Article Text
PDF
Abstract
Background In the USA, the epidemiologic features of type 1 diabetes are not well-defined across all 50 states. However, the advent of large nationwide insurance databases enables the investigation of where type 1 diabetes cases occur throughout the country.
Methods An integrated database from a large nationwide health insurer in the USA (Clinformatics Data Mart Database) was used, from 2001 to 2017. The database contained longitudinal information on approximately 77 million people.
Results The incidence of type 1 diabetes was greatest in areas of low population density across the 50 states. Individuals in the lowest population density areas had rates that were 2.28 times (95% CI 2.08 to 2.50) that of persons living in high-density areas. This association was consistent across various measures of rural status (p<0.001 for population density; p<0.001 for per cent rural as defined by the US Census Bureau; p=0.026 for farmland). The association between rural areas and the incidence of type 1 diabetes was evident across all four general regions of the USA.
Conclusions The predilection of type 1 diabetes in rural areas provides clues to potential factors associated with the onset of this autoimmune disease.
- geography
- diabetes
- epidemiology
- epidemiology of diabetes
- neighborhood/place
Statistics from Altmetric.com
Introduction
Type 1 diabetes mellitus is a chronic autoimmune disease that often begins during childhood and exhibits considerable geographic variation worldwide.1 The highest incidence rates are found in northern Europe, especially in Scandinavia.1 While there are known genetic factors involved in the aetiology of this disease, such as specific human leucocyte antigen haplotypes,2 it has long been suspected that environmental factors also play a role.3 A considerable body of evidence has implicated gastrointestinal viruses as possible contributors to pancreatic beta-cell destruction and the onset of this disease.3 4 In addition, there is recent evidence implicating untreated drinking water as the source of the enteric viruses that are associated with type 1 diabetes in children.5 Although incident cases of type 1 diabetes in the USA are estimated to be 64 000 annually,6 little is known regarding geospatial correlates of its incidence across all 50 states.
The advent of large healthcare databases containing both geographic and health data now affords the opportunity to study the distribution of type 1 diabetes in the USA. Therefore, a study was designed to assess the residential place where the onset of diabetes occurred. Because there is evidence that (1) enteric viruses may be involved in the pathogenesis of this disease3 4 and (2) enteric viruses occur more often in untreated water sources which are more common in rural areas in the USA,5 7 8 the association between residence in rural areas and the onset of type 1 diabetes was evaluated.
Methods
An integrated relational database from a large nationwide health insurer (Clinformatics Data Mart Database; OptumInsight, Eden Prairie, Minnesota, USA) was used. Members in the database (approximately 77 million people) were enrolled in private health insurance. Their health information contained date-specific outpatient and inpatient services, diagnoses and procedures, pharmacy records, laboratory data and demographic data. Data were extracted from 01 January 2001 to 30 June 2017. Incident cases of type 1 diabetes were identified using a published algorithm to extract diagnoses and insulin use.6 No restrictions were made by age or sex. This study was reviewed and declared exempt by the Institutional Review Board.
Incidence rates of type 1 diabetes (cases per 100 000 person-years) were calculated for each residential zip code (ie, area in which the person resided at the time of the patient’s first diabetes diagnosis or at the time of enrolment for those without diabetes). The congruence between zip codes and the US Census Bureau Zip Code Tabulation Areas (ZCTAs) was determined using the Missouri Census Data Center’s crosswalk file.9
US Census Bureau data were used to describe characteristics of each ZCTA. Three methods were used to determine rural status. The first was population density (number of people per square kilometre) for each ZCTA, which was then categorised into deciles. The second was the percentage of households that were in rural areas within each ZCTA. Rural areas were places of less than 2500 persons and were not considered urban by definition of the Census Bureau.10 The third was the percentage of people who lived on farms in each ZCTA. Farms were designated through the determination of whether US$1000 or more of agricultural products were produced and sold at that place.11
Additional covariates included the percentage of the ZCTA that contained children aged 0–15 years, the percentage that was of the white race and the percentage that was of Scandinavian ancestry. Because of prior evidence of type 1 diabetes occurring at higher latitudes,12 another covariate was proximity to Canada. That is, ZCTAs were chosen within states that shared a border with Canada or that were above the 46th parallel north in latitude. These were ZCTAs within Alaska, Washington, Idaho, Montana, North Dakota, Minnesota, Wisconsin, Michigan, Ohio, Pennsylvania, New York, Vermont, New Hampshire and Maine.
For the statistical analyses, zero-inflated negative binomial regression was used to assess the association between rural areas (explanatory variable) and incidence rates of type 1 diabetes in the area (ie, number of new cases of type 1 diabetes offset by the natural logarithm of person-years). The most urban areas formed the reference category. Each of the three rural status variables was modelled within a separate regression. Unadjusted incidence rate ratios (IRR) were calculated with 95% CIs, as well as IRR adjusted for age, race, Scandinavian ancestry and adjacency to Canada. In secondary analyses, results were constrained to youth, aged 0–19 years. Analyses were conducted using Stata/MP V.15.1. Alpha was set at 0.05, two-tailed.
