Post by Patti on Sept 1, 2006 14:42:41 GMT -5
"Megan Blewett, a brilliant high school student who spent much of the summer doing research at the Broad Institute of MIT and Harvard, gave a flawless presentation of a study she had recently completed.
Her paper examined whether Multiple Sclerosis (MS)-the most common primary neurological disorder of young adults-belongs in the same category as Lyme Disease, West Nile Virus, and other zoonotic diseases, which are caused by infectious agents that can be transmitted between animals and humans."
www.gismonitor.com/news/newsletter/archive/archives.php?issue=20060824&style=web&length=short
GIS Monitor
August 24, 2006
GIS for Public Health
Blewett first began to examine the relationship between MS and Lyme after being struck by the similarity in distribution maps of the two diseases. Epidemiological and biochemical similarities suggest a common bacterial basis for MS and Lyme. If there is a zoonotic influence in MS, Blewett reasoned, geostatistical analysis (inferential statistical techniques combined with data visualization and cartographics) should show that MS has the same geographic distribution as similar zoonotic diseases-and that the latter are good predictors of the former. As a control variable, Blewett used accident/injury, because this condition should be unrelated to a bacterial distribution.
Blewett obtained mortality counts from the Centers for Disease Control and Prevention (CDC), population data from the 1990 Census, and population elevation data from the MIT Geoserver and the U.S. Census Gazetteer. She constructed a database of the elevation of each county seat, which she then gave to ESRI for use by other researchers.
From the start, Blewett had to contend with, in her words, "the current lack of standardization in health geographics data." Because Lyme is a reportable illness, she had assumed that she could readily find a standardized Lyme incidence dataset online. Instead, she had to contact (and follow up with) every single state epidemiologist.
She entered the data into an Excel spreadsheet for each of the 3,141 counties in the United States for each of the years from 1992 through 1998-in the process creating the most comprehensive dataset available on Lyme incidence at the county level. "This process took several months," she told me, "though I am currently making this dataset available to other researchers."
To compare the distributions of different diseases, Blewett had to compile a database of their incidence and of their associated environmental variables. She began with a search for an authoritative source of incidence and prevalence data for Lyme, MS, breast cancer, ALS, and accidents/injuries. Death certificates are filed at the state level and coded in a standard way across all states.
The information is then collected from the states by the National Center for Health Statistics, which publishes it along with other vital statistics. Blewett determined the appropriate code for each of the five diseases of interest, then used these codes in DataFerrett to extract the state of residence for those who died of those diseases in the United States in 1998.
To calculate the incidence variables for each state, she added to this data the population from the 1990 Census and the 1998 deaths from DataFerrett. She used the same method to obtain data at the county level. To this state and county data, Blewett added the number of new Lyme cases reported each year from 1992-1998, the centroid latitude, the centroid longitude, and the population elevation, then averaged the centroid latitude and longitude over all counties in a state to calculate the state value. She repeated this process to calculate each state's population elevation.
The results of Blewett's statistical analyses support geographically the proposed connection between MS, Lyme, and related diseases, while the geographic distribution of the control variable is very different. Her correlations and regression analysis also show a clear geographic co-occurrence of MS and Lyme but no such relationship with the control variable.
This suggests a common cause for MS and Lyme. The strong association of MS with motor neuron disease (ALS) and the weaker, but significant, association with breast cancer, also suggest a possible common environmental mechanism for these diseases.
In her paper, Blewett also explains the results of her statistical analyses using biochemical principles. This summer she studied the chemical structures of compounds that bind uniquely with Neuregulin, a protein that appears to play a role in many of the diseases that her maps had showed to be geographically correlated. "I find it interesting," she told me, "that the epidemiological commonalities I found hold up on the biochemical level as well. This seems to give credence to the hypothesis that a common agent may be at work in MS and related diseases."
