3dsas4

Hot Spots for Lead Poisoning over the past 10 years  (produced in response to a local news release, in 2018)

.

The Newest Priorities in Health Care must become Spatial Modeling.  But to advance to this levels within our current system, a number of major changes still need to be made.

The first of these changes in developing the knowledge based for successful spatial modeling to commence within health care.  In nearly every aspect of health care and epidemiology taught in academia, spatial modeling takes second stage to standard 75 year old methods of dealing with public health and the improving health care within the system.

The businesses with the greatest potential for engaging in the more advanced method of analyzing population health–the insurance companies–have never taken on the task of implementing GIS into their day to day analytic processes.  As a result, they remain dependent mostly upon older formulas and methods for improving programs.  And they perform these tasks at very slow rates, so slow, that the progress made in health care practices is never a result of their engagement in health and disease monitoring.  Most advances in health care come from the technological and chemical/pharmaceutical and treatment advances made by the businesses devote to medical equipment.  In the meantime, insurance companies, the allies of health care industry, have done little to advance the quality of care we receive.

 

 

Based on ICDs and/or Lab results, Rare Diseases, Genetic Diseases, Genome and Congenital Disease mapping are standard parts of the new Disease and Diagnosis Tracking system

.

Now, some want to argue that their engagement in prevention and intervention programs is proof that my claim is wrong.  However, most of these processes are required of the health insurance industry, not a product of their own creativity and design of new practices.

If insurance companies had taken on GIS when that was first possible, sometime around the Windows 3.1 to 3.2 period, ca. 1995, they would right now be completely engaged in spatial health analysis, and some of the more advanced companies be engaged in the 3D and prediction modeling of population health, at such levels that urbanization programs and programs designed to improve rural population health could have been established by now.

 

Picture2

The Ranking of Medical GIS Implementation within a Care Setting

Back in 2012, I evaluated the levels at which health care businesses were engaged in spatial analysis.  I designed a 1 through 10 scoring system for a program, using the various applications of spatial analysis processes and tools to the programs they provide.

Level 1, score 1 engagement was one step over doing nothing at all–it consisted of manually mapping, but mostly tabulating your data of health, much like the public health institutions and agencies have been doing since this process was first put into practice around 1840.  Talk about being behind, this puts many insurance companies more 150 years behind their potential and analysts and inventors or new research processes,

Level 3 to 4 is about where many smaller businesses engaged in health insurance now work.  They are probably trying to define high risk populations, income and race-ethnicity groups, and using census data to locate the highest risk areas, and then, developing intervention programs targeting these health issues.  The most basic of these issues are the standard we always hear about–the ones that just won’t be reversed and go away–namely, the low income related poor health/high cost for care related problems, the simplest of unhealthy human behaviors like not engaging in preventive health care behaviors, like exercise, deciding to cease the use of drugs or engage in smoking, reducing your STD risk, developing early recognition programs that work for preventing breast and colon cancer.

 

 

But maybe the management of the most basic human behaviors is not the best route for health insurance companies and programs to take when trying to reduce the costs of disease and immediate and long term care.

Another problem the typical health care programs, companies and even leaders have in designing an better program, is broadening the focus of what they do.  For decades, the same preventive health plans have been at the forefront=smoking, drugs, drinking, nutrition and diet,  lack of exercise, work and unemployment related stress and their impacts upon domestic life and personal safety within the home settings.

PMIhumansphere

But wait a minute, I was just talking about some of these earlier processes–ah yes–the major problem is not that these programs we design do not work; the main problem is that we have failed to redirect out focus into more diseases and broader programs for prevention that better fit many local population health needs, not just the same handful to dozen or two health problems.

Now, again, some are going to argue with how I am wording this criticism.  Annual HEDIS reviews of most plans cover thousands of health care related statistics (if you accept the simple monitoring of an RX list as representing over a thousand metrics being monitored.)  Similarly, annual reviews sponsored by NCQA, as an example, require successful program to engage in anywhere from a half dozen to two dozen or more monitoring-intervention focused health care, disease prevention programs.

BUT, (and yes that’s a big BUT!), these are pretty mush the same programs again and again.  There were similar programs in place decades ago, like the annual anti-Tuberculosis campaign that took place nationwide around holiday season, or Jerry Lewis’s Easter Seals program, or the numerous cancer prevention programs that were developed.  Some of these programs worked, others remain important to the national public health programs in place.

ChildhoodImmunizations

An example of Analytics is the Immunization Surveillance program.  Case outbreaks are evaluated historically and monitored.  Outbreaks lead to updates and reevaluation of prior GIS analyses performed.  Both ICD related to the disease and refusal to vaccinate are monitored.  Cluster analysis is performed and simple analysis of area by population features.  Clustering and immunization refusal can be linked to local SES and/or cultural patterns.  This helps to define the ranges and density in frequency of refusals required for critical levels to be reached, for an outbreak to begin, or the disease to most likely be spread to a new site.  Risk factors can then be analyzed for regions, based upon outbreak numbers and risk.

Reaching Level 6 – Comprehensive Analytics

During the early 1900s, we had the excuse of lack of EMR and no GIS, for not engaging in super-active public health improvement programs nationwide.  We don’t have that excuse today.

We need program that don’t just monitor some token (mostly symbolic and obviously linked to generations of ineffectiveness) public health problems, like the several highest cost programs nationwide devoted to asthma and diabetes prevention.  In the 60’s, diabetes programs focused upon genetically produced Diabetes Type 1 patients; today this program has to serve tens to hundreds of thousands more people with the human behavior-induced (obesity-linked) form of diabetes onset.

 

 

In the 1960s, it looked as though epilepsy had a chance of being cured in a lifetime, according to Parke-Davis. Large contributions were made to the Epilepsy-Find a Cure campaigns.  These campaigns are not longer common to the TV screen, which in part is due to more than 50 years of ineffectiveness.  Sure, medications have improved, and some treatment-preventive methods make for safer long term outcomes.  But, the unknown possibly gene-based reason for onset, has only delayed progress in treatment, or at least caused it to come to a standstill.  So instead, those who have epilepsy, must deal with life in terms of two things–improved quality of life, including increased engagement in preventive self-administered care practices, improved interactions with friends and neighbors at the community level, and hopefully (but unlikely) reduced cultural based reactions to the person with the diagnosis of epilepsy (it is still to find a job that will allow for seizures to result in needed breaks or days off).  By first hand experience (I dealt with this for easily twenty to thirty years, including the controversial surgery before its time), social knowledge and cultural definitions still remain a stumbling block in allowing epileptics to reach their fullest human potential.

So, it is up to any new programs now being developed to fully assess all of the health features of their populations, not just engage in the needed programs devoted to asthma treatment, mental health [programs, diabetes and lifestyle, or even programs that ensure patients remain devoted to whatever programs, activities and prescription products that are needed for them to stay healthy.  Current programs speak about having these programs, but due to cost, knowledge base and staffing, only those select few examples exist.

3dsas5

For a certain ICD, or medical condition such as a lab value, or count /prevalence of individuals with a certain risk, can be evaluated based upon: Number of Patients, Visits engaged in for the ICD or risk activity measure,  and numbers of procedures (complexity of practice) a patient goes through to receive care for the problem.   TB cases for example, evaluated for average number of visits per patient provides a more useful rendition of the Patient-Visit-Care data, about the clustering or patients and cases, relative to cost and demands upon the system.  Plans can be made based on the small clusters of cases with the largest numbers of visits per case average, in other words–those regions and cases with the highest cost.  To document the need for change, and later success, a thorough review of the total population for its top ICDs and cost/visit relationships are required.  For example, when the epidemiological transition modeling of the ICD data is applied, as in the below figure, evidence showed us that TB patients demonstrated the highest visits per patient over time, but have also undergone one of the most successful reductions in visits per year, over the past 12 years.

Epitrans_BlackMalesEpitrans_BlackFemales

To fully understand the health of its population, an insurance company or managed care system must produce complete reviews of its patient’s health status, annually, and to apply this same process retrospectively with patients that have been in their care for years.   On many of my pages about spatial analysis of health, I defined these various processes.

For a company or business to go above that Level 3-4 that I claim they exist at, they can easily reach what I am calling that Level 5.5–which is characterized by full engagement in environmental and population health monitoring, and experimenting with spatial analysis and the applications of spatial mapping software to the program.   To be a successful Spatial Analyst centered program, a company must have developed the knowledge base and programming needed to engage in spatial analysis of anything and everything within the EMR.  That constitutes the level 6 position in my 1 to 10 scale scoring system.

