Brian Altonen, MPH, MS

Opportunities for researching exceptionally large populations (millions to hundred of millions of people) for health related stopics are rare.  The majority of exceptionally large population studies focus on the individual and his/her financial history and activities as a consumer.  Unlike economic information, researchers of medical information has the addition HIPAA regulations to contend with.  Population health information compromises HIPAA whenever both place and very rare medical data are revealed in a single presentation.  Normally individuals are shielded somewhat from identification whenever a health study involved fairly common diseases.  But when diseases exist in whioch the cases are just a few of a kind, the chances for revealing an individual’s medical history becomes more controversial. 

6 Childhood Abuse Types

Countering this problem with the release of specific information about someone’s health is the impact that exceptionally large population studies provide for us researchers.  With exceptionally large population studies, even when cases are especially rare, the anonymity of any information pertaining to them is concealed or shielded due to the large areas used to perform such analyses.  HEDIS and NCQA somewhat correct for this problem by defining fairly large areas for their regional analyses.  The same actions can be taken for exceptionally large  population studies involving medical and pharmacal data stored and minimally used by large corporations, insurance companies, and overseers of these large businesses or health related non-profit agencies.

Whenever the opportunities arise to perform detailed health reviews of large populations, these databases need to be fully taken advantage of.  In recent years, and perhaps throughout much of the history of this data gathering experience, these companies engaged in information utilize that data almost totally with the goal of improving upon corporate revenues.  So much activity is engaged in trying to understand the consumer population, even including cost related to health status, that very little time is spent utilizing this data to its fullest extent.  It is not unusual for a company to have more than 100 experts trained in evaluating costs and quarterly earnings, and just one or two individuals, if any, devoted to population health.  Some of the corporations with the potential to engage in such important work do have processes in place enabling studies to be performed on medical history and drug use, but always with the inclusion of cost-related analyses.  This focus on costs has the effect of resulting in methodologies rich in numerous systems related errors.  Cost becomes the focus of any review that is performed on a program, with the goals of relating one series of costs to another, such as when two sequential years are evaluated for increases or decreasese in costs and incomes accrued due to disease-specific drug uses.  Even when prevalence is calculated for such studies, the focus of the outcomes relate to how to reduce costs, reduce financial waste, and increase corporate earnings for both the insurer and the company performing these tasks.

Digestive Tract Cancers

Unfortunately, most of the practices engaged in by such agencies are primarily information focused and provide valuable data such as sums, averages, ranking and frequencies.  Once again, health is not the focus of these presentation, only finances, and therefore, valuable medical information is unfortunately ignored or severely underutilized.   

Additional evidence for this udnerutilization of health data is the old-fashioned way in which health measurements are still being performed on the exceptionally alrge datasets.  More refined methods of analyses need to be developed and be employed to take advantage of the rare opportunity of analyzing an exceptionally large dataset.  When we research a disease for example, knowing the average age of a cluster of individuals with the disease, or the average number of days of hospitalization due to such a disease, or the related costs linked to this care, do not provide any useful population health data.   Even when attampts are made to provide more detail about the particular illness or condition, results tend to be provided in  a way that is interesting but lacks much field use or application.  Reporting the frequency of STDs for example in individuals under 30 years of age does not allow us to model the causes for disease based upon specific age related behaviors.   We cannot for example differentiate pre- and post-high school graduation patients using these 5 year age range increments; we have to employ either a model that defferentiates this 0-17/18 yo group, or develop a more concise way of modeling this information.  

The Taxonomy or Nosology of Epilepsy

In the past, 5 year age groups are normally employed due to small population counts (<250,000).  With exceptionally-large datasets, the limitations of utilizing 5-year age groups become quite clear due to the frustrations these outcomes can result in.  For example, with a review of anorexia nervosa we might learn that the the average women  has a high prevalence of this in either the 10-14, or 15-19, or 20-24 age group.  When we research these cases using smaller age increments, we learn the more valuable finding that the peak age for anorexia for our particular population is 17.  Likewise, by using detailed age analysis techniques, we could learn that the peak age for a specific class of medications related to treating chronic lifelong disease states is often 57 years of age, +/- 1 to 2-years depending upon male or female group status and type of chronic disease medication.    In terms of smoking behavior and risks, a 1-year analysis would reveal the peak incidence/prevalence age of smoking for males to be around 16-18, and for females 43.

Common Psychologic and Psychiatric Diseases or Disorders

For these reasons, a specific set of methodologies were developed for researching population health in one-year increments using the age-gender population pyramid format.    The methodology presented in this section should be the standard approach taken by major companies with exceptionally large datasets on health.  Unfortunately it is not. 