Blewett is careful to point out several possible confounding factors and spurious correlations. People diagnosed with chronic illnesses often move to Florida or California, because of the weather in those states, or to East Coast states with better healthcare. States with higher rates of diagnosis sometimes display lower death rates because, with experience, doctors in those areas often are more familiar with treating the disease.
To protect the privacy of individuals, federal statistics lump together data for counties with fewer than 100,000 residents; this excludes mostly rural counties, which have a higher incidence of Lyme transmitted by ticks in wooded areas. Finally, MS and Lyme are commonly misdiagnosed and people's state/county of residence may not be a good indicator of where they were exposed to a disease.
Blewett found ArcGIS 9.1, without any extensions, sufficient for mapping the distributions of diseases. She used Excel to create her datasets and SPSS to calculate the descriptive and inferential statistics and to run the correlations and the multiple regressions. She saved the SPSS file as a Dbase IV file, then opened it and saved it in ArcGIS to use in her cartographic analyses.
"Overall," she told me, "I have been extremely satisfied with the ArcGIS software. This technology has given me insight I would not have gained using more conventional biochemical / statistical techniques. In the future, I want to explore the ArcGIS statistical package and other ArcGIS extensions. Operating individually also means that cost is an issue; ESRI was very generous in supplying me with the tools I needed. I especially enjoyed presenting at the ESRI user conference."
What are Blewett's plans for future research? "One of my focuses now," she says, "is collecting more datasets. I hope to compare MS distribution to bird migratory patterns, tick density, and population density, among other factors. I am still searching for a nationwide comprehensive tick distribution dataset."
Will GIS remain one of her key tools? "GIS holds the potential to answer questions few other approaches can. I hope to continue to use GIS to investigate the epidemiological overlap between MS and Lyme. Since the cause of Lyme has already been identified as a spirochetal bacterium, looking at commonalities could eventually reveal the cause of MS as well. Future steps include using standard deviational ellipses to both compare disease distributions and to investigate the spread / saturation of the diseases over time.
I often worry that medical research today is not sufficiently goal-oriented. Scientific fields of study are becoming increasingly narrow and complex. GIS allows us to see the big picture and I hope to use it to guide any future biochemical investigation of these disorders."
Her paper examined whether Multiple Sclerosis (MS)-the most common primary neurological disorder of young adults-belongs in the same category as Lyme Disease, West Nile Virus, and other zoonotic diseases, which are caused by infectious agents that can be transmitted between animals and humans."
www.gismonitor.com/news/newsletter/archive/archives.php?issue=20060824&style=web&length=short
GIS Monitor
August 24, 2006
GIS for Public Health
Blewett first began to examine the relationship between MS and Lyme after being struck by the similarity in distribution maps of the two diseases. Epidemiological and biochemical similarities suggest a common bacterial basis for MS and Lyme. If there is a zoonotic influence in MS, Blewett reasoned, geostatistical analysis (inferential statistical techniques combined with data visualization and cartographics) should show that MS has the same geographic distribution as similar zoonotic diseases-and that the latter are good predictors of the former. As a control variable, Blewett used accident/injury, because this condition should be unrelated to a bacterial distribution.
Blewett obtained mortality counts from the Centers for Disease Control and Prevention (CDC), population data from the 1990 Census, and population elevation data from the MIT Geoserver and the U.S. Census Gazetteer. She constructed a database of the elevation of each county seat, which she then gave to ESRI for use by other researchers.
From the start, Blewett had to contend with, in her words, "the current lack of standardization in health geographics data." Because Lyme is a reportable illness, she had assumed that she could readily find a standardized Lyme incidence dataset online. Instead, she had to contact (and follow up with) every single state epidemiologist.
She entered the data into an Excel spreadsheet for each of the 3,141 counties in the United States for each of the years from 1992 through 1998-in the process creating the most comprehensive dataset available on Lyme incidence at the county level. "This process took several months," she told me, "though I am currently making this dataset available to other researchers."
