Our results are based on a population based follow-up study on Danish national registers. We used data from the Danish Civil Registration System (CRS) to identify singleton births in Denmark born January 1st
1979 through December 31st
1,548,746 (see for a flow chart of missing variables) and their next of kin (mother, father, siblings, mother's parents, and mother's siblings). All live born children and new residents in Denmark are assigned a unique civil personal registration (CPR) number, allowing accurate linkage of individual level data between registries.
Information on birth outcomes, such as gestational age at birth and birth weight, were obtained from the Danish Medical Birth Registry which has been computerized since 1973. Annual information on maternal education, residence, and income was retrieved from the Fertility Database at Statistics Denmark, available since 1979. The date of conception was estimated by the date of birth minus gestational age, and the prenatal exposure period included from 12 months before conception until the birth of the child. We categorized children as exposed to bereavement during prenatal life if their mothers lost a child, husband, sibling or parent during the prenatal period, ascertained via the death registry; all the remaining children were included in the unexposed cohort. Cohort members were followed from birth until first diagnosis of diabetes, death, emigration, or December 31st 2004, whichever came first.
Diabetic status of children and their parents was ascertained from the National Hospital Registers, which holds information on all discharges from Danish hospitals since 1977; outpatients have been included in the register since 1995. Diagnostic information was based on the Danish version of the International Classification of Disease (ICD) 8th revision, from 1979 to 1993 (249) and 10th revision, from 1994 onwards (E10). We also looked for additional cases from the Pharmaco-epidemiological Prescription Database, in operation since 1995, using the ATC codes for insulin analog drugs (A10A), though there was almost 100% overlap with these cases and those found in the hospital register. If an individual has a diagnosis for both type-1 and type-2 diabetes, they were classified as type-1 since the later is more often misclassified than the former. Also, type-1 diabetes is a more serious disease than type-2 and therefore probably better diagnosed.
We estimated incidence rate ratios (IRRs) from birth using log-linear poisson regression models and used person-years as the offset variable. The analysis was performed using PROC GENMOD in SAS version 9.1.3. Unadjusted and adjusted models were generated. Adjusted estimates of IRRs (aIRR) included maternal age (≤18, 19–34, 35–40, 41+); residence (Copenhagen, cities with over 100,000 inhabitants, and other); income (1st, 2nd, 3rd, and 4th quartile); maternal education (primary, secondary, high); parental diabetes status (any hospitalization diabetes ICD code); marital status (married, not married); gestational age at birth (<32 weeks, 32–36 weeks, 37+ weeks); birth weight (<1500 grams, 1500–2500 grams, 2500+ grams); sibling order (1,2,3,4+); calendar year (1979–1989, 1990–1999, 2000–2004) and offspring's sex (male, female). Age, residence, calendar period, maternal education, maternal income and parental marital status were treated as time-dependent variables as they were extracted from the birth year of the offspring.
To examine the association between timing of the bereavement and diabetes, we categorized exposed children by prenatal trimesters. To examine the relationship between type of bereavement and diabetes, we categorized exposed children by relationship to the deceased. We further categorized the cause of death by traumatic death (unexpected causes: ICD-8 codes 7950–7959, ICD-10 codes R95–R97; motor vehicle accidents: ICD-8 codes 8100–8230, ICD-10 codes V01–V89; suicide: ICD-8 codes 950–959, ICD-10 codes X60–X84; and other accidents and violence: ICD-8 codes 800–807, ICD-10 codes V90–V99, W00–X59, X85–Y89); and death from other causes. This study was approved by the Institutional Review Board at the University of California at Los Angeles. Individual written consent was not obtained from study participants as there was no record linking the subjects to identifying information.