Results
There were 43 827 individuals with incident type 1 diabetes in the database with a residential zip code. The incidence of type 1 diabetes was greatest (48.7/100 000 person-years) in areas of low population density (<5 persons/km2). The incidence was lowest (21.1/100 000 person-years) in areas with the greatest population density (≥1577 persons/km2). Using rural/urban as defined by the Census Bureau, the incidence rate of type 1 diabetes was 23.3/100 000 person-years in areas that were less than one-third rural, 28.2/100 000 person-years in areas that were 33%–66% rural and 38.7/100 000 person-years in areas that were greater than 66% rural. With regard to farmland, the incidence rate of type 1 diabetes was 24.2/100 000 person-years in areas that were <5% farmland, 36.7/100 000 person-years in areas that were 5.0%–9.9% farmland and 43.5/100 000 person-years in areas that were ≥10% farmland.
The increasing incidence of type 1 diabetes with rural areas remained evident after adjustment for age, race, Scandinavian ancestry and residence near Canada (table 1). Individuals in the lowest population density areas had incidence rates that were 2.28 times (95% CI 2.08 to 2.50) that of individuals in the highest population density areas. When using the Census Bureau’s definition of rural, individuals in the most rural areas had a 63% increased risk of developing type 1 diabetes compared with those living in the least rural areas. Additionally, people living in farmland areas had a 70% increased risk of developing type 1 diabetes compared with those living in the least farmland areas. When all three rural variables were forced into the final adjusted model as continuous variables, all remained statistically significant (p<0.001 for population density; p<0.001 for per cent rural; p=0.026 for per cent farmland).
Incidence rate ratios for type 1 diabetes by rural areas in the USA¶
The association between rural areas and the incidence of type 1 diabetes was evident across all four general regions of the USA (figure 1). The differential in incidence rates between rural and urban was the greatest in the Northeast region (IRR 2.23, 95% CI 1.92 to 2.60 for >66% rural compared with <33% rural). For the Central region, the IRR was 1.55 (95% CI 1.46 to 1.65) for >66% rural compared with <33% rural. The corresponding IRR for the South was 1.62 (95% CI 1.53 to 1.71) and for the West was 1.71 (95% CI 1.55 to 1.89). Individuals living in rural areas were more likely to develop type 1 diabetes, regardless of the general region of the country.
Incidence rate ratios for type 1 diabetes by rural areas and regions in the USA. Adjusted for % 0–15 years of age, % white, % Scandinavian and adjacency to Canada. Baseline category was <33% rural.
When the analyses were restricted to youth only (aged 0–19 years), population density, percentage rural and percentage farmland remained significant (p<0.001 for each).
Discussion
Type 1 diabetes occurs more frequently in rural areas of the USA. This association is independent of age, ancestry and northern latitude of residence. Studies in Finland,13 Sweden14 and the UK have found similar associations,15–17 although there are suggestions that socioeconomic factors may potentially modify these effects. A study in Poland, however, did not report evidence of a difference in incidence by rural status—but did find a north-south gradient.18
The underlying reason for the observed association between rural areas and type 1 diabetes in the USA is speculative. Various gastrointestinal viruses have been implicated in the aetiology of type 1 diabetes.3 4 19 Previous research has indicated an association between enteric viruses in untreated water supplies and type 1 diabetes.5 Untreated drinking water from rural domestic wells can periodically be contaminated with enteric viruses,7 8 although areas serviced by public water sources can occasionally become contaminated due to sporadic flooding and overflows.20 However, whether this oral-faecal route may play a role in the onset of type 1 diabetes is exploratory at this point. Environmental contributors to the onset of type 1 diabetes appear complex and may be modified by genetic factors.1–3
In conclusion, people living in rural areas in the USA are more likely to be diagnosed with type 1 diabetes than those living in urban areas. Location plays a role in the aetiology of this disease.
What is already known on this subject
Incidence rates of type 1 diabetes are highest in northern European countries.
In these northern European countries, cases tend to occur more frequently in rural areas.
What this study adds
Using a nationwide database containing 77 million people in the USA, incidence rates of type 1 diabetes were significantly greater in areas of low population density, in areas that the Census Bureau designated as rural and in farmlands.
The association between type 1 diabetes and rural areas was evident across all four general regions of the USA: Northeast, Central, South and West.
People living in the lowest density areas had two times the risk of developing type 1 diabetes as people living in urban areas in the USA.
It is important to identify where the onset of type 1 diabetes occurs because this may offer clues to the aetiology of this disease.
References
Footnotes
Contributors MAMR designed the study, acquired the data and conducted the data analysis; wrote the article and revised it for publication; and is responsible for the content of this article.
Funding This study was funded by the National Institutes of Health (grant UL1TR000433) to the Michigan Institute for Clinical & Health Research.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.