The purpose of dual mapping, or the combination or raster data with rasterized point data, is to produce the most accurate representation of the earth’s surface with regard to any spatial medical feature. These spatial features rely upon people to exist, and the environment to be produced. This environmental setting may be purely natural, or a figment of our imagination as people with a certain amount of personal psychology and cultural influence always influences how we feel about our self, and how we react to our physical and/or health state.
Dual mapping merges the traditional point mapping technique into the very broadly applicable grid mapping technique described elsewhere on this blog site. The purpose for doing this pertains to need to employ spatial error adjustments whenever mapping disease or health and using these maps to develop interventions. With square grid cells, it is possible for errors in location to be produce for where certain interventions need to take place. To prevent these errors, we reassign the grid data back to the true point data, forming a hybrid of these two outcomes and a final product that resembles the true point-area distribution patterns, making use of broader area results to draw local conclusions.
The applications of this method are several:
- Assigning environmental causes as either natural or demographic related
- Developing more effective interventions
- Prediction modeling
This technique represents the fourth level or generation of a research project I have engaged myself in over the years. There is a value to seeing smooth surface models depicting disease patterns, and a value to being able to utilize this data effectively for public health purposes. The spatial techniques that rely upon raster imagery techniques allow for full area assessments, not just point by point analysis. The outcomes presented here are useful across all space, and can be modified to take into account increases in the distribution of people spatially, as well as increases in numbers and density for very specific spots and areas. In some cases, once the behavior of a disease is well understood, we can simulate changes in a region and predict how a particular disease state might behave in that region, one of the best outcomes for this kind of work.
There is one reason I decided to develop a dual mapping algorithm. When mapping the spatial distribution of disease patterns, I noticed that the traditional belief we have about mapping diseases isn’t always true. In my beginner’s statistics course I used to teach what the book says, which is that looking at numbers doesn’t tell us much of anything. It tells us a lot about the events that are happening but doesn’t give us much of a perspective on what those numbers mean. So, to determine what these outcomes mean, we compare them with the status quo, the same events as they occur elsewhere. We compare them to some standard and try to determine if the place or group of people we are analyzing is greater or lesser than the others, the world as a whole for example.
But sometimes numbers mean something to us, even when they are very big or very small. It means something the say that 3 million times a year mothers pay to have their kids participate in a regular doctors visit for preventive care purposes, and that 90% of these take place involving just one particular kind of activity, whereas 495 our of 500 of those who don’t agree that they should bring their kid to the doctor’s office, live only in one small part of the country. The number 500 out of 3 million represents a very small percentage . . . 5/30,000 = 1/6000 = 0.00017 or .017% in fact. Barely an amount that might make us concerned.
Normally, we might ignore this number and write it off in the business community were it having to deal with sales, or resistance to purchasing something, or rates of returns of a damaged or unreliable product. After all, 99.983% are satisfied. But because 495 of the 500 noted are in just one area, that can have some meaning, and when viewed on the national map, we find it represents a hundred times more likelihood for happening in just one place, than any other place in the country.
Next, if we review that statistic of such small numbers temporally, again and again, and see that it represents a recurring problem, now we have a problem that we can benefit from by paying attention to it. We know there is this place where this same thing happens again and again, and through the mapping of these numbers can demonstrate that no program has taken the time to try to eliminate that recurring problem. The risk of not engaging in prevention of further activity is easy to understand–someone’s kids are going to suffer from this lack of sociocultural intervention. Even though to proceed with such a program means that a half a million dollars will have to be spent targeting such a small area, the business part of the health care world has decided that these few lives of children that could be potentially lost are not worth the amount it would cost to intervene. What family would accept that idea if that child was theirs?
