Journal: Drug Safety

Title:

A Comparative Assessment of Observational Medical Outcomes Partnership and Mini-Sentinel Common Data Models and Analytics: Implications for Active Drug Safety Surveillance

Running Header:

Comparison of Mini-Sentinel and Observational Medical Outcomes Partnership Common Data Models

Authors:

Yihua Xu, PhD 1

Xiaofeng Zhou, PhD 2

Brandon T. Suehs, PharmD, PhD 1

Abraham G. Hartzema, PharmD, MSPH, PhD 3

Michael G. Kahn, MD, PhD 4

Yola Moride, PhD, FISPE 5

Brian C. Sauer, PhD, MS 6

Qing Liu, MS 2

Keran Moll, MS, PhD 1

Margaret Pasquale, PhD 1

Vinit Nair, BPharm, MS 1

Andrew Bate, PhD 2

Affiliations:

1. Comprehensive Health Insights, Humana Inc., Louisville, KY, USA

2. Pfizer Inc., New York, NY, USA

3. College of Pharmacy, University of Florida, Gainesville, FL, USA

4. Department of Pediatrics, University of Colorado, CO, USA

5. Faculty of Pharmacy, Université de Montreal, QC, Canada

6. University of Utah, UT, USA

Corresponding author:

Brandon Suehs

Comprehensive Health Insights, Inc.

515 W. Market St.

Louisville, KY40202, USA

Phone: (+1) 512 419-9849

Fax: (+1) 210 591-6542

Email:

Table 1. High Dimensional Propensity Score parameters specified for Mini-Sentinel and Observational Medical Outcomes Project implementation

Parameter / OMOP / MSa
Cohort / New user / New user (Min)
Washout period / 180 day / 180 day
HOI / Incident / Incident (MULT)
DOI/Comparator Data / Drug era / Drug era (created by MP3 )
Gap period for Era table/Treatment episode / 30 day / 30 day
Risk period / Fixed Risk Period: 30 days from index date
Variable Risk Period: exposure duration plus 30 days from the exposure end date / ITT 30 day/AT 30 day
Covariate selection / Empirical / Empirical (HDPS only)
Propensity score (PS) application / PS adjustment using logistic regression
Stratified on PS deciles (sensitivity analysis) / Stratified on PS deciles (full cohort)

Abbreviations: HOI, health outcome of interest; DOI, drug of interest; HDPS, high dimensional propensity score; PS, propensity score; ITT, intent to treat; AT, as treated.

a, The MS CDM does not have drug era built in the CDM structure; however MP3 has an algorithm to aggregate drug exposures to treatment episodes (similar to OMOP Drug Era) with user supplied gap period options.

Min (minimum incidence) -- identifies and reports metrics only for the first valid incident treatment episode during the query period, but uses ALL prior observed claims to define incidence. That is, incidence is the first exposure during the query period with no evidence of prior exposure using all available data.

MULT (multiple incidence) -- identifies and reports metrics for all valid incident events during the query period, and uses claims observed during the specified WASHPER (wash-out period) days to determine incidence. That is, incidence is defined as any event during the query period with no evidence of an event during the prior WASHPER days.

ITT (Intention to Treat): A fixed follow period of 30 days from index date; AT (As Treated): follow up period included the drug exposure period (following index date) plus 30 days.

Note: ITT and AT are used in this table based on the use of terminology in the MS software package and documentation. ITT and AT in this setting are not to be confused with use of the terminology in other settings (e.g., randomized clinical trial).

HDPS only: empirically (i.e., HDPS) selected covariates only for the PS model.

Table 2. Examples of data fields not included in one or both common data models

Only In OMOP / Only In MS / Not in either CDM / Source Table / Data field details
Dx_Codetype / MS Diagnosis / ICD Code Versioning (09 = ICD-9-CM; 10 = ICD-10-CM; 11 = ICD-11-CM; OT = Other)
PX_Codetype / MS Procedure / Procedure code type flag (ICD9, ICD10, CPT. HCPCS, etc.
procedure_type_concept_id / OMOP Procedure Occurrence / A foreign key to the predefined concept identifier in the vocabulary reflecting the type of the procedure. Different from MS PX_codeType
relevant_condition_concept_id / OMOP Procedure Occurrence / A foreign key to the predefined concept identifier in the vocabulary reflecting the condition that was the cause for initiation of the procedure. Note that this is not a direct reference to a specific condition record in the condition table, but rather a condition concept in the vocabulary.
Admitting_Source / MS Encounter / Source of admission (home, Ambulatory Visit, ER, etc.)
Discharge_Disposition / MS Encounter / A = Discharged alive ;E = Expired; U = Unknown
Discharge_Status / MS Encounter / Home, Other hospital, nursing home, etc.
DRG / MS Encounter / 3-digit Diagnostic Related Group
DRG_Type / MS Encounter / DRG Code Versioning (1 = CMS-DRG [old system], 2 = MS-DRG [current system])
drug_exposure_end_date / OMOP RX Exposure / The end date for the current instance of drug utilization. It is not available from all sources.
drug_type_concept_ id / OMOP RX Exposure / A foreign key to the predefined concept identifier in the vocabulary reflecting the type of drug exposure recorded. It indicates how the drug exposure was represented in the source data: as medication history, filled prescriptions, etc.
stop_reason / OMOP RX Exposure / The reason the medication was stopped, where available. Reasons include regimen completed, changed, removed, etc.
refills / OMOP RX Exposure / The number of refills after the initial prescription. The initial prescription is not counted, values start with 0.
prescribing_provider_id / OMOP RX Exposure / A foreign key to the provider in the provider table who initiated (prescribed) the drug exposure.
visit_occurrence_id / OMOP RX Exposure / A foreign key to the visit in the visit table during which the drug exposure is initiated.
relevant_condition_concept_id / OMOP RX Exposure / A foreign key to the predefined concept identifier in the vocabulary reflecting the condition that was the cause for initiation of the drug exposure. Note that this is not a direct reference to a specific condition record in the condition table, but rather a condition concept in the vocabulary.
payer_source_value / OMOP Payer Plan Period / The source code for the payer as it appears in the source data.
plan_source_value / OMOP Payer Plan Period / The source code for the person's coverage plan as it appears in the source data.
family_source_value / OMOP Payer Plan Period / The source code for the person's family as it appears in the source data.
MedCov / MS Enrollment
DrugCov / MS Enrollment
