Use Case

A collabratobe study was conducted invoving five participating sites of OneFlorida to study the association between OUD and several relevant clinical risk factors among individuals receiving at least one opioid prescription.

Each participating site extracted EHR records between 01/01/2012 and 03/01/2019 for patients who had opioid prescription (including Codeine, Fentanyl, Hydromorphone, Meperidine, Methadone, Morphine, Oxycodone, Tramadol, Hydrocodone, Buprenorphine), and no cancer or diagnosis of opioid use disorder before their first prescription. Among these patients who were exposed to opioid, a case of opioid use disorder is defined as having first diagnosis of opioid use disorder within 12 months after their first prescription and a control is defined as having no diagnosis of opioid use disorder in the entire time window.

Logistic regression chosen to model the outcome variable and the following risk factors: age, race, gender, and insurance type, Race/ethnicity, and having at least one diagnosis of anxiety, alcohol use disorders, depression, sleep disorders, rheumatoid arthritis, other pain conditions, cannabis-related disorders, nicotine-related disorders, other psychoactive disorders, cocaine-related disorders.
Estimated odds ratio are shown in the following figure…

We applied the robust-ODAL method to study the risk factors of acute myocardial infarction (AMI) in a population with pharmaceutically treated major depressive disorder using data from five insurance claims databases from Janssen Research and Development at the Johnson & Johnson. The databases have been converted to the OMOP Common Data Model.

• The outcome, AMI, was defined as the occurrence of the respective diagnosis codes in an inpatient or emergency room setting.
• The risk factors we included in the logistic model: obesity, alcohol dependence, hypertensive disorder, major depressive disorder, type 2 diabetes, and hyperlipidemia.

Comparison between the log odds ratio estimates from the ODAL (yellow), robust-ODAL (green), meta-analysis (blue), and pooled analysis (black) with data from OHDSI network for AMI as the outcome and CCAE as the local site.

Data are heterogeneously distributed across the five datasets. As a consequence, it is believed that fitting a joint logistic regression model across all sites might lead to bias as it ignores the difference between the sites. And the estimates from the pooled analysis are possibly biased. Our proposed robust-ODAL algorithm is designed to account for such heterogeneity and as a result, it is shown to have the widest confidence interval, which properly reflects the potential impact of heterogeneity.

We used ODAC to study the risk factors of acute myocardial infarction (AMI) and stroke in a population with pharmacologically treated major depressive disorder using data from four different US insurance claims databases from Janssen Research and Development at the Johnson & Johnson. The databases have been converted to the OMOP Common Data Model.

• Both AMI and stroke were defined as the occurrence of the respective diagnosis codes in an inpatient or emergency room setting. We only counted the first occurrence of each condition per patient in order to preserve independence.
• For AMI we included known observed risk factors: age, gender, alcohol dependence, hyperlipidemia, hypertensive disorder, depression, obesity, and type II diabetes.
• For stroke the following known risk factors were included: congestive heart failure, coronary arteriosclerosis, hyperlipidemia, ischemic heart disease, renal failure, hypertensive disorder, transient cerebral ischemia, and type II diabetes.

(A) Estimated log hazard ratios with 95% confidence intervals for risk factors of acute myocardial infarction using pooled analysis (red), meta-analysis (green), and One-shot Distributed Algorithm to fit a multicenter Cox proportional hazards model (ODAC) (blue). (B) Estimated log hazard ratios with 95% confidence intervals for risk factors of stroke using pooled analysis on the combined dataset across all sites (red), meta-analysis (green), and ODAC (blue).

A collabratobe study was conducted invoving five participating sites of OneFlorida to study the associations between ADRD and several clinical risk factors among individuals who were 65 years of age, and had no ADRD diagnosis before 2014/03/01 (i.e., the index date) were considered as “at-risk population” and were included in our analysis. The outcome of interest is the time to the first diagnosis of ADRD.

We identified a set of risk factors from the literature and extracted the factors from patients’ medical records in OneFlorida. All records before the index date (2014/03/01) were taken into consideration in this analysis. Factors such as demographic variables (age, race, gender, and insurance type), vital signs (body mass index [BMI], lipid panel), smoking status, selected clinical diagnoses, and medications were included. Since the laboratory test results (e.g., complete blood count) had high frequencies of missing values (>50% in the total study population), these factors were removed. Besides, the clinical diagnoses that were made in <1% of the total study population were removed to minimize potential bias introduced by the small sample size.

Cox regression model chosen to model the outcome variable and the risk factors.

The above figure displays the estimated log hazard ratio and 95% confidence interval for each risk factor using the three methods: ODAC (green), meta-analysis (light pink), and pooled estimate (black). ODAC estimates are closer to the pooled estimates for 11 out of the 13 risk factors we considered in the Cox regression model.