The infant mortality rate of a state or country is used a measure of the overall health of that region . The infant mortality rate is defined as the number of deaths of children under one year of age per 1,000 live births, and although this definition is standard worldwide, the variability in definitions and reporting standards for live births and fetal deaths makes comparisons between and among regions problematic . Prior research has noted a difficulty comparing rates of Infant Mortality (IMR) in the United States to rates in other countries due to variation in reporting of births at the early stages of viability . Some European countries require a minimum gestational age of 22 weeks or a birth weight threshold of 500 g to register a live birth, while the United States and Canada register higher numbers of infants weighing less than 500 g; this results in higher reported infant mortality rates .
Comparisons between and even within states in the United States suffer from similar difficulties; the number of live births along with the number of deaths of pre-viable infants differ due to regional reporting differences . When these data are used as the basis for policy making, they could potentially result in the mistargeting of resources from states with the greatest need to states whose levels of need may be overinflated based on an artifact of the reporting standard. In addition, the definitions create challenges when developing and delivering targeted interventions because they obscure the line between those infants and mothers who would benefit from postnatal interventions to prevent infant mortality (e.g., those born at gestational ages > 22 weeks) and those who require prenatal interventions (e.g., those with peri-viable gestational ages or fetal deaths) to prevent miscarriages, fetal death, and extremely preterm birth .
While the Centers for Disease Control and Prevention (CDC) promulgates a national definition of live birth and fetal death, some states include clarifying statements of what constitutes a fetal death while others do not . This results in major differences in fetal death reporting requirements by state . For example, in New York, all fetal deaths -- regardless of gestational age -- are reported. In Ohio, all fetal deaths of gestation ≥ 20 weeks are reported, while in Kansas, fetal deaths of weight ≥ 350 grams are reported . As a result, the non-live birth of an infant weighing 300 grams at 19 weeks of gestation would be reportable in New York and not in Kansas or Ohio. In contrast, if that infant were born at 20 weeks of gestation, the death would only be non-reportable in Kansas based on the weight requirement. Infant mortality and fetal death rates may be distorted in states where rules create a documentation requirement that suggest a false dichotomy between the two criteria, or a documentation gap where a birth event neither meets the criteria of an infant live birth nor a fetal death, as in the latter Kansas case. To our knowledge, the only available collection of these state-level definitions and requirements is from 1997 . Updates to fetal death definitions and further specification of live birth definitions that are consistent across states are necessary to prevent documentation differences.
There are also differences between states in documentation of deaths from pre-viable births. While no infant born at 21 weeks and 5 days gestation or less has ever lived past infancy in the United States, numerous infants at these early, pre-viable, gestational ages are classified as live births based on the definition criteria. There is evidence that all U.S. states have publicly reported infant mortality data from pre-viable gestational ages; however, these rates differ between states due to the proportion of pre-viable live births reported . Further investigation regarding these statewide differences and the effects on infant mortality rates is necessary to understand why there is a need for new reporting standards.
This paper uses Ohio Vital Statistics birth records derived from an evaluation of the Ohio Infant Mortality Reduction Initiative (OIMRI) and national data from the Centers for Disease Control and Prevention to explore these challenges and suggest changes that would improve our ability to compare state-level data.
Ohio birth records and death records from Ohio Vital Statistics were used to examine deaths within Ohio. Records were linked using birth certificate numbers. Due to data availability and the scope of the original evaluation these analyses were completed for, only data for years 2008-2015 are used from these Ohio datasets. National data for all states and Washington DC from 2007 to 2017 were acquired from the CDC Wonder database . Variables used included number of live births, number of infant deaths, and gestational age at birth. Descriptive statistics including counts, percentages, birth rates, and infant mortality rates were calculated.
In Ohio birth record data, 554 (6.83%) of deaths from 2008-2015 were from infants under 20 weeks of gestation, and 1,461 (18.01%) of the infant deaths were from pre-viable live births (< 22 weeks), despite accounting for under 0.2% of all live birth records (Table 1). The infant mortality rate for all linked records in our Ohio dataset was 7.16 per 1000 births. However, when infants born at less than 22 weeks were excluded, the state's infant mortality rate became 5.87 per 1000 births.
