The following are examples of how you may integrate this with other programs.
If we view this methodology as the link between regular GIS and the old fashioned non-GIS methods of Quality Assurance reviews, we can actually use this methodology to improve out local outcomes and to provide almost immediately the data that are needed to determine how a program is going.
This methodology can be used to detect the following, often in very short time:
- Migration routes for diseases inland for new in-migrating diseases from other countries.
- Places where the people are not fully engaged in disease prevention steps, such as places where skipping visits are common, or failing to engage fully in a therapeutic regimen, etc.
- The locations of the poorest social or community settings where infrastructures need to be strengthened by constructing local clinics and the like.
- Towns and cities where the sociocultural paradigms include large groups that avoid care due to language barriers.
- Places where older cultures remain active and where due to cultural beliefs, certain social practices linked to disease are clustered or of higher density.
- Landscapes where natural ecology makes the region a likely source for a new in-migrating foreign born zoonotic disease.
- Places where foreign born diseases cluster due to recurring travel patterns. .
- Physiographic provinces and ICDs for which latitude or elevation actually make a difference
- The most important ecological sites for a major potential epidemic inducing vectored diseases.
- Inland urban settings where disease often by pass the ocean edge surveillance points.
- Rural settings where people are not only living on the fringes of American culture but also engaging in alternative means of living and/or practicing sectarian lifestyle or religion of potential threat to national security, or even defining small groups that appear utopian and militaristic.
- Settings where child abuse and child sexual abuse exist more that elsewhere.
- The most important physical and sociocultural public health issues at culture rich settings like the Pacific Northwest or New York.
- Places where outbreaks of violent behavior are more common in parts of this country.
- Where Tex-Mex cross border relationship are most linked to child sexual abuse.
- The nature of vaccinated disease outbreaks and the types of population settings they tend to recur in.
- The part of northern country border where unexpected tropical disease outbreaks occur and why.
- What foreign born diseases are most closely linked spatially to primate research center facilities.
Example 1. Immunization refusals as the first example.
With the immunization refusal outcomes your notice the two major urban centers of the Pacific Northwest are where the majority of cases are recorded. You have just run the SAS tool for data mapping, and decide to rerun it with a filter in there for areas with amount more than a predetermined amount, say neighborhoods with 50 or more cases reported based on a 25 mile grid pattern. You rerun the program, get the results, and then export these for importation into your GIS.
Once the data is imported into a standard GIS, you can now query for the worst sites, or map them according to some outcomes grouping pattern such as the use of quartiles or natural breaks. You can then query for where specifically the worst sites are, set up a buffer for 5 or 10 miles around these places, and then analyze the buffered region for clinical data and the typical demographic data, especially age-gender-ethnicity and SES information. This buffer region can also be used to find the clinics in that area participating in one or more of your programs, and send letters specifically to those places informing them of last month’s findings, your plans to return to this query in 4 to 6 months, and recommendations on how to publicize this public health issue within the immediate area.
A problem with this method is it has that “Big Brother” look to it. It ends up that skipping immunizations may be a public health concern, but only in areas where this behavior is very prevalent is there reason to express concern about it. In the other studies of immunized disease patterns that exist due to failure to immunize, only a few diseases that we vaccinate again demonstrate a worrisome prevalence in certain urban locations. These also demonstrate a distribution that is highly correlated with population density, a and areal SES features such as median income and data out there detailing the amount or level of poverty in a region such as the GINI coefficient.
In the Pacific Northwest, there is a high amount of activity related to “alternative” or complementary medicine. One of the basic belief systems in such settings is the belief that children should experience diseases the way they did a century or more earlier, as some form of “natural process.” Adding to this belief system was the development of a health care ideology focused on ADD or ADHD, and the links that some tried to establish between the measles vaccine, which utilized a mercury salt to improve shelf life and stability, and the development of more children with this diagnosis. The result of this was a number of “natural medicine” families and parents who refused to promote the use of the measles vaccines or any other on their children. For the moment, this means mostly that their children are at risk but due to “the herding effect”, the bulk of the population around them is not and so even if their child becomes ill, the likelihood of transfer of this disease resulting in an outbreak is fairly low.
Comparing the distributions of the diseases themselves with the spatial distribution of immunization programs and those who resist these services shows us that the disease itself exists in certain places due to population density impacts and not refusal patterns.
