In this section, I will present the criteria to distinguish between Type 1 and Type 2 diabetics. I will include definitions and descriptive statistics of the variables included in the study, and I will also provide the design of the four sets of regressions that I performed.
2.1 Criteria for Distinguishing Type 1 from Type 2 Diabetics
Using a set of approved criteria to distinguish between Type 1 and Type 2 diabetes has been very important to my study, as I started getting correlation when I separated the data by the two main diabetes types. The lack of correlation when the groups are combined and the overall correlation when their data is run separately against economic and medical variables confirms that the circumstances of Type 1 and Type 2 diabetics are different from each other.
To diagnose patients as diabetic, doctors typically use an A1C test or a combination of two glucose tests (measuring glucose in the plasma and administering an oral test; Seley 2009). A score of 6.5 or more indicates the effects of diabetes. To distinguish between Type 1 and Type 2 diabetes, doctors have traditionally considered age and BMI (body mass index; Fourlanos et. al 2006 and Utzschneider et. al 2004). However, new research suggests that a diagnosis based on BMI is proving to be an inaccurate diagnosis in some cases. Some patients originally diagnosed as Type 2 are now being recognized as Type 1 and vice versa.
For example, researchers have discovered that some adults originally diagnosed as Type 2 diabetic have characteristics of Type 1 diabetes (Golden et al. 2010), and children can be misdiagnosed as well (Brunk, et al. 2011). Researchers involved in a study of obese teenagers by the American Diabetes Association indicated the importance of performing lab tests to measure autoantibodies and C-peptide levels that will either identify or rule out Type 1 diabetes. The results of such lab tests are not available for patients in the MEPS data set, but I can distinguish between Type 1 and Type 2 diabetics with reasonable certainty using the available data.
ICD-9 diagnostic codes were used by doctors in 2009 for medical billing purposes. For example, a code of 250.00 indicates Type 2 or undefined, 250.01 indicates Type 1 diabetes without complication, and 250.02 is Type 2 diabetes without complication – uncontrolled. Other decimals, for example, indicate complications such as ketoacidosis, or renal, ophthalmic or neuropathic manifestations (ICD9Data.com 2017). In the 2009 MEPS Study, ICD-9 codes were provided in HC128, a separate spreadsheet from the main HC129 spreadsheet. The study includes 3-digit codes such as 250, but, unfortunately, specifying decimals attached to the codes were not included for privacy reasons. Instead, I relied on the self-reported as diabetic column to distinguish between diabetics and nondiabetics, but I utilized the 3-digit ICD-9 column as a check. When patient IDs were compared in each, the columns matched up and confirmed themselves.
By utilizing information available in the dataset, I was able to determine with reasonable certainty whether identified diabetic patients were Type 1 or Type 2 diabetic. In addition, I compared my results to national statistics, which estimate that Type 1 diabetics make up 5-10% of all diabetics. My results assigned between 6-7% of diabetics in the dataset as Type 1 diabetic.
Table 2.1 lists the seven criteria that are presented in the order in which they were applied, with the first criteria given higher importance than the last. I have also included a description of each. Figure 2.1 is a flow chart that illustrates the execution of the criteria upon the dataset in the order of importance.
The MEPS survey is self-reported, and often, patients did not fill out all the data fields requested, or the data was otherwise not provided. If patients did not include enough information or the information was conflicting to the point that the data was inconclusive, the patient was removed from my study. The following are a few examples of missing data and how I would classify them using the criteria:
Example 1: A patient has indicated that they are diabetic, they are taking insulin, but they did not indicate what year they were diagnosed. However, their BMI is less than 27.5, so the patient was classified as Type 1 diabetic.
Example 2: A patient has indicated that they are diabetic, but they did not indicate whether they are taking oral medications or insulin. However, they were diagnosed over the age of 55 and their BMI is greater than 27.5. There is a chance that they could be GADA Type 1 diabetic, which often presents similar characteristics to Type 2, but the instance of GADA Type 1 diabetes is rare, so for the purposes of this study I classified them as Type 2 diabetic.
Example 3: A patient self-reports themselves as diabetic, current age of 50, and age of diagnosis as 49, but does not provide any BMI information and does not indicate whether they are taking oral medication or insulin. The patient is likely Type 2, but without sufficient information, the patient is excluded from my study.
