Juan Acuna Zarinah Ali Barbara Gladders Priscilla Guild Marian MacDorman Bill Sappenfield Tri Tran

1
Stage 1 State Assessment of
Vital Records Reporting
State Infant Mortality Toolkit
Data Reporting Subcommittee

• Juan Acuna
  Zarinah Ali
  Barbara Gladders
  Priscilla Guild Marian MacDorman
  Bill Sappenfield
  Tri Tran

2
Stage 1 State Assessment of
Vital Records Reporting
State Infant Mortality Toolkit

• Thought State infant mortality rates and trends
  may not be real and may be artificially
  impacted by vital records reporting.

3
Vital Record Information System
Public Health Surveillance System

• Assess/Monitor Data Reporting and Quality
• Reporting and processing issues
• In-state trends
• Multi-state comparisons

4
Overall Assessment
Change in Reporting or Processing

• Change in reporting regulation, process, or
  training?
• Change in quality processing?
• Change in health care system?

5
Reported Vital Events
Conception
1 Year
Live Birth
Gestation
Infancy
20 wks
4 wks
Infant Death
Fetal Death
Feto-Infant Death
6
Fetal Deaths In-State Trends

• Reporting
• Change in reporting regulation, process, or
  training?
• Clarification of viability or gestation?
• Change in quality processing?
• Change in abortion reporting?

7
Fetal Deaths In-State Trends

• Data Analysis
• Number and percentage of unknown birthweight and
  gestational age
• Percentages and mortality rates by birthweight
  and gestational age
• Percentage of all deaths 20-27 weeks gestational
  age

8
Five-year Fetal Mortality Rates by Birthweight
Delaware, 1989-2002
9
Graph Notes

• Initially, the lt500 gram fetal mortality rate was
  lower than the rate for 500-749.
• lt500 gram fetal mortality rates increased
  steadily from 1989-1993 through the 1995-1999
  time period, at which time it decreased slightly
  and held steady over the next few time periods.
• Though the fetal mortality rates for 500-749
  grams showed a similar trend, the increase was
  not as large.
• The 750-999 rates had a slightly decreasing trend
  over time.

10
Five-year Percent of Fetal Deaths lt28 Weeks
GestationDelaware, 1989-2002
11
Graph Notes

• Delawares reporting requirements for fetal death
  are 350 grams or more, or in the absence of
  weight, 20 wks or more gestation.
• Although the percent of fetal deaths lt28 weeks
  has increased from 1989-2002, it is below 73 as
  estimated in the NSFG, and varied widely by race.
  Although each group displays an increasing trend
  over time, the black percentage came close to 73
  before decreasing.

12
Percent of All Fetal Deaths by Gestation,
Louisiana, 1995-2002
13
Graph Notes

• This graph shows the percentage of all fetal
  deaths over time. The green line is lt20 weeks
  gestation. The pink line is 20 to 23 weeks
  gestation. The yellow line is 24 to 27 weeks
  gestation.
• As can be seen in this plot, the percentage 20 to
  23 weeks gestation shows a cyclical pattern and
  may be increasing over time.
• The other percentages by gestation are not
  changing substantially over time.
• State reporting requirements make the percentage
  lt20 weeks artificially low.

14
Fetal Deaths Multi-State Comparison

• Percentage of all deaths 20-27 weeks gestational
  age

15
Number of States by Percent of All Fetal Deaths
20 to 27 Weeks Gestation, 2000-02
16
Graph Notes

• This graph shows the percentage of all fetal
  deaths gt20 weeks that are 20 to 27 weeks
  gestation. The peak is at 45 to 50.
• This is a highly skewed distribution to lower
  percentages which is suggestive of some factor(s)
  affecting the distribution.
• Generally most fetal deaths in the US should be
  between 20 to 27 weeks if not earlier. The
  National Survey of Family Growth estimates the
  percentage at 73. This suggests underreporting
  of fetal deaths by all states.

17
Percentage of All Fetal Deaths 20-27 Weeks
Gestation
18
Graph Notes

• This graph compares the percentage for states for
  two periods 2000 and 2002. Light blue marks are
  for states that report all fetal deaths. Yellow
  marks are for states that have gestation and/or
  birthweight requirements. The lines represent
  the best linear relationship for the marks.
• As can be seen in this plot, those states with
  gestation and/or birthweight requirements have
  lower percentages.
• The r square value describes the linear
  relationship of the two percentages. Values
  approaching 1 or -1 are highly correlated. In
  this situation, as you would suspect, the
  percentage for 2000 is closely correlated with
  the percentage for 2002.