 

 

Infibulation or Female Genital Cutting/Mutilation is the first Culturally-linked “health” practice analyzed across the system and mapped, before initiating a full scale monitoring program, which includes the development of a plan and policy or set of rules required to produce a HIPAA compliant database (5 levels of compliance or use were defined, from full internal to full external data sharing), the definition and use of different parts of that database: a) name and registry data, levels 1 and 2 use only, b) visit history (temporal data and visit type only, for analyzing counts, frequencies, and ratios), c) procedures history (lab procedures and all other events done in care ranging from education, to diagnostics, to treatment and rx, etc.), d) results/outcomes of special testing (in general), e) Special lab results, microbiology, etc., f) pharma hx and such, g) ICD primary research ICD and allied or related ICDs, h) lifespan ICDs (diagnostics, procs, before during and after the medical experience).  Others are pulled based upon purpose and request.

Example of this I see at work every day, include:

  • Pulling and massaging the EMR data and then mapping this data (within a half-day or less), in order to see where possible risk areas or subpopulations exist, engaging in this in response to a query from a PCP or some news report (i.e. by development of lead exposure >20 maps, or the maps of kids whose parents refused immunization after the recent measles outbreak in Rockland County.)
  • The development of standard base Population datasets, for immediately comparing and contrasting prevalence and incidence for a topic, again “on the fly”, or looking at recent changes in patient types in the ER due to a new flux of immigrants.
  • Finding the hots spots for a given bacteria or disease type that demonstrates seasonal and cultural / SES related patterns of re-eruption, based upon the last ten years of EMR data, in response to the recent legionnaire’s clusters detected, or the new adenovirus outbreak, or concerns that medication resistant TB may still be able to re-emerge in the future due to certain immigrant-rich communities.

 

 

This is the model developed for testing all of the EMR data, which includes notes (open text) and is evaluated as a non-structured data NLP process, as a NLP-process merged with a structured data search and extraction process, and a purely structured data search and extraction process (the methods most often employed).  NLP is show to increase the amounts of pulls of researchable data from two-fold to ten-fold, depending upon the research questions.  Seven data mining and extraction sql searches were merged into a single data pull run to accomplish this task; the total process takes place typically in three to six queries placed around the initial raw data master query.  The Data transformation process (deidentifying the data, making the dataset HIPAA compatible at one or more levels) includes a removal, change or recoding of name (removed), patient_id (replaced with a math equation defined fake_id), address (replaced by lat-long, if needed), zip code (may be lat-long instead), facility (unique letter or unique letter-number coded), provider (recoded and deidentified), certain dates (i.e. birth and death dates, may be substituted with age or replaced by just year, or year and month),specific rare ICDs (placed into clusters or groups).

In effect, there is no excuse for the current status of GIS implementation in health care.  At the ESRI level, some changes were made administratively with this goal and the related tasks in mind.  There are many examples of GIS implementation that can be produced.  And the Health GIS news is always sharing examples of new programs being initiated.  Yet, none of these programs demonstrate the level of sophistication I am talking about with amount and types of use for GIS.

GIS needs to be implemented as a preventive health tool, a prediction tool, a cost analysis tool, an administrative planning tool.  It can be used to evaluate, routinely, patient-visit practices.  It can be used to see if there are peaks in product use or whether a change in medical claims policies (such as defining a new ICD10 resulting in improved DMDD care, for Community A vs Community B) actually changes the practices that are happening.

With the programming and methodology for GIS use established, any program can transfer the data from ongoing research programs into the spatial analysis aspects of health care.  In theory, annual reports can even be implemented, for monitoring just Hispanic or African American Care related needs , or evaluating if changes are happening at the Culturally-related, Culturally-linked vs. Culturally-bound disease patterns level (which I have explain in detail several times, elsewhere).

FemaleSpouseAbuseMap1Closeup

Most of the programming that is required for the spatial analysis of health exists in some form or fashion across the IT world. Yet, to date, most of the agencies functioning as part of the health care system have failed to develop programs that successfully analyze and report on all that we need to know about health.

In the recent past, a number of Medical GIS people have argued with this statement. In part, they are correct because for each of their systems, some part of an ideal Medical GIS program has been perfected and is ongoing.

But this statement is not about those successful parts of Medical GIS that are put into play. To produce a successful program, an agency or company or corporation has to be able to implement spatial analysis at all levels. This not only includes the basic descriptive analytics that many places are engaged in. It also includes full scale analysis of the population that is managed, its ICDs, its health habits, the Charlton, Elixhauser, and other standardized scores for evaluating patient health risk, the costs for care on an individual in relation to age, ICD, SES, and procedure basis, the frequency of use of the system at all clinical and administrative/preparatory levels, the changes in numbers that occur over time, the development of machine logic NLP review processes for physicians’ and patients’ notes designed to accompany CPT and diagnostic code generated output, prediction models for any or all of the above, the semi- to fully automated regression analysis methods.

 

 

The other concern of high priority has to relate to HIPAA and patient confidentiality. To deal with this, five tiers of data development and processing were designed, to make data compliant with internal, external, and combined internal-external EMR data users. This includes the design of algorithms that produce effective deidentifiers (fake ID substitutes), and recode certain data forms that otherwise could still the possible identification or a given individual. Degree of data change for each of these levels is focused on things like exact DOB/Deathdate, exact location information, even exact ICD identificiation in some cases (aggregates are used instead where appropriate).

Before implementing this new GIS, a number of alternative spatial analysis processes were developed and tested during the past three tears, using EMR data (examples of which I have posted), using SQL and SAS (non SAS GIS) to develop, utilize and test a variety of new spatial methods to extract data and transform it, making it HIPAA compliant, well before importing it into SAS or GIS. This new GIS demonstrates some of the benefits of still managing complex EMR data within the GIS Workstation and Reporting environment.

 

 

Lead Poisoning Cases, Children < 5 yo

Due to the recent reports published in the news about falsifying documents pertaining to health safety inspections performed in low income settings, a means for evaluating the results of Lead Exposure blood tests performed on children under 5 at the time of testing was developed. Only unsafe values above 20 were checked, and mapped for this first run.

This same process may now be used to evaluate any datum or binomial query about NLP evaluated contents of an EMR note, be it of structured or non-structured data form. Three queries were developed based upon structured data defined expectations, a combined structured-non-structured result that may be considered a “hit”, such as a procedure and note indicating the result, and a completely non-structured text entries based NLP data review.

 

 

The next level of this work is to design a standard way to report upon the entire histories of a given set of patients. Four key databases, which are independently analyzable, yet linkable, have been identified for this process. They define the patient and population (1 unit), the ICD history (5-7 units per patient avg), the Visits schedules and times (20:1 ratio, by type of visits as well), and the Procedures or Care Events (labs, imagery, educational steps taken, discussion types, interventions) linked to Visits type-time relationships (50-2000:1 patient ratio). In addition, special tests are identified as well to be of value in separate reporting, such as microbial/infectious disease history, psychological/and mental health history, etc.

The ultimate goal of any Medical GIS program, implemented by an institution or healthcare plan, should be the development of an effective combined intervention-prediction modeling healthcare analytics program, designed to complement and add to the other analytic programs you may already have in place.

 

 

H+H1

The current Healthcare Management programs nationally have remained in this unfortunate state of limbo over the past decade.  I make this statement based on a series of studies I performed in 2012, the results of a survey I posted for 3 years regarding GIS implementation by facilities and programs, and the current state of progress depicted by published Medical Geography articles.

There are some new processes underway, but in spite of their increasing number, not a single facility has emerged as the leader in this field.  That is because all research groups involved remain in the experimental stage in their work.

This is depicted in the above process defining flowchart I established around ten years ago, for scoring where a program of may be placed in my 1-10 scale for evaluating GIS-Remote-Sensing-Population Health levels of research, and applying GIS to health, at a large combined complete EMR database, and agency or corporation level.

Back in 2010/2012, some might recall I successfully mapped the entire United States using a national database of data related to 50 million – 120 million patients, depending upon the sets of data being analyzed, mostly focused on age-gender relationships and several thousand ICD groupings, for the most important ICDs in US health care.

