Cs Certificate of Live Birth , Philippines

							Child Mortality
       HSERV/EPI 539

      Stephen Gloyd, MD, MPH
 Professor, Dept of Health Services
Director, International Health Program
Child Mortality: Outline
Background
Approaches to Estimating Mortality
 –   Direct Methods
 –   Indirect Methods
Verbal Autopsy
                Background
Why do we need mortality data?
  a.       Broad indicator of social development –
           especially trends over time
  b.       Targeting interventions
  c.       Measuring impact of interventions
       –     (e.g., mortality targets in Child Survival programs such
             as measles immunization assessment)
  d.       Important to governments (especially MOH) for
           planning, advocacy

  Remember the complex etiology of overall mortality
 Common indicators of
 reproductive mortality
IMR – infant mortality rate
CMR – childhood mortality rate
U5MR – under five mortality rate
PMR - Perinatal mortality rate
MMR – Maternal Mortality Ratio (not
   today)
5 yrs
Mortality




            Birth
                    Childhood
4456 + 1476 = 5932   U5MR ~ 5932/55,000 =
                     108 deaths/1000 live births
           x   x x
                     x




Zimbabwe                 X = WO, UNICEF
          Estimating Mortality
          using direct methods
Vital Records
     a.    slow
     b.    incomplete, often not accurate
     c.    occasionally vital registration works (e.g., India)

Note: Vital records data can be used (and adjusted)
     a.    If fraction of omitted deaths is constant (Egypt, Mexico)
     b.    If numerator & denominator equally underestimated
     c.    Usually, pop with highest IMR omitted
          Direct Methods (2)
Sentinel sites
    - demanding, not necessarily representative

Note: ALWAYS CHECK OTHER SOURCES FIRST!
  a. To see if you need to do another survey
  b. Even if you do a survey, will need to compare
       results from other sources
  c. Results of a Mortality surveys should not stand on
       own
              Direct Methods (3)
Sample Surveys
  a.       Births / Deaths in past yr (or past 2, 5 years)
  b.       Serious underestimate of prevailing MRs
  c.       Not from poor fieldwork...more inherent in method
            Omission of deaths (Early deaths, stillbirths often not too
             significant to mother or midwife)
            Poor understanding of question
            Reference period often shortened/deaths placed outside
            Biases due to differential rates of omission
            More child than adult deaths omitted
  d.       Studies in West Africa
            picked up 2/3 of adult deaths, 1/2 of under-5 deaths
    Sample size issues
If you do direct estimates of child mortality:
What is an adequate sample size with (for example)
•    IMR of 100/1000 live births
•    100,000 pop
•    4500 births/yr, 450 deaths/yr
Remember -
      Bias in representation of population
      Seasonal variation
     Multiround sample surveys
Baseline survey follow up rounds at 6, 12 months
•     Women asked about pregnancy status each round
•     Probe to account for HH structure not accounted by births deaths
Usefulness
•     Can provide good estimates, but not secular trends
•     Can measure SES, health service correlates
•     Can use verbal autopsy methods
•     (Example, Cebu (Philippines), Hunduras
    Surveillance Systems
•    Done with small populations, requiring detailed
     recording of each vital event
•    Good for field trials, interventions
•    Very expensive
•    Less and less representative
•    Can use mortuaries, etc
        Indirect Methods

1.   Chidren Ever-Born /Children Surviving
     (CEB/CS) - "Brass method”
2.   Birth history method using life table analysis
3.   Previous birth technique
         Chidren Ever-Born/Children Surviving
                     (CEB/CS) –
                   "Brass method”
Information needed:
a.       Total ever-born
b.       Total ever-died
c.       Age of mother
          Good for mortality trends (less useful for recent trends) usually
           2.5-5 years before survey
          Usually converted into life tables (survival to age 1-20)
           CEB/CS - continued
b. If you obtain info from:
         Older women – get mortality from more distant past
         Younger women - get more recent mortality
      o     In general, method not good for very recent estimates
      o     Tends to select more older women, not representative of
            younger (poorer) population subset
      o     MR sometimes is in error if age dist of women gives different
            age pattern of mortality
      o     Life table estimates based on models (assumptions) of
            consistent patterns over time
                 Birth History Method –
                  Life Table Analysis

a.       Information needed:
           Live births, DOB, Sex
           Numbers surviving, dead DOD (5-15yrs)
b.       Demands more time, good knowledge of dates & ages
c.       Experienced interviewers, longer survey (20-30min)
d.       Fewer interviews needed (3-4 births/woman)
e.       Complex analysis using life tables
f.       5-year recall estimate relates to avg 2.5 yrs
g.       Results similar to PBT
h.       SBH - Short Birth History (2-3 births)
With CEB/CS or Birth History

