussions, we started by looking at simplistic models based in excel and then we ended our last seminar by discussing critiques of comprehensive models. We discussed how various models worked in the 1970s, 1990s, and 2000s.
The purpose of this discussion is to more clearly articulate the strengths and weaknesses of modeling heat-related hospital admissions and the ambient air temperature. This includes the following sub-topics:
Heat-related adverse health outcomes are related to prolonged high ambient air temperatures (McGeeehin & Mirabelli, 2001). With a projected 3°-5°increase in average temperature by 2100, Massachusetts’ populations will likely be exposed to more and more extreme heat events and hot days (Davis, Knappenberger, Michaels, & Novicoff, 2003). Certain populations experience the adverse effects of heat waves and hot ambient air temperatures more intensely than others. During the 1995 and 1999 heat waves in Chicago and France’s 2003 heat wave, heat-related fatalities were highest among those who lived alone and did not leave the house on a daily basis (Klinenberg, 2003; Poumad`ere, Mays, Le Mer, & Blong, 2005). Socially isolated situations are naturally exacerbated by other vulnerabilities such as diabetes, obesity, and age; people living alone who are 65 and above who are particularly vulnerable (Reid, et al., 2009). Populations with limited access to technological adaptations, like air conditioning, are particularly vulnerable to hot days (Executive Office of Energy and Environmental Affairs, 2011). With historically cool summer temperatures, communities in Massachusetts often have limited access to air conditioning and other technological adaptations to mitigate the risk of hot weather.
Boston and the commonwealth of Massachusetts do not have a comprehensive heat wave plan. Boston does have a ‘Heat Alert Plan’ that is activated when the temperature is above 86 degrees Fahrenheit and 68% humidity for two consecutive days. The plan includes extending swimming pool hours, home care agencies are advised to set up phone trees, cooling centers While, like Boston, some towns in Massachusetts have cooling centers, that is often as far as they will go with heatwave planning. Cooling centers are places where people can sit in an air-conditioned spaces and drinking cooled water.
We postulate that criteria (e.g. number of consecutive hot days, percent of population 65 and above) can be used to determine an excess risk of hospitalizations for the state of Massachusetts due to high ambient air temperatures.
Based on the methods in the Metzger et al. article on heat and mortality in New York City, different climatic measures could be used as the independent variables in the model (Metzger, Ito, & Matte, 2010). A heat index, made using the high temperatures and precipitation of a day and town, will be the primary independent variable and the daily weather data set will originate from the National Climatic Data Center (National Climatic Data Center, 2014). Other variables of interest may be the year, month, and day of week, which will be included as a timestamp. Additionally, it may be valuable to look at the number of consecutive days above a certain temperature. This would involve adding lag variables and transforming the primary independent variable in order to capture this effect. Additional exposure variables will need to be considered such as the percent of population living alone, age, childhood asthma prevalence, levels of obesity, diabetes prevalence, and more. The likely source for such confounder estimates may be problematic because of varying levels of analysis.
The outcome of interest is the increase in the hospital admissions data for the increase in temperature-precipitation ratio (or the 1 day increase of heat events as defined by the SSC). De-identified hospital admissions data will be accessible from the Massachusetts Department of Health. Using the ICD-9-CM codes 992.0-992.9 (excessive heat or light effects), E-code E900.0 and E900.1 (weather conditions caused excessive heat and due to man-made conditions). Hyperthermia could be defined at ICD-10 code X30 or T67 and we would exclude manmade heat exposure, ICD-10 W92. This identification of outcomes is based on New York City’s Heat Illness and Deaths study (Metzger, Graber, & Razvi, 2013).
Metzger, K. B., Ito, K., & Matte, T. D. (2010). Summer Heat and Mortality in New York City: How Hot Is Too Hot? . Environmental Health Perspectives , 118 (1), 80-86. http://demo.indiaenvironmentportal.org.in/files/Summer%20Heat%20and%20Mortality%20in%20New%20York%20City.pdf
Optional Readings: (Prioritize the Reid, et al. article)
Reid, C. E., O’Neill, M. S., Gronlund, C. J., Brines, S. J., Brown, D. G., Diez-Roux, A. V., et al. (2009). Mapping Community Determinants of Heat Vulnerability. Environmental Health Perspectives , 1731-1736. http://www.jstor.org.libproxy.mit.edu/stable/pdfplus/40382459.pdf (also on Stellar)
Klinenberg, E. (2003). Heat wave: A social autopsy of disaster in Chicago. Chicago, IL: University of Chicago Press. http://www.jstor.org.libproxy.mit.edu/stable/pdfplus/40193896.pdf (also on Stellar)
1) Given the exposure and outcome variables mentioned above, what type of model could represent this data well? Are there other variables that may be valuable to include in this model?
2) What would be the major weaknesses in this type of model? How could we improve it?
3) What would be the benefit of creating this type of model? How could it be used to increase these benefits?
Other Background Readings:
Back to 11.522 home page.