Little is known about the incidence of insulin-depen- dent diabetes mellitus (IDDM) within the United States. Estimates have been produced through sur- veys such as the National Health Interview Survey and by review of school records.(1, 2) Only two studies have been completed in the United States which employed hos- pital records to estimate the incidence of diabetes in chil- dren. The first, from the Mayo Clinic(3) in Rochester, Minne- sota, reported an annual diabetes rate of 14/100,000/yr for school-aged children (aged 6-18 yr of age) during the years 1945-69. Although case ascertainment was excel- lent, the rates were based on only 55 cases less than age 30 at diagnosis and there was no separation of IDDM (type I) and non-insulin-dependent diabetes (type II). The second study by Sultz (4) in Erie County, New York, examined the inci- dence of various childhood diseases and reported an in- crease in IDDM from 6/100,000 in the late 1940s to 11/ 100,000 in the early 1970s. The rates were below those of the Mayo Clinic and many other studies outside of the United States. Furthermore, most of the childhood chronic diseases examined by Sultz had a similar increase, sug- gesting that the results may have been an artifact of missing cases in the early years.
To examine the incidence of IDDM, a registry has been devBeloped for Allegheny County, Pennsylvania, through re- view of hospital records and surveillance of pediatricians for the years 1965-76. Allegheny County is a large urban community with the city of Pittsburgh being the major popu- lation center.5 In 1970, there were 563,457 individuals below age 20 in the county. The nonwhite population repre- sented 11% of the under 20 population and over 95% of these were black. The Pittsburgh IDDM Registry is the first population-based diabetes registry which includes a large nonwhite population.
The current report describes the development of the reg- istry and presents incidence by age, race, and sex for the years 1965-76.
Development ot the Pittsburgh IDDM Registry. The Pitts- burgh IDDM Registry is an ongoing registry designed to ascertain all new cases of IDDM under the age of 20 who are Allegheny County residents at time of diagnosis and who were diagnosed after 1964. The current report describes a 12-yr period of 1965-76. Ascertainment of the cases was accomplished by retrospective review of hospital records. To be included in the registry, the patient had to be on insu- lin therapy at hospital discharge. Since children, as verified later, are almost always hospitalized at onset or later, it was felt that a hospital-based survey would identify virtually all cases.
Allegheny County has one large children's hospital (Chil- dren's Hospitai of Pittsburgh) and 25 general hospitals that possibly would see young diabetics. All but one of the hos- pitals were surveyed (permission was not obtained for one small general hospital). Few cases were likely to have been missed by this omission because of its size. In addition, we reviewed physician records from two large private practices devoted primarily to diabetes.
Monthly listings of discharges as compiled by the Hospi-
tal Utilization Project (HUP) were carefully inspected in
Subjects identified only from readmission records were included only if verification of residence was obtained from municipal telephone books for the year of diagnosis. Hospi- tal records at the time of diagnosis were obtained for over 90% of the cases.
A few of the hospitals were not members of HUP in the earlier years. In these hospitals, attempts were made to use manual systems to identify the diabetics. Occasionally, listed records could not be retrieved. However, we believe that most of these problems were minimized because dia- betic children are likely to be readmitted to a hospital in the county and would, therefore, be subsequently identified. We have estimated the underascertainment rate and it is pre- sented in the results section.
It is important to note that over 50% of the IDDM cases are diagnosed at Children's Hospital. Another 20% diagnosed at other hospitals are seen later in Children's Hospital's dia- betes programs. Thus, although there were 26 hospitals in the county, over 70% of the diabetics are eventually seen at Children's Hospital and could be identified through review of that hospital's records.
To check the completeness of ascertainment, all pedia- tricians (approximately 150) in Allegheny County were re- quested to identify children under their care who met the cri- teria for inclusion in the registry. In the county most children are seen by pediatricians. They were asked to provide the name, birthdate, residence, date, and hospital of diagnosis for each child. There was a 92% response rate with 1 82 dia- betic cases identified. Several large practices refer all dia- betics to Children's Hospital and thus their patients were not
We also attempted to validate the registry through the identification of diabetics within the school systems. For reasons of privacy, little information was obtained and the school surveillance was discontinued.
Frequency of cases (numerator). Cases were classified according to age, race, and sex for the period of diagnosis between 1965 and 1976. As presented in Table 1, over the 12 yr, there were 937 cases diagnosed where full informa- tion was obtained on sex, race, and age at diagnosis. Two additional cases were excluded because race was not re- corded on the medical records. All cases identified as non- white were black. There were many more white children de- veloping the disease than nonwhite and the largest number of cases occurred in the 10-14-yr age group. Table 2 pre- sents the breakdown of the number of cases by year. This varied from 56 in 1 967 to 89 in 1 969 with no systematic trend discernible.
Population (denominator). To determine the incidence rates, the population of Allegheny County was estimated for each year. Estimates for 1965-69 were linearly interpolated from the 1960 and 1970 censuses.(5) Estimates after 1970 were provided by the Allegheny County Health Department. Temporal trends in incidence rates. The 12 yr were di- vided into 3-yr periods to yield more stable estimates of the incidence rates, particularly for the nonwhites where cases were few.