In the more recent work, instead of size of population, I took advantage of a combination of population size and time details to evaluate care features.  This allowed me to see relationships that exist between types of care, and the details of how patients and physicians engaged in that care.  When considered together, the amount of care analysis reviewed at the total (gestault) level for care per patient, per healthcare problem, per personal health status, etc, ranged from just 1 to 21 years of care, with 1-16 years care as  a better assumption, due to lack of completion of data entries per patient, at a per patient per year “continuous care” related level.

So, of about 12-16 million people reviewed, 50-120 M combined patient-years of data were reviewed.  Multiple that value times 7 to get and estimate on the numbers of visits reviewed, and times 40-100 to get the total numbers of health care activities evaluated at the clinical level, minus the details, for whatever practice activities that physicians were engaged in.

With such a huge dataset evaluated during the recent stage of this work, a number of important types of studied emerged.  These are some examples.

PhysicalCulturalGroupSubclasses2

PROJECT 1

Epidemiological transition, amongst Religious Groups. 

Goals:

  1. Develop general religious group maps of the region
  2. Develop three tables per group and gender:  2015, 2010, 2005
  3. Produce maps for specific health behaviors/ICDs/etc.

Major Groups:

  1. Non-committed (agnostic, atheistic)
  2. Muslim
  3. Jewish
  4. Asian/Cultural
  5. Christian (merge two subgroups)
  6. Natural Theologians
  7. Unknown (include?)

 

Epitrans_BlackMalesEpitrans_BlackFemales

PROJECT 2

Recategorizing ICDs

Goal

Recategorize ICDs into well know subgroupings, and then develop newer categories for major comparisons

i.e. emergent vs non-emergent, chronic disease versus non-chronic, genomic/genetic related vs others

Steps:

  1. Produce tables of emergent vs non-emergent (with or without gender; replace gender binomial with E vs non-E.  Perhaps relate to Race ethnicity religion groups
  2. Produce maps of the same
  3. Duplicate the first two for Chronic vs Acute
  4. Duplicate the steps 1, 2 and 3 for genomic and genetic/congenital, versus the rest

 

SSanalysis

PROJECT 3

Applications of QGIS

Goal

Produce a variety of QGIS maps for the region, for different levels of data mining, restructuring, analysis and evaluations

Steps

  1. Demonstrate utility of this process across all forms of EMR data, including basic ICD, to Lab results, to risk related lab data features
  2. Produce maps of Height, Weight and BMI from data already pulled
  3. Reassess and remap for age-gender, add race-ethnicity-religion (RER) if time allows.

 

FGMmapping

PROJECT 4

Cultural Ethnicity Influences

Goal

Identify Cultural-Ethnicity related ICDs of the four type previously defined by past studies (Cultural/Geographic Infectious disease and genetic), Culturally-linked (possibly genetic. systems related), culturally-bound (mostly behavioral health related), culturally-related (high risk due to race-ethnicity in the US).

Steps

  1. ICD analysis
  2. Groupings
  3. Tables and/or mapping

FemaleSpouseAbuseMap

FlowchartStatus

This slideshow requires JavaScript.

 

I am current developing a NYC GIS research process for evaluating 15 to 20 years of health data.  Unlike most medical GIS programs, the data in this system is quite complex, and includes standard medical data along with case and practice specific structured and unstructured data components.

A preliminary review of the billions of datum rows in the EMR for this project suggests that there is an average of about 1.5  to 2.2 million patients appearing in the EMR due to new healthcare activities per year.  Over the years, the total number of patients found in this EMR, which consists of 8 separate parts, is approximately 20 million people.  However, there is considerable overlap in the patients from one network to the next, and it is estimated that in fact this data set probably represents 11 million to 15 million people.  Only a small percentage (about 7-15%) of these patients have 10 plus years of care.  In spite of the lack of a full lifespan’s worth of care for a single patient, these patient groups with 10 years of continuous care can be used to define standard processes involved with health care, of any sort.

The EMR’s components may be group for lengthy reviews as

  1. patient data (mostly structured),
  2. Patient temporal diagnosis (ICD9 or 10) data (chronic and acute, date of onset, completion, etc.)
  3. visits data (structured, mostly numeric, date and time data; used to evaluate type of visit, time/date relationships, length of stay or service, and place(s) of care and their sequence, temporally),
  4. procedures data (structured for the most part; well defined billable actions such as labs, xrays),
  5. events data (structured and non-structured text and notes data; activities may be interpreted like procedures, but they are usually done by the physician as additional care processes, and usually not billed, such as administering a psych test, providing the patient with a pamphlet on STDs),
  6. outcomes of results data (structured and non-structured; highly detailed findings; includes activities engaged in with these items, such as residents reviewing the xray, or students evaluating the blood sample and providing their diagnosis for review by their mentor)
  7. additional outcomes or results data (focused on a unique topic such as pharma, sample testing, histology evaluation results, genomics, or bacteriology)

 

Approximately 72,000 health care procedures are detailed in this EMR.  There are well over 300,000 events (many educational materials) per facility.  Since patient visits are temporal, they are crucial to developing any pattern for evaluating patients health care processes.  The visit types most often used are: office visit , emergent care, inpatient, regular visit, ambulatory surgery (other surgery is in inpatient), other visit (for xray, other unique item), referrals, communications by phone or mail about patient’s history.

A typical average visits per patient per year ration is about 6 to 10 per year.

A typical average procedures engaged in for care, per visit is about 40 to 100, increasing with age and complexity of patient’s health.

A typical inpatient stay can last about a week and consist of 1,000 to 10,000 procedures (with repeated ids), and each procedure generating as may as 1000 lines of data includes structured results, and non-structured additional information (who was engaged, notes on related issues, etc.)

The following diagram is used to depict these datum relationships

basicpatientevaluation

A collection of visits for a single problem may be modeled as follows:

leafstemmodel

Each visit has measurable components and time elements (PVEP = Patient-Visit-Event-Procedures)

measuringavisit

The following is an example of a lifetime of care model (each oval is a visit):

Case4

In March 2018, GIS was added to this patient care-disease modeling program.

— an update on this site’s stats — 

00_aNYCHealthGIS

Step 1.  Social, Societal, Demographic and Population Base Maps developed

H+H1

Step 2. Early application of data to QGIS

H+H2

Geography of Site Visitors for from 2009 to 2018 (so far)

baMPH_01

 

All Time most popular sites:

 

baMPH_02

Popular in 2018

baMPH_04

New Mapping Technology videos visited on this site (most are in Youtube):

baMPH_03

H+H2

For more on this technology, see the SAS generated videos produced by programming mapping in SAS (not SAS GIS) about ten years ago; this technology is complemented by the use of the new GIS workstation projects.

For examples of these visits, search through these blog posts, or see the collections of videos put on file at:

Topic:

Childcare

Migratory Disease Patterns

 

NPHG (National Population Health Grid)

My programming for this work is kept available at the following site, rarely referred to and/or visited:

Managed Care Innovations (SQLs)

https://wordpress.com/view/managedcareinnovations.wordpress.com

00_HowtoimplementMedGIS

The analysis of electronic medical records and the operation of a geographic information system are two very unique sets of skills, which when combined require time and finesse to get to an advanced level with the analyses.

In many ways, the complexity of each as stand-alones are comparable.   There are multiple levels of format, complexity and relationship to medical data, which in this EMR system includes “dimensional” databases devoted to

  1. Patient/Demography,
  2. Visits,
  3. Procedures or Events (treated as at least two levels in my models),
  4. Results, outcomes or findings attached to the prior two, and
  5. Location (x, y and z)
  6. Time (t).

In a traditional GIS, this is comparable to the different “layers” that may be overlaid for analysis, and the “dimensions” at which maps or images can be produced and evaluated using this method.

00_TraditionalGISlayers

To develop an effective data storage, surveillance, analysis and presentation tool, equal amount of time and effort need to be spent at the EMR and GIS end.  You comprise value and quality of spatial produce when too much time is spent on developing an effective EMR, without engaging adequately the GIS-related spatial analysis and presentation potentials of the program.

In traditional GIS classes, twenty years ago, the ability for researchers to use GIS to perform their analyses, besides the traditional Excel with Add-ons, or SPSS, or Stata, or SAS, or S+ to perform the analyses.  The early impressions were that GIS could very well make these older, very traditional analysis programs obsolete.  However, most recent changes in GIS and Statistical Analysis tools demonstrate the inability of any of these traditional software brands to hybridize their technology with other brands out there.

