Table 5 GAPPS quality assessment of epidemiological parameters in global estimates using adapted version of GRADE: STILLBIRTH

Defi nition

For international comparison stillbirth rates refer only to late fetal deaths (>1000g or >28 weeks gest). ICD-10 defines a fetal death as «death prior to the complete expulsion or extraction from its mother of a product of conception, irrespective of the duration of pregnancy; the death is indicated by the fact that after such separation the fetus does not breathe or show any other evidence of life, such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles» without specification of the duration of pregnancy. The denominator is all live births plus late fetal deaths.

Stillbirths in the last trimester (late fetal deaths) occurring during the time of labour, but excluding major congenital abnormalities. In verbal autopsy data «fresh stillbirth» is used as a surrogate marker for intrapartum stillbirth. The denominator is all live births plus late fetal deaths.

Multiple classification systems are in use. A comparable system to map the results of verbal autopsy data on cause-of-death with more complex classification systems is urgently needed.

Systematic global estimates available? Source and date

Stanton, Lawn et al, Lancet 2006

WHO MPS, 2006

GAPPS, GBD, CHERG, WHO (Stanton, Lawn et al in process)

Lawn, Shibuya, Stein WHO Bull, 2005

GAPPS, GBD, CHERG, WHO (Stanton, Lawn et al in process)

   N of countries with VR data used (note if used as reported, new analysis or adjusted)

44 countries with adult VR coverage > 90%; rates adjusted by model coefficient for VR to allow for under reporting

102 countries with SBR data used as reported;

33 VR (plus 11 from EuroPeristat, some of which are VR-based)

Not in Vital registration

Not in vital registration

   N of countries with survey-based estimate used (note if used as is, new analysis or adjusted)

1 country, adjusted by model coefficient

102 countries with SBR data used as reported;

In process -

Not in current national surveys

Only 1 national surveys with SB COD data

   N of countries where model-based estimate is used (basis of model)

Model-based estimates are used for 128 countries; based on a random eff ects model

88 countries - based on average SBR:ENMR ratio for that region

(in process)

52 countries with data, 141 countries with estimate based unadjusted reported data by country or if no country data on median for WHO subregion (14 subregions)

(in process)

   Types of data inputs used in the model

VR, published studies from community and facility-based studies, household surveys, unpublished datasets

Historical data from 12 High Income Countries were used to calculate a ratio of SBR:ENMR, and was applied to generate SBR from ENNMR for high mortality settings

VR, published data from community and facility-based studies, household surveys, unpublished datasets

National registries, published data from community and facility-based studies, unpublished datasets

National registries, published studies from community and facilities, unpublished datasets

   N of observations included in estimation dataset

323 observations resulting from searches of 33,714 citations, plus household surveys and additional unpublished datasets

102 country estimates, plus historical trend data from 12 developed countries

437 observations

73 observations resulting from searches of 13,496 citations

~70 datasets

   Median year of input data

Median year for High Income Countries = 1998; Median year for Low Income Countries = 1990

Acceptable date range not specifi ed

2000

1995

2000 (range 1981-2008)

   Variability in outcome measurement methods

Yes

Yes

Yes

Yes

Yes, includes clinical assessment; medical records, ICD codes, Verbal Autopsy results

   Limitations re: population representativeness

Yes; ~20% of observations from sub-Saharan Africa and South Asia rely on data from hospital studies (likely biased)

No, all observations used were national in scope

Yes; ~41% of observations from both sub-Saharan Africa and S/ SE Asia rely on data from hospital studies (likely biased)

Study based data often from non representative populations

Yes, ~68% of low income country obs from hospital data

   Generalizability to population of interest (ie., match between burden of disease and geographic distribution of data )

73% of observations in dataset from Low Income Countries; 24% from sub-Saharan Africa; 21% from SE and S Asia

47 Low Income Countries had estimates based on the 1.2 SBR:ENMR ratio from historical High Income Countries

56% of observations from high income countries

Important gaps in the input data, especially: China, central Africa and central Asia

~40% of observations from low income countries

   Is there systematic equity assessment

No

No

No

No

No

   Global estimate

SBRate = 24 per 1000 births N of stillbirths = 3.2 million

SBRate =24 per 1000 births N of stillbirths = 3.3 million

(in process)

1.02 million

(in process)

   Range

Range of SBRates: 19 – 30 per 1000 Range of N’s of stillbirths: 2.5 - 4.1 million

Wide uncertainty assumed, but not quantified

(in process)

0.66 million - 1.48 million

(in process)

   Consistency between estimates if more than one series

Good consistency at global, reasonable at regional but poor at country level

to compare when done

none to compare with

none to compare with

Overall summary of quality of data input

Moderate for High Income Countries, low for Low Income Countries

Moderate/low for High Income Countries, very low for Low Income Countries

Moderate/low for High Income Countries, very low for Low Income Countries

Overall summary of estimates Quality according to standards for global estimates

Moderate - transparent methods but limited by input data and by adjustments made to 18 countries which increases global total of stillbirths by approximately 1 million

Moderate to Poor - limited by input data, not fully transparent inputs, the output for high mortality settings is dependent on ENMR (some of which are model-based) and multiplying all the ENMRs by 1.2, which increases the global total of stillbirths by approximately 1 million ;

Moderate - limited by input data and median-based method which may be less sensitive than results from a regression-based model

Priority areas to improve measurement now

1. Increase consistency in use of definitions (weight/gest age cut-off s) 2. Increase quantity and quality of SBR data in vital registration and national audit data 3. Increase quantity and quality of SBR data from household surveys 4. Increase dissemination of country level best estimates of SBRs

1. Agree on a simple, consistent classifi cation system 2. Increase quantity and quality of stillbirth time and cause-of-death data in vital registration and national audit data 3. Increase quantity and quality of SBR data from verbal autopsies 4. Increase dissemination of country level best estimates of SB time and cause-of-death