The current methods in place are taken primarily due to their past achievements, and the fact that they are considered to be “traditional and proven”.  These older methods are also very much old fashioned.  They utilize  methodologies and formulas designed for small data sets with just a few areas of focus.  You cannot easily apply the same methods and formulas to analyzing diabetes and heart attack incidence and prevalence relative to drug utilization histories, without making the necessary modifications required to make one research method to be applicable to the second research proposal.  The methods used to analyze pre- and post-partum visits for a women less than 25 years of age in relation to childbirthing and neonatal health are distinctly different from the methods used to analyze childhood asthma treatment or management of seizure control.  The truth is, public and population health research consists of uncountable numbers of studies that can and should be performed.  When performing these studies, is there any method that can be applied to link these various different research methodologies and areas of focus together into a common baseline feature.  This method of documenting population form and structure provides the link needed for merging all highly filtered population reviews into a common theme–the health of the entire population based on its complete age-gender distribution of disease patterns, visit activities, lab results, preventive medicine activities, and other electronically documented health related procedures.

Common Age-Gender Asymmetries

When these formulas were constructed, their inventors never expected these formulas to be applied to exceptionally large datasets.  The datsasets these formulas were developed with and designed to analyze were typically much less than 1 million.  In fact, they were usually less than 50,000.  This is because the manpower requirements for engaging in clinical analyses dictated these limitations.

During the 1960s, science and social science were pretty much separated from each other when it came to statistical analysis methods.  Science considered itself the more rubust field, studying large numbers of objects being measured, trying to develop equations that were meant to be used for such large numbers, so long as they could be instrumentally measured and documented.  Nobody sat in a particular palce for a long period of time and jotted down all of these tens of thousands of findings, these were automatically recorded and later reviewed by people and their calculators, or computers.  

The social science and psychology took a more “personalized approach” to gathering and analyzing research results.   These researchers were more likely to measure wualitative features and make use of non-parametric methods for gathering data and performing analyses.    Whereas scientists like to count the number of eggs that were lain and hatched, sociologists and psychologists made note of the numbers of healthy behaviors these babies were engage in, bordering on almost subjective reviews at times in order to define different kinds of behavior.   Although their styles and mathematical methods of calculating results were different, their purpose and intentions were identical.

They tend to review medical conditions, but due to their limitations are often not effective at displaying detailed population age, gender, ethnicity linked health features.

This methodology corrects for this problem with the current statistical methods.  It is not designed to replace the old methods already being implemented.  It is meant to be used as a population health reporting tool, which none of the current methods accomplish. 

Gender or Culturally Specific Asymmetries

Applications to Population Health Research

There are multiple uses for this approach to analyzing population health that must be taken into account.   The minute details of the 1-year pyramids that are produced using this methodology make it possible for clinical workers to develop programs that are more effective at reaching their population target, and due to the ability of this methodology to define multiple age peaks, allow for the development of a more effective age-gender specific intervention or treatment program.   

When we review the ICD Population Pyramids for the following ICDs, we see how much of a role both of these approaches to evaluating this data can play in a health acre system.

For example, a standard way of reporting incidence-prevalence might be to report rates in 5 or 10 year age bands.  This information is helpful background, but doesn’t do much to inform us on how to approach the health concern under review.  For diabetes, asthma, smoking, we use small populations, relate them to larger national populations or averages, and then determine how this might impact an intervention program being developed. 

Drug Dependency

Using the plans for developing an anti-smoking campaign as an example, employing typical methods for reviewing your data, you may notice that there is a high prevalence of smoking in the 15-19 and 20-24 year old groups, and so develop an intervention program targeting these two age groups.  The 1-year age-gender pyramids tell us that there are age-peak differences between genders, and that the primary age to be active is 13-19, a 6-year period rather than a 10 year period, saving us 40% of the costs should a mailing of intervention packets be required. 

To prevent the highly controversial genital infibulation practices within this country, we can plan a mailing for the entire African-American population, targeting those from the central part of Africa where these activities might take place in the US.

Organ Transplants

The following questions were answered, and/or resolved using this methodology:

1. At which age (as a 1-year value) are male patients most likely to be smokers? female patients?
2. What is the narrow age range for cannabis abuse patients?  How does this compare with cocaine abuse?  General sedatives abuse?
3. What is the peak age for male patients to be documented as being engaged in an uncontrollable gambling behavior?
4. Which forms of suicide are women more like to engage in than men?  Which forms are predominantly male practices? 
5. What are the two peak ages for children interested in engaging in some form of fire starting activities?  How are they gender distributed?
6. Which psychiatric conditions mostly or almost completely impact male populations? female populations?
7. Is non-compliance with prescription drugs more male or female in nature, and at which two ages does it peak?
8. Premenstrual Syndrome has two age peaks.  What are they?
9. Which fracture type shows the most age-gender dominance with most cases occuring in 65+ years old women?
10. Traumatically-induced amputations are more apt to occur in which age-gender groups for the upper half of the body (fingers, thumb, hand)?  the lower half (toes, foot, legs)? [one is childhood, the other adult]
11. Which fracture type afflicts middle age populations the most?
12. In terms of age-gender distributions based upon 1-year age brackets, how is schizophrenia unique from most other psychiatric disorders?
13. What other form of detection (hint: tissue or joint emergency/urgent care problem) serves as an important indicator for shaken baby syndrome, and demonstrates a unique statistical age-gender distribution directly linked to this syndrome? 
14. With sickle cell disease in an active form, which gender has a longer longevity?  Which has more equal age-gender distribution: being a carrier or expressing the disease?
15. What is the major age-related difference between thallasemia cases and sickle cell cases?
16. Conditions linked to food consumption, and behaviors linked to age and gender, include the following: alcohol polyneuropathy, alcoholic gastritis, alcoholic fatty liver, alcohol-induced cirrhosis, acute alcohol hepatitis, alcoholic cardiomyopathy, nutritional cardiomyopathy.  Which age range and gender is expected to dominate the prevalence of these disease types?
17. Lymphatic drainage problems impact which gender the most? at which age ranges (young, middle age, older employed, retired)
18. Asbestosis is very male dominant, pneumoconiosis is slightly male dominant.  Why the difference?
19. What are the four age groups in which African Infibulation practices are most likely to be documented in the US? which of these more than likely are taking place in the US and therefore are preventible in the US health care system?
20. Of the following three digestive tract obstructions– intussussception, paralytic ileus, and volvulus–which impacts male patients > 70 years the most, with risk increasing as they get older?
21. Which gender develops intestinal conditions more than the other due to radiation therapy?
22. Myocarditis is only a male-dominant disease at what narrow (20 yr) age range?
23. Osteoporosis and Gout have nearly mirror images in terms of age-gender distributions.  Which is female dominant and which is male dominant?
24. Of kidney, heart, heart valve, skin, bone, cornea, lung, liver, bone marrow, stem cells, pancreas, intestines
    • Which Organ Tissue replacement is most likely to be evenly distributed across all age brackets? 
    • which is least likely? 
    • which engage the oldest population? 
    • which exclude older people? 
    • which is exclusively middle age? 
    • which is almost exclusive to the younger population? 
    • which one, due to the symmetry of the underlying medical history, is also symmetric in terms of tissue replacement by gender?   
25. Which of the following liver problems are male dominant? female dominant?  Viral Hepatitis, Chronic Autoimmune Hepatitis, Chronic hepatitis, Biliary Cirrhosis.
26. Which one of the following inhalation syndromes or conditions has evenly distributed impacts for male versus female populations? silicosis, talcosis, inorganic (metal-graphite) dust, organic (flax, cotton, cannabis) dust, chemical fumes/vapors.
27. Which of the four cardiac valvular disorders has the greatest impact on patients under 20 years of age?
28. In terms of prevalence based on claims records with body text (not just header identifiers for data), what is the major difference between asthma and chronic bronchitis?
29. For the following–Renal osteodystrophy, hydronephrosis, renal calculi (lower or upper tract), and cystitis (all forms)
    • Which one demonstrates the greatest relationship to aging and old age?
    • the least? 
    • which is male dominant?
    • female dominant? 
    • neither? 
30. Spinal Bifida demonstrates a claims-related female peak at about 15-17 years of age, whereas for male patients there is a dual peak, the first at 15-17 years and the second during the mid to late 20s.  Why this distinct claim related difference?  (Hint: this recurrance occurs with congenital anomalies like transposition of great vessels and most mental retardation ICDs)
31. Of the four Digestive Tract ulcer types–Peptic, Gastric, Duodenal and Esophageal–which displays the most age-specific male dominance?
32. Many psychogenic diseases appear to impact male patients several to ten years before female patients, with peak ages 10 to 15 years younger for those afflicted first.  Which of the following does not demonstrate this psychosomatic effect?  Post-gastric surgery syndromes/side effects, irritable bowel syndrome, psychogenic constipation.
33. Transfusion related acute lung injury (TRALI) impacts which gender the most in old-age populations?
34. Which of the following STDs is/are male dominant in terms of visits and claims?  female dominant?  [Gonorrhea, Early stage Syphilis, Late stage Syphilis, Chlamydia]
35. Graft Host disease is more likely to occur in which gender and for which three 5-year age bands?
36. Diabetic Glaucoma has a older age of onset in geriatric patients for which gender?
37. Diabetic bone change afflicts which gender the most?

.

.