Ascertainment of EMR-based Clinical Covariates Among Patients Receiving Oral and Non-insulin Injected Hypoglycemic Therapy

Association of Clinical Covariates With Non-insulin Diabetes Medication Initiation Using Electronic Medical Records (EMR)

Status

Completed

Phases

Unknown

Study type

Observational

Source

ClinicalTrials.gov

Registry ID

NCT02140645

Enrollment

166613

Registered

2014-05-16

Start date

2014-05-31

Completion date

2015-03-31

Last updated

2017-02-08

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Diabetes Mellitus, Type 2

Brief summary

The objective of this study is to identify EMR-based clinical covariates and quantify their association with the prescribing of each specific type 2 diabetes (T2DM) medication under investigation. This will include an assessment of how well these covariates are captured through claims data proxies, and their potential to confound comparative research of T2DM medications.

Detailed description

Purpose:

Interventions

DRUGlinagliptin

non-randomized

Study design

Observational model

COHORT

Time perspective

RETROSPECTIVE

Eligibility

Sex/Gender

ALL

Age

18 Years to No maximum

Healthy volunteers

Inclusion criteria

* Dispensing of an oral or non-insulin injected hypoglycemic medication between May 2011 and June 2012 * Diagnosis of type 2 diabetes mellitus * Presence of electronic medical records (for the EMR-based subset)

Exclusion criteria

* Age \<18 at T2DM medication initiation * Missing or ambiguous age or sex information * At least one diagnosis of type 1 diabetes mellitus * Less than 6 months enrolment in the database preceding the date of the first dispensing * Prior use of the index drug

Design outcomes

Primary

Measure	Time frame	Description
Missing EMR (Electronic Medical Record) Characteristic: Smoking	Up to 20 months	The missing EMR characteristic smoking defined as current, unknown, versus past/never smoker. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic smoking was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: Duration of Diabetes	Up to 20 months	The missing EMR characteristic duration of diabetes defined as \>7, 5-6, 3-5, 1-3, \<1 (in years) in duration. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic duration of diabetes was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: Duration of Diabetes (Continuous)	Up to 20 months	The missing EMR characteristic duration of diabetes defined as starting year/starting age of diabetes. Linear regression models were ran using a prioritized list of claims-based covariates as predictors and the value of select EMR-based clinical characteristics duration of diabetes as continuous outcomes. The estimated value represented is actually prediction accuracy defined by R-squared.
Missing EMR Characteristic: BMI (Body Mass Index)	Up to 20 months	The missing EMR characteristic BMI defined as not obese, overweight, obese, severe obesity. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic BMI was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: BMI (Continuous)	Up to 20 months	The missing EMR characteristic BMI is BMI value. Linear regression models were ran using a prioritized list of claims-based covariates as predictors and the value of select EMR-based clinical characteristics BMI as continuous outcomes. The estimated value represented is actually prediction accuracy defined by R-squared.
Missing EMR Characteristic: HbA1c (Hemoglobin A1c (Glycosylated Hemoglobin))	Up to 20 months	The missing EMR characteristic HbA1c defined as value in 6 months prior to and including index date. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic HbA1c was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: eGFR (Glomerular Filtration Rate)	Upto 20 months	The missing EMR characteristic eGFR defined as value in 6 months prior to and including index date. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic eGFR was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: Total Cholesterol	Up to 20 months	The missing EMR characteristic total cholesterol defined as value in 6 months prior to and including index date. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic total cholesterol was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: Systolic BP (Blood Pressure)	Up to 20 months	The missing EMR characteristic systolic BP defined as value in 6 months prior to and including index date. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic systolic BP was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Missing EMR Characteristic: Diastolic BP	Up to 20 months	The missing EMR characteristic diastolic BP defined as value in 6 months prior to and including index date. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic diastolic BP was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Binary EMR Characteristic: Neuropathy	Up to 20 months	The missing EMR characteristic neuropathy defined as participants with any note of diabetic neuropathy. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic neuropathy was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Binary EMR Characteristic: Nephropathy	Upto 20 months	The missing EMR characteristic nephropathy defined as participants with any note of diabetic nephropathy. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic nephropathy was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Binary EMR Characteristic: Retinopathy	Up to 20 months	The missing EMR characteristic retinopathy defined as participants with any note of diabetic retinopathy. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic retinopathy was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.
Binary EMR Characteristic: Pancreatitis	Up to 20 months	The missing EMR characteristic pancreatitis defined as participants with any note of prior pancreatitis. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic pancreatitis was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.

