Make checkbox fields explicit

[pipliggins]

No description provided.

[github-actions]

Summary of the missing optional fields for the ccp-ghana parser:

table	missing	total_fields	percentage_coverage
subject	14	56	75.000000%
visit	16	53	69.811321%
observation	23	70	67.142857%

SUBJECT                                         
dataset_id                                      False
sex                                             False
preterm_infant                                  False
has_asplenia                                    False
has_tuberculosis                                False
has_chronic_respiratory_disease                 False
diabetes_type                                   False
has_apnoea                                      False
has_inflammatory_bowel_disease                  False
has_rare_disease_inborn_metabolism_error        False
has_tuberculosis_past                           False
has_comorbidity_other                           False
vaccinated_covid19                              False
vaccinated_covid19_dates                        False
VISIT                                           
dataset_id                                      False
phase                                           False
treatment_oxygen_mask_prongs                    False
treatment_antifungal_agent_type                 False
treatment_anticoagulation                       False
treatment_steroids                              False
treatment_immunosuppressant                     False
treatment_cpr                                   False
treatment_offlabel                              False
treatment_respiratory_support                   False
treatment_colchicine                            False
treatment_immunoglobulins                       False
treatment_delirium                              False
treatment_delirium_type                         False
treatment_monoclonal_antibody                   False
treatment_pacing                                False
OBSERVATION                                     
acvpu                                           False
clinical_classification_critical_illness_scale  False
total_fluid_output_ml                           False
oxygen_o2hb                                     False
clinical_frailty_score                          False
inability_to_walk_scale                         False
blantyre_coma_score                             False
mid_upper_arm_circumference_cm                  False
fio2_percent                                    False
richmond_agitation-sedation_scale               False
riker_sedation-agitation_scale                  False
mean_arterial_blood_pressure_mmHg               False
anorexia                                        False
bleeding                                        False
confusion                                       False
cyanosis                                        False
feeding_intolerance_pediatrics                  False
hepatomegaly                                    False
irritability_pediatrics                         False
lung_sounds                                     False
lymphadenopathy                                 False
severe_dehydration                              False
heart_sounds                                    False

[ekamau]

looks good!

[pipliggins]

@sadiekelly sorry to bug again... I thought this covered all the bases for checkbox fields by looking for fields with ___x in the name, but having looked again at some of the data files it's not clear to me that a lot of the symptom data wasn't also recorded as checkbox-like - e.g., the Guinea, Ghana & Uganda-v2 (these are just the first ones I checked) seem to have a data structure that suggests they use 0 as a default for admission symptoms/comorbidities, even though there are 'unknown' as options.

Any insights?

[sadiekelly]

@sadiekelly sorry to bug again... I thought this covered all the bases for checkbox fields by looking for fields with ___x in the name, but having looked again at some of the data files it's not clear to me that a lot of the symptom data wasn't also recorded as checkbox-like - e.g., the Guinea, Ghana & Uganda-v2 (these are just the first ones I checked) seem to have a data structure that suggests they use 0 as a default for admission symptoms/comorbidities, even though there are 'unknown' as options.

Any insights?

hi @pipliggins for those databases you've mentioned, it is fine to treat these not as checkboxes - they were either radio button or dropdown lists for the Yes/No/Unknown options meaning that one option should be selected and if none were selected then the cell in the csv for that data item would just be blank, there would not be a default entry for these.
I think your approach of identifying checkbox fields by the ___x pattern sounds sensible, this will definitely cover all redcap databases that used checkboxes. I think other database formats didn't tend to use checkboxes, it was quite a redcap thing. I can do a quick look through the data dictionaries to find non-redcap setups and look for any rogue checkboxes if that would be helpful?

