c_zerovac: Zimbabwe, Chad, and CotedIvoire

[wengxyu1030]

Zimbabwe

The code was problematic on the missing condition, corrected for all the waves: https://github.com/Data-Whale-DC/WB-HEFPI-MICS5/blob/6b46fe08ac3a11bec32b44d2b156383f4ea0449d/MICS5_do_c_vaccination.do#L371-L379

The sub-variables it depends on: c_bcg, c_dpt1....are showing inconsistency between the different data sources (DHS vaccination rate is lower than MICS in almost every vaccination type), which contributed to the gap between DHS and MCIS for c_vaczero, please see the detailed screenshot below:

[wengxyu1030]

Chad

Similar case for Chad.

[wengxyu1030]

TrinidadandTobago1987

The code and data reference makes sense: https://github.com/wengxyu1030/DHS-Recode-I/blob/7d637716e7ba26a5fce51105bd2218da9d42ef87/7_child_vaccination.do#L67-L71

But the reulst is 100% because except those who confirmed taking none of the vaccinations, others are registered as missing at least for one of the vaccination:

c_vaczero	Freq.	Percent	Cum.
1	145	1.85	1.85
.	7,692	98.15	100.00
Total	7,837	100.00

[wengxyu1030]

CodedIvoire1998

[robin-wang]

Finally catching up on this...

[robin-wang]

Zimbabwe

The code was problematic on the missing condition, corrected for all the waves: https://github.com/Data-Whale-DC/WB-HEFPI-MICS5/blob/6b46fe08ac3a11bec32b44d2b156383f4ea0449d/MICS5_do_c_vaccination.do#L371-L379

The sub-variables it depends on: c_bcg, c_dpt1....are showing inconsistency between the different data sources (DHS vaccination rate is lower than MICS in almost every vaccination type), which contributed to the gap between DHS and MCIS for c_vaczero, please see the detailed screenshot below:

@wengxyu1030 Response on Zimbabwe : Given the mild and stable differences, propose to recommend keeping only DHS per previous discussions. Just to confirm on the underlying logic of vaccination vars - we are not suffering from missing condition issue for other vaccination indicators? Additionally, I am not seeing updates on c_vaczero missing values. Could you please enlighten me on the changes?

I attach the version I had below.

[robin-wang]

Chad

Similar case for Chad.

Thanks much. i see polio and zerovac being affected to a great extent by the 2019 kink. Could you please remind me what the updated values are Cheers!

[robin-wang]

CodedIvoire1998

I checked report for Cote d'Ivoire 1998 and Cote d'Ivoire 1994. 1998 has much higher vaccination rates than 1994, showing the opposition of what the DHS section demonstrated. 1998

1994

THEN, I LOOKED AT [V007], AND WWWTTTFFF 1998birth has v007 == 1994, and 1994birth has v007 == 1998.... hence the reverse in trend.... @wengxyu1030

[wengxyu1030]

Zimbabwe The code was problematic on the missing condition, corrected for all the waves: https://github.com/Data-Whale-DC/WB-HEFPI-MICS5/blob/6b46fe08ac3a11bec32b44d2b156383f4ea0449d/MICS5_do_c_vaccination.do#L371-L379 The sub-variables it depends on: c_bcg, c_dpt1....are showing inconsistency between the different data sources (DHS vaccination rate is lower than MICS in almost every vaccination type), which contributed to the gap between DHS and MCIS for c_vaczero, please see the detailed screenshot below:

@wengxyu1030 Response on Zimbabwe : Given the mild and stable differences, propose to recommend keeping only DHS per previous discussions. Just to confirm on the underlying logic of vaccination vars - we are not suffering from missing condition issue for other vaccination indicators? Additionally, I am not seeing updates on c_vaczero missing values. Could you please enlighten me on the changes?

I attach the version I had below.

Thanks for your review @robin-wang The code on missing should be all up to date and correct, I guess the version you are referring to is not the main branch? Please confirm and share the link you are referring to if not consistent.

[wengxyu1030]

Chad Similar case for Chad.

Thanks much. i see polio and zerovac being affected to a great extent by the 2019 kink. Could you please remind me what the updated values are Cheers!