This is exemplified by the recent change in SAS from the traditional SAS (8.* to 9.*) to SAS Enterprise, without the traditional programming atmosphere offered by the older products. (Hopefully the new SAS will improve, but for now it’s overall value has been reduced by one or two application related levels).

When we look, for example, at the years long attempt to add a GIS option for SAS, in the form of SAS-GIS, the results of this product were quite upsetting, due to the quality of the output and figures, no necessarily the meaning and value of the analytics itself.  SPSS is now compromised by its level of complexity, and breakdown into multiple sections of the software subscribed separately.  Like a number of drawing programs that did the same, obtaining these forms of software became impossible for smaller groups, making it necessary for more affordable, more productive tools to be developed.

00_StepsforcreatingaNYCHealthGIS

Due to these recent changes, there may never be a perfect mapping environment for adequate population health surveillance and analysis, in fact, a form of EMR-GIS that is equally valuable and applicable across potential EMR-GIS settings, be they linked to big business or small business, large or small EMRs, insurance companies or much smaller institutional healthcare settings, large area or small area focused operations, large npos or small npos with minimal funding to support their goals and plans.

Transformed Data

Next, it helps to interpret these two large parts of the EMR-GIS system (EMR and GIS) in view of their smaller parts.

The EMR currently in use within my system makes use mostly of SQL and SAS.  This two-tiered method of pulling and then evaluating data was successfully developed and implemented to perform the prior Big Data spatial health projects that were posted several years ago, when the national health data was analyzed (varying from 40-120M U.S. patients, 1-2 billion records per year, using SQL in a Teradata datapull work environment), then exported, filtered and turned into quantifiable location data, and then mappings using a SAS polygon-grid mapping program that I invented (it took only 15-30 minutes to produce a national map).

04_DataPINRequirements

In this newer system, more data crunching and redefining is done as part of the initial datapull process.  All data are geocoded, and made HIPAA compliant as part of this initial datapull.  This means a number of basic features of the data have to be maintained, such as no personal names, SSN, patient or member IDs, phone number, exact address, etc.  These are all changes in the initial data pull process.

For example, for research purposes race and ethnicity data are converted to the U.S. Census standards, and missing or unknown data meaning converted to the appropriate subgroups (for example, as AfrAm or BL, Wh, As, SoAs, Al/NatAmer, HI/Pac Isl, . . . “N/A”, “unknown”, “no response”, “refused to answer”, “other” may be coded together or as their own unique subgroups).  As another example, ten to twelve groups were defined for “religion”, referred to as Religion Groups (file column name “Relig_Grp”), and grouped as [I am being quite non-specific and incomplete here]: Christian, Christian-related (sects), Jewish, Muslim, Buddhist, Hindu+, Agnostic, Atheistic, Natural Theologians, . . . None, Other, . . . Unknown, Not noted . . . etcetera.  When appropriate, all locations/specific places, proprietary and specific names are also recoded or eliminated.

The end product of this data pull is “transformed data”, which has four levels of HIPAA related compliance.  This initial pull generally results in Level 1 data–which means that it would be difficult for an HIT individual to trace it back to the actual individual, without knowledge of the SQL transformation programming that was used.  This data may then have to undergo further transformation or aggregation, depending upon it uses and needs.

Transformed Data Levels

  • Level 1 is intended for internal use and may include generally acceptable information, for example, a list of patients with names, DOBs, MRNs, addresses and phone (never SSN) numbers to contact. 
  • Level 2 is no name, no address (lat-long instead), and preferably DOB converted to decimal age, for example, internal studies that are part of the system/network, but may not be fully active or engaged at the ally facilities. 
  • Level 3 includes the above, plus recodes or removes all facility identifiers and PCP info, as needed, converting these to unique identifiers to something that can be decoded later when needed; this is intended for external use (but may include location or facility for certain outside npo activities such as quality assurance or improvement checks and program grading projects). 
  • Level 4 is aggregate data (i.e. adequate for unmonitored course or college level training), with all of the above features, and further limitations applied when needed in complete compliance with location related features, as defined by NIH PHI guidelines. 

03_Requirements

 

Software

Up to this point, some form of data storage process and a GIS are mentioned as requirements.  A SAS may be used as a substitute for the GIS, assuming the programming I promoted here and elsewhere– areal (i.e. zip. census block) and grid (namely square or hexagon) spatial analyses, without basemapping — could be implemented (no SAS-GIS add-in is needed).  In many cases, the data pulls are done using some internal software and/or sql.  In the above figure, other steps are required to implement some standardized surveillance-analytics program.  They define the most basic requirements.

Setting aside the selection process for a GIS for the time being, knowing your potential data and information resources for one of the most essential parts of this process.

01_NYHealthDataLayers

Data

Again, using the NYC setting as the example, there are several sources for basic information data and spatial data available for setting up a surveillance analytics spatial workstation.

03_SpatialData_Resources

The better know sources for data for spatial analysis work are the GIS companies and resources, with ESRI perhaps the better known, and a number of Federal, State, regional, agency related resources serving as the primary sources for the actual GIS spatial (point, line, polygon) shapefile data.  Knowledge related resources or datasets (directly or indirectly spatial) that can be linked to shapefiles comprise the rest in the above listing.

 

06_GISdatatypes_Spacevsother

The following sites were used to access the base layer and background mapping data for establishing this EMR-GIS.

  • NYC Oasis Basemap (to locate sections for a study; to review features of that section): http://www.oasisnyc.net/map.aspx
  • NYC Open Data: https://opendata.cityofnewyork.us/
  • NYC Open Data, datasets browsing page : https://data.cityofnewyork.us/browse
  • NYC Planning (department of transportation) maps and baselayers/data: http://www1.nyc.gov/site/planning/data-maps/open-data.page
  • NYC Map Tiles (background or basemaps, i.e. older maps digitized): https://maps.nyc.gov/tiles/
  • NY Orthoimagery maps (also base maps): https://orthos.dhses.ny.gov/  see also http://gis.ny.gov/gateway/mg/napp_download.htm
  • USGS Earthexplorer: https://earthexplorer.usgs.gov/
  • Landsat,  downloads/purchases:   https://landsat.usgs.gov/landsat-data-access
  • Landsat general page :   http://www.landsat.com/
  • Zip Codes (some ESRI links): https://www.zip-codes.com/zip-code-map-boundary-data.asp

 

06_GISdatatypes

NYC Open Data site is the source for most of the shapefiles and spatial data, which are used to link EMR data to.  There are spatial and non-spatial (location or area related) data available at this site.  Most of this data is reliable and useable (can somehow be linked to a GIS, either by lat-long-x,y coordinate, or place name/zipcode/etc.).

02_NYOrthoimagery

The Orthoimagery page provides rasterized datasets that in some ways are comparable to the use of Landsat imagery (though not exact as LS or NDVI ,etc), and may be evaluated using some of the same remote sensing methods or strategies.

 

00_NYHealthData

Medical GIS

To date, there has been a number of barriers preventing the adequate application of GIS within the health sector to the implementation of a facility or program based GIS devoted to monitoring, surveillance, intervention and other health program management activities.  At the institutional and insurance company level, this barrier exists due to the lack of knowledge and experience within the Medical Records or EMR data management system.  There have been numerous examples of GIS utilization attempted here and there throughout the system.  Still, to date, there is no clearcut leader in the field making use of a combined EMR-GIS data warehouse management practice or procedure.  For nearly twenty years, EMR-GIS practices have remained at the experimental level, when interpreted using the process I developed years ago just before analyzing the national U.S. EMR data for the first time at the 50M to 100M patients level almost 10 years ago (see nationalpopulationhealthgrid.com).

The result of this disengagement of GIS expertise in the field is literally the stagnation of health management at the local and regional government program, insurance program level.  Is it possible that the limits have been reached for singularly trained people, forming a team of varying experts, who when even working in groups are unable to take their team performance to the next level of achievement?

The barrier to health improvement has often been related to this lack of GIS implementation.  The combination of changes in software, hardware, storage technology, data analysis speeds, data build and restructuring speeds, have partially limited the ability of “experts” to make any long lasting changes.  Since all of these parts of the EMR data technology undergo changes and development at rapid speeds, by the time a process is developed for such a program, the tools and information have changed, the older knowledge base is outdated, desires to patent or own a particular process get outdated (half of the 17 years patent rights may be gone), and the health of the people may have even changed, making certain areas of focus no longer applicable.

