Weight for age zscore mismatch

[mahalakshme]

Related ticket: https://avni.freshdesk.com/a/tickets/3216

Steps to reproduce:

• On doing the following,

select * from jss.individual_phulwari_growth_monitoring where individual_id=1299188 order by id desc;

you will find that the 'Weight for age z-score' is -8.5 for the first encounter visit.

• But when using these values in the WHO provided anthro software here the value is -6.71
  
• The above discrepancy exists when zscore is less than -3 or greater than 3

Preliminary Analysis:

A different formula is used for calculation of zscore is less than -3 or greater than 3
For the formula check page no:300 in here: 924154693X_eng.pdf or check compute_zscore_adjusted method in the file z-score-helper.R in the repository here

Why this needs to be fixed:

• For the org in focus though for grading/categorising the severity for the child, the above fix doesn't matter, the org mainly focusses on childern where zscore is less than -3 and hence they need reliable values to understand the improvement, to gain confidence to continue the program and to answer their funders.

Acceptance criteria:

• Fix according to the above formula. After fixing, the values should be inline with the expected values in the Who anthro analysis.xlsx sheet attached by the client in the ticket.
• Round the zscore value to 2 digits instead of current one.

[vinayvenu]

It looks like the calculation is slightly different for zscores beyond the -3 SD range (see attached document page 301). This might be the reason for the discrepancies.

I am not sure how important the exact zscore really matters for those beyond the -3 range since there are no real decisions being made at this level from a primary health perspective. Either ways, it is good to correct the calculations and verify that it exactly matches the document for zscores beyond +-3SD

924154693X_eng.pdf

[mahalakshme]

@vinayvenu I didn't add 2 decimal points round-off to this story since looks like we have done it intentionally. The comment in our code says Weight/height measurements are available only for the nearest single digit, so the final figure can only have one degree of accuracy. So need to make sure before adding it to a card.

[petmongrels]

.

[mahalakshme]

@petmongrels The below data can be used to test for cases where zscore is greater than 3
For id 1110369 in jss.individual_phulwari_growth_monitoring, existing "Weight for age z-score" value is 8.7, but looks like actual is 11.08

For id 1233198, existing is 11.6, but looks like actual is 27.08

The weight values in the above entries are not realistic/seems wrong. But this should not be an issue to verify the correctness of zscores.

[petmongrels]

it seems this tool uses decimal month age as well, which I haven't implemented - so the numbers are slightly off. but I think it is fine, as the numbers are quite close to adjusted zScores now.

[mahalakshme]

@petmongrels how much off it is? The LMS values in WHO website itself is available only for whole months and not for decimal months. I think the same is used in anthro R package which is probably used in anthro WHO calculator exe. wondering we have missed somewhere else like rounding of months? Were the values off for less than -3 as well?

[petmongrels]

@mahalakshme
from the example in this story.......
For id 1233198, existing is 11.6, but looks like actual is 27.08.........our actual value now is 27.57

there is nothing in code that I can make out.