Table 1: Contribution of gestational age to the infant mortality rate in Ohio, 2008-2015. View Table 1
In CDC Wonder data for 2007-2014 for all states and Washington DC, all U.S. states have publicly reported infant mortality data from pre-viable gestational ages (Table 2) . The percentage of all infant deaths from < 22 weeks of gestation ranges from 8.3% in West Virginia to 24.0% in Rhode Island (Ohio's percentage was 18.4%). While infants born at less than 22 weeks of gestational age accounted for 0.57/1000 live births in Alaska, the rate in Washington DC was more than three times as high at 2.05/1000 . (Ohio's rate was 1.59/1000). These differences seem to correspond with overall infant mortality rates in these states. Of the 10 states with the highest proportion of pre-viable live births, 6 were in the top 10 highest rates of infant mortality in this database. If the infant mortality rate from those with pre-viable births and deaths is subtracted from the infant mortality rate, the calculated infant mortality rate decreases for all states. For the states with the highest rate of pre-viable live births, this decrease is disproportionately large.
Table 2: Infant mortality rates and rankings by state, 2007-2017. View Table 2
If states were ranked based on these CDC WONDER data, the change in rank by removing pre-viable live births would change Ohio's ranking by 5 spots, from 44 to 39. Of those states with the 10 highest rates of pre-viable live births reported, 8 would have improvements in rank, while the 10 with the lowest rates of pre-viable live births would all decrease or stay the same in rank.
Ohio-specific data show that pre-viable births have a large impact of the infant mortality rate. Subsequent analysis of CDC Wonder data shows that the proportion of pre-viable live births and their effect on the infant mortality rate differs substantially among states. These findings are supported by the work of Goyal and DeFranco . Using the US National Center for Health Statistics data from > 2000 US counties, Goyal and colleagues found variation by US region in the proportion of early gestation births (17-20 weeks of gestation) categorized as fetal deaths versus live births with a subsequent death . They found that for every 1 point increase in the fetal death percentage, there was an associated 0.02 point decrease in a county's infant mortality rate; they concluded that the variability in reporting requirements makes county-to-county comparisons challenging. In another study, DeFranco and colleagues found racial disparity in the infant mortality rates of pre-viable infants in Ohio . Births at gestational ages between 16 and 22 weeks accounted for 45% of Non-Hispanic Black infant mortality and thus may explain part of the racial disparity in infant mortality in Ohio . These researchers also raised the concern that this phenomenon can be obscured when pre-viable infants are included in state rates instead of separately examining these rates.
When one considers the implications of infant mortality rates on strategic priorities within each state, between states, and internationally, it becomes increasingly important that data enable an apples-to-apples comparison [9,10]. When comparing the United States to countries in Europe, often births at pre-viable gestational ages are excluded and the disparity between the United States and other European countries decreases . We submit that the national surveillance definition of infant mortality should be changed to exclude pre-viable live births. We acknowledge that picking a specific gestational age cutoff is difficult as there is no clear cutoff that determines viability, and with improvements in neonatal intensive care, this is a moving target. However, for a surveillance definition to be accurate and allow for reasonable comparisons between states, it is vital that state-to-state variability in measurement and documentation be minimized. In addition, the reporting of fetal deaths should be standardized across states and adjusted to avoid a documentation gap. As state definitions have not been collected and documented since 1997, these definitions should be standardized, updated, and disseminated. We note that we are not proposing that doctors cease collecting pre-viable birth data, rather we submit that post-hoc analysis of data on infant mortality should be focused on those cases where medical technology exists to intervene.
Additionally, this approach would allow for the ongoing and expanded collection of pre- and peri-viable infant mortality data, which could offer important insights into contributing factors and the health of the community. More clearly differentiating the infant mortality rate, fetal death rate, and pre-viable infant birth rate will allow intervention programs to be properly tailored to both populations and regional challenges. Areas with high rates of pre-viable births and fetal deaths may have very different risk factors and educational needs than those with high infant mortality rates. Interventions in regions with rising infant mortality rates could focus on improving perinatal and postnatal care while in areas with rising rates of fetal death and pre-viable births, interventions intended to prolong pregnancy might need to be a stronger focus. As infant mortality rates are frequently compared among states, further standardization of infant mortality rate calculations would make such comparisons much more appropriate as policy-makers continue their efforts to reduce infant mortality in the United States.
The OIMRI Evaluation was funded by the Ohio Department of Health (ODH) and administered by the Ohio Colleges of Medicine Government Resource Center. The views expressed in this paper are solely those of the authors and do not represent the views of ODH. This study includes data provided by ODH which should not be considered an endorsement of this study or its conclusions.
All authors contributed to the design, analysis or interpretation, writing and editing of this work.