Why don’t Parents Immunize? A poliomyelitis victim of the 1940s to 1950s.
(click above question for link to video)
The following are the most basic uses this technique can provide in population health analyses. (Note: some of these applications are covered elsewhere on this site as well and so may seem redundant.)
1. Defining the major goals and actions for a program.
To understand where to begin an intervention program, we need to know where the bulk of the problem exists. The above image depicts the spatial distribution of the refusal to immunize a child by a parent while both were engaged in a well-visit process.
There are several peaks worth noting, but the two major peaks are on the west coast, in the cities of Seattle followed by Portland, OR. The inland migration of this behavior is seen more in the Pacific Northwest than in the southwest, inferring a sharing of traditions and philosophies well into the interior of the country in the states of Washington, Oregon, Idaho and perhaps Montana. Also worth noting perhaps is the central Rocky Mountains small peak in the Utah vicinity, an area where cultural upbringing again appears to play a role in increasing the inappropriate refusal of care practices being researched.
Heading east, there is a ridge in the middle of the country, involving the cities along the Mississippi River and their suburban or nearest neighbor settings. The peaks in this ridge are about as high as those where the overall population is incredibly more dense, from Ohio through Buffalo, across to New York City.
If this behavior were randomly distributed, engaged in as a by-product of population size and density and nothing more, we’d expect a massive peak to exist in the New York setting, with diminishing sizes as one headed westward. What we in fact see is an increase in this atypical behavior as we head west, with the amount of this behavior increasing substantially as distance from the east coast is increased. Due to population differences across this country, being that the US population is not equally dispersed between the Midwest and Far West, we see a somewhat exponential effect of distance over space with regard to the degree to which this unhealthy behavior is engaged in.
Cultural differences in terms of knowledge base and belief account for these behaviors for the most part, with the more likely adaptation of alternative teachings and philosophies found as one heads westward away from the “East Coast” or “Status Quo.” Such a spatial relationship could exist as well for other community or social health related practices, such as reliance on alternative medical care programs or engagement in alternative preventive health practices as a decision made by the patients, not the clinicians.
2. Comparing two sets of data regarding related or unrelated events documented in EMRs.
Continuing on the theme of immunizations research, the 3D mapping method can be used to perform a simple correlation analysis of two events, such as the increased numbers of cases in an area relative to the numbers of bad health behaviors or activities.
The following 3D rendering of Chicken Pox cases in the country demonstrates the commonness of a parent allowing a child to experience this form of the Pox as a substitute for being vaccinated against its potentially fatal ally Small Pox. This behavior is fairly diffuse in the population, and represents a form of passive avoidance when it comes to deciding whether to have a child immunized or not. behavior.
The next two images depict the distribution of variola vaccine refusal, the option a kid goes through should he/she not bear a history of chicken pox, versus the actual Small Pox events or history of which that took place in this country. In a single area, at most 3 people are noted to have experienced small pox (located near the Tex-Mex border).
When we compare this with the refusal behaviors, the two obviously do not spatially relate. The Pacific Northwest center of childhood immunizations refusal is still noticeable, along with unexpected secondary peaks in the Chicago region and perhaps Denver region. There are also clusters of refusals noted in western New York-Ohio region and on the southern half of the peninsula of Florida.
The dissimilarity between refusing to allow a child to be immunized and the outbreak of that disease that could be immunized against is less when we review cases for measles and rubella. The measles vaccination is in theory the most common refused immunization action due to the purported effect of organically based mercurial compounds included in the vaccine. A public fear was generated about 15-20 years claiming that this cause autism in children. Fearing their child might develop this condition, measles vaccinations were refused for a while. Years later this fear was proven not to be true, but the social fear of the vaccine remains. The maps above show that measles is perhaps peaking in the Midwest, and not at all where the refusal to immunize is greatest. Rubella shows a more population density distributed set of results, with cases densely packed into clusters in the Eastern U.S., in contrast to the large peak of refusals found in the Pacific Northwest.
Mumps, being a bit more common in its occurance, shows a distribution even more like a standard U.S. population density map. Again, a Midwest peak exists however, now suggesting the possibility that an in-migration of this disease could be happening by way of the Mississippi River route. Pacific Northwest cases demonstrate a peak as well, but again, whether or not this is linked to mumps vaccination refusals of young children remains uncertain.