[pipliggins]

@sadiekelly sorry to bug again... I thought this covered all the bases for checkbox fields by looking for fields with ___x in the name, but having looked again at some of the data files it's not clear to me that a lot of the symptom data wasn't also recorded as checkbox-like - e.g., the Guinea, Ghana & Uganda-v2 (these are just the first ones I checked) seem to have a data structure that suggests they use 0 as a default for admission symptoms/comorbidities, even though there are 'unknown' as options.
Any insights?

hi @pipliggins for those databases you've mentioned, it is fine to treat these not as checkboxes - they were either radio button or dropdown lists for the Yes/No/Unknown options meaning that one option should be selected and if none were selected then the cell in the csv for that data item would just be blank, there would not be a default entry for these. I think your approach of identifying checkbox fields by the ___x pattern sounds sensible, this will definitely cover all redcap databases that used checkboxes. I think other database formats didn't tend to use checkboxes, it was quite a redcap thing. I can do a quick look through the data dictionaries to find non-redcap setups and look for any rogue checkboxes if that would be helpful?

Hi @sadiekelly if you have the spare capacity that would be great! The well filled-in radio fields make spotting these easily a bit more difficult, good to know.

[sadiekelly]

@sadiekelly sorry to bug again... I thought this covered all the bases for checkbox fields by looking for fields with ___x in the name, but having looked again at some of the data files it's not clear to me that a lot of the symptom data wasn't also recorded as checkbox-like - e.g., the Guinea, Ghana & Uganda-v2 (these are just the first ones I checked) seem to have a data structure that suggests they use 0 as a default for admission symptoms/comorbidities, even though there are 'unknown' as options.
Any insights?

hi @pipliggins for those databases you've mentioned, it is fine to treat these not as checkboxes - they were either radio button or dropdown lists for the Yes/No/Unknown options meaning that one option should be selected and if none were selected then the cell in the csv for that data item would just be blank, there would not be a default entry for these. I think your approach of identifying checkbox fields by the ___x pattern sounds sensible, this will definitely cover all redcap databases that used checkboxes. I think other database formats didn't tend to use checkboxes, it was quite a redcap thing. I can do a quick look through the data dictionaries to find non-redcap setups and look for any rogue checkboxes if that would be helpful?

Hi @sadiekelly if you have the spare capacity that would be great! The well filled-in radio fields make spotting these easily a bit more difficult, good to know.

hi @pipliggins I've checked through the data dictionaries, most are REDCap format so will be covered by the ___x. Of the couple that are not, generally checkboxes were not used. I think predicovid may have been one, and the checkbox fields there are not mapped so no changes needed. Otherwise looks like the checkboxes have been accounted for so all good!

[pipliggins]

@sadiekelly sorry to bug again... I thought this covered all the bases for checkbox fields by looking for fields with ___x in the name, but having looked again at some of the data files it's not clear to me that a lot of the symptom data wasn't also recorded as checkbox-like - e.g., the Guinea, Ghana & Uganda-v2 (these are just the first ones I checked) seem to have a data structure that suggests they use 0 as a default for admission symptoms/comorbidities, even though there are 'unknown' as options.
Any insights?

hi @pipliggins for those databases you've mentioned, it is fine to treat these not as checkboxes - they were either radio button or dropdown lists for the Yes/No/Unknown options meaning that one option should be selected and if none were selected then the cell in the csv for that data item would just be blank, there would not be a default entry for these. I think your approach of identifying checkbox fields by the ___x pattern sounds sensible, this will definitely cover all redcap databases that used checkboxes. I think other database formats didn't tend to use checkboxes, it was quite a redcap thing. I can do a quick look through the data dictionaries to find non-redcap setups and look for any rogue checkboxes if that would be helpful?

Hi @sadiekelly if you have the spare capacity that would be great! The well filled-in radio fields make spotting these easily a bit more difficult, good to know.

hi @pipliggins I've checked through the data dictionaries, most are REDCap format so will be covered by the ___x. Of the couple that are not, generally checkboxes were not used. I think predicovid may have been one, and the checkbox fields there are not mapped so no changes needed. Otherwise looks like the checkboxes have been accounted for so all good!

Fab, thanks for checking this @sadiekelly!