Low Birth Weight, Vital Records, and Infant Mortality

(1)

COMMENTARIES

1143

black populations as valid reasons for a renewed

push to reduce infant mortality. No group has a

greater reason for concern and action than the

American Academy of Pediatrics. We would like

for every Academy Fellow to make himself or

her-self knowledgeable on this subject and become an effective advocate for further change on the local

scene.

REFERENCES

MARTIN H. SMITH, MD, FAAP

President

American Academy of Pediatrics

1. Wegman ME: Annual summary of vital statistics-1985.

Pediatrics 1986;78:983-994

2. American Academy of Pediatrics, Task Force on Infant Mortality: Statement on infant mortality. Pediatrics 1986; 78:1155-1160

3. Preventing Low Birth Weight, report of a study by a com-mittee of the Institute of Medicine. Washington, DC, Na-tional Academy Press, February 1985

4. Southern Regional Task Force on Infant Mortality: Final Report-1985. Washington, DC, Southern Governors

Asso-ciation, 1985

Low

Birth Weight,

Vital

Records,

and Infant

Mortality

About one tenth of all infant deaths occur in

babies weighing less than 500 g at birth, almost all

of whom die very shortly thereafter. In 1983, when

the United States reported 3,638,933 live births, 4,368 of them were less than 500 g; that year there

were 26,507 neonatal deaths. This means that

slightly more than 0.1% of all live births

contrib-uted to 17% of neonatal mortality. Given this order

of magnitude, any change in the numbers relating

to these tiny babies can have a disproportionate

influence on reported infant mortality and on

in-terstate comparisons. Two questions promptly arise. How accurate and meaningful are the data

regarding babies born weighing less than 500 g?

What can be done to decrease the deaths in this

category?

Wilson et a!’ call attention to how the number of

very low birth weight infants reported by a state

may be affected by the state’s definition of a live

birth. The World Health Organization (WHO)

def-inition, which the US Public Health Service

rec-ommends that all states follow, specifies, in essence,

that any product of gestation which, after complete

separation from the mother, shows any one of four signs of life (breathing, heart beat, voluntary

mus-cle movement, or umbilical cord pulsation) is to be

reported as a live birth. Within our federal system,

however, each state is legally empowered to adopt

its own precise definition and regulations. Not all

are in accord with the Public Health Service

rec-ommendation.

In fact, as Wilson et al’ point out, only 33 states

have adopted the WHO definition in all its details.

Many state regulations differ from the specific

wording of the WHO definition in only a minor

way, but Ohio has departed significantly. Ohio

re-jects the definition of a live birth as “any product

of gestation” and establishes a minimum of 20

weeks of gestation as a criterion for a live birth. The definitional question is further complicated

by changes that may take place in state law. In

Michigan, for example, a new Public Health Code

adopted in 1978 led to a new State Health

Depart-ment regulation on “live birth” promulgated in

1981. In this revision, many of the details of the

WHO definition that had been previously included

were omitted. There is now neither reference to

duration of gestation nor specification of signs of

life. Nevertheless, authorities in Michigan believe

there have been no real changes in reporting

prac-tice since the new regulation took effect.

Nationwide data for 1983 analyzed by Wilson

and her colleagues show that there was indeed

substantial variation among the 50 states, both in

the rates of birth of infants weighing less than 500

g and in the proportion these babies constituted of

the state’s total neonatal deaths. On the other hand, the variation seemed to be essentially independent

of the legal definition of a live birth. The Ohio

rates, for example, indicate that approximately 195 infants were born at less than 500 g, an unlikely

figure under a 20-week gestation rule. There were

other inconsistencies. Because the proportion of

low birth weight infants among the black

popula-tion is regularly twice that among whites, one would

expect states with a large black population to have

a higher rate of babies with birth weight of less

than 500 g. Mississippi, however, is reported to

have had a 1983 rate of 1.2 births of babies less

than 500 g per 1,000 total births, whereas Vermont

had a rate of 1.6. Yet, in Mississippi 48% of all 1983

births were black as against 0.3% in Vermont.

All this lends support to the inference that what

is likely to be more important than differences in

at Viet Nam:AAP Sponsored on September 7, 2020

www.aappublications.org/news

(2)

1 144 PEDIATRICS Vol.