Can use a HH Survey Approach
Be careful using standard EPI surveys:
     Sampling not suitable
     Families with kids 12-23 months; small, biased
      sample
     Need sample of women/births, and deceased
      children
With CEB/CS or Birth History
Use of methods with Modified EPI survey
          EPI, Diarrheal M&M, and CM
          Increase sample of HH
          Interview all women of reproductive age
          Return to HH when eligible woman not there
       o      need adequate training of field workers
       o      need strong survey team, with leader who has training & experience
       o      ensure protocols strictly followed
Standard: One eligible woman/HH
          3000-4000 HH enough to provide reasonably precise estimates
          Cost of CDD/MMT survey about $20,000
          Extra cost of Mort - preparation and sample size
          Avoid free standing survey
x   x



        x
    X = WO, UNICEF
X



      X
x
    x



        x
  Previous Birth Technique (PBT)
                          Brass and Macrae
Information needed:
            Numbers surviving, dead DOD (5-15yrs)
            Doesn’t require memory of dates of births/deaths
            Assume high mortality/fertility, 30 month birth interval
            Approximates under-2 MR
         o     Note: Second-to-last child approximates under-5 mortality
            Mortality estimates centered on period 12-18 months prior
             to survey
            Best as index to follow trends vs comparing to life table,
             IMR, U5MR - absolute level may be misleading
            Few simple questions at time of delivery
              PBT - continued

In Maternities:
            Most simple (timing, resources, logistics)
            Coverage bias- estimates refer to hospital pop
         o     Note: Mali (‘85) home-birth 1st month MR 3.4x hosp-birth MR &
               2x clinic-birth MR
            Need stable (and reasonably high) utilization of hosp
             birth
               PBT - continued
In HH Surveys:
           Population-based estimate
           Timing problem
           Resources required for large sample size
           Approaches:
        o     Include all women of reproductive age (not just with kids)
        o     Standard-full DOBs/DODs history
        o     Short cut- Date of last LB and whether alive
        o     Survival of Previous Born Child
           If added to EPI survey
        o     Same rules re: Increasing sample size
        o     Including all women
             PBT - continued
In Well-Child Clinics
        Coverage issue (as above)
        Known that previous births of kids still alive have lower
         MRs than mothers whose youngest child has died
        Likelihood of younger child dead when predecessor is
         also dead is 2x as high as when child is alive (one can
         adjust rates using this)
        Better with younger kids, born < 2 yrs ago

*SIMPLE MESSAGE:
    - Most obvious method is not always the best!
                  Verbal Autopsy
Cause Specific Mortality
   Difficult to diagnose in wealthy countries
     MDs differ in diagnostic practices
     Significant discrepancies between clinical & autopsy dx
     Underlying vs immediate causes

   Measuring Cause-specific MR worse in DCs
     no MD/health worker contact at death
     no death certificate - poor vital registration
     single cause difficult with multiple causation of death
Verbal Autopsy - continued
   Used for years (at least 60s & 70s)
   Assumption:
     certain diseases have characteristic signs/symptoms which can be
           recognized and recalled by a lay observer

   These characteristics are sufficiently distinctive to
    differentiate diseases of interest
   Try to replicate conventional medical history but
    lack supporting PE, special exam
 Verbal Autopsy - continued
 Certain diseases are amenable to VA:
     NN: Tetanus
     Child: Measles, Diarrhea, ALRI, Injury
     more difficult: malaria, meningitis, pertussis, tb,
      AIDS
     Those diseases with relatively distinct
      syndromes may work
       VA Data Collection
Reporting
      every 2 weeks
      annual
      adjunct to vital records
Interviewer
      same ethnic group
      education
      sex (esp in reproductive-related issues)
      interobserver variation not known
More VA Data Collection
Death-interview interval
     weeks to months best (some purposely wait)
     9-12 months at latest
Respondent
     mother best (age is an issue)
Questionnaire
     screening followed by algorithm
     some use unstructured interview
     timing of symptoms helpful
     age-specific questionnaire
     local terms useful
     open vs. closed ended
More VA Data Collection
 Supplemental Information
      health record card
      make sure it is right one
 Diagnosis and Coding
      WHO CODES
      MD or epidemiologist or interviewer
      more than one MD
 One vs. Multiple Causes
      infections usually over-reported, nutrition under-
       reported
      post-measles deaths (rule is 6 weeks post)
      short vs. long list of diseases
     VA Specific Diseases
a.       Neonatal Tetanus
          3-30 days old
          poor cry/suck
          stiffness and convulsions
          confused with NN hemorrhage, etc
b.       Diarrhea
          definition of Diarrhea acute vs chronic
          def of dehydration
          more than one sign (4 good)
c.       Measles
          rash/fever
          duration > 3 days
          child > 5 mo/old
More VA Specific Diseases
 d. ALRI
      difficult
      age variation of sx
      simplified clinical dx (resp rate) not helpful
      cough > 4 days
      SOB > 1 day
 e. Injuries
      often overlooked
      external cause/nature of injury
      likely to be most accurate
Not Satisfactory for VA
 a. Malaria
      difficult since non-specific
 b. Pertussis
      in outbreaks ok
 c. Malnutrition
      KW may be dx by mother, not MAR
       AIDS
      chronic diarrhea, wt loss
      confused with malnutrition, tb
       Validation of VA
Gambia 76% (hospital-home dx immediate)
– 88% one month after death
– 96% with 3 MDs independently coding
Philippines
– 43% multiple illnesses
– Tetanus 100% correct
– Measles 98% sensitivity 90% specificity
– ALRI 86% sensitivity low Specificity
End