Brian Altonen, MPH, MS

Note: this page and neighboring pages are from older teaching materials used for a lab on GIS and the corresponding lecture/discussion series on ‘GIS, population health surveillance, epidemiology and public health’.

***************************************************************************

An Integrative Health Care GIS

The purpose of GIS is to make the best and most complete use of your data by adding a spatial and temporal perspective to the implementation process.  GIS provides us with research results that cannot be produced by any of the standard SAS and non-SAS population health research methods current employed by nearly all health care provider groups, agencies and companies.  For the most part, GIS has been required only for measuring the amount of services provided out of the allowable radius from a given primary care provider facility or region.  This information is reported annually to federal agencies in order to improve the availability of care services for all patients under their management.

GIS can also be used as a tool to engage in regular surveillance activities involving numerous potential public health physical and social disease indicators.  The lack of this use of GIS is due mainly to the lack of adequate training at the administrative and basic non-clinical staffing levels for nearly all health care facilities.  Even within systems where a GIS is in place, the use of this tool is primarily for research purposes, focused on special topics and very unique populations.  GIS is typically not use to evaluate the health of an entire population, either under any form of care from the health care provider or as an annually eligible member whose personal health information is incorporated into the annual quality of care reviews most of these accredited agencies are obligated to engage in.

This presentation of how to employ GIS as part of a standard system requires a redistribution of responsibilities of GIS to the office of Quality Assistance, Quality Improvement, or any other office responsible for monitoring and improving the care institution’s population health statistics.  For this use of GIS, the primary population health researchers responsible for improve the services provided and engaging in a regular or annual reporting process are also responsible for developing an effective GIS that can be utilized for routine and ad hoc queries regarding special topics and overall population health.  As a standard part of this GIS, the routine HEDIS/NCQA processes must also be able to be integrated with the standard regular population health reviews engaged in by the institution.

To accomplish the development of such a system, the system must be able to meet all of the requirements for effectively completing the tasks at hand required for the 5 basic steps that define this Integration of GIS with Health Care Decisions process.

The five basic steps required to integrating GIS into a work environment where it previously did not exist are as follows.

1. Acceptance
2. Planning
3. Action
4. Results
5. Maintenance

Acceptance.

Accepting the need for GIS is required for obvious reasons.   First, there has to be a function it serves in the overall process of the work environment and job responsibilities.  These functions can vary, but tend to include such things as:

• satisfying client’s needs
• satisfying internal corporate needs
• answering questions critical to performance of the corporation
• developing new corporate work environment strategies designed to improve short term and long term outcomes
• provide a method for evaluating success and designing plans for change.

All of the above are the major benefits of implementing a GIS in the work environment.  Corporations that don’t have GIS, or are operating some form of GIS related work, but only meeting one or a few of the above needs, are not reaching their fullest potential.

Five years ago, a Texas company came to me with the idea of mapping diseases nationally, in order to secure a contract with the CDC for managing this epidemiological surveillance system.  They wanted to know my insights into how to do this, involving such large datasets in diverse formats used to provide information for a variety of different types of software programs.  The data quality issue was my main concern, but setting that aside, developing a way to spatially review diseases temporally and spatially was the main goal.  The simplest systems for doing this had already been developed, and are still in use, namely the State by State, County-based statistical system, with point data overlain, and on occasion Metropolitan and Town area statistics generated based on Census data for these predefined Metro areas, and Incorporated Town areas.  To change this coarse overview of the country into a review that is more spatially refined, a smaller area method had to be developed.  The problems with town and village manners for mapping data pertain to accurate address matching problems.  This data was more detailed and such, but the points erratically places about the surface of a research area, making it tougher to develop a reliable TIN model of the places evaluated, stretched across this entire country.  This is why the grid method was defined.  Grids produce equally distributed points where data can be assigned, and used to produce isoline (contour) images of case/metric distributions and engage in accurate kernel density analyses of the country as a whole, on down to its smallest regions defined by the grid cell size selected.

The company for which this method could have been applied never managed to get past the second stage of the project.  This was because the management lacked both the knowledge base to engage in this project, and therefore never fully accepted its possibility of moving ahead at full force.

Similar behaviors are apparent in other technology based industries that consider themselves futuristic in their approaches to solving certain problems.

During the late 1980s, I worked for a stock market investment company (the term for the time was “futurist”), overseeing the technological advanced purportedly being made by the big biotechnology firms for the time, such as Calgene and Cyrus.  Seven major companies, engaged in the action part of this process were evaluated for the probability of success with their applications of the science needed to successfully produce a pharmaceutically applicable end product.  Only one of these companies succeeded, and lasted well beyond the seven year half life then predicted for these industries–Calgene, for effectively making good use of its Agrobacterium tumefaciens cell culture methodology.

Planning.

Once a decision is made to begin a particular program or project, you are now placing yourself/your team into a potential loop of activities based upon:

Planning-Action-Results-Maintenance

During the last stage, new planning has to always be kept in mind.