A self-reported dataset, the 2009 MEPS study includes more than 36,000 observations (patients; see Table 2.2). Each observation was weighted to represent the United States’ population. About
one-third of the patients could not be included in the study because of a lack of self-reported data that prevented them from being compared accurately with the others, but the slimmed down data set still represented the intended population.
About 29.1 million individuals, or 9.3% of the population, have diabetes in the United States (American Diabetes Association 2016). 9.5% of observations in the self-reported MEPS study, which included individuals 18 and older, indicated that they had diabetes. Among the weighted numbers, only 6.41% identified themselves as diabetic. In the study, no children under the age of 18 identified themselves as diabetic, but we know that there are children in the United States who have diabetes.
To improve overall accuracy, I removed from the dataset all subjects under the age of 18, including nondiabetics (13,387 observations). Doing so adjusted the total percentages among the weighted numbers, and this action explains the discrepancy in how the weighted numbers compare to national averages. For purposes of analysis in the descriptive statistics section, I separated the observations into five age groups, as shown in Table 2.3.
Notice that there are fewer Type 1 diabetics in the Age ≥ 65 category, and this is because many have traditionally died before making it to retirement. This statistic is changing as improved insulins and glucose meters help patients to better monitor their diabetes More patients are able to utilize an insulin
pump and sensor, the latest models of which work together as an artificial pancreas and greatly improve the quality of life of patients.
Demographic Variables. In all four of the regression sets, I ran certain demographic statistics as independent variables in every analysis that was performed. Those variables are listed with their descriptions in Table 2.4. I performed the regression analyses in the statistics software STATA, which requires that nominal data be in binary format – in ones and zeroes – and I formatted the MEPS data accordingly. For example, I separated the race data into separate columns for each of six racial groups, with the seventh, Caucasian, serving as the null value.
Variables included in the MEPS data that are indicators of socioeconomic status (see Table 2.5) were tested in separate computations alongside the demographics listed above. Impoverished individuals may have limited access to healthcare services for a variety of reasons. It may be as simple as a lack of transportation. Others may not be aware of what is required to maintain a healthy A1C level. Some poor dietary habits can be rooted in a lack of education or the inability (or refusal) to consume fruits and vegetables. Others are the result of food deserts, when residential units in the inner-city that are located more than a mile from the nearest grocery store (Levine 2011).
Table 2.6 includes descriptions of variables that indicate a patient’s access to health insurance, and includes insurance type, which is utilized in the fourth regression set of diabetics only (including
private insurance, Medicaid, Medicare, and Tricare/ Champ VA). Tricare is insurance for uniformed service members, retirees and their families, and Champ VA is insurance for disabled military veterans. People with Type 1 diabetes are disqualified from entering the military, but if the onset of the condition occurs to an active member of the military they aren’t automatically sent home (Hieronymus and
Rickerson 2012). Medicare covers people over the age of 65 as well as those who are disabled or who have end-stage renal disease. Diabetics who are on dialysis have end-stage renal disease (Mayo Clinic 2017-2) and can be under the age of 65.
The fourth regression set also uses a variable regarding whether diabetics were treated for other priority conditions other than diabetes, such as heart disease, skin conditions, or asthma, which are also expensive to treat and maintain. Additional variables include whether a patient received recommended tests in 2009 which detect the onset of secondary symptoms, including four A1C tests, at least one neuropathy test and at least one retinopathy test. Although other tests are used, the tests mentioned were selected for my study because of the available MEPS data.
2.3 Descriptive Statistics.
The following figures provide a comparison of the descriptive statistics between Type 1 and Type 2 diabetics, as well as nondiabetics in each age group for the variables. Actual numbers are provided in Appendix A.
The number of years that a patient has lived with diabetes is more indicative than age of the onset of conditions such as retinopathy or nephropathy. Figure 2.2 compares by age group the mean number of years that individual patients have lived with the disease. This variable was added to the third regression set to compare it with current age. I wanted to demonstrate whether the number of years with the disease or age plays a more significant role in the advancement of diabetes.
In Figure 2.3, women (represented as 1 in the data) were slightly more likely to be Type 1 diabetic than men (represented as 0 in the data), especially those age 50 and younger. In the age 18-24 group, women were a whopping 95 percent of Type 2 diabetics, although the high percentage was likely due to the small sample size, with only nine individuals with Type 1 and seven with Type 2 diabetes. Women were also more likely than men to be diabetic in the age 25-40 group, and men were only slightly more likely to be Type 2 diabetic than women in the 41-50 age range.
