Resolve Measure 8 and Component 11 data definition and quality issues.

[grosscol]

Problem

The original definition of C11 is sum(contra_choice != "unknown" & contra_choice == imm_method). That is: count the number of cases where contraception choice was not unknown and contraception choice matched the immediate method.

The problem is that it doesn't account for how the values of the fields have been interpreted by the analyst or abstractors. 90% of the counts for the above definition of C11 also occur in a row where counseling is FALSE. This leads me to believe that the value of "other" was used in the data set when it should be "unknown".

Current Mitigation

The following definition is currently used for C11 to handle the lack of counseling. Which impacts Measure 8:

C11 = sum(counseling & contra_choice != "unknown" & contra_choice == imm_method)

Solution

This needs to be addressed in the following ways:

Definitions

• The meaning of the contra_choice values needs to be defined.
• The valid values of contra_choice when counseling is FALSE need to be defined.

Data Corrections

• A transform for the existing data needs to be approved: e.g. when counseling is FALSE, change value of contra_choice to "unknown".
• A confirmation that the existing entries of "other" in the imm_method field are correct.

Alternatively: Data could be re-uploaded with conformant use of other in contra_choice and imm_method fields. Effectively fixing the data issue at the source.

[grosscol]

Removing the mitigation as it may not be a valid assumption. Per the conversation in slack:

zachll  3:05 PM
OK, this reminds me that Michelle had said that often providers were skipping the step where they checked a box to indicate that counseling was done
and instead they would just document the preferred contraceptive
so it could be that if there is documentation, we infer that counseling was done

grosscol  3:05 PM
So maybe that fix isn't a valid one.

zachll  3:05 PM
right
3:06
one option is to just ignore the relationship
between counseling and documentation

grosscol  3:06 PM
Seems reasonable.

zachll  3:06 PM
let's do that for now

grosscol  3:06 PM
Okay.  Retabulate the data without the restriction on counseling documented.

[grosscol]

Update the component 11 criteria to check that contraception provided was "0-3 days"

grosscol  4:55 PM
imm_method == contra_choice & contra_prov == '0-3 days' gives 126 matches.

zachll  4:56 PM
on its face that is a lot better

grosscol  4:57 PM
Okay.  I'll change component 11 to include that requirement.  That will change the numerator of measure 8.

...

zachll  4:58 PM
great, that makes a lot of sense and I had not noticed that when we wrote the logic, but it should be there