19
So What?

• Assumptions
• The real percentage of 20-27 wks is 73
• Your states percentage is 35 or 35 deaths of
  100 deaths
• (35 ?) / (100 ?) 73
• Answer 143 additional deaths

20
Live Births In-State Trends

• Reporting
• Change in reporting regulation, process, or
  training?
• Clarification of viability or gestation?
• Change in quality processing?

21
Live Births In-State Trends

• Data Analysis
• Percentages by birthweight and gestational age
  especially lt500 grams and lt24 weeks gestational
  age
• Changes in distribution
• Focus on very premature and very low birthweight
  births
• Overall and race categories

22
Five-year Percent of All Live Births by
Birthweight Delaware, 1989-2002
23
Graph Notes

• The percent of live births lt500 grams declined
  slightly in the mid-90s, but then increased from
  .23 to .30.
• The percent of births 500-749 grams has remained
  fairly stable.
• The percent of births 750-999 grams followed the
  same pattern as the lt500 gram births, though the
  increase was not as large.

24
Five-year Percent of All Live Births by
Gestation Delaware, 1989-2002
25
Graph Notes

• The 1998-2002 time period has the highest
  percentages of births lt22 wks, 23-24 wks, and
  26-28 wks gestation.
• The largest increase was at lt22 wks gestation.
• The percent at 25 wks showed a decrease.
• Because almost all babies born at lt22 wks
  gestation die, small changes in the percent of
  these births can have a disproportionately large
  effect on the infant mortality rate.
• Like the birthweight graph, there is a noticeable
  dip in the percentages during the 1994-1998 time
  period.

26
Percent of All Live Births by BirthweightLouisian
a, 1995-2002
27
Graph Notes

• This graph the percentage of live births by
  birthweight over time.
• There is a clear spike in the percentage of 500
  to 749 g live births in 2000 that we have not
  been able to explain. For the purpose of trend
  analysis, lets ignore the spike.
• Generally the percentage of births increased in
  almost all categories until around 2000 with
  almost all categories showing a decrease in 2002.
• The general increase in the percentage of lower
  birthweight adversely affected the states infant
  mortality rate over this time period.

28
Live Births Multi-State Comparison

• Percentage of lt1,500 gram live births that are
  lt500 and 500-749 grams
• Percentage of lt32 week live births that are lt24
  and 24-27 weeks gestational age

29
Number of States by Percent of All Live Birth
lt1,500 Grams, 2000-02
30
Graph Notes

• This graph shows the number of states by
  percentages of birthweight of all births lt1500
  grams. This limitation adjusts for differences
  between states in the percentage of births lt1500
  g.
• Generally, states have a smaller percentage of
  births lt500 g than 500-750 g. Both distribution
  have somewhat a normal distribution.

31
Percentage of All Live Births lt1,500 Grams,
2000-02
32
Graph Notes

• This graph compares the percentage for states for
  two birthweight categories lt500 g and 500-749 g.
  Each mark represents both categories. The line
  represents the best linear relationship for the
  marks.
• The line is very flat and the R squared value is
  close to zero showing that these values are not
  correlated. In fact, states with a larger
  percentage of 500-749 g have a large percentage
  of black births. This is not true for lt500 g.
• These percentages should be correlated suggesting
  that some factor is affecting the percentage for
  lt500 g live births underreporting.

33
Number of States by Percent of All Live Birth lt32
Weeks, 2000-02
34
Graph Notes

• The patterns are similar between birthweight and
  gestational age.
• This graph shows the number of states by the
  percentage for all births lt32 weeks. This
  limitation adjusts for differences in the
  percentage of births lt32 weeks.
• Generally, states have a smaller percentage of
  births lt24 weeks than 24-27 weeks. Both
  distribution have somewhat a normal distribution.
  

35
Percentage of All Live Births lt32 Weeks, 2000-02
36
Graph Notes

• The patterns are similar between birthweight and
  gestational age.
• This graph compares the percentage for states for
  two gestation categories lt24 wks and 24-27 wks.
  Each mark represents both categories. The line
  represents the best linear relationship for the
  marks.
• The line is very flat and the R squared value is
  close to zero showing that these values are not
  correlated. In fact, states with a larger
  percentage of 24 to 27 wks have a large
  percentage of black births. This is not true for
  lt24 wks.
• These values should be correlated suggesting that
  some factor is affecting the percentage for lt24
  wk live births underreporting.