Being married has been shown to improve an individuals’ health, particularly among males. The improvement is partly the result of having improved access to quality and affordable health care, since married individuals have access to both their own and their spouses’ health care provider (Gomez, et al. 2016). As illustrated in Figure 2.4 (where single, divorced or widowed is 0 and married is 1 in the data), nondiabetics were more likely to be married than diabetics in most age categories. Type 1 diabetics were the least likely to be married, which affects their socioeconomic status and access to health care.
Not many individuals diagnosed with Type 1 diabetes have traditionally lived to be age 51-64, but newer medicines and lifestyle recommendations have made it possible in recent years for individuals with this disease to live longer, given their access to and ability to afford the needed healthcare and prescriptions. For example, Type 1 diabetics who can afford insulin pumps, glucose monitors can significantly delay the onset of secondary symptoms and thus reduce their overall diabetes-related expenses (Huang, et al. 2010). Pancreas transplants are also performed, usually in conjunction with another organ transplant, and they can effectively cure a patient for up to 10 years.
Race. The 2009 MEPS race data for total nondiabetics (see Figure 2.51) lined up closely with the projected U.S. population for 2010 (2009 U.S. Census; see Table 2.7). However, diabetes does not occur among races proportionally. In fact, in recent years, the Center for Disease Control and Prevention (CDC) reported in the 2014 National Diabetes Statistics Report that new cases of diabetes are skyrocketing, particularly among U.S. minority populations (see Table 2.8).
Although Caucasians (whites) still make up a majority of total diabetics today, the numbers are changing. According to the CDC (2014), new cases of diabetes in 2000 affected 7.6% of Non-Hispanic whites, 9% of Asian Americans, 12.8% of Hispanics, 13.2 percent of Non-Hispanic blacks, and a staggering 15.9% of the small Native American population (24.1% of which are living in southern Arizona reservations).
The Hispanic population is on the increase in the United States, and the U.S. Census (2017) predicts that it will grow substantially in the next 40 years. By 2050, the group is expected to comprise almost 30% of the U.S. population, nearly double their current ratio. Cases of both Type 1 and Type 2 diabetes among this population are also on the rise from previous years.
The MEPS variable indicating race originally had values of:
1 – Caucasian (white)
2 – African American
3 – American Indian or Alaska Native
4 – Asian
5 – Hawaiian or Pacific Islander
6 – Multiple races reported
For purposes of conforming to STATA’s requirements, I divided the original race column into five binary columns, with Caucasians holding the null value and African Americans, Native Americans, Asians, Pacific Islanders, and Mixed Race each having a column where their respective value was one. Hispanics had separate columns in the raw data, and I added it as a sixth option for race. 394 individuals classified as mixed race, and of those, 52 indicated Hispanic origins, 141 listed as African in origin also identified as
Hispanic. Of the Native Americans, 136 listed some Hispanic ethnicity, 18 Asians included some Hispanic ethnicity, and 12 Pacific Islanders included some Hispanic ethnicity.
Table 2.5.1 displays the races of observations from all age groups combined. All race categories comprised percentages that were proportional to their populations. Native Americans are only proportionally represented among Type 2 diabetics and only have trace representation among Type 1 and nondiabetics.
Neither Type 1 nor Type 2 diabetes is believed to be genetically tied to a specific race. In Bao, et al. (2013), when analyzing multiple studies that sought to identify whether there was a genome-wide association (racial origin) of the risk of diabetes, the results indicated that genome-wide association did not provide additional foresight from traditional lifestyle indicators, such as an elevated BMI, elevated blood pressure, and a family history of diabetes, to identify the likelihood that someone would develop diabetes. Researchers (Charatan 2001) have found that eating a healthy diet and getting regular
exercise can reduce the risk of developing Type 2 diabetes by half, despite any genetic predisposition to the disease. Any racial correlations can be explained by cultural traditions and community lifestyles.
In the 18-24 age category (Table 2.5.2), Hispanics comprise 33 percent of Type 1 diabetics and nearly 65 percent of Type 2 diabetics.
In the 25-40 group (Table 2.5.3), the number of Hispanic Type 2 diabetics is still elevated at 24.2%. African Americans also comprise 25.7% of Type 2 diabetics and 19.8% of Type 1 diabetics. It is
unknown what causes Type 1 diabetes, but there does not appear to be a racial component to the disease. In all age groups, cases of Type 1 diabetes appear to match proportions of racial populations.