78 No. 6 December

1986

regulations per se are differences in local practice, either in applying the legal definition or in putting

the correct information on the birth certificate. For

example, rates might be affected considerably if,

despite a state’s adoption of the precise WHO

def-inition, all babies who did not breathe were

rou-tinely considered stillborn because no great

diii-gence was exercised in searching for a heart beat or

movement of voluntary muscle. Such a practice

would, of course, compound the problem of

inter-state or intercity comparison. As I have pointed out

repeatedly in the “Annual Summary of Vital

Sta-tistics”2 rates in some states and cities are affected

seriously because a border is crossed to obtain

med-ical care.

Another aspect of the accuracy problem is the

question of how the weight is recorded on the birth

certificate; an example might be reading 5 lb as 500

g. Some light is thrown on the situation when birth

and death records are routinely matched, because

birth certificates provide information on the

char-acteristics of baby and parents, not ordinarily

pres-ent on the death certificate. Michigan, like a

num-ber of other states, searches for a birth certificate

when a death is reported in the first year. In

addi-tion, whenever a birth is reported of a baby

weigh-ing less than 1,000 g, an investigation is undertaken

if a death certificate does not turn up in the

ordi-nary course of events. Michigan experience suggests

that as many as 5% to 7% of infants whose birth

weight is reported at less than 1,000 g actually have

higher weights. An even larger proportion of these

errors may occur within the less than 500-g group.

Thus, without careful follow-up, the number of

infants weighing less than 500 g may be overstated,

making it seem that a higher proportion of births

are in this weight group.

A constant difficulty in interpreting infant

mor-tality data is underreporting of births, deaths, or

both. In a recent editorial in the American Journal

of Public Health, Kleinman3 reviewed

underreport-ing from a historical perspective. For many years,

underreporting of births in the United States was

so prevalent that statistical adjustment was

re-quired to approximate the true situation more

closely. This practice was discontinued in 1959, but

David4 suggests that the problem has not been

entirely resolved. He noted that in seven southern

states the proportion of nonwhite births in

hospi-tals more than doubled between 1950 and 1967. One

might think that this change, in addition to being

conductive to greater availability of medical care,

would have led to more reporting of births (the

denominator of the fraction “infant deaths/live

births”) and, consequently, a lower infant mortality

rate. Yet, the infant mortality rate in nonwhite

infants declined only slightly during the period. One

possible explanation is that, because low birth weight infants born outside the hospital often die

quickly, and may be buried without an undertaker,

they may not have been reported as births or deaths. David’s suggestion that the move into hos-pitals led to increases in both the numerator and

denominator of the infant mortality fraction is

supported by the observation that, although 8.37%

of nonwhite infants in the seven states were

re-ported with a birth weight of less than 2,500 g in

1950, the proportion increased to 13.34% in 1967,

an increase of 62%.

Kieinman3 believes that underreporting of

out-of-hospital births and deaths may also be a serious

problem in groups like undocumented aliens. He

conjectures that if even 5% of births to Latino

mothers in California were not reported and 25%

of them died-not an unreasonable

assumption-the rate of fetal and infant deaths in Mexican-born

mothers would have doubled. Furthermore,

un-derreporting is not limited to out-of-hospital births,

as shown by recent studies in Kentucky and

Geor-gia.5’6

These observations, as well as the paper by

Wil-son and collaborators, reinforce the need to take

into account differences in population composition

and in accuracy of vital records when comparing

different states or countries. If there is systematic

underreporting or inaccuracy of data for any

seg-ment of the population and that segment has, in

fact, a higher death rate, the total figures will be

distorted.

One way to derive useful information, rather than

comparing two states for a given year, is to study

trends over a number of years.7’8 If it is assumed

that there has been no major change in procedures

or practice, trend analysis may lead to valid

conclu-sions for one state as well as in comparison with

others. By analyzing trends over the period 1968 to

1983, Kleinman7 has identified nine states that had

infant mortality trends less favorable than the

na-tional average.

Finally, the other issue-what can be done to

decrease mortality among low birth weight

in-fants-is thornier. Identifying the size of the

prob-lem through the use of vital records is a beginning.