To compare the distributions of different diseases, Blewett had to compile a database of their incidence and of their associated environmental variables. She began with a search for an authoritative source of incidence and prevalence data for Lyme, MS, breast cancer, ALS, and accidents/injuries. Death certificates are filed at the state level and coded in a standard way across all states.
The information is then collected from the states by the National Center for Health Statistics, which publishes it along with other vital statistics. Blewett determined the appropriate code for each of the five diseases of interest, then used these codes in DataFerrett to extract the state of residence for those who died of those diseases in the United States in 1998.
To calculate the incidence variables for each state, she added to this data the population from the 1990 Census and the 1998 deaths from DataFerrett. She used the same method to obtain data at the county level. To this state and county data, Blewett added the number of new Lyme cases reported each year from 1992-1998, the centroid latitude, the centroid longitude, and the population elevation, then averaged the centroid latitude and longitude over all counties in a state to calculate the state value. She repeated this process to calculate each state's population elevation.
The results of Blewett's statistical analyses support geographically the proposed connection between MS, Lyme, and related diseases, while the geographic distribution of the control variable is very different. Her correlations and regression analysis also show a clear geographic co-occurrence of MS and Lyme but no such relationship with the control variable.
This suggests a common cause for MS and Lyme. The strong association of MS with motor neuron disease (ALS) and the weaker, but significant, association with breast cancer, also suggest a possible common environmental mechanism for these diseases.
In her paper, Blewett also explains the results of her statistical analyses using biochemical principles. This summer she studied the chemical structures of compounds that bind uniquely with Neuregulin, a protein that appears to play a role in many of the diseases that her maps had showed to be geographically correlated. "I find it interesting," she told me, "that the epidemiological commonalities I found hold up on the biochemical level as well. This seems to give credence to the hypothesis that a common agent may be at work in MS and related diseases."
Blewett is careful to point out several possible confounding factors and spurious correlations. People diagnosed with chronic illnesses often move to Florida or California, because of the weather in those states, or to East Coast states with better healthcare. States with higher rates of diagnosis sometimes display lower death rates because, with experience, doctors in those areas often are more familiar with treating the disease.
To protect the privacy of individuals, federal statistics lump together data for counties with fewer than 100,000 residents; this excludes mostly rural counties, which have a higher incidence of Lyme transmitted by ticks in wooded areas. Finally, MS and Lyme are commonly misdiagnosed and people's state/county of residence may not be a good indicator of where they were exposed to a disease.
Blewett found ArcGIS 9.1, without any extensions, sufficient for mapping the distributions of diseases. She used Excel to create her datasets and SPSS to calculate the descriptive and inferential statistics and to run the correlations and the multiple regressions. She saved the SPSS file as a Dbase IV file, then opened it and saved it in ArcGIS to use in her cartographic analyses.
"Overall," she told me, "I have been extremely satisfied with the ArcGIS software. This technology has given me insight I would not have gained using more conventional biochemical / statistical techniques. In the future, I want to explore the ArcGIS statistical package and other ArcGIS extensions. Operating individually also means that cost is an issue; ESRI was very generous in supplying me with the tools I needed. I especially enjoyed presenting at the ESRI user conference."
What are Blewett's plans for future research? "One of my focuses now," she says, "is collecting more datasets. I hope to compare MS distribution to bird migratory patterns, tick density, and population density, among other factors. I am still searching for a nationwide comprehensive tick distribution dataset."
Will GIS remain one of her key tools? "GIS holds the potential to answer questions few other approaches can. I hope to continue to use GIS to investigate the epidemiological overlap between MS and Lyme. Since the cause of Lyme has already been identified as a spirochetal bacterium, looking at commonalities could eventually reveal the cause of MS as well. Future steps include using standard deviational ellipses to both compare disease distributions and to investigate the spread / saturation of the diseases over time.
I often worry that medical research today is not sufficiently goal-oriented. Scientific fields of study are becoming increasingly narrow and complex. GIS allows us to see the big picture and I hope to use it to guide any future biochemical investigation of these disorders."