Documents/bpData/README.md

In OHDSI/Legend: Large-Scale Evidence Generation and Evaluation in a Network of Databases

Results from a sensitivity analysis in PanTher using blood pressure data to show balance.

The script used to generate these is here.

Description of files below. Note that all analyses are done assymetrical. So if there is data for target A and comparator B, there will not be data for target B and comparator A.

BalancePsMatching.csv

Computes standardized difference of mean before and after matching on propensity scores that did not use blood presure.

Primary key:

• targetId / targetName
• comparatorId / comparatorName
• covariateId / covariateName (BP diastolic or BP systolic)

Use these columns to select single or combination treatments:

• targetType (Single or Combination)
• comparatorType (Single or Combination)

Most relevant values:

• beforeMatchingStdDiff is the standardized difference of mean before matching
• afterMatchingStdDiff is the standardized difference of mean after matching

BalanceAdjustBpPsMatching.csv

Computes standardized difference of mean before and after matching on propensity scores that did use blood presure.

Primary key:

• targetId / targetName
• comparatorId / comparatorName
• covariateId / covariateName (BP diastolic or BP systolic)

Use these columns to select single or combination treatments:

• targetType (Single or Combination)
• comparatorType (Single or Combination)

Most relevant values:

• beforeMatchingStdDiff is the standardized difference of mean before matching
• afterMatchingStdDiff is the standardized difference of mean after matching

HrsData_all_PsMatching.csv

Computes the hazard ratio either matching on a propensity score that does include blood pressure, or matching on a propensity score that does not include blood pressure. Results are reported both with and without empirical calibration.

Primary key:

• targetId / targetName
• comparatorId / comparatorName
• outcomeId / outcomeName
• type (Orginal, meaning without blood pressure in the propensity model, or Adjusting, meaning with blood pressure in the propensity model)
• estimate (Uncalibrated or Calibrated)

Use these columns to select single or combination treatments:

• targetType (Single or Combination)
• comparatorType (Single or Combination)

Most relevant values:

• rr is the estimated hazard ratio
• ci95lb is the lower bound of the 95% confidence interval of the hazard ratio
• ci95ub is the lower bound of the 95% confidence interval of the hazard ratio
• p is the two-sideed p-value against the null hypothesis of no effect

BalancePsStratification.csv, BalanceAdjustBpPsStratification.csv, HrsData_all_PsStratification.csv

Same as above, but using PS stratification instead of matching.

OHDSI/Legend documentation built on Dec. 29, 2020, 3:52 a.m.