It is known that instances of Type 1 diabetes are on the increase worldwide, particularly among children. Patterson et al. (2009) reports that the number of diagnoses in Europe among children age 15 years and younger is increasing an average of 3.9% each year, and prevalence is predicted to rise
from 94,000 in 2005 to 160,000 in 2020. This trend may explain why in this study cases of Type 1 diabetes are higher among minority races in the younger age groups.
Figure 2.6 compares patients’ subjectivity to the poverty line based on annual household income, where the poverty line equals 1. The darker colored dots represent the mean, and the corresponding bars represent -1σ to 1σ. When looking at all age groups combined, you can see there was little variability between the Type 1, Type 2 and nondiabetics regarding where patients stood against the poverty line. However, differentiation was more apparent when the data was broken into age groups. In almost every age group, diabetics were poorer than nondiabetics, and Type 2 diabetics were poorer than Type 1 diabetics in many age groups. It is unclear from the data whether diabetic patients were experiencing poverty at the time of diagnosis or if poverty had developed post-diagnosis.
Type 1 diabetics had more education than Type 2 diabetics overall and in most of the age categories. Type 1 diabetics and nondiabetics also had similar levels of education in every age range except ≥ 65 and the total, as illustrated in Figure 2.7. However, if someone’s stated number of years of education was less than zero, they were removed from the data set, as a negative number indicates an answer such as “I refuse to say” or “I don’t know.” A patients’ level of education is important to this study, as education plays a significant role in socioeconomics. As a result, I removed 396 individuals from the population sample.
Over the course of their lifetimes, Type 1 diabetics will spend more out-of-pocket for prescriptions, in particular, but also for all medical expenses, than Type 2 diabetics. It should be noted that the MEPS data only represent patients’ experiences in 2009. In Figure 2.8, patients who were not able to receive financial care were strongly linked to the inability to afford the care. In all age groups combined, Type 1 diabetics had the highest instance of this challenge, especially in the age 25-40 age
group, where they had more than double the percentage (12%) of patients with this problem than any other segment.
In Figure 2.9, only about half of patients who said they were delayed in getting care identified cost as the reason for the delay. The data, when split into age groups, varied, but when all age groups were combined, the data was consistent between Type 1, Type 2 and nondiabetics, with Type 1 and Type 2 diabetics only slightly higher than nondiabetics.
Patients who are not able or are delayed in receiving care are less likely to maintain healthy A1C levels and are thus more likely to experience an accelerated onset of secondary symptoms. The appearance of secondary symptoms quickly increases medical expenses for all diabetics, as it often involves a hospital stay and can require expensive procedures.
Patients who balance their carbohydrate and protein intake, eat the recommended amount of fruits and vegetables, and get regular exercise can delay the onset of secondary symptoms such as blindness, foot amputations, heart disease and stroke. These events skyrocket medical expenses that are attributed to the treatment of diabetes and can lead to the inability to work (Secrest et al. 2011).
In Figure 2.10, which discusses patients who self-reported that they were unable to get prescriptions, patients acknowledged that their inability was due to financial reasons. However, more Type 2 diabetics said they were experiencing this challenge than Type 1 or nondiabetics.
In Figure 2.11, Type 1 diabetics were more likely to be delayed in getting prescriptions than
Type 2 diabetics, especially in the 25-40 age group. These are individuals who are able to get prescriptions but are not able to get them in a timely way. Type 1 diabetics acknowledged that financial reasons were the cause for the delay in getting prescriptions for both the 18-24 and 25-40 age groups. In all age groups combined, Type 1 and Type 2 diabetics were significantly more likely to be delayed in getting prescriptions than nondiabetics.
The financial reasons unable or delayed variables were assessed at a value of one for each of the following reasons listed in the MEPS:
- Could not afford care
- Insurance company would not approve/cover/pay
- Doctor refused family insurance plan
All other possibilities were given a value of zero. The “unable to get” and “delayed in getting care,” as well as “unable” and “delayed in getting prescriptions” variables were adjusted to values of one for yes and zero for no.