The pertussis or whooping cough illustrates a population density derived case flow pattern, with Midwestern peaking slightly higher than that of the New York to Carolinas megalopolis setting. The in-migration route along the Mississippi River is better demonstrated by the map, with some peaking seen in interior Texas and within Utah.
This next behavior demonstrates the lowest relationship between the disease, poliomyelitis, and refusals to immunize. Unlike the prior diseases, polio possibly has some ecological features influences its behaviors. Even if this is not mentioned in the literature, this relationship has to be suspected due to the very strong spatial relationship it has to the central U.S., at the northern end, in very close association with the Great Lakes. This suggests a natural ecology to the disease even before it infects people.
A tremendous amount of historical documentation on poliomyelitis can be found in early 20th century medical literature. During the 19th century, epidemics and various types of diseases possibly linked to polio were described as well, although often as fevers, nervous fevers, and the like. The first solid differentiation of a fever as being different from the several generic classes defined (constant or continuous, typhus, remittent, intermittent, yellow or bilious, and ague) possibly came during the 1860s. Prior to then, polio is hard to differentiate except through the details about its long term consequences, the fatality of which made it seem like a constitutional disease.
In all, these maps provide some very helpful insights into the diseases which children expected to be are immunized against by the age of two. The following summaries can be claimed based on this review of the above images:
- Small Pox is deadly, but rarely influenced by immunization refusals due to its scarcity. Its diffusion tends to be more random in nature and it’s usually found with low areal counts.
- Measles and Rubella also do not follow the rules of increased prevalence of the disease due to refusals to immunize children. Each has its own longitudinal pattern (east versus midwest versus west coast dominance or peaking), the human ecological reasons for such remain uncertain.
- Mumps and Pertussis are the most common infectious diseases to strike children regardless of where they are, across the United States. Mumps seems to peak in the Midwest. Pertussis peaks along the eastern megalopolis, probably due to its specific manners of spread in relation to the numbers and density of people.
- Poliomyelitis is the most ecologically defined disease pattern, and favors the Great Lake setting, with some clustering seen as well along the heavily populated Mid-Atlantic population setting.
The most important insight these maps provide deal with poliomyelitis. Reasons for the clustering of cases around the Great Lakes needs to be explored some more. Ecological reasons for this clustering stand out the most. Family-linked genetic histories for the settlers of this part of the country might explain it as well. In-migration seems unlikely, for if such were the case the midwest and Pacific Northwest should present unique peaks as well.
3. Choosing your “Weapons”, or Intervention Strategy
The above figure shows the relationship between the two separate studies performed, the first on refusal to immunize and the second a review of actual diagnoses of immunized disease events.
The distinct difference in how the refusals are distributed when compared with the actual cases that occur demonstrates the importance of the herding effect. We’d expect most of the immunized cases to prevail in the Pacific Northwest, yet they don’t. This is because of the very low population of people who are against immunizations in general. Even in the Pacific Northwest, where this behavior is many times more frequent than the rest of the country, we don’t see outbreaks happening due to these events.
In a review of Diphtheria for example, the outbreak that did happen was seen in the Rick Mountain region, not the Pacific Northwest. The midwestern outbreak was removed (explained and shown in the video on this presentation) in order to demonstrate the major recurring peak for this disease that exists in the southeast.
Diseases like Mumps and Rubella occur more where the population densities are greatest in the nation, not where theoretical risk due to refusals is at a peak. Polio has its greatest history as a disease around the eastern half of the Great Lakes region, a very ecological setting suggesting there are local physiographic, climatic and topographic reasons for its unique prevalence or simply history of diagnosis in that region.
An example of applying the NPHG mapping locally is demonstrated by this more focused review of the Pacific Northwest refusal patterns. As expected, the distribution of this behavior is a product of population density for the most part, although there might be some areas where it is more prevalent. If we increase the resolution of the data we’d be able to detect where these communal settings are, which surprising are distributed in both rural and immediate borderland urban-suburban setting and closed communities. In theory, these communal settings and their clusters of non-immunized children pose a threat to adjacent neighborhoods. No doubt, some of them are being monitored unofficially as well based upon my own episurveillance experiences of these settings.
The center for this behavior in Oregon is between Salem and the Eugene-Rosendale area. For Washington, it is near the centroid for the Sea-Tac area. In the eastern half of both states, clusters prevail around Spokane and LaGrande, Yakima and possibly The Dalles.
(3.5 min. video link of above incidence maps)
4. Planning Mega-Intervention Programs
Defining the best places to begin a program makes the best use of NPHG for a Medical GIS approach to this epidemiological public health issue. It is very easy to produce a map, even a detailed map with more input than the traditional techniques now in use, and to in turn not follow through on the findings that are presented.
The detailed map tells us specifically where to spend money on intervention programs. We can apply both prevalence and numbers mapping to the same projection and produce a hybridized version of the above data.
This hybrid approach to risk mapping has some major benefits that to date have been excluded from enarly all applications of GIS in use for surveillance purposes.
A hybridized-N-Prevalence, or N-IP, or N-PI, or N-squared version of any of the last three, will fine tune any programs being established. If you need to know where the most certain highest risk area is without even taking into consideration income, ethnicity and the like, the hybrid model of the disease and behavior patterns is the way to go.
Another routine process to take advantage of is the kriging method. Kriging takes into account the nearest neighbor effect and adjusts risk according to nearest neighbor event or case density. The following shows where the refusals happen, associated with a krig analysis outcome.
With this method, we see high N sites down to the exact lat-long, but also see density of these peak areas. Even though the 3D maps to the left do not display a risk for the east coast that well, we know this risk exists due to the krig results on the right. This can be done for N, Nsq, IP and PI maps and these four maps merged into a single product using a specific formula with weights assigned incrementally based on final N and IP/PI results.
The major rule to follow here, due to the nature of the formulas and algorithms used to generate these results, is that the more demographically based a spatial pattern is, the more value that can be assigned to this hybrid mapping approach, in terms of finding peak areas of bad health, poor compliance, greater likelihood for outbreaks, etc..
I have also applied these techniques to cost related analyses and type of health care coverage (MCD, MCR,CHP) and use it to develop a formula grading the different forms of non-compliance seen in certain SES, sociocultural or low income areas. Extremely well focused outlier behaviors tend to produce the greatest peaks in metrics like delayed refills, deference of care, etc. Suburban, intercity areas demonstrate the highest rates of patient-generated fraud; the proximity of additional services to take advantage of is the determining factor with such cases (i.e. fraudulent refills of pain reliever medications by patients 25-50 years of age). Due to the Southeast higher density of 65+ year olds, we find the two extremes in compliance behavior peaking in the same part of the country–prescription drug compliance for LTC medications and refills, versus non-compliance or change of PCP behaviors at a peak in this region as well. In the case of Severely Mentally Disabled patients, there is a twin peak age for onset and first diagnosis for men, versus a single peak age of women receiving their onset/first diagnosis. This gender-specific delayed treatment behavior is universal in the health care field, but differs in intensity regionally across the United States.
In my study of regions (the states used to define each region are for the most part concealed for partially concealing places and results), there were a number of very distinct regional differences noted, such as the overwhelming number of older age patients in the Florida region, and the matching of their disease patterns for ages 65-85, with the same medical problems suffered by younger working age populations residing at the NY-CT border region.
In the above examples, we expect cold-induced hemagglutination to be a northern latitude event, but find this difference in distribution is not clearly indicated. We find the northern mid-Atlantic states to be where the most childhood and young adult-age may be happening. As just mentioned, AMI is as much a risk in the older working class of NY-CT as it is a reason for visits demonstrated by retirees in the Florida area. Not too surprising, the “healthiest” parts of the US according to ICD reviews are in the Rocky Mountains further westward. But this is only for the Rockies and westward as a region. Numerous behavioral ICDs demonstrate peaks in these areas as well, showing SES and sociocultural behaviors can also play into the overall health of these regions.
When it comes to social and physical health of children, we find that evaluating every childhood medical condition, in as detailed a fashion as this review of immunizations and immunization refusal took place, is essential to improving the overall program out there for improving childhood health and schooling. We can find specific areas to target for each of these social concerns, and save money by developing what are potentially more effective and successful intervention programs using these NPHG methods.
link > 25sec Video < link
on rotating map of the distribution of Homeless Kids identified in my last national study of a selected dataset
Shaken Baby video on distribution of cases
To be Continued . . .