Looking at the other side of this problem–the disease itself. We find the rarest infectious disease has some of the most costly programs out there to try and eradicate it. Small pox is a disease that has devastating effects if it actually occurs. The national evidence shows that the pattern by which it occurs is random, and only in one or two events for any given area, popping up in unpredictable patterns, demonstrating absolutely no aggregation, for now. The goal of public health care is to zero out these events in some way. But if we take a very close look at the sources for those diagnoses or claims for the disease, we find they are mostly due to unimmunized people migrating into the country, either carrying the disease or who come in contact with a disease carrier and subsequently get the disease as a result. We cannot stop in-migrating, therefore, we can never fully eradicate the small pox disease like we want. And if we did, that would only be for a short period of time, because the rest of the world has not eradicated their small pox problem. So, in order to fully eradicate this problem, we have to engage ourselves globally, not regionally, nationally, or politically.
Relating this same logic to another infectious disease with the same needs and requirements, such as measles, we find that the disease has a pattern that is dependent upon population density. This disease has cases that follow the status quo, and its related infection cases occur according to population density patterns. When we compare this to the numbers of people that refuse to engage in it, we find that single peak scenario again existing. The disease itself abides by people’s behaviors and susceptibility as a population/aggregates of people phenomenon, whereas the refusal to immunize your kid once again portrays as a very regionally specific cultural phenomenon.
The Small pox represents a situation in which hybridized approaches to eradication and/or prevention can be used. We use population density and disease rates at the small area level to determine how to continue our intervention programs for measles. Meanwhile, we also need to intervene on those who are against the immunize schedule in general, and so we produce a program designed to attack the problem this one particular region has with the socially accepted practice of refusing to engage in a specific public health program decided to minimize, not eradicate, the recurring measles problem that exists nationally.
In essence, what this means is that realistic goals need to be set for testing a program’s success, and then the upper limits of a successful program have to be better understood and added to the plans on how to best deal with a particular infectious disease public health issue. In the case of small pox, we know there are occasional eruptions of cases, which show no relationship to the refusal to immunize, so another theory or paradigm as to why it recurs in the diagnostics mapping algorithm has to be better understood, whereas with measles, we can tell this is a population density related disease pattern that has unfortunately human behavior ecology related causes for concern in just one part of the country. This means that a different program needs to be generated to isolate and disable that one particular high risk area.
The hybridization mapping process means that you map the regular ecological statistics along with the human behavioral statistics, assigning a weight to each to determine where the places in need of the most aggressive interventions are located. To do this, you assign a value to your findings with spatial analysis, in such a way that it can be compared with any other risk factor analyzed, and then the two overlain to produce a new map with the weighted results of these two metrics. You do this because it is easy to do an can be automated. With grid mapping, there are not special steps that have to be undertaken to make two seeming unrelated events, in terms of how the numbers are derived, able to the correlated for purposes of dealing with interventions.
We can add other metrics to this approach to analyzing health, by applying the grid method, in which the numbers part of this analysis is many times easier to perform. We can conceptualize cost for example, and add that to the schematic. We can map out places where the disease occurs, and how it impacts the work force, how it costs industries money in terms of performance, output and worktime, or we can use it to map out long term costs that people experience due to lack of participating in a prevention health care activity. Instead of childhood infectious diseases, we can relate this to breast cancer screening rates versus later stage diagnosis or death rates related to breast cancer. We can use the same model to define the risk for chlamydia relative to population density, average ages of particular segments of the population, median income related information, and types of community health and insurance programs serving each area.
Dual or hybrid mapping demonstrates why we want to make use of a grid modeling technique at the small area level to map diseases. When performed at a small area level per grid cell, the limits of accuracy disappear due to the standard 3-5% errors in place and location problem, and errors in analytical method problem. We traditionally assume a certain percent error in our programs, and always report that. The best use of an areal mapping technique to analyze spatial information is to chose your areal size with the following rules in mind:
- assign the size of the smallest area based on common sense applications or error assumptions. You know that side of the street may be off, so use street size as you limiter for example.
- model your spatial analysis in G stats and such to determine critical cell sizes, because too small an area can sometimes means the ‘too many zeroes’ problem could creep in.
- realize that the main reason to use this is to develop a krigged isolines rendering of the areas you are researching, and that you are looking for peaks and valleys in whatever is being researched. This information is then used to design the next intervention strategies.