Pandas' table_pivot Generating the wrong number of columns

I have a table of patients as shown bellow:

The id column is the identifier of a patient.
The exam column is the code representing a specific medical examination.
The value column is the numerical result of that examination.

What I want to do is pivot the table so that, each row corresponds to only one patient and the columns the exams. Like this:

I looked it up on how to do it, and the pandas library have the function pandas.DataFrame.pivot_table that does exactly what I needed.

So, after giving context to the problem here is the question. My original table have 884 unique exams codes, meaning the after the pivot it should result in 884 columns. But that is not the case, after I pivot the table, it shows the correct number of rows but only 557 columns.

I checked pandas' documentation and searched for someone with the same problem but I couldn`t find it.

Anyone knows what might be the problem ?

-> Original table
*The name of the columns were simplified in the example shown above

->Pivot code
table = pd.pivot_table(results, values='valuenum', index='subject_id', columns=['itemid'] )

-> Table after pivot

Ps: Sorry for the poor English, I'm a non native speaker.

What might be going on:

Just guessing here, but I know that null values can mess with pd.pivot_table and cause values to fall out of the index or the columns. Maybe adding dropna=False to your pd.pivot_table call might fix the problem?

Examples showing how dropna=True might be the problem:

# Full data
data_full = {
    "id": [1, 1, 1, 2, 2, 2],
    "exam": [100, 110, 120, 90, 100, 120],
    "value": [34.2, 23, 4.2, 114, 54.2, 6.7],
}

# I am going to pick on the example of id: 2, exam: 90, value: 114
data_null_id = {
    "id": [1, 1, 1, pd.NA, 2, 2],
    "exam": [100, 110, 120, 90, 100, 120],
    "value": [34.2, 23, 4.2, 114, 54.2, 6.7],
}

data_null_exam = {
    "id": [1, 1, 1, 2, 2, 2],
    "exam": [100, 110, 120, pd.NA, 100, 120],
    "value": [34.2, 23, 4.2, 114, 54.2, 6.7],
}

data_null_value = {
    "id": [1, 1, 1, 2, 2, 2],
    "exam": [100, 110, 120, 90, 100, 120],
    "value": [34.2, 23, 4.2, pd.NA, 54.2, 6.7],
}

datas = [data_full, data_null_id, data_null_id, data_null_exam, data_null_value]

# For each of the contrived scenarios I made above
for data in datas:
    # Build a DataFrame of that scenario
    df = pd.DataFrame(data)
    # Pivot as you do in your example
    pivot = pd.pivot_table(df, values='value', index=['id'], columns=['exam'])

    display(pivot)

You will see that the exam is omitted even if the value or the id is null. However, if you calculated a unique count on exam beforehand, it would not agree with the number of columns in the pivot table.

Possible Fix

If you change the pivot declaration to pd.pivot_table(df, values='value', index=['id'], columns=['exam'], dropna=False) you will see that the column is not dropped.

I hope this helps! Sorry that I can't know exactly what's going on without the data itself.