37
So What?

• Assumptions
• The real percentage of lt500 g live births is
  0.30
• Your states percentage is 0.20
• The real mortality rate is 900/1000
• (.30-.20) .900 .09
• Answer 1 per 1,000 IMR increase

38
Infant Deaths In-State Trends

• Reporting
• Change in reporting regulation, process, or
  training?
• Change in the linkage of infant deaths to live
  births?
• Change in quality processing?
• Change in follow up of lt750 gram live births with
  delivery hospitals?

39
Infant Deaths In-State Trends

• Data Analysis
• Number and percentage of unlinked death records
• Number and percentage of unknown birthweight and
  gestational age
• Percentages and mortality rates by birthweight
  and gestational age
• Percentage and mortality rates by state of death

40
Infant Deaths In-State Trends

• Data Analysis
• Mortality rates of infants who die soon after
  birth
• Feto-infant deaths and mortality rates
• Compare fetal, infant, and feto-infant mortality
  rates

41
Five-year Infant Mortality Rates by
BirthweightDelaware, 1989-2002
42
Graph Notes

• From 1989-1993 the mortality rate of infants born
  lt500 grams decreased until 1993-1997. This low
  point may reflect a problem with the data. The
  rates then began to increase and continued
  through 1998-2002.
• Of note, the rates were lowest at the same time
  DEs overall IMRs were at their lowest.
• Rates for the 500-749 gram infants followed a
  similar decreasing and then increasing pattern.
• The 750-1499 group initially decreased and then
  held steady.

43
Five-year Mortality Rates 24 to 27 Weeks
Gestation Delaware, 1989-2002
44
Graph Notes

• The infant mortality rates decrease steadily from
  89-93 to 93-97 and then increase.
• Fetal mortality rates shows the opposite trend.
  The rates increase from 89-93 to 93-97, and then
  decreasing through 1996-2000 where they level
  off.
• Because of fetal mortality rate trend, the trend
  for the combined feto-infant mortality rates is
  similar to the infant mortality trend but
  dampened.
• The circle highlights the time periods where
  decreasing fetal mortality rates corresponds to
  increasing infant mortality rates. This can
  signal a shift in reporting from fetal deaths to
  live births.
• However, when the feto-infant mortality rate is
  limited to infant deaths lt2 days, the rates do
  not increase in later years suggesting it is not
  reporting but an increase in older infant deaths.
  

45
Infant Mortality Rates by Birthweight Louisiana
1995-2002
46
Graph Notes

• This graph shows the trends in infant mortality
  rates by birthweight.
• The rates for lt500 g and 500 to 749 g births
  decreased over time and dramatically decreased in
  2000, then the rates almost increase to initial
  levels.
• Other rates did not change substantially over
  time.

47
Percent Linked Infant DeathsLouisiana
48
Graph Notes

• This graph shows the percentage of infant death
  certificates linked to birth certificates over
  time.
• The percentage in LA has increased over time
  slightly but is below national levels of 98 to 99
  percent.
• These lower percentages of linked certificates
  artificially lowers the states mortality rates
  in linked analysis exactly by the percentage of
  unlinked certificates.

49
Infant Mortality Rates for lt500g Live
BirthsLouisiana US
50
Graph Notes

• This graph shows the infant mortality rates for
  lt500 g live births for LA and US.
• During the early 1990s the states rates are
  close to US rates. LA rates decrease over time
  and substantially dropped in 2000. The rate then
  returned to almost national levels.
• The states rates dropped to levels that are
  unbelievably low and suggests substantial
  underreporting. This explains much of the
  decrease in the states overall infant mortality
  rate.

51
Reported IMR and Adjusted IMR (US lt500),
Louisiana, 1995-2002
52
Graph Notes

• This graph shows LAs infant mortality rate over
  time. The reported rate is the traditional
  infant mortality rate reported by the state. The
  adjusted rate adjusts the states rate assuming
  the infant mortality rate for lt500 g live births
  were the same as the US average rate.
• Although the states adjusted rate for decreased
  over time and then increased, the amount of
  change was substantially less. The dramatic drop
  in 2000 also did not occur. This suggests that
  much of the states trend in infant mortality
  rate is related to lt500 gram mortality.

53
Infant Deaths Multi-State Comparison

• Infant mortality rates of lt500 and 500-749 gram
  live births
• Infant mortality rates of lt22, 22-23 and 24-27
  weeks gestational age

54
Number of States by Percent of Out-of State
Infant Deaths, 2001-02
55
Graph Notes

• This graph shows the number of states by the
  percentage of out-of-state infant deaths.
• The number of states by percentage of
  out-of-state infant deaths is highly skewed with
  most states having a small percentage of out-of
  state deaths (except for lt1).
• 5 states have between 10-20 of births
  out-of-state and 1 state has more than 20.