Table 3 presents the average yearly age-adjusted rates for the four time periods and age and sex combinations. The incidence rates were age-adjusted using the age distribu- tion of the 1970 total Allegheny County population as the standard and 95% confidence intervals were determined.(6) We have included in the appendix the yearly age-specific rates for the grouped time periods. The large number of age-specific rates followed a similar pattern to the age-ad- justed rates.
As shown in Table 3, the age-adjusted rates did not ex- hibit a consistent pattern for the 12-yr period for any race and sex combination. Examination of the rates and confi- dence intervals over the 12 yr indicates that the variations over time were likely chance occurrences.
Overall incidence rates. Since no consistent temporal pat- tern appeared, the data were collapsed across the total time
Age of occurrence of diabetes. Table 5 presents the mean age at onset of diabetes for the race and sex classifications. None of the comparisons was significantly different. How- ever, when the pattern of the age of occurrence was exam- ined for the white population (Figure 1), there were earlier age-specific peaks for females than males. The peak ages of onset were 1 0 and 1 2 for females and 1 3 for males. Simi- lar findings have been noted in other studies.(7,8) Nonwhites were also examined but the population size was too small for valid conclusions.
The analysis of the Pittsburgh IDDM Registry revealed that the yearly rate of occurrence for diabetes in individuals less than 20 yr of age ranged between 10/100,000 for black males and 1 6/1 00,000 for white males. Table 6 presents the incidence estimates for IDDM in other population studies so that our results can be compared with the previously re- ported literature. Included are the number of diabetics iden- tified in each study. Our incidence rates are higher than most others. The differences in rates among the studies may
It has been suggested that the incidence of IDDM is in- creasing in the United States.(16) These conclusions have been primarily based on Sultz's data mentioned previously.(4) There have been fears that an increase might be the result of the widespread use of measles, rubella, and mumps vac- cines. In Allegheny County, vaccination programs began in the early 1970s with prompt and ongoing immunizations of most of the pediatric population. There was no increase of IDDM incidence from 1965 to 1976, providing little evi- dence to suggest that immunization programs caused an in- creased incidence of diabetes.
Recent theories have been set forth concerning racial dif- ferences in the etiology of diabetes.17,18 These theories have been promulgated on the assumption that IDDM in blacks is rare. Estimates have been provided that whites are five _ 95% No. of 95% Cl/100,000 Population cases Rate/100,000 Cl/100.000 15 7.98 4.02 7,672 2 8.69 12.04 29 14.28 5.19 7,901 2 8.48 11.72 34 16.28 5.48 7,662 1 4.46 8.64 21 12.10 5.19 6,148 3 15.84 18.16 99 12.79 2.52 29,383 8 9.08 6.29 2.982.57 9.30 6.56 15 9.11 4.61 7,320 0 0.00 0.00 32 16.43 5.69 8,119 1 4.14 8.08 42 19.45 5.88 8,251 2 7.89 11.07 37 19.48 6.29 6,911 2 9.83 13.50 26 16.48 2.88 30,600 5 5.45 4.77 16.572.90 5.83 5.12 12 7.47 4.23 8,165 1 4.34 8.25 40 24.45 7.57 6,866 3 15.61 17.06 49 23.81 6.66 8,254 3 12.02 13.66 26 13.28 5.10 7,314 1 4.60 8.98 27 17 50 3.04 30,600 8 8.70 6.03 17.863.11 9.41 6.45 4 2.75 2.72 8,444 0 0.00 0.00 30 23.09 8.25 5,684 2 11.99 16.43 4s 23.70 6.94 7,916 7 29.76 21.94 28 13.85 5.12 7,780 4 16.97 16.71 6 053.04 29.824 13 14 64 7.93 6.673.20 15.84 8.63 times as likely to develop diabetes as blacks.(19) The studies have been hospital-based prevalence studies. There are questions concerning the representativeness of the popula- tions. In our data whites were slightly more likely to develop diabetes than blacks. Instead of a white/black ratio of 5: 1 as has been suggested, we estimate that the ratio is at most 1.5:1. However, an interesting pattern of racial differences appeared. These were primarily centered in the younger age groups in which the ratio of white to black approached 2: 1. It is important to note that the number of nonwhites was small, thus the confidence intervals were large. The data, however, suggest that whites have a slightly higher rate of IDDM than nonwhites, particularly in the younger age groups.(/P>
The white/black difference appeared to be real, although not a large effect. Furthermore, the major racial differences were in the younger age groups, suggesting that there may be something unique about susceptibility to diabetes in younger children. In addition, the difference in age at onset for males and females suggests that the diabetes may be in- fluenced by maturation or growth.
The rates of occurrence of IDDM in other areas of the United States in future registries must also be determined to estimate the magnitude of the disease and to study risk fac- tors that are possibly related to incidence. Future epidemio- logic studies in diverse populations may lead to identifying etiologic factors in the occurrence of insulin-dependent diabetes mellitus.
ACKNOWLEDGMENTS