Both Type 1 and Type 2 diabetics often experience challenges in the workforce because of their medical conditions. Administering insulin to themselves during work hours or needing to take an extended break if they experience a sudden blood sugar low may seem to some as beyond “reasonable accommodations.” But diabetics are protected under the Americans with Disabilities Act Amendments Act (ADAAA) of 2008, which established protections for individuals who have chronic medical conditions (American Diabetes Association 2014). ADAAA applies to all employers with 50 or more employees. Unfortunately, diabetics may find themselves getting passed up for promotions when in competition with fellow employees, but diabetics, even those with secondary symptoms, statistically have not been found to have less productivity in the workplace than nondiabetics (Lavigne et al. 2003). They rely on quality and affordable insurance from their employer.
Figure 2.12 shows the instance of patients on food stamps. In general, Type 2 diabetics were closer in line with nondiabetics than Type 1 diabetics, who were less likely to be on food stamps. Type 1 diabetics’ highest instance of food stamp use was in the age 18-24 group at 17 percent. Type 2 diabetics’ highest instance of food stamp use was in the 25-40 age group, at more than 30 percent. Nondiabetics’ highest instance of food stamp use was at about 12 percent in both the 18-24 and 25-40 age groups.
Figure 2.13 illustrates the monthly monetary value of food stamps and displays both the mean and standard deviation. Although Type 1 diabetics were generally less likely to be on food stamps, they
were issued the greatest value of food stamps in the 18-24, 25-40 and 41-50 age categories, where the means of each were about $400 per month. The standard deviation for Type 1 diabetics was also higher than for the other categories.
Figure 2.14, which compares family size, indicates that diabetics tended to have smaller households than nondiabetics overall. The total number of Type 1 diabetics had larger families than the total number of Type 2 diabetics, but breaking down the statistics by age groups provided additional information. Type 2 diabetics tended to have larger family sizes in the Age 25-40 and ≥ 65 groups, and nondiabetics tended to have larger family sizes in the 41-50 and 51-64 age ranges. The only age category where Type 1 diabetics had significantly larger family sizes was in the Age 18-24 group, when patients may still be living with their parents and siblings.
Figure 2.15 indicates that for the percentage of subjects who are insured, there are many more who are not insured, regardless of whether they have diabetes or not. The “ever had insurance” variable was the combination of three columns from the original data corresponding to the three rounds of the survey. The same was done for “employee insurance” and the three initial “employer offers insurance” variables. Uninsured for all of 2009 was adjusted to one for yes and zero for no.
Because of the exorbitant medical costs that they experience, most diabetics must have access to affordable health insurance to have any chance of sustaining a healthy lifestyle. As a result, they, or a member of their household, traditionally have had to work full time for an employer that provides health insurance, or they have had to receive government sponsored medical insurance such as Medicaid. Figures 2.16.1 through 2.16.4 indicate the percentage of individuals in the data set who were enrolled in various types of insurance in during 2009, including Tricare/Champ VA, Medicaid, Medicare and private insurance. In some cases, patients may have enrolled in multiple types of insurance. For
example, they may receive private insurance through their employer and also qualify for Medicare. In 2009, many self-employed individuals and those who work part-time were not able to qualify for affordable insurance because of pre-existing conditions, but since then, the Affordable Care Act has improved the access to health insurance for many (Shaw et al. 2014).
The MEPS data did not specify between those who were enrolled in Tricare, for active U.S. military and their families, or Champ VA, for disabled U.S. veterans. A column that represented the entire data set of all insurance types, as well as the uninsured, indicated if a patient was presently active in the U.S. military. There were less than 1% in all age groups, and they were all nondiabetic. Type 1 and Type 2 diabetics are disqualified from entering the military, but becoming diabetic during their service is not an automatic cause for dismissal. Another column indicated whether a patient had been honorably discharged from the military, which also represented the whole data set, and there was a proportional representation of Type 1, Type 2 and nondiabetics included.
Diabetics dedicate significantly more of their budget to medical expenses than nondiabetics, which means they typically have less expendable income. In addition to quarterly doctor’s visits and expensive oral or intravenous medications, diabetics may need to pay for needles, syringes, blood glucose meters and test strips. In impoverished situations, patients sometimes choose to pay for the needs of another family member at the expense of paying for part or all of their own medications.
It should be noted that a majority of diabetics in the study were treated for other priority conditions (see Table 2.9) in addition to diabetes. These conditions may be related to diabetes such as high blood pressure or a stroke, or they may be unrelated, such as a chronic skin condition or asthma. Priority conditions include a group of medical conditions that are selected for their prevalence, expense, or relevance to policy, including hypertension, heart disease, high cholesterol, emphysema, chronic bronchitis, diabetes, cancer, arthritis, asthma, attention deficit/hyperactivity disorder, and stroke (Agency for Healthcare and Research Quality 2017). I have included them as variables in the fourth regression set to acknowledge that patients’ medical expenses could be spent on other conditions besides diabetes. Priority conditions are used by the MEPS for statistical purposes to track the expenses associated with specific chronic medical conditions that merit their attention. Some examples of priority conditions are shown in Table 2.9.