@robin-wang By logic, the updated code would lead to zerovac and full_immu should all be missing because there's no data for measle vaccination. Polio 2019 code seems fine to me, but it's in french, your help is appreciated.

[wengxyu1030]

CodedIvoire1998

I checked report for Cote d'Ivoire 1998 and Cote d'Ivoire 1994. 1998 has much higher vaccination rates than 1994, showing the opposition of what the DHS section demonstrated. 1998

1994

THEN, I LOOKED AT [V007], AND WWWTTTFFF 1998birth has v007 == 1994, and 1994birth has v007 == 1998.... hence the reverse in trend.... @wengxyu1030

You are a genius!!!

[robin-wang]

Chad Similar case for Chad.

Thanks much. i see polio and zerovac being affected to a great extent by the 2019 kink. Could you please remind me what the updated values are Cheers!

@robin-wang By logic, the updated code would lead to zerovac and full_immu should all be missing because there's no data for measle vaccination. Polio 2019 code seems fine to me, but it's in french, your help is appreciated.

Understood, measles missing for 2019, hencezerovac and fullimm should be missing for 2019. For polio, all good. but need decision with TTL.

      Polio1             2                3 

2014 76 66 50 Value 80 70 56

2010 56 42 25 Value 60 48 32

2004 78 60 36 Value 80 63 39

2000 90 78 51 Value 92 79 52

---

From Chad2019 report, they fooked up. Those without vaccination cards were completely ignored, and they diplomatically apologised for this.... Let's remove all 2019 indicator for Chad...

Malheureusement, l'application CAPI a été programmée avec une instruction de saut qui par erreur n'a activé que cet ensemble de questions pour les enfants qui ont participé aux trois premiers événements de vaccination répertoriés (sur un total de 18 événements de vaccination). Ces trois éléments étaient chronologiquement les plus éloignés dans le temps et ne s'appliquaient donc qu'aux enfants dont leurs âges sont près de trois ans. Cette erreur n'a pas été détectée tout au long de la collecte des données, ignorant en fait complètement tous les jeunes enfants sans documents de vaccination disponibles. Étant donné que les indicateurs standards reposent à la fois sur les documents de vaccination disponibles et sur le rappel de la mère, les tableaux standards ne peuvent pas être produits. Les estimations de la couverture vaccinale présentées dans les tableaux de cette annexe ne représentent donc que le sous-ensemble des enfants âgés de 12 à 23 mois pour lesquels les carnets de vaccination ont été consultés. L'utilisation de ces taux de couverture pour tous les enfants dans tout le pays conduira à des conclusions biaisées. Les données collectées sur les enfants avec des documents de vaccination sont cependant conformes à la norme et cette annexe décrit ces résultats. Comme ces estimations n'ont pas de comparaison directe avec les données collectées précédemment, cette annexe présente également les taux recalculés de la précédente enquête EDS- MICS 2014/15 au Tchad, c'est-à-dire de 2014-15. Une telle analyse plus approfondie peut être entreprise par des experts en données de vaccination, comme il se doit pour toute enquête. Le but de cette annexe est de présenter des données comparables, non d'évaluer la qualité d'une comparaison ou de mesurer des tendances.

[robin-wang]

【Vaccination indicators - denominator update】 Per discussion with Sven, I am making underlying changes to vaccination indicators throughout, namely, using h1 (DHS for example) vaccination card as the foundation for indicator denominator.

Sample: *c_measles child Child received measles1/MMR1 vaccination gen c_measles =. replace c_measles = 1 if (h9 ==1 | h9 ==2 | h9 ==3)
replace c_measles = 0 if h9 ==0
replace c_measles = 0 if missing(c_measles) & !missing(h1)
replace c_measles = . if h9 > 3

LOG for noticeable changes to vaccination indicator values: DHS-VI

• Gabon2012: c_dpt1 drops from 99.58% to 78.66%; c_polio1 drops from 67.11% to 65.97%, less drastic.

• [ ] To confirm the valid use of [h1] vaccination card

• [ ] To confirm large value changes between original methodology and the updated one

[robin-wang]

For both c_zerovac and fullimm, since we have addressed missing values in individual vaccination indicators, Sven proposed in meeting that we are to keep the current logic :

If anything misses, then the aggregate indicator goes to missing