At present, treatment of extremely low birth weight

infants is long drawn out, very costly to limited

resources, and, in the case of infants weighing less

than 500 g at birth, uniformly unsuccessful. Much

the more promising avenue is prevention. Yet, the

proportion of pregnant women in the United States

benefiting from so simple and proven a technique

as routine prenatal care actually decreased for both

white and black mothers between 1981 and 1982.’

(3)

COMMENTARIES 1145

Prevention is much less glamorous than use of

modern neonatal technology, involves a

socioeco-nomic as well as medial approach, and is by no

means cost free. The long-term results, however,

are well worth the investment. Attention certainly

must be paid the improving accuracy and

compar-ability of vital statistics, but we need even more to

attack the underlying reasons for deaths that need

not occur. In most of the other countries of the

world that have infant mortality rates lower than

that of the United States, the proportion of babies

born at low birth weight is distinctly lower than in

the United States. This success should stimulate us

to greater action.

REFERENCES

1.

MYRON E. WEGMAN, MD

University of Michigan

School of Public Health

Ann Arbor

Wilson AL, Fenton LI, Munson DP: State reporting of live births of newborns weighing less than 500 grams: Impact on neonatal mortality rates. Pediatrics 1986;78:850-854

2. Wegman ME: Annual summary of vital statistics, 1984.

Pediatrics 1985;76:861-871

3. Kleinman JC: Underreporting of infant deaths: Then and now. Am J Publw Health 1986;76:365-366

4. David RI: Did low birth weight among U.S. blacks really increase? Am J Public Health 1986;76:380-384

5. McCarthy BJ, Terry J, Rochat RW, et al: The underregis-tration of neonatal deaths: Georgia 1974-1977. Am J Public Health 1980;70:977-982

6. Centers for Disease Control: Birthweight-Specific neonatal mortality rates-Kentucky. MMWR 1985;34:487-488 7. Kleinman JC: State trends in infant mortality, 1968-1983.

Am J Public Health 1986;76:681-687

8. Zemach R: Comments on ‘state trends in infant mortality.’

Am J Public Health 1986;76:688

Accumulating

Evidence:

Using

Meta-Analysis

to Carry

Out

Research

Reviews

in

Pediatrics

In pediatric research, as in most fields, new

find-ings generate the most excitement. Summarizing

old ones has traditionally been a task assigned to

research assistants. But as research results

accu-mulate, should we make stronger efforts to see what

a body of research, taken overall, tells us? And

would systematic methods to do this improve on

the ad hoc or idiosyncratic efforts of the past?

In this spirit, several books and articles have

recently appeared describing procedures called

meta-analysis.’9 They urge both researchers and

policymakers to improve the quality of research

reviews by making them more systematic, and they

outline simple but concrete techniques for doing

this. The authors of these books and papers express dismay at the poor quality of most research reviews

and offer suggestions for strengthening them. Their

underlying theme is that planning a review brings

up certain decisions, and how they are made will

ultimately drive a review. What are these key

de-cisions?

WHAT QUESTION IS A REVIEW ORGANIZED TO ANSWER?

In my experience, there are at least three: (1) For

a treatment or therapy or drug, what is its effect on

average? (2) Where and with whom is a treatment

particularly effective? (3) Will it work here? What

are practical guidelines for implementing a

treat-ment in a particular place?

An example of how a research review emphasizes

different answers depending upon the question is

recent work by Hauser-Cram and Shonkoff.’#{176} They

summarize results of 31 studies of the value of early

intervention for disabled infants and their families.

All of them focused on children less than 3 years of

age, who were raised at home or in a foster home

(not a residential facility) and who had a physically

identifiable handicapping condition. Among these

studies, the value of early intervention “on the

average” is more than .5 SD. A highly positive

result.

Does the review offer any insights about how to

organize such interventions better in the future?

Can we learn from past experience? Hauser-Cram

and Shonkoff’#{176} examine the question of where and

with whom these programs work especially well,

and the results are interesting. Interventions

de-signed for children categorized as “developmentally

delayed” offer significantly greater benefits than

those designed for mentally retarded children. The

smallest positive impact comes from programs for

children with orthopedic handicaps.

Finally, the reviewers ask whether an early

inter-vention program works well in a specific place or

at a certain time-what do the studies teach us?

The answer is, quite a lot. For example, they find

an interaction between severity of disability and

(4)

1986;78;1143

Pediatrics

MYRON E. WEGMAN