Example 2. A review of Well-Visits as a Factor important to Improving Immunization History
In my well visit study of kids and mothers initiated in 2004/5, and finalized in 2006/7, it was determined that well visit scheduling was the major factor preventing kids from meeting their immunization requirements by the age of two.
To analyze this spatially, the data pulled must be on well visits that meet the typical requirements, and the amounts and types of immunizations provided for each patient. These immunizations have to be coded as part of a well visit, or not, and if they are not part of the well visit activities, you have to determine what activities are they a part of such as a walk-in or urgent care claim, or an emergency room visit, or some other unrelated medical history query of the child.
These can easily be evaluated without GIS and without SAS. But to engage the use of SAS in this process, you need to produce two programs to grid map the data according to two sets of criteria separately–well visits and immunizations. By coding that data, you determine degrees of completion for each, and can even subcategorize the data further into grid maps of each well visit and each immunization type. The grid data for these in turn can be reclassified into values that define the risk, assigning a high risk score to individuals whose child had no visits, or no immunization, and well as failure to complete certain immunization series.
The two maps can in turn be multiplied, and then the end product produced, along with a krigged map depicting the richest areas for these poor behaviors. You can also assign a value to well visit behavior, ranking it higher than immunization patterns for example, assigning people who don’t conform with the recommended visits as families/kids of higher risk for multiple reason. This new hybrid map that you produce by multiplying the two grid maps makes for outcomes that are very realistic, and testable with regard to intervention outcomes. Done only in SAS, these NPHGs provide us with the answers and solutions in much less time than having to integrate this into a GIS.
Example 3. Where do these problems truly exist?
Let’s say you have effectively implemented programs to circumvent the problems you uncovered with well visits and immunizations, and now you wish to find those other areas out there where risk can become a major problem if not dealt with in time.
You have grid mapped scores for immunization behavior and missed well visit opportunities and want to see if there are areas that are high risk to a correlation of these two metrics in one area or place.
You can simply multiply the two datasets together and map the entire dataset. To increase your resolution of this image, you can square it or cube it to increase the height of the 3D data (z-values) depicting highest risk spots. Another way to evaluate this is to quantify and reclass the information based on percent distributions. The map that is produced may look the same, but unlike previous maps can be compared with other metrics, other datasets with completely different outcomes ranges.
Percentile rankings are comparable between two hard to related details about health. For example, events related to immunization problems can be in the 1/100 to 1/1000 category of incidence, whereas well visit non-compliance might be at the 1/10 range. To correlate them to each other, you provide them an identical range metric–percentage–and then cross compare, cross multiply, or correlate the two percentages. With mapping, this is done spatially by comparing and contrasting two surfaces. By treating the maps like surface features, you overlay one on the other mathematically, to see where the two accentuate each other in the same place. You then search for areas which are 75% non-compliant when merged together, and design specific neighborhood specific intervention programs to better target and impact those particular social group settings.
In the Oregon-Washington area, this method helps to define those small communities (utopian like settings) with the highest risk of impacting the neighborhood or vice versa. These settings can then be places of a special list for ongoing surveillance, and their neighborhood clinical settings informed of any underlying increased risks to health that these behaviors might generate.
Video Link: http://youtu.be/W1d8fBxz5V4
Example 4. Identifying a Nidus or Nest for Disease.
A number of years ago, a series of diphtheria cases presented in the Salt Lake City area of Utah. It seemed possible that there was a spatial pattern here in need of review.
The way the diphtheria nidus was identified was by its abnormally large peak on that one 25 mile grid cell. So the first thing you can do is re-pull the data from just that area from your original dataset, and re-run it using a smaller grid pattern. Next, you evaluate max and min data for dates, and then use this to produce a formula that assigns integer values to each day, beginning with day 0–the earliest case. Then, you assign a value to that day by assigning it a percentile place in the overall time span for this series of cases. For example, in a theoretical example in which the epidemic lasts 100 days from day 0, people about 15 days into the time frame are assigned a value of 100-15=85, 30 days into it 100-30=70, 45 days into it 100-45=55, etc. This value is assigned to each case, and then the cases are averaged on a per grid cell basis, or even mapped as individual points.