A healthy ratio for activities related to the planning stage is

70 : 20-25 : 5-10.  [status quo:new stuff:innovations]

Seventy percent of the company work is status quo, 20-25% is “new stuff” and 5-1o% is geared toward experimentation and innovation.

Efficient systems maintain this ratio, inefficient systems only succeed in accomplishing tasks in the first and part of the second series of activities.  Companies without any of the innovation, are potentially doomed for failure.

The role of planning is to assure yourself movement into the third stage–Action.  Without well defined goals and objectives developed as part of the planning process, it seems unlikely that much success can ensue and be measurable.

With one of the bacterial cell culture biotechnology companies I reviewed, the company had this goal or developing a new form of reverse osmosis to concentrate medicinal agents being produced.  But the technology and natural laws were not fully assessed, or even eluded to in the annual stock market reports and in personal conversations with the scientists engaged in this process.  The hope was to develop something in a year or so, but no well defined route had been developed, and in less than 18 months, between the first annual report, which received numerous support from inventors, and the second report, which stated that progress was made but not products developed, and the third report, which demonstrated failure to perform, the company went under.  This was because investments died out after the second report.   A normal amount of time towards measurable productivity is under a year.  If a company needs 15 months to get even the most basic processes working and performing properly, the project should not have even been initiated.  In developing software algorithms for reporting to your client for example, in some new fangled way, you should be able to produce a reliable, accurate and near final report in under a year, giving you another year to test its progress and results before marketing and selling it, and taking responsibility for your company’s results.  Teams which spend 6 months rehashing the same issues, the same algorithm, with not results after 9 and then 12 months, should not be engaged in the project from that point on, and the project manager should be reassigned or encourages to seek out new work.

The most important parts of this phase are:

• Goals
• Objectives
• Well defined, measurable indicators of success.
• Preliminary evidence proving these goals and objections are realistic and achievable.

Each of these processes is very well delineated in the standard Health Assessment activities like Quality Improvement Activities (QIAs) and Performance Improvement Projects (PIPs).  The same logic applies to corporate projects.

Action.

How to engage in the Action step is quite obvious.  Follow your predefined plans, fine tune them, produce some end products that can serve as measures of success.  Then make whatever changes have to be made in order to:

• improve results
• speed up processing
• apply the results in such as way as to produce automated reporting, or some other very useful method of making full use of this evaluation process.
• make corrections and changes as needed
• retest the new process
• repeat these last steps in a loop
• demonstrate change and accomplishment as a quarterly attribute of the process, not a bi-annual or annual process.

In terms of GIS projects, this means that GIS has to be put to use, and well defined end products developed and routinely generated, scored and improved upon.  All steps should be automated as much as possible, and their speeds of progress increase 4 fold following several weeks or stages of use.

Maintenance.

Continue the above process, adding new technologies, new processes, extensions to the program underway, in order to increase its productivity and further develop its utility.

Innovations are developed as a part of this process as well.

Very valuable innovations qualify as synthetic genius applications (more on the meaning of that later).

In a GIS work environment, this means the following should be viewed as potential indicators of improvement and success:

• Implementation of projects with much larger datasets
• Production of maps with greater spatial details (higher resolution)
• Production of maps with displays of statistical data, not just descriptive data
• Production of maps with displays in new forms, utilizing original data, not standard base data
• Development of a useful, applicable end product
• Generation of standardized reporting methods containing these GIS end products
• Facilitation and speeding up of the reporting process.
• Making better use of increase time slots available to initiate experimental, research applications for methods already developed–the innovation stage.

The Marketing Process.

Once all of the above steps are complete, businesses need to market their success.  This means that management has to understand of logistic of what is being developed, its applicability to the world at large, and how this method is more cost-saving, time-saving or in some other way very important and valuable to the field as a whole.

Without the required knowledgebase, management, CEOs, Directors, VPs, Senior Managers, cannot sell or market this end product that effectively.  If they cannot make sense of the content of the product, they cannot explain why it works and why it is good.  This is where GIS implementation currently fails in the business sector.

Two mistakes are made with Managers selected for GIS programs.

First, the managers selected are old timers with primary engineering spatial drawing skills, but lacking in a form foundation of understanding spatial statistics

Second, the managers selected are not GIS trained or experienced in the use of GIS.  If they haven’t developed a map on their own, with their own dataset that they developed, they cannot understand or explain how and why this technology is good.

This ‘black box thinking’ (knowing what it looks like on the outside, but nothing about what’s on the inside), increases the likelihood for loss of support internally and externally, and on behalf of potential clients considering the use of this new technology.

For the implementation of GIS into a corporate world, Maintenance cannot be maintained, and converted into some form of marketable success, without adequately trained and experienced management.

Examples of Application

With the above theme and processing of business intelligence data in mind using GIS, we can develop very interesting methods of analyzing populations and predicting their future needs and behaviors as patients or consumers.

