[an error occurred while processing this directive]
Period linked file
To create the period linked files for each year 1990-2003, death certificate numbers of infants (under one year of age) who were Indiana residents were matched and linked to their corresponding birth certificates. Accuracy of linking was determined by measuring consistency between various items common to both the birth and the death certificates such as date of birth, infantís first name, infantís last name, infantís sex, and motherís maiden name. This link created a single record, containing information found on both the birth and death certificates, from the two previously separate records.
Throughout 1990-2003, 98.2 to 99.4 percent of the Indiana resident infant death records were successfully matched with their corresponding birth records (Table I). Of the matched birth/infant death records, those who were not Indiana residents at the time of birth were excluded, and the remaining matched infant deaths (residents of Indiana at both the time of birth and the time of death) comprised the final Indiana period linked file for each calendar year. As a result, the overall yearly infant mortality rates (IMR) in this report, which are based on the linked birth/death file, are slightly lower than the rates reported in the general vital statistics mortality reports, which are based on the vital statistics mortality files that include all infant deaths who were Indiana residents at the time of death (Table II, Figure I).
Due to the consistently high percentages of death records that could be linked to their birth records throughout the 1990 to 2003 period (98.8%), weighting of data at the State level was not considered necessary. Individual counties, however, varied in the proportion of infant death records that could be linked to their corresponding birth records. In Marion County, only 0.8 percent of infant death records could not be linked to their birth records, compared to 3.6 percent of infant death records that count not be linked in Lake County (Figure II). Therefore, there is more underestimation of IMR in Lake County compared to Marion County due to the less successful linking of records in Lake County.
Infant mortality rates
Infant mortality rates are calculated by dividing the number of infant deaths in a period (numerator) by the number of live births in the same period (denominator), times 1,000 or 100,000. For example, the 2003 period linked file contains a numerator file that includes all Indiana infant deaths in 2003 that have been linked to their corresponding birth certificates, whether the birth occurred in 2003 or in 2002, and a denominator file that includes all Indiana live births during 2003.
The IMR is subject to random (or chance) variation in the number of births or deaths involved. Rates based on fewer than 20 deaths are considered unstable, because their 95 percent confidence interval is about as wide as the rate itself (1). Therefore, in this report, IMRs based on less than 20 deaths are not reported.
Random variation and stability of rates
In this report, the number of infant births and deaths represents complete counts for the State of Indiana. Therefore, the reported infant mortality rates are not subject to sampling error. However, when rates are compared over time, between areas, or among various subgroups, the number of events and the corresponding rates are subject to random variation. That is, the rate that actually occurred may be considered as one of a large number of possible different outcomes (rates) that could have arisen under the same circumstances (2). As a result, rates in a given population may tend to fluctuate from year to year even when the health of that population is unchanged. The simplest method for addressing the issue of random variation is the computation of 95 percent confidence interval. This interval indicates that it has 95 percent probability of including the true rate.
Random variation in rates based on a relatively small number of events tends to be larger than that for rates based on more frequently occurring events. A useful rule is that any rate based on fewer than 20 cases in the numerator (infant deaths in this report) will have a 95 percent confidence interval, which is about as wide as the rate itself (1). For example, in an area with 20 deaths out of 1,000 live births, it can be said that the true rate is within 20 +/- 10 per 1,000, which is not precise information. For this reason, in this report, infant mortality rates based on fewer than 20 deaths are not reported. One way to deal with the stability problem is to combine several years of data to increase the number of events, reduce the effect of random variation, and improve the reliability of the mortality rates.
When the number of events, in this case the number of infant deaths, is large, the relative standard error (RSE) is small, and the binomial distribution is used to estimate the 95 percent confidence interval. When the number of events in the numerator (infant deaths) is less than 100, the confidence interval for the rate can be based on a Poisson distribution (3). The formula for RSE of the IMR is:
where D is the number of deaths, B is the number of births, and SQRT denotes square root of the expression in parentheses.
