Airflow Automated Validation Checks

Dataset Validations

Checks on Output of the Build

test name	test description	tables checked
validation_check_roid	checks that ROIDs are unique	prod_combined_all, prod_combined_abridged
validation_check_roid_vs_ros	checks that the count of ROIDs matches expected count	prod_combined_all, prod_combined_abridged
validation_all_payers_have_rates	checks that non-null rates exist for all payers	prod_combined_all, prod_combined_abridged
validation_all_networks_have_rates	checks that non-null rates exist for all networks	prod_combined_all, prod_combined_abridged
validation_most_hospitals_have_rates	checks that non-null rates exist for >95% of hospitals	prod_combined_all, prod_combined_abridged
validation_billing_code_types_have_rates	checks that non-null rates exist for expected billing code types	prod_combined_all, prod_combined_abridged
validation_bill_types_have_rates	checks that non-null rates exist for expected billing types	prod_combined_all, prod_combined_abridged
non_outlier_median_canonical_rate	checks that non-outlier median canonical rate is between 1000 and 4000	prod_combined_abridged
non_outlier_median_canonical_percentage_of_state_avg_medicare	checks that non-outlier median canonical % of med is between 1 and 1.7	prod_combined_abridged
no_negative_rates	checks that there are no negative non-outlier canonical rates	prod_combined_abridged
no_rates_gtr_20m	checks that there are no non-outlier canonical rates > 20m	prod_combined_abridged
non_outlier_coverage_gtr_30_pct	checks that non-outlier coverage > 30%	prod_combined_abridged
raw_rate_types	for each raw rate type, use trace_raw_id to find the matching rate in core rates / hospital rates	prod_combined_all
transform_rate_types	reproduce each transformation using trace_raw_id to find the matching rate in core rates / hospital rates	prod_combined_all
impute_rate_types	verify each imputation calculation using fields in prod_combined_all	prod_combined_all

Sample Analyze Report Creation

For each provider org below, re-create an analyze report using data-only and check that results are within a reasonable range of expected values.

• commonspirit_health
• mass_general_brigham
• providence
• intermountain

---

Pipeline Validations

Checks at Various Stages in the Build

Rate Object Space:

test name	test description	tables checked
validation_check_roid	checks that ROIDs are unique	tmp_rate_object_space
ros_billing_code_types	checks that expected billing code types exist	tmp_rate_object_space
ros_bill_types	checks that expected bill types exist	tmp_rate_object_space
ros_taxonomy_groupings	checks that expected taxonomy groupings exist	tmp_rate_object_space
ros_payers	checks that expected payers exist	tmp_rate_object_space
ros_networks	checks that expected networks exist	tmp_rate_object_space

Raw Data After Mapping to ROS:

test name	test description	tables checked
validation_check_roid	checks that ROIDs are unique	tmp_raw_mrf_payer_rates, tmp_raw_mrf_hospital_rates
validation_all_payers_have_rates	checks that non-null rates exist for all payers	tmp_raw_mrf_payer_rates, tmp_raw_mrf_hospital_rates
validation_all_networks_have_rates	checks that non-null rates exist for all networks	tmp_raw_mrf_payer_rates, tmp_raw_mrf_hospital_rates
validation_most_hospitals_have_rates	checks that non-null rates exist for >95% of hospitals	tmp_raw_mrf_payer_rates, tmp_raw_mrf_hospital_rates
validation_billing_code_types_have_rates	checks that non-null rates exist for expected billing code types	tmp_raw_mrf_payer_rates, tmp_raw_mrf_hospital_rates
		tmp_raw_mrf_payer_rates, tmp_raw_mrf_hospital_rates

Raw Data Combined:

test name	test description	tables checked
validation_check_roid	checks that ROIDs are unique	tmp_int_combined_raw
validation_check_roid_vs_ros	checks that the count of ROIDs matches expected count	tmp_int_combined_raw

After Combining Benchmarks and Accuracy:

test name	test description	tables checked
validation_check_roid	checks that ROIDs are unique	tmp_int_accuracy_brit
validation_check_roid_vs_ros	checks that the count of ROIDs matches expected count	tmp_int_accuracy_brit

SubDag Output:

test name	test description	tables checked
validation_check_roid	checks that ROIDs are unique	tmp_int_combined, tmp_int_combined_abridged
validation_check_roid_vs_ros	checks that the count of ROIDs matches expected count	tmp_int_combined, tmp_int_combined_abridged
validation_all_payers_have_rates	checks that non-null rates exist for all payers	tmp_int_combined, tmp_int_combined_abridged
validation_all_networks_have_rates	checks that non-null rates exist for all networks	tmp_int_combined, tmp_int_combined_abridged
validation_most_hospitals_have_rates	checks that non-null rates exist for >95% of hospitals	tmp_int_combined, tmp_int_combined_abridged
validation_billing_code_types_have_rates	checks that non-null rates exist for expected billing code types	tmp_int_combined, tmp_int_combined_abridged
validation_bill_types_have_rates	checks that non-null rates exist for expected billing types	tmp_int_combined, tmp_int_combined_abridged