The current trend in statistical analysis of Big Data in the business world is to review past information for large numbers of people and use this to try to draw conclusions about a given population, be it a general population without much regard to specifics, or a highly specific group or class of people that your company or industry is targeting.  Analyzing large groups, like those 55 and older, for trends in need, costs, specific product lines, specific risks for catastrophic events, are often carried out using some of the more recently established statistical formulas out there, such as Monte Carlo modeling and Bayesian analysis of variations in responses for given classes and subclasses of diseases, groups of people, places, socioeconomic status levels, etc.

Reviewing the specific population of interest, and then deconstructing its data and coming up with new patterns by pulling that data back together into some theoretical model, is the way that data can be evaluated for later applications to GIS.  The development of specific datasets, demonstrating patterns, from the bottom up, is the argument behind reconstructing data for specific ICDs, specific claims types and features, specific ethnicity/race groups, specific age groups, etc.  With spatial analysis, you reconstruct your place from the data that is available, and by mapping it, define where certain findings that you are in search of exist.

Environmental Patterns.  For example, the Hispanic population has two aspects to it that are important.  The first is that there are tradition US Hispanic cultural settings, versus the non-US Hispanic  groups.  Second, each of these have their subtypes, some of whom are Hispanic in lifestyle, language and family living patterns, the other very non-Hispanic/United States in nature, and even these have those traditional original Hispanic families who resided in their place for more than three or four centuries, of Spanish-Mexican descent, versus the Euro-American in-migrated families only on this continent for 200 years or less (for ex. Colorado, New Mexico, Arizona, Nevada, Oklahoma, Texas families).  There are unique ICD patterns that each of these subgroups of the US Hispanic population have.  These groups will probably have significantly different diabetes and asthma chronic disease findings, and most certain cardiac, neuropsychiatric, and culturally-bound ICD differences.  Recreational drug activities, social misconduct, interpersonal relationship patterns are very different between these groups as well.  Criminal behaviors, exposure to poor living environments, engaging in disease-related eating habits, having a likelihood of being exposed to certain infection and microorganismal diseases are going to be different as well.  For example, the Southern North and Middle American and South American disease due to Cryptosporidium may not seem different between these two cultural groups or heritage settings, but both the prevalence of and the agents responsible for other culturally-linked disease patterns such as Pinta (Treponema carateum) from Central Mexico, and Chiclero’s Ulcer (Leishmania mexicana) from Yucatan will be very different from each other, as well as from the Cryptosporidium map of distributions.  The reason for a Pinta peak in a certain part of the US might imply certain cultural or familiar reasons not normally considered during the initial phase of a disease ecology review.

Cultural Patterns.  Sleep disorders are prevalent around this country.  Nocturnal sleep apnea is prevalent usually due to an obesity related condition or problem.  Nocturnal deaths during sleep, caused by sleep itself, are infrequent to rare, and for Laotian, Cambodian, and some Thai populations occur due to the stress of moving away from the home country and into the U.S.  This syndrome impacts primarily the oldest members of this group, but has enough of a psychosomatic component to its origins to result in middle age cases perhaps developing as well a generation or two later.  American physicians diagnosed these nocturnal deaths as a result of a cardiac disease, when in fact the cardiac cause was secondary to the emotional-psychiatric cause induced by the drastic change in living conditions the elders endured.  This difference is not only seen by what ICD is assigned to the condition, but also how it can be prevented and reduced in terms of mortality rates.  Exclusion of the cultural model for this disease pattern results in a method of treatment that might end up costing more in long term treatment costs and result in earlier onset of deaths in patients.   Form a marketing level, the proper culturally-defined treatment of this “ICD” (no true ICD and exact exists for this even in ICD10) presents catastrophic illness related costs from becoming a part of the scenario.  A detailed demographic population study and density of the ethnic groups at risk using GIS will enable a public health epidemiologist to target counseling and intervention activities more precisely in order to prevent these events from continuing.  This same practice can be engaged in form most if not all other culturally-bound, culturally-linked and culturally-prevalent disease diagnoses (my culture and medicine section details how to differentiate diagnoses within these groups).

SES and Poverty patterns.  Low income and poverty have impacts on family health and interactions between family members.  The best evidence for this are several ICDs linked to these living conditions.  GISing disease can allow for high risk areas to be defined for these problems.  When we look at the national trends for such behaviors as child abandonment, spouse abuse, child abuse, etc., we see peaks that have to be reassessed to determine if they are part of a recurring theme, or are simply a result of random local surges in prevalence rates for given regions due to very local reasons.  This use of GIS and ICDs tells us how the analysis and intervention programs are to be developed for a given “hot spot.”  For example, it will tell us if the program has to be temporary or permanent, or targeted towards specific neighborhoods or specific chains of events taking place in the local economy, such as higher gas rates, increased unemployment rates, certain times of the year, etc.