Fix weight for height z score calculation

[mahalakshme]

Current status:

weight for length data on WHO website has 2 separate set of standard data based on age. Age is not considered in Avni wflh zscore calculation. Some of the LMS values used are of ages 0 to 2 years and some are of ages 2 to 5 years.

Acceptance criteria:

• Incorporate age also in wflh calculation in addition to gender and height currently used
• For LMS values not defined for a particular age - say age > 2 years and height < 65 cm - return undefined

Out of scope:

• Implementation team need to make sure if rules work as expected - for APF when checked since they use both WFA and WFH can be handled - not an issue.
• correcting existing data

Impact analysis:

Using of data from 2 to 5 years leads to a higher negative value(more critical grade). Hence mixing of both data would have led some children below 2 years to be categorised as more critical and some children more than 2 years as less critical.

In a couple of values where I applied the formula the difference of zscore between correct and incorrect is .1 to .2. So the children in the border value used in the standard to classify would have impacted.

In a couple of implementations I checked, weight for height is used along with weight for age for categorisation of child status. This would have again reduced the impact of this issue.

Hence we can chose not to correct the existing data for now considering the complexity and its impact.

Testing:

Verify using the tables and charts mentioned here

Old - Ignore:

• • Reasons: We dont differentiate height and length for child in our app. And according to WHO growth computation, The construction of the weight-for-length (45 to 110 cm) and weight-for-height (65 to 120 cm) standards followed a procedure similar to that applied to constructing the length/height-for-age standards (see section 3.1). To fit a single model, 0.7 cm was added to the cross-sectional height values. This was the average difference found between length and height in 1625 children aged 18 to 30 months measured for both length and height. After the model was fitted, the weight-for-length centile curves in the length interval 65.7 to 120.7 cm were shifted back by 0.7 cm to derive the weight- for-height standards corresponding to the height range 65 cm to 120 cm.
• • Across heights, the difference of weights between both tables is only .1 to .3
• • We can display in UI that for these values z-scores are approximate and not accurate.
• • We can always ask for feedback

[vinayvenu]

• Don't we have the Windows package that we can install to figure out test cases?
• -3SD for WFA girls is around 80 cm. Someone who is 65 cm at this age must be considered out of scope for our calculator
• What else is being done in the anthro R package?

[mahalakshme]

@vinayvenu

1. In the windows exe application - Anthroplus when I checked yesterday with Taqi, it didn't have weight for height value in the result. It had only WFA, length for age and BMI for age

2. https://github.com/WorldHealthOrganization/anthro/blob/96e8fe016f4e2ed79eb187f08278b336bafbb7b5/R/z-score-weight-for-lenhei.R#L17 - anthro R package method link - some logic is used to figure out m,l, and s values

[mahalakshme]

@vinayvenu
Not necessarily I see this holds '-3SD for WFA girls is around 80 cm. Someone who is 65 cm at this age must be considered out of scope for our calculator' and we have above such entries in db. consider in JSS individual with id, 57083, - though her age is greater than 4 years, height as per height for age is less than -3 SD. and its WFH zscore is 1.05.

[vinayvenu]

@mahalakshme lets discuss separately and finalize this story.

[vinayvenu]

Self note: Potential impact in growth charts

[himeshr]

@mahalakshme, the calculated values for wfh zscore do not exactly match the chart.
Even after we use 2 separate tables for ageGroups 0-2 years and 2-5 years, for boys /girls.
Is #1247 complete.? Because, i think the formula used for wfh is not correct for all scenarios and the AC here does not talk about logic correction..

[mahalakshme]

@himeshr I think then you can just comment on the differences you see here - just egs. It can be dealt in a separate card.

[himeshr]

@himeshr I think then you can just comment on the differences you see here - just egs. It can be dealt in a separate card.

okay.. Sample Error scenario examples are as follows:
Reference: https://www.who.int/tools/child-growth-standards/standards/weight-for-length-height

boys => 2 to 5 years
--------------------
((6.5, 68.5).wfh, -3.01); // -3
((7.0, 68.5).wfh, -2.04); // -2
((7.6, 68.5).wfh, -1); // -1
((8.2, 68.5).wfh, -0.06); //0, correct
((9.0, 68.5).wfh, 1.06); // 1
((9.8, 68.5).wfh, 2.04); // 2
((10.7, 68.5).wfh, 3.03); // 3


girls => 0 to 2 years
--------------------
((5.8, 67.0).wfh, -3.01); // -3
((6.3, 67.0).wfh, -2.03); // -2
((6.9, 67.0).wfh, -0.97); // -1
((7.5, 67.0).wfh, -0.04); // 0
((8.3, 67.0).wfh, 1.05); // 1
((9.1, 67.0).wfh, 2.01); // 2
((10.0, 67.0).wfh, 2.96); // 3

girls => 2 to 5 years
--------------------

((6.2, 68.5).wfh, -2.95); // -3
((6.7, 68.5).wfh, -2.02); // -2
((7.3, 68.5).wfh, -1.02); // -1
((8, 68.5).wfh, 0); //0, correct
((8.8, 68.5).wfh, 1.03); // 1
((9.7, 68.5).wfh, 2.05); // 2
((10.7, 68.5).wfh, 3.03); // 3

[himeshr]

For validation check following test cases:
Compare the z-score returned by app with simple table references available here

I. Girls, age 0-2 years
II. Girls, age 2-5 years
III. Boys, age 0-2 years
IV. Boys, age 2-5 years
Weight values corresponding to median, with heights 65, 70, 75

[AchalaBelokar]

• tested senario maha@APF on staging with jija individual
  weight and h.ods