Click on a row to explore the results for that model. When you wish to explore a different model, then select the new result row and the tabs will be updated.

Demo Video

Predictors of diagnostic transition from major depressive disorder to bipolar disorder: a retrospective observational network study

Development Status: Completed


This shiny application contains the results of the internal and external validations of a model developed to predict 1-year risk of transition from major depressive disorder to bipolar disorder

During manuscript development and the subsequent review period, these results are considered under embargo and should not be disclosed without explicit permission and consent from the authors.

Below are links for study-related artifacts that have been made available as part of this study:

Protocol: link


Below is the abstract of the manuscript that summarizes the findings:

Background: Many patients with bipolar disorder (BD) are first diagnosed with major depressive disorder (MDD), and there are high prospective diagnosis conversion rates from MDD to BD. It is unknown whether MDD is a BD comorbidity or it is an earlier stage of BD.

Methods: IBM MarketScan® CCAE, MDCR, MDCD, Optum EHR, and Optum Claims databases were analyzed. Patient inclusion criteria were: age >10 years; >= 1 year of observation prior to MDD onset; >=1 year of observation after MDD (or BD diagnosis transition earlier); no BD, schizophrenia, schizoaffective disorder, or psychotropic use prior to MDD. Cyclops regularized logistic regression models were developed on one year MDD-BD conversion with all standard covariates from the patient level prediction package. A simple model was then derived, and time-to-conversion Kaplan-Meier analysis was performed up to a decade after MDD, stratified by score range. External validation of the model was performed across the OHDSI network.

Results: The model internal AUC ranged 0.633-0.745 depending on the database. Nine variables predicted one year MDD-BD transition: age at MDD onset; MDD severity; psychosis; comorbid mental disorder; anti-anxiety drugs; pregnancy; substance misuse; self-harm thoughts/actions. Each predictor was assigned a score with a higher score meaning greater conversion rate to MDD over time in the Kaplan-Meier models. Year-by-year AUC estimates trended stable to improving with time.

Conclusions: Our approach produced a simple, clinically understandable model that validates well across multiple international data sources. Some validation datasets had comparable AUCs to those of the training sets, and some were too small to assess – further network study sites will be incorporated into the validation as IRB approvals are obtained.

Study Packages

  • Model validation: link

The preliminary research results on this web-based application are from a retrospective, real-world, observational study in support of this activity and will subsequently be submitted to a peer-reviewed, scientific journal.

Data Information

The following databases were used in this study:

Database Name Country Type Years
Clinformatics Optum® De-Identified Clinformatic Data Mart Database – Date of Death (DOD)     USA     Claims     2000-2019
AUSOM     Ajou University School of Medicine Database Korea     EHR     1999-2018
CCAE     IBM MarketScan® Commercial Database     USA     Claims 2000-2019
CUIMC     Columbia University Irving Medical Center Data Warehouse USA     EMR 1990-2020
VA-OMOP     Department of Veterans Affairs USA     EMR 2009-2010, 2014-2020
JMDC     Japan Medical Data Center     Japan     Claims     2000-2019
MDCD     IBM MarketScan® Multi-State Medicaid Database     USA     Claims 2006-2019
MDCR     IBM MarketScan® Medicare Supplemental Database     USA     Claims 2000-2019
optumEhr     Optum® de-identified Electronic Health Record Dataset     USA EHR     2006-2019

All databases obtained IRB approval or used deidentified data that was considered exempt from IRB approval.

AUC per year

Select the databases to plot the AUC by year


Conversion from MDD to bipolar over time

Select the database to plot



Model Settings: help

Population Settings: help

Covariate Settings: help






Model Table