Countries

United States

Participant flow

Recruitment details

Existing data cohort design using data from the MarketScan database from May 2011 through December 2012. 492963 potential patients were identified in the database, but after removing patients who violated inclusion and exclusion criteria 166613 patients were actually analysed in the study.

Participants by arm

Arm	Count
Linagliptin 1 Patients had a recorded diagnosis of type 2 diabetes mellitus (T2DM) using an oral and non-insulin injected glucose-lowering medication Linagliptin subjects matched to any other DPP-4 (Dipeptidyl peptidase-4 inhibitors).	243
Any Other DPP-4 Patients had a recorded diagnosis of type 2 diabetes mellitus (T2DM) using an oral and non-insulin injected glucose-lowering medication DPP-4 (Dipeptidyl peptidase-4 inhibitors).	3,041
Linagliptin 2 Patients had a recorded diagnosis of type 2 diabetes mellitus (T2DM) using an oral and non-insulin injected glucose-lowering medication Linagliptin subjects matched to pioglitazone.	205
Pioglitazone Patients had a recorded diagnosis of type 2 diabetes mellitus (T2DM) using an oral and non-insulin injected glucose-lowering medication pioglitazone.	530
Linagliptin 3 Patients had a recorded diagnosis of type 2 diabetes mellitus (T2DM) using an oral and non-insulin injected glucose-lowering medication Linagliptin subjects matched to second generation sulfonylurea.	150
Second Generation Sulfonylurea Patients had a recorded diagnosis of type 2 diabetes mellitus (T2DM) using an oral and non-insulin injected glucose-lowering medication second generation sulfonylurea.	3,050
Total	7,219

Withdrawals & dropouts

Period	Reason	FG000	FG001	FG002	FG003	FG004	FG005
Overall Study	Not linked	5,489	61,654	4,236	13,833	3,286	70,896

Baseline characteristics

Characteristic	Linagliptin 1	Any Other DPP-4	Linagliptin 2	Pioglitazone	Linagliptin 3	Second Generation Sulfonylurea	Total
Age, Continuous	56.1 Years STANDARD_DEVIATION 11.6	54.8 Years STANDARD_DEVIATION 11.2	55.2 Years STANDARD_DEVIATION 11.7	54.7 Years STANDARD_DEVIATION 11.1	54.7 Years STANDARD_DEVIATION 11.8	54.8 Years STANDARD_DEVIATION 11.7	54.9 Years STANDARD_DEVIATION 11.5
Gender Female	118 Participants	1310 Participants	101 Participants	194 Participants	74 Participants	1332 Participants	3129 Participants
Gender Male	125 Participants	1731 Participants	104 Participants	336 Participants	76 Participants	1718 Participants	4090 Participants

Adverse events

Event type	EG000 affected / at risk
deaths Total, all-cause mortality	— / —
other Total, other adverse events	0 / 0
serious Total, serious adverse events	0 / 0

Outcome results

Primary

Binary EMR Characteristic: Nephropathy

The missing EMR characteristic nephropathy defined as participants with any note of diabetic nephropathy. The associations between claims-based covariates and missingness on EMR characteristics were investigated by estimating a logistic regression model (and multinomial logistic regression, depending on the number of categories for the EMR characteristic) for each EMR characteristic where an indicator for missing the EMR characteristic nephropathy was the dependent variable and all claims-based covariates were included as independent variables. The estimated value represented is actually prediction accuracy defined by C-statistics.

Time frame: Upto 20 months