The formula for 95 percent confidence interval based on binomial distribution is:
The formula for 95 percent confidence interval based on Poisson distribution using Table III is:
where Dadj is the adjusted number of infant deaths used to take into account the RSE of the number of infant deaths and live births and is computed as follows:
L (Dadj) and U (Dadj) refer to the values in Table III corresponding to the value of Dadj.
Comparison of two infant mortality rates
If either of the two IMRs to be compared (R1 or R2) is based on less than 100 infant deaths, first compute the 95 percent confidence interval for each rate and then check to see if they overlap. If they do overlap, the difference is not statistically significant at the 0.05 level. If they do not overlap, the difference is considered statistically significant.
If both rates are based on 100 or more deaths, the following Z-test is used to test for significance:
If Z is equal to or greater than 1.96, the difference is statistically significant at the 0.05 level; and if Z is less than 1.96, the difference is not statistically significant.
In this report, the statistical analyses were performed only on change in IMR between the two periods of 1990-1994 and 1999-2003 in various categories.
Race and Hispanic origin
The racial and ethnic designation used in this report is that of the mother from the birth certificate. The linked file provides more accurate data for computing the IMRs by race and Hispanic origin compared to that reported in the general vital statistics mortality report (4). In the linked file, the race of the mother from the birth certificate is used both in the numerator and the denominator of the IMR. In contrast, in the general vital statistics mortality report, the race information in the denominator is that of the mother from the birth certificate, while the race information in the numerator is the race of the decedent recorded on the death certificate as reported by an informant or on observation. As a result of this difference in the method of reporting race data in the linked file and general vital statistics mortality file, race-specific IMRs from these two data files may be slightly different.
Birth-weight and gestational age edits
Birth-weight and gestational age edits were performed according to National Center for Health Statistics (NCHS) guidelines (5). Birth weights below 227 grams and above 8,650 grams were considered as unknown. Birth weights within the acceptable range of 227 to 8,650 grams were checked for consistency with gestational age. Gestational age in completed weeks was calculated from the date of last normal menstrual period to the date of birth. The clinical estimate was used for 2.8 percent of the records in which the date of last menses was missing (1.8%) or when the computed gestational age was either out of the acceptable range of 17 to 47 weeks (0.8%) or inconsistent with the birth weight (0.2%).
Age-appropriate education of mother
In this report, motherís educational attainment is reported as the years of completed education regardless of motherís age as well as an age-appropriate education. A motherís education was considered appropriate if she had completed high school or had the appropriate number of grades for her age. For example, for a 17-year-old mother, at least 11 years of education was needed to be considered age-appropriate. For teenage mothers, the age-appropriate education is presented to correct for the effect of motherís age on years of education.
Prenatal care utilization
In addition to initiation of the prenatal care as a measure, different prenatal care indices have been developed as alternative measures based on the timing of the prenatal care, the number of prenatal visits, and the gestational age of the infant at birth. In this report, the Kotelchuck Adequacy of Prenatal Care Utilization (APNCU) Index is used (6). This Index includes the adequate plus care category for those women with unexpectedly large number of prenatal visits given the gestational age at delivery and the month prenatal care began.
The Kotelchuck Index does not assess the quality of the prenatal care, simply its utilization. The prenatal care utilization is considered inadequate if care is initiated after the fourth month of pregnancy regardless of the number of visits. Once prenatal care is initiated during the first four months of pregnancy, then the index is classified into inadequate, intermediate, adequate, or adequate plus if the ratio of actual-to-expected number of visits is less than 50 percent, 50-79 percent, 80-109 percent, or 110 percent and more, respectively. The adequate-plus category mainly includes women who are considered at high risk and receive extra prenatal services and have a disproportionately high share of low birth weight infants compared to other categories (7). The prenatal care index is considered unknown for mothers who have missing information/inconsistent values on the initiation of prenatal care, the number of visits, or gestational age, and for mothers with duration of gestation below 20 weeks and above 44 weeks (considered unacceptable values according to Kotelchuck Index).