Implementing a GIS at the healthcare level, in particular within the private business or hospital/facility levels, enables more directly targeted, patient and doctor implemented changes to be made.  Whereas at the insurance level, the same achievement is theoretically possible, the one or two steps away from the patient-doctor interaction that an insurance company places itself, and the frequent discontent patients suffer due to the lack of helpful or adequate coverage (even lack of coverage) insurers provide, severely hamper any ability of the insurance company to have a timely impact on patients’ health.  To implement a GIS at the caregiver’s facility level ensures the facility of its right and ability to make improvements.

01_NYCHealthImprovement

If we look at this issue as a similar series of different care programs were implemented in the past, we see a parallel here.  In terms of intervention rates for patients receiving some form of preventive care, such as childhood immunization, passive programs with insurers that fail to interact with patients (i.e. PPV) have much lower rates than HMOs, which is turn have lower rates than Managed Care programs, in which the provider and patient are regularly evaluated and scored for their more interactive relationship.  Enabling a program to evaluate its population provides its leaders directly with opportunity to make decisions that immediately speed up or slow down certain parts of the healthcare process.  They need not wait for feedback from their patients’ insurers, and in general, they can do nothing if they rely upon last year’s (or if lucky last quarter’s) reported posted by the regional public health review.

Finally, it is important to note that the variety of measures that a program can engage in is much greater when the program itself carries out these activities, instead of waiting for regional agencies and scorers of programs to determine what limited measures to use to evaluate an institution or facility’s performance.  An effective combination of EMR evaluations and GIS monitoring and surveillance can carry out such processes as fast as on a daily or live basis, instead of retrospective.

The benefits of implementing an internal GIS include enabling a program to surpass its competitors, even the smaller programs may supercede the prior successes of their much larger competitors.

 

Summary

To succeed in the implementation of an EMR-GIS program at the institutional level (not just the limited research level), the following processes need to take place.

  • learn the required skills, implement them, and develop the required work habits;
  • define rules and regulations, and establish/publish policies;
  • ensure HIPAA compliance, meet related NIH research and PHI requirements;
  • set up an IRB capable of handling all of the above processes;
  • produce a task force comprised of experts in these fields;
  • review and test the EMR data, including routine error analysis checks;
  • document/detail the Levels 1 through 4 requirements of data transformation, and define the pre-Level 1 data limits (i.e. ‘no SSN release, ever’, etc.)
  • implement a GIS–first at point-vector level, and then at raster and imagery levels
  • define the EMR, GIS and combined EMR-GIS-analytic processes (flowcharts)
  • establish some regular analysis and outcomes reporting standards (mimic your HEDIS, and then some), and semi-automate to fully automate these processes
  • engage in structured and non-structured text analyses EMR data analytic processes
  • develop and initiate qualitative, quantitative and combined research programs
  • apply the EMR-GIS tools and methods to searching for new grant opportunities and identifying unique population related needs for your program
  • develop a big data GIS reporting “Atlas” that can be regularly produced for your facility/facilities and the appropriate parts of the program
  • produce a time comparison report of the ICDs and therapeutic/diagnostic results performed on your facilty or institution’s population, comparing three periods of time, in order to define the changes in ICD rankings that have taken place over time [last year, vs. 6 yrs ago, vs. 11 yrs ago) at the patient, visit, visit:patient ratio (VPR) levels. (Also consider repeating this for special subgroups, i.e. just child age diseases, or just chronic disease rates, or just infectious diseases.)

 

 

00_aNYCHealthGIS

 

 

Picture2

For the past few years, researchers accessing the electronic medical records system have been most devoted to very basic forms of observation, surveillance, monitoring, and reporting.  GIS has been theoretically applied for the most part, or to experiment with this analytic process, to supplement processes already underway for quality improvement activities, and/or to use GIS to produce basic spatial expressions of the data researchers are working with.  The best use of GIS is to apply it to explain why certain things happen, for predictive modeling, and to evaluate change at some fairly sophisticated, detailed level of analysis.  Yet for the most part, we primarily envision and apply GIS in health research to explain something that happened, not why and where it will happen.

The public health and quality improvement practices have already developed GIS in order to monitor and report upon such basic public health data as STD rates, or watching for infectious disease outbreaks, or monitoring the HIV incidence for suspicious sub-populations that serve as some nidus for some outbreaks.  Public health and population health management programs employ it for reporting and planning purposes, such as for evaluating and recording childhood immunizations and to intervene where changes are most needed, or to study 18 to 24 year old Chlamydia rates in young people who are sexually active and demonstrate high rates for unexpected pregnancies.

Health improvement programs may use GIS to report on annual diabetes well visit rates or to show spatial relationships that might exist for high and low A1c, LDL and BP areas.  In the background, some parts of the New York City GIS teams have been able to provide potential researchers with helpful baseline spatial data to use for developing new spatial surveillance systems, by providing important supporting spatial datasets for this work., such as relating the placement of clinics, offices or similar service facilities, to increase the engagement of patients with these programs.

A number of years ago, I graded the level of accomplishment an program had in spatial epidemiology applications as typically as high as level 5.5 to 5.75.  This was based upon the different practices these groups normally engage in with GIS, using it to record, develop a history of, experiment with, and research how to implement this form of practice for basic health and safety concerns.  This also has attached to it the supposition that for the most part, health care programs engage GIS at only some basic level.  To determine just how much a program is engaged in the use of this tool, one need only reflect on how many measurable factors or details contained within an EMR system are being analyzed.  Is the EMR being used to its fullest extent?

An example of the 3D Mapping of cases using SAS 8.* and 9.* (non-SAS-GIS), applied for surveillance purposes since 2007

For the most part, GIS in health has been very slow in advancing in the actual implementation of this means for reporting during the past ten years.  We are certainly more familiar with the potential uses of GIS and its possible applications outside the already well-established realms define by environmental health, public health, population health, and quality of service/intervention teams.  For the most part, these projects remain single examples of what is being done.  Very few programs engage GIS at the level of big data reporting, such as mapping all 999 ICD9s at the spatial-temporal-age-gender-race level, per program,  per facility, per larger unit (i.e. insurance programs, companies, NIH funded, SES focused programs) responsible for providing that care.

That is now changing within the population health surveillance and research activities I engage in on a regular basis.

For the past 2+ years, time was spent exploring the complexity of a complete EMR.

When we all first learn about ‘big data’ what we see as examples do not accurately demonstrate the details, length and complexity of health data that resides in the most basic, first or second generation residing within an EMR system.  There are not just 1, 2 or 3, or even 4 to 6 special tables within which all data are recorded.  Each group of data that can be placed within an EMR forms its own table, with multiple rows entered per unit of activity or metric being evaluated.  These multiple row tables are modified or reconstructed into one-to-one or one-to-a-few formats.

When all patient data are entered, for example, these data which are stored initially as rows, get converted to columns, with patient identifiers (or its numeric assignment) as the index column(s) for this work.   Each patient can then have columns that depict name, gender, dob, dod, mother’s name, address, state, zipcode, race, religion, insurer(s), etc.  In the system I use, these tables are called dimensions and provide the most important personal, family and demographic data that exist for any given patient.

When a patient interacts with the health care system, there visits happen.  Some programs call these interactions between health care staff, and another entity involved with the patient –such as patient, parent, other provider, previous care giver, other facility.  In this review, I term these actions “visits”, but it has other common names elsewhere in the QOC/QI system.

This second dimension of care, the Visits, have only a few basic elements that define them, such as location (coded even down to the extreme, such as bed in a room), date, time (at day-hour-minute-second level) that something starts, ends or happens, time of closure or completion, etc.  The study of the Visits Dimension for a patient’s care process provides the dimensions needed to correlate events over space and time, allowing for a review of practitioner or systems logic, and identifying situations where changes may need to be made, through rule-setting, policy, procedure, assignment of place for the event to happen, implementation of different programs for poor performance teams, groups or places.  Without even looking at what practices were performed for a patient in the health care setting, we can see where further investigation may need to be made due to higher failures or death rates are seen for a given program.  The details of what were done and which of these went wrong haven’t even payed a role yet in the health care process.

The third level or dimension of care pertains to the details of what events ensue during a given visit.  The definition of each step in the care process also enables a time element to be defined for the care process.  This means the patient may come in a time1, see the MD at time2, received an injection at time3, be seen by a specialist at time4, undergo and MRI at time5, be evaluated at time7, be admitted for inpatient care at time8, and then undergo nearly a thousand more time-defined processes over the next few days of treatment, recovery, and then discharge.  Other temporal processes that can be evaluated here include time till initiation, overall time elapsed, time to recovery, and even post-inpatient time in relation to unwanted readmission events.

measuringavisit

FIGURE 1

In this review of the care processes, what happens with a “visit” are generally interpreted as events or procedures.  Events are what happens to a person, that typically is considered part of the care process.  Procedures are practice related events that typically involved additional skills and are often coded with a procedure identifier because that identifier may be linked to the cost of care and billing.  As a general rule, events are not charges, procedures are.  Events carried out by a clinician are considered in defining the bill for the visit.  Procedures carried out by the clinicians and/or technicians are often charged per routine, not per visit.  But like always, we have exceptions–such as procedures that are free but documented as part of the visit.  One of the most common of these within the system I operate are the Vital Signs taken and related medicate history questions asked and entered during each visit event.

The value of Procedures and Events coding is that the kinds of services being offered are considered, along with their relation to the overall timing and sequence of activities engaged in for the care process.

The “bread and butter” of all health care processes are the results of these procedures (and sometimes events).  “Results” is the term applied to these datum elements here.  And results are typically more than just the “result” of a test.

Typical results entered into a data warehouse include such datum as (with semicolons as separators): Yes; T; 1; 20; 3.45; “2,4,5,3,6,1,8” ; Complete; French; 13450; John Smith Sr.; “>150”; “168/96”; phq9; “above normal range”; Dr. Chase”; etc.

Over the past few years, extensive reviews were carried out for the size and numerica relationships between these four core “dimensional” datasets–patient. visit, event/procedure, and results–a general accounting of these figures, for just one visit and its linked events, is about 1:7-10:40-400:4000-40,000.  I term this ration P;V;E or P; R.

Patients are their own unique number, Visits are their own unique number, but for a health related happening (a diagnosis linked to the visits), one patient may have 7-10 visits per year related to it (directly or indirectly) per year.  Each visit in turn results in various events and procedures (vitals taken, labs ordered, educational materials provided, referrals give, etc.); even the most basic, simplest visit, such as a 9 month old well visits, will have Procedures entered for several immunizations, several health and safety checks with the mom, height and weight measures, pulse, an overall health evaluation visual exam of the kid done for scoring the child’s development, etc. etc.  Therefore, 40 to 400 events (educating the mother about breast feeding) and procedures (labs, health metrics) are not atypical to any system.

The key to understanding each program, each system, requires a complete evaluation of these different measurables, numerically and percent wise, to see what the norm is for the system, and to see how its various subcomponents perform and document the same duties. So, for a single institution, we might assumed that all follow a protocol, and that each one could have different time related findings, but all within institutional standards, such that all of the products of that type of visit are the same (i.e. vitals documented and entered, immunizations that are due were completed, all of this occurring in less than one hour.

“Results” is the next dimension, but the data content of this is actually best considered multidimensional.  The basic format of results data should be qualitative or quantitative, structured or non-structured, parametric or non-parametic.   Different institutions may store subparts of these data into separate places, such as grouping all pulses into one dataset, or all lab measures and results into a single laboratory results file, or all xrays taken into a single xrays database, with dates, times, procedure taken, amount of energy administered, time in and time out, results, initial interpretation, final interpretation, etc.

Results are any outcome of happening linked to a procedure or sometimes event.  Therefore, results can also be evaluated as relating to any of several groups of data entries:

  • process or procedure related info, such as exposure time, amount of xray administered, test tube/sample number, type of test, numeric sequence of sample taken, frequency, drug administered Y/N, type of test administered, units of measurement, US or metric values,
  • true results, like positive diagnosis (structured or non-structured), amount of energy read, size of nodule noted, amount of radioactive substance detected within tissue, estimated cells per cc,  percentile ranking of  height, LDL, BP, and id number for organism identified
  • events, activities, notes, that follow and/or relate to those results , such as normal range, maximum range allowed, viewed by PCP #, diagnosed approved by department head (Y/N), reliability of results (0/1), event closed or not (0/1).
  • general or non-specific non-structural data, such as words, text, impressions, notes of normal range, etc. entered as free text into a cell designed for this (Comment, Other, or Note cell ) and used by the practitioner to provided additional notes, which may or may not be specifically related to the procedure at hand.

basicpatientevaluation

FIGURE 2

The data evaluation, up to this point, focuses on just the Visit as the chief event, or research and analysis unit.  We can look at one visit and all that happens in relation to it,  be it a well visit, an inpatient stay, an emergency event followed by hospitalization, a referral to a specialist, a meeting with a social worker.  You can analyze the time component, the sequence, and/or the length of time until a certain point is reached (how long until the MRI was done?).

The data evaluation can also assess all of these events, over time, for a single patient, in relation to his/her medical history, and onset of new diagnoses or ICDs.

In the following model, the sequential visits related to single problems are assessed, such as diagnosis of heart disease, leading up to valve replacement.  Each one of the processes as defined in Figure 1 above, is presented by an oval in this figure.

leafstemmodel

FIGURE 3

Over time, these processes of care escalate and can have a cascading effect on patient health care needs.

therapeuticprocesses_fig4

FIGURE 4

In the more complex, lifespan models, all of the diagnoses and actions taken to care for someone may be placed into this model, to define lifelong related population health processes and individual health care experiences.

therapeuticprocesses_casestudy

In an application of this model to a personal medical history model, to review cost of care to the overall results of that care, for someone with a long history of epilepsy, the following cost analyses models were developed.  They predict the relationships of rising costs for care in patients well controlled, not controlled, and those who underwent some intervention care (such as neurosurgery), versus those who didn’t. [These were all covered earlier; the arguments for costs depicted here are found on those blog pages].

20_lifetimecosts

 

The Spatial Modelling Dimension

The next level of implementing this process for evaluating health care involved the application of these above statistical processes to data that may be linked to GIS research processes.

All data in an EMR, structured or not, parametric or not, numeric or text, can be converted to fully quantitative data by adding simple several spatial elements to the project.

The common comparisons between facilities and clinics, or health care between races and neighborhoods, for examples, are informally spatial in nature, and more formally best referred to as geographic, since latitude and longitude, distances, time, and spatial relationships are not a part of their formal numbers based evaluation processes.  By add the location-distance relationship, such as through the use of centroids, or space area analyses, or patient place (lat long) data, any and all health EMR data becomes quantitative in nature.  A fully text based, non-structured, content analysis, or 50 people undergoing a rare experience, is made spatial by adding lat-long to their analyses (although this one metric alone benefits more by other non-parametrics such as race, gender and age).

Health care analyses that become replicable and semi- to fully-automated in EMRs analyses can also be semi-automated or manually interpreted using GIS.  The values of this application for GIS to healthcare monitoring are fairly easy to visualize.

By implementing a GIS at this time for this surveillance program, a second process for evaluating health spatially is now in operation.*

This process of spatially evaluating data in SAS was developed a few years ago.  The means for producing videos of these 3D models of health in an urban area were perfected, in SAS Basic and SAS Graph (no SAS GIS was developed).

0_8_Multiple3Dmodeling

Which is fortunate, since SAS GIS has recently been turned over to a new workstation format for spatial analysis using SAS–SAS Enterprise with ERSI ArcGIS extension.  At the institutional level, this doubles the cost for implementing such a program at the QI/QOC level for Managed Care programs, like the ones I have worked with.

The spatial SAS methods applied serve in the analysis and projection/display process, with the animation of results that can be developed for rotating 3D model imagery the major benefit of this spatial analysis method.  [see Below]  We can further improve upon this by smoothening out the shapefile centroid data used to produce these models, by converting irregular shapefile (zip code) data into more regular square cell grid data (the algorithms for this I presented numerous times elsewhere).  We can further smoothen these presentations with a hexgrid modeling algorithm I developed (also detailed elsewhere; no example here, for now).

 

With the addition of a regular GIS workstation to the analytics process for evaluating 20 years of 11 million people’s health data, this work environment enables higher levels of the above (see initial figure) scoring system to be reached–Levels 6 and then Level 7.  Because the data pulls and reconfiguration are based upon automated or semiautomated, often SQL and then SAS macro processes, it is possible to run these evaluations for numerous new types of studies: for example, re-evaluating past reserch projects and questions across the system, by focusing on any form or group of ICD, labs, diagnostics, psych test results, demographical, Age-SES-Race-Ethnicity-Religion (SAS-RER) grouping, neighborhood (latlong), NYC healthy area polygon, nearest office visit (location theory/distance), inpatient stay pattern, Log reg / Kaplan Meier derived life expectancy patterns.

The following is an early example:

H+H1

 

H+H2

NOTE: Pb = Lead, Px = poisoning, Hx = history.  This is for 0-9.99 year olds
Future postings will review these processes in more detail, cover the theory, review the programming and statistical methodologies,  and provide various types of examples.
*Thanks to my research assistant Terrence Calistro for installing and developing the GIS.

.

The ongoing “discussion” regarding the development of a new health care program has only distracted the majority of health care giver away from meeting specific needs of a community and more towards how even the most basic needs of a population can best be met.

Even though the last health care system or program (and still current for the time being) was “problematic”, to say the least, some of the valuable directions that program took us were geared towards better managing “managed care” practices and guiding health care providers in directions other than along some superficial, impractical route toward making patients healthier.

The problem with the last (and still current for now) system is that even though it managed to increase the coverage of previously unmanaged  populations, it also increased the cost for other receiving their care, and even disenrolled a significant number by making them ineligible for continued service.

The most important victims of this failed health care administration process, initiated by PPACA and driven further in the wrong direction by the plans being shared right now, are those who have been inadequately covered much of their life, and those with medical issues that are not at the center of the whatever actions these new programs will be taking.

 

.

Long term care related needs, or in particular chronic disease care related needs, are once again the main argument being addressed.  During the 1970s, this is why many people could not be managed well or completely.  The existence of a prior diagnosis enabled health care insurers to charge more to these people for coverage, and to even refuse payments for illnesses and health events related to diseases they did not already manage.

The logic for this argument was never fully put into writing, but in a contemporary sense, these companies can now argue that they refuse to cover chronic diseases diagnosed prior to coverage, because the patient was not enrolled in their care once the disease developed.  The implication here is that, had the patient been enrolled in their system, then that patient would be in great or better health, theoretically, and the needs of the patient presently wouldn’t exist, due to their “incredibly great preventive care program”.

.

.

Now of course, such programs did not exist at all back in the 70s, nor do they really exist today.  There are care management programs and services out there in the health care systems, there are programs focused on specific diseases that serve to manage their patients better.  But these programs work at some “minimalist” level, for the most part.  We in health care still leave it up to the patient to take responsibility for his or her health, which still is the best way to manage these problems that cause lifelong disabilities and progressive disease problems.

.

Still, the main health related issues suffering the worst from the current indecisiveness of congress and the other untrained politicians making decisions, in areas that require IQs another 25 to 50 percentile above their own, pertain to the forgotten less commonly monitored, severely under researched disease and health problems that exist in the current population.

The most likely patients to suffer the worst consequences of what is now taking place in Congress and the US healthcare system are the members of the youngest generations.  the types of problems they suffer with regard to health happen, because the physical and mental health of their parents are once again put at risk.

These under researched medical problems, which have not been adequately analyzed at the spatial level nationally except by myself, are the focus of these 3D Population healthgrid maps published on this posting.

.

.

The ability of our standard healthcare insurance programs to produce these insightful maps, like the many regional Blue Cross-Blue Shields out there, the various Aetnas and Cignas, has been around for nearly twenty years now.  I first demonstrated the ability for a program to produce this type of mapping 8 years ago on this site, and not that the math related abilities to produce these results at the software-hardware systems level, have existed for about 12 years, perhaps 15 years. The fact that no companies produce these products in large amounts suggests that either–they are not wise when it comes to the best use of their technology, or they are hampered by major companies designed to produce these kinds or tools (Cerner and Soarian) and their lack of skills at this level as well, or finances are simply preventing them from developing this fairly simple method of HIT production.

.

.

The requirements for these 3D models are basic, There are no GIS programming or programs or tools and extensions required to develop a 3D modeling of you population’s health.  This tool may be used to identify the highest risk areas.  It may be used to define exactly where the most important interventions need to be directed.

.

.

The current squabbles between pro-PPACA and pro-Trump enthusiasts are wasting time trying to get a system up and running, again, at some mid-to-late 1970s discriminatory health care level.  This form of ‘selective managed care’ is better off referred to as “Survival Care.”  It increases the pressure on the patients to make choices about what to include in their coverage, and what not to pay into, in spite of the risk that a new disease could later impact your life.  For the child, since the child makes no choices about what he or she needs the healthcare coverage for, the child is a passive victim to this new form of healthcare mismanagement.

.

.

The lack of adequate physical care coverage and the lack of adequate salary or income to meet your special needs as a patient, increases the likelihood that other socially-induced problems will erupt in greater numbers, in spatially denser patterns.

.

.

As the children of this new form of mismanagment age, they will in turn cost the system more.  Whether or not this impacts the mortality of certain diseases needs to be seen.  It will however increase the underlying comorbidity factors that can exist in a given population.  These victims of infancy and being young children will age, and the related mental or behavioral health prevalences for whatever their lifestyle progresses towards will resurface as new problems, new diseases.

 

.

This is the consequence of the problems that US health care is now facing, due to each and every president who has failed to manage the most guilty — the insurance companies that refuse to take the steps they need to take, no matter what president defines those steps for change.

.

If the plans don’t change, “Trump Care” is going to flunk out, in much the same way as PPACA failed.    Just like their precedents failed – – – the “Advantage” programs, “Managed Care” ideals, the pre-managed care HMOs, all of which failed because they ultimately were rebelled against and/or modified by past and present health insurance companies.

 

The number of kids less than 6 years of age who experienced skull fractures in this country, since 2000.  Skull fractures in young children symbolize the results of many social problems that exist, ranging from lack of attentiveness to your young child, to unnecessary accident or injury, to neglect, to abuse.  With multiple causes. they may be unpreventable, but they can at least be reduced in number.  This social problem is one that health insurance can help to improve, but requires considerable change on behalf of society itself.  Most importantly however, is the fact that more than likely not one insurance company has reviewed the distribution of this problem in large areas, such as the state or regional level.  Too much time is spent bickering on thousands of unneccessary public health problems, like this one, due to in adequate care and service provided at the health insurance industry level.

How is it that the White House, congress, HMOs planners, and the bulk of the health insurance industry can lead us in circles?  With the establishment of the HMP concept in 1971, by Pres. Richard Nixon, the plans were to improve the health insurance for everyone in the United States.  However, health insurance coverage has always been a product made available mostly to the employed, in particular those who earn more and work for large companies.

Health insurance for small companies has never experienced an advantage over the past 45 years.  Nor have better programs evolved due to this initial development of Health Maintenance Organizations or HMOs.  Similar failures exist in nearly every one of the programs for health insurance that was approved and managed.  The programs with the least failures illustrate how we assign value to a new plan.  The plans with the least non-compliant insurance companies demonstrate a “win” for this system; the the plans with the greatest number of successes.

The inadequacies of government and health insurance industries are directly responsible for the societal dissatisfactions that current exist, and are one of the major contributing factors leading to the violence and life threatening behavioral problems that currently exist.  The Columbine incident serves as a primary example of this.  At first, we cannot clearly see how such a localized event may relate to this nation’s “personality.”  Poor health insurance is one of numerous triggers that impact how people behave and “feel” about life in general.  Differentiation of kids into the “rich and poor”, “have and have nots”  leads to psychological unrest, which in turn feed into whatever other social failures exist in a community setting.  This does not imply that insurance companies are the direct cause for these tragedies; but the lack of attention insurance companies pay to dealing with these kinds of social disruptions, and developing programs to treat them or tack action against them which is well targeted, is a consequence of their inability to spatially evaluate these important social/societal problems.

In the recent changes taking place in the nation’s health insurance “plans”, while a consider number of previously uninsured were subsequently enrolled, a large number of previously enrolled recipients lost their coverage.  Its like trading pewter for tin; one is more highly favored than the other–for now.

A major problem that persists due to this bickering and changing of plans is the customers–the “members” according to insurance companies, and “patients” according the health care providers–suffer as the consequence of governmental-corporate arguments.  When much of the time is spent by congress trying to develop yet another “new plan”, little or no time is spent repairing the damages that already exist.  In this case, these “damages” are the increasingly ill-health that the American people are demonstrating, to nearby health facilities and to insurance companies less and less willing to sponsor them, and facilitate a transition to a stabler, perhaps even healthier, state of living.

Adult sexual abuse is a major problem in this country, which can be directly related to the unwillingness of insurance companies to develop productive programs.  Their unwillingness to be required to cover individuals with chronic disease opens the door for refusing coverage to people with other medical problems, that appear to be “unphysical” at first.  But if the many socially defined conditions and behaviors have underlying genomic, physiological, endocrine, genetic, or epigenomic (impact of environment on our genetics expression), this makes it possible for these companies to refuse care to families with significant biologically, drug-abuse induced misbehaviors, including child and spouse abuse, or LGBTI-related health risks.   Is society ready for the gene argument to be used to explain why the insurance companies can refuse to cover anyone’s health care, for any given medical diagnosis reason?

This current decision being made about health care will delay the development of the overall health care system itself, due to the “control” and “power” insurance companies have on Congress, and White House politics and Congress have upon the development of governmentally funded healthcare coverage.  Remember, the issues currently at hand with the new Trump Health Insurance plan pertain to those who are in need of coverage, many in desperate need.  Illegal immigrants get disqualified for never having paid into the system, nor legally tried to become a part of the system.  (Its like joining a club without paying the dues–should other paying “members” let this continue?).  But the less engaged “innocent bystanders”, who are “victims of their health”, get punished the worst by this.

An abused child, malnourished by neglectful parents, may never receive the care required to reach his or her 18th year.  I produced this may “one generation of pre-school children ago” an applied a formula that emphasized the cities where this form of abuse was the greatest.  As noted earlier, the fact that insurance companies are unaware of this can be blamed on their disrespect for the public and their focus on increasing earnings, instead of health.  But the fact that the government too is engaged in this lack of respect or interest in public health enables these initial niduses for this new social epidemic to continue and spread into other communities.  It would be very easy to study these “hot spots” for malnutrition, study comparable areas without as much abuse, and determine the causes for this social problem. But the current system remains too focused on financial aspects to enable this part of the healthcare system to be improved.

The next steps taken by congress to disqualify people from health insurance coverage, or demand increase financial contributions on behalf of those that want it or need it, will ultimately reduced a system with lesser customers. (Obamacare disqualified the borderline poor who were disabled, impacting enrollment in Medicaid Medicare and Handicapped Disability programs by 8 to 10 percent in November 2015.)  The cost for care will become a deterrent to even the most commonly sick.

Back in 1970, my mother had to deal with the refusal of Blue Cross-Blue Shield (BCBS) to manage my epilepsy needs, which included an annual visit with the neurologist, coverage of the cost for my prescription drugs, my annual EEG, my CAT scan (then a new technology), my meetings with social workers and counselors, my occasional emergency or critical care visits.  When compared with the cost for my annual physical for school, BCBS covered less than 10% of my healthcare needs.  So why even enroll?

Imagine for a moment a society where those eligible for treatment due to drug abuse are only the employed, able to afford health insurance.  The amount of opioid abuse in this country could mean that 50% or more of those experiencing this problem will not be fully managed unless effective programs are put in place.  This means that insurance companies unwilling to cover you for this past history, also set the stage for your future demise, should such a habit become hard to break.  This also mean, the only surviving opioid abusers (with that inherited genome), could become the middle to upper class workers, able to afford insurance, able to afford the care for their abuse provided by the insurers.  The lower class (other genome bearers) may ultimately die off, allowing social Darwinism to become the deciding factor for how addicted the future generations of kids, produced by these workers, will influence society and its healthcare needs.

Another unfortunate set of conditions influenced by the transition of care that appears to be happening is the enhanced neglect of care provided to certain victims of human behavioral conditions.  Over the years, child abandonment has developed small hot spots or niduses within large urban settings.  The statistics I used to analyze this, for the most part, came from the pre-Detroit Failure, pre-BLM social disruption years. The impacts of poverty now may be accentuated in regions where poverty was already having an impact on the health of specific communities.  Interestingly, the Obamacare and Trumpcare systems both failed to address this growing problem in certain urban settings.  They perpetuated bad practices already developed just before the passage of PPACA, and initiated in pieces by the everchanging HMO plans since 1971.

This map again depicts the core spots where abuse and neglect lead to the abandonment of newborns by placing them in predefined spots (hospitals, police stations, firehouses) where they may be transferred to a reputable care facility.  These are the 2000 to 2010 stats on this newly evolved “diagnosis” entered into the medical records.  It would be interesting to see how much these cases have increased, and what new niduses were developed since 2010.

It is possible that the currently change in healthcare insurance program will, like the failures before it, also be short lived, assuming it is passed.  Theoretically, it is important to remember that congress is the cause for its passage, but Insurance Companies themselves the cause for its evolution into this state of failure that it is experiencing.

Insurance companies do have an unwritten social obligation, which they don’t recognize, to provide coverage for healthcare in as many ways possible, so as to reduce the rates of sickness and death that result from poor personal and professional health care management.  But when insurance programs direct their attention primarily to their investors and selves, as financial officers looking for success, they engage in a unique form of evolutionary selection, akin to an offshoot of the “Soylent Green” scenario.  You need to “feed” your consumers enough to keep them engaged, but prevent them from dying off completely — otherwise, you lose your need for existence.

Children abusing children–peak areas in this country.  Are there are programs devoted to this cause in these given areas?  Probably not.  I doubt anyone in the insurance business is familiar with this above 2010/2012 finding.

Like other social “diseases” or misbehaviors, insurance companies have spent minimal time trying to deal with behavioral health.  Their focus is on physical health, or conditions that can be successfully treated by basic physical means.  Like what was said earlier, society has given rise to a new form of physical disease due to the genomic and epigenomic philosophies.  Genetics somehow relates to our health, knowledge of the genome will allow for these relationships to be “uncovered” (or so they think), and by understanding human genetics makeup and gene expression, we can turn any social or psychological condition into a physical condition with potential “physical treatments”.

Other examples of Social Disease and the inability of insurance companies to manage their patient populations follow.

Homelessness in 2010 (V and E codes derived); new niduses have no doubt been developed.

Congenital Tuberculosis – peak areas influences through immigration and the lack of adequate testing and screening; kids are born with TB due to inadequate plans.  These are all cases, not just newborns (when they are diagnosed) or kids <18 yo, and so may represent some young adults as well, who were born here and had TB.

Beri beri – a disease of malnutrition and another sign of regional mismanagement by health insurance companies and governments.

Rickets – another disease malnutrition — this time CHRONIC!

Insurance companies and the past plans have done little to properly manage suicide.  In the last ten years, suicide by children and especially teens has become a priority according to many suicide programs.  Yet, it is likely none of these programs have reviewed the distributions of age-group related cases at the small area level.

Pedestrian accidents, which involve being hit by a car

 Based on V-codes, E-codes; another overlooked preventable status of care and life style.

Mapping all of the ICDs, V-codes and E-Codes requires an accelerated mapping program.  This program was actually developed between 2005 and 2010, tested, and used to produce these national maps.  Some studies were performed as well on multiple age-groups and cultural backgrounds.

In spite of the possibility of developing such a system, none of the major providers of EMR data use and integration have designed a program equivalent to this version that I developed.  To date, these companies include all of those engaged in EMR utilization, even the largest ones providing numerous agencies with EMR related BI and ERP support devoted the healthcare and Population Health.

An early test of my modeling algorithm.  In the first months, I developed the “Three-Tier” model for overlaying one set of conditions or features over another.  This was used to compare the distribution of two similar ICDs, one culturally-related, the other culturally-linked (genetically linked, but not culturally bound; I covered these types of diseases and their classification on several other pages).  

The goal of such programs should be to produce 1000s of map per day and design focused reporting tools on place, race, ethnicity, religion, language, culture, age range, country of origin, in-migration patterns, land use patterns, SES, genomics, epigenomics, etc.  The current system of health insurance is extremely inadequate for managing the complex health of an increasing complex US population.

The value of Medical GIS is the answer these problems, assuming it is properly employed.  Current products, because they require stepwise development and data development and process, are still inadequate for meeting these needs.  The government and health insurance programs, and their healthcare reform antics, are the primary cause for this delay in technological advancements.  We are repeatedly sidetracked by arguments focused on eligibility for health insurance coverage, not on population, community, family and personal health related needs.