56
Number of States by the Percentage of Linked
Infant Deaths, 2001-02
57
Graph Notes

• This graph shows the number of states by the
  percentage of linked infant deaths for two
  categories of events same state versus
  out-of-state. And, the percentage of linked
  infant deaths is higher among those born in-state
  versus those born out-of-state. This can be seen
  by the predominantly yellow bars lt98.
• This means that out-of-state events can be harder
  to link which means these deaths could be
  excluded from linked birth-death analyses.

58
Out-of-State Infant Deaths, 2001-02
59
Graph Notes

• This graph compares the percentage for states for
  two categories percentage of out-of-state death
  and percentage unlinked for these deaths. Each
  mark represents both categories. The line
  represents the best linear relationship for the
  marks.
• The line again is very flat and the R squared
  value is close to zero showing that these values
  are not correlated. In fact, states with a large
  percentage of out-of-state events do not
  necessarily have a high unlinked percentages.

60
So What?

• Assumptions
• The percentage of out-of-state infant deaths is
  10
• The percentage of unlinked infant deaths is 50
• .10 .50 .05
• Answer 5 lower overall IMR

61
Number of States by Infant Mortality Rates,
2000-02
62
Graph Notes

• This graph shows the number of states by IMR for
  lt500 g and 500-749 g live births.
• The IMRs are somewhat skewed to lower mortality
  rates suggesting some factor(s) are involved.
• When you compare the mortality rates for
  published rates for 500 to 749, many states
  experience lower rates. This is suggestive for
  underreporting. Although we dont have a
  standard for lt500 live births, many of these
  rates are unbelievably low.

63
Infant Mortality Rates, 2000-02
64
Graph Notes

• This graph compares the IMRs for states for two
  birthweight categories lt500 g and 500-749 g.
  Each mark represents both categories. The line
  represents the best linear relationship for the
  marks.
• The line is angled up and the R squared value is
  .375 showing some correlation. These values
  should be highly correlated. This is again
  suggestive that the lt500 gram mortality rates are
  likely underreported.

65
Number of States by Infant Mortality Rates,
2000-02
66
Graph Notes

• The patterns are similar between birthweight and
  gestational age.
• This graph shows the number of states by IMR for
  3 gestational age categories.
• The IMRs are somewhat skewed in opposite
  directions depending on the gestational ages
  suggesting some factor(s) are involved.
• When you compare the mortality rates for
  published rates, many states experience lower
  rates. This is suggestive of both underreporting
  and likely misclassification of gestational age.
  Although we dont have a standard for lt500 live
  births, many of these rates are unbelievably low.

67
Infant Mortality Rates, 2000-02
68
Graph Notes

• The patterns are similar between birthweight and
  gestational age.
• This graph compares the IMRs for states for two
  birthweight categories lt22 wks and 22-23 wks.
  Each mark represents both categories. The line
  represents the best linear relationship for the
  marks.
• The line is angled up and the R squared value is
  .41 showing some correlation. These values
  should be highly correlated. This is again
  suggestive that the lt22 week mortality rates are
  likely underreported.

69
So What?

• Assumptions
• The real IMR of lt500 g is 900/1000
• Your states IMR is 700/1000
• The states percentage of lt500 g live births is
  0.30
• (.900-.700) .30 .06
• Answer 0.6 per 1000 IMR increase

70
Potential Reporting Problems by
Birthweight Below
1.0 Standard Deviation
Fetal Deaths Fetal/Infant Deaths
Infant Deaths Infant Deaths/Births
Live Births All Three
71
Map Notes

• The map reports those states that report values 1
  standard deviation below the mean for states for
  fetal deaths, live births and/or infant deaths.
• No clear geographic pattern emerges and the
  problems do appear to correlated. In fact, the
  reporting of one measure is not linearly
  correlated to the reporting of the other
  suggesting that these issues are not correlated.

72
Suggested References

• Fetal Death Reporting
  Martin and Hoyert
• Reporting lt500 Gram Live Births Wilson,
  Fenton, Munson
• Importance of Looking at Very Low
  Birthweight-Specific Mortality MacDorman,
  Martin, Mathews, et al.
• Gestational Age Reporting Alexander,
  Tompkins, Cornely
• Reliability of Birth Certificate Data
  DiGiuseppe, Aron, Ranbom, et al.