Tutorial

The best way to interact with the network is to use the Buyer extension objects and its associated methods.

Querying health datasets

Executing a query : For executing a query, we get the api key and ip address from the web dashboard and then create a Buyer instance and then call a query method on that buyer instance with a SQL query string and then print recieved query accuracy and results.

from fortress_sdk import Buyer

api_key = "buyer_key"  # get this from the fortress web dashboard
ip_addr = "127.0.0.1"

# Initialize a buyer instance
buyer = Buyer(api_key, ip_addr)

# Initiate a query
sql_query = "select count(*) as numpeople from public.condition_era_death"
result, accuracy = buyer.query(query=sql_query)

sql_query = "select count(*) as people_per_condition from person"
result, accuracy = buyer.query(query=sql_query)

# execution result and accuracy of the query
print(result)
print(accuracy)

NOTE!! : Related to SQL querying rules

• allowed: Group by, selection of count, sum/min/max/avg of limited columns, of raw columns that are the group-by columns, any where conditions

• not allowed: join, selection of columns that aren’t in the group by, arbitrary sum/min/max/avgs

Pretty print query history

###########################################
####        Print Query History        ####
###########################################

buyer.print_query_history()

+--------------------------------------------------------------------------------+
query ===> select count(*) as numpeople from public.condition_era_death
result ===> [['numpeople'], [30297]]
accuracy ===> within: 0.1% of the true value with 95% probability

+--------------------------------------------------------------------------------+
query ===> select count(*) as people_per_condition from person
result ===> [['people_per_condition'], [2000000]]
accuracy ===> within: 0.0% of the true value with 95% probability

Get queryable table names

print("----------------------TABLE NAMES-----------------------")
columns = buyer.get_columns()
print(columns.keys())

['CONDITION_ERA', 'CONDITION_ERA_CONDITION_OCCURRENCE', 'CONDITION_ERA_DEATH',
'CONDITION_ERA_DEVICE_EXPOSURE', 'CONDITION_ERA_DOSE_ERA', 'CONDITION_ERA_DRUG_ERA',
'CONDITION_ERA_DRUG_EXPOSURE', 'CONDITION_ERA_EPISODE', 'CONDITION_ERA_MEASUREMENT',
...
...
'PROCEDURE_OCCURRENCE_VISIT_DETAIL', 'PROCEDURE_OCCURRENCE_VISIT_OCCURRENCE', 'SPECIMEN',
'SPECIMEN_VISIT_DETAIL', 'SPECIMEN_VISIT_OCCURRENCE', 'VISIT_DETAIL',
'VISIT_DETAIL_VISIT_OCCURRENCE', 'VISIT_OCCURRENCE']

Get queryable table columns

print("----------------------TABLE COLUMNS-----------------------")
columns = buyer.get_columns()
print(columns["PERSON"])
print(columns["PERSON_VISIT_OCCURRENCE"])

-----------------------COLUMNS--------------------
['birth_datetime', 'care_site_id', 'day_of_birth', 'ethnicity_concept_id',
'ethnicity_source_concept_id', 'ethnicity_source_value', 'gender_concept_id',
'gender_source_concept_id', 'gender_source_value', 'location_id', 'month_of_birth',
'person_id', 'person_source_value', 'provider_id', 'race_concept_id', 'race_source_concept_id',
'race_source_value', 'year_of_birth']

['admitted_from_concept_id', 'admitted_from_source_value', 'birth_datetime',
'care_site_id', 'day_of_birth', 'discharged_to_concept_id', 'discharged_to_source_value',
'ethnicity_concept_id', 'ethnicity_source_concept_id', 'ethnicity_source_value',
'gender_concept_id', 'gender_source_concept_id', 'gender_source_value', 'location_id',
'month_of_birth', 'person_id', 'person_source_value', 'preceding_visit_occurrence_id',
'provider_id', 'race_concept_id', 'race_source_concept_id', 'race_source_value',
'visit_concept_id', 'visit_end_date', 'visit_end_datetime', 'visit_occurrence_id',
'visit_source_concept_id', 'visit_source_value', 'visit_start_date', 'visit_start_datetime',
'visit_type_concept_id', 'year_of_birth']