Figure 2.17 indicates the percentage of patients who were treated for priority conditions in 2009, and it’s clear that a majority of diabetic patients are treated for other conditions that are important to maintain and can be expensive over time.
Figure 2.18 provides the percentage of diabetics who received all the required tests in 2009 to detect the onset of secondary symptoms. The tests include four A1C tests, one neuropathy test in the feet and one retinopathy test in the eyes.
2.4 Design of Regression Sets
My thesis project consisted of four sets of regression analyses. The dependent and independent variables used in the first set are listed in Table 2.10. The dependent variables were the population samples for Type 1 and Type 2 diabetics as well as nondiabetics, and the independent variables included the demographics and socioeconomic indicators. The demographics were run individually for Type 2 diabetics only, and they were combined and run with each individual socioeconomic indicator variable against all population samples.
Table 2.11 shows the dependent and independent variables for the second regression set. The dependent variables were the appearance of secondary symptoms in the population samples for Type 1, Type 2 and nondiabetics. Additional dependent variables were comparisons between Type 1 & nondiabetics and Type 2 & nondiabetics to determine if having diabetes makes an individual more likely to develop these symptoms. Nephropathy (kidney failure) and retinopathy (blindness), which are unique to diabetics, were run on Type 1 and Type 2 patients only against the independent variables including
the demographics and socioeconomic indicators. For these tests, each socioeconomic indicator was run individually along with the demographic variables combined.
Increased risks of cardiovascular problems and stroke are also associated with diabetes, so those factors were run as dependent variables in the regression tests against all population samples. These other dependent variables were run against the demographic variables combined.
The tests that the American Diabetes Association recommends on an annual basis, which include four A1C tests (one each quarter), a foot test for detecting the presence of neuropathy, a dilated eye test for detecting the presence of retinopathy, a cholesterol test and the flu vaccine were added as independent variables and run together against nephropathy and retinopathy for both Type 1 and Type 2 population samples. The tests were run along with the demographic variables combined, as well as the decimal value of the poverty line or one of the socioeconomic indicators.
The third regression set, shown in Table 2.12, is similar to the second set but adds an additional independent variable, the number of years with the disease. It was run against the Type 1 and Type 2
population samples only. This variable was generated by subtracting the current age of the patient from the reported age of diagnosis. Patients who did not list an age at diagnosis were not included in the set, which comprised approximately 10% of Type 1 diabetics and 5% of Type 2 diabetics. Since this data is self-reported, the age at diagnosis is approximate. Each regression in the set was run with age and without age to see whether age or the number of years with the disease has a stronger influence on the development of secondary symptoms. It was also performed to detect any differences due to the possibility of age and years with diabetes being correlated with each other.
Nephropathy and retinopathy were run against the socioeconomic indicators, in combination with all the demographics. The other secondary symptoms were only compared with the demographic variables combined. The three recommended tests used to detect the onset of secondary symptoms were run against the decimal value of the poverty line, as well as the demographic variables combined. In both the second and third sets, the test results reported for 2010, 2009, 2008 and before 2008 were included.
The independent variables for the fourth regression set, shown in Table 2.13, included whether patients have received the recommended tests, including four A1C tests, one neuropathy test, and one retinopathy test. The three tests are also combined as a fourth binary category. These variables were run against the population samples for Type 1 and Type 2 diabetics, as well as Type 1 & 2 combined.
Independent variables included the patients’ insurance type, such as private insurance, Medicare, Medicaid, or Tricare/ Champ VA. It is possible that some patients were enrolled in more than one insurance type. Additional independent variables are whether patients were treated for a priority condition, including a diabetes priority condition or other priority condition. The demographic variables were also included as independent variables and were run together along with each of the other independent variables
A fifth regression set included the appearance of secondary symptoms as the dependent variables and used Type 1 and Type 2 diabetics as two separate population samples. The independent variables included the insurance types as well as the demographics. This test was completed to answer the second research question regarding whether there is one insurance type more likely to delay the onset of secondary symptoms.