What you expect to get with a perfect epidemic is a peak in the middle, followed by a reduction in pillar sizes as you move away from the nidus on the 3D map image. In situates where hierarchical diffusion is occuring, there is one nidus standing tall, a few distanced cases standing shorter on the map, with each these surrounded by still smaller pillars of third and fourth generation infectious cases. By viewing all of this spatially, you may know or deduce immediately what the causes are and how to intervene.
The hierarchical pattern with multiple secondary peaks indicates a higher risk diffusion process. To effectively intervene you have to intervene with the spread to the next generation nidus. Diseases that simply spread from the original nidus outward and back and forth across the same region are much easier to develop and intervention for.
In either of these cases, the use of a GIS serves more as information gathering tool for developing and carrying out your intervention program. The SAS 3D modeling portion of this work served to identify not only what and what, but also how this problem behaviors temporally and spatially, something a simple glance at a page may only occasionally produce fairly accurate results for should outside agencies not familiar with the surrounding communities become engaged.
Example 5. Disease Ecology.
GIS enables a good spatial epidemiologist the opportunities to integrate local land use and ecological data with medical data. In the case of animal host diseases, disease passed by insect vectors, diseases that are water born, and diseases that are due to exposure to particular environmental features, one can easily prepare for certain types of recurring analyses using a combined GIS and SAS approach.
In the above image, Chicago Illness is displayed, a condition produced by a local bacterium Leptosporidium, associated with dogs, and which has fungal-like ecological and behavioral patterns. Lake edge effect is clearly visible with this disease pattern. Large area surveillance demonstrates it has minimal distribution outside the immediate region.
The advantages of using regular GIS instead of SAS-GIS is several fold. SAS-GIS, although it integrates well with the database extraction tools, is a slow, awkward way of producing mapped data that, unless you have an adequately SAS and series of SAS-GIS extensions, requires many times more the effort than traditional GIS.
The SAS-GIS technology is worthy of implementing for occasional reporting processes, but for reporting on several dozen to several hundred ICD, V-codes, E-code events or other claims and service records per report, SAS-GIS is, to put it nicely, too cumbersome, time consuming and apt to failure or misprogramming. Even traditional GIS has its limits for repeated excessive mapping on a daily basis. With the raster tools out there, the possibility of producing 10,000 images per day is reasonable. With traditional GIS, even with extensions and add-ons meant to facilitate series of analyses, the rate at which end products are produced and printed out in a manner that is worth presenting is incredibly low for most traditional GIS in a standard desktop or laptop related workplace.
Pure SAS, without SAS-GIS or the extensions needed to automate the process, is not only inexpensive, relatively speaking, but dozens to hundreds of time more productive at the per output product level. For a standard workday, this methodology produced 370-450 images per run, time 15 to 20 runs per day, when not automated. That totals to about 400×17.5 = 7000 images per day, 750-800 per hour. If these images are for video use, that means you have 15 to 20 videos produced per day. If these images are for single display per metric tested and object displayed, a standard report of 25 to 50 images, 10 per page, can be produced manually each day; automating this process enables the same process to occur in just one to two hours.
Generating a disease ecology map requires SAS be used to generate the initial data for implementation of mapping with. If the mapping is produced by SAS, that process is commenced and the project is done, at least as far as the mapping goes. If that mapping is done with GIS, another period of time is required to export and then import the data from one tool and then import it into another. The value of disease ecology and topographic data therefore has to be taken into consideration with this process.
In the case of census/land use mapping in relation to medical codes (i.e. emergency visits related to childhood fractures, accidents, motor vehicle events, accidental amputation, falls, dislocations, etc.), all in relation to place, income level, land use codes, etc., GIS becomes more manageable in the overall analysis process than SAS. In the case of ER visits related to MVAs or suicides, GIS is probably more research-friendly than SAS, due to its additional spatial data visualization features.
Whatever the case, the role of SAS to GIS to other tools remains about the same. SAS is useful for data processing and high scaled analyses, with some added non-GIS mapping features already emphasized throughout this research methods evaluation. SAS sometimes provides just a little more flexibility that SQL in this case, more ability to work with the data and engage in more rugged analyses.
Disease ecology work (natural ecology or human ecology based) is really an atypical procedure when it comes to traditional SAS use in the medical/health care workplace. But SAS generated datasets are important material for more effective GIS processes to be applied to disease ecology work. In this case, SAS provides the product needed for GIS to work.