ohdsi / legend

An R package implementing Large-Scale Evidence Generation and Evaluation in a Network of Databases (LEGEND).

OHDSI Large-Scale Evidence Generation and Evaluation in a Network of Databases

Under development. Do not use.

License

The Legend package is licensed under Apache License 2.0

Development

Legend was developed in R Studio.

Development status

Under active development.

legend's Issues

Issues with positive control synthesis for Hydralazine

When T = Hydralazine , signal injection did not work properly for on-treatment window. This is probably due to the fact that most exposures to hydralazine have length = 0.

When comparing mono-therapy to duo-therapy, the attrition diagram is missing the first few steps

PDF download fails

WHen using the LegenMedCentral, the PDF fails to download

(I may be reporting it in wrong repo) (for the shiny app would be more appropriate but it combines multiple apps)

see

Report absolute risk (incidence) difference

Discrepancies between protocol and code for Hypertension

1. R package uses the default covariate settings (all drugs, drug classes, condition, condition classes, procedures, observations, etc. from documentation ); Protocol specifies covariates as demographics, conditions, condition aggregation, drugs, drug aggregation, Risk scores: Charlson comorbidity index -- does not include procedures, measurements, or observations

Median follow-up in LEGEND data model

To my knowledge, current LEGEND data model cannot provide 'median follow-up' duration in each data source. We need to consider to add median follow-up duration to the model.

how to consider a patients would have more than one AMI events with more than 180 day gaps in hypertension Lancet article

hello, i am confuse about how to deal with the situation that one patient might occur two or more primary outcome events with a difined gaps such as 180 days. i do not know how to handle this in cox regression you mentioned in Lancet hypertension article published in 2019. thank you for your answer

Add pt counts to covariate table 1

We should update the code to log counts (as well as percentages) of covariates for Table 1s (may need to update LEGEND data model); several journal guidelines have "insisted" on these numbers.

Corrections to HTN protocol

Section 8.4.1 inclusion rules Dx's need to be update to HTN (not depression).

java.sql.SQLException: [Amazon](500310) Invalid operation: value too long for type character varying(17)

Error:
java.sql.SQLException: Amazon Invalid operation: value too long for type character varying(17);

SQL:
INSERT INTO legend_study.legend_feasability_depression_attition (
exposure_id,
target_id,
comparator_id,
sequence_number,
description,
subjects)
SELECT exposure_id,
target_id,
comparator_id,
sequence_number,
description,
subjects
FROM (
-- Restricted to common period: take final number
SELECT epcs.cohort_definition_id AS exposure_id,
epcs.target_id,
epcs.comparator_id,
CAST(4 AS INT) AS sequence_number,
CAST('Restricted to common period' AS VARCHAR(255)) AS description,
epcs.num_persons AS subjects
FROM #ep_cohort_summary epcs
) temp

Finding outcome definitions, their concept sets, and how to translate those to ICD-10 code

i cannot find the SNOMED ID the articiles showed

PS distributions appear reversed for some comparisons in the viewer

See e.g. T = Amlodipine, C = Furosemide

Warning: All coefficients (except maybe the intercept) are zero. Either the covariates are completely uninformative or completely predictive of the treatment. Did you remember to exclude the treatment variables from the covariates?

Hello Marc,

When running assessPropensityModels(...), there is such an error:

Warning: All coefficients (except maybe the intercept) are zero. Either the covariates are completely uninformative or completely predictive of the treatment. Did you remember to exclude the treatment variables from the covariates?

It seems there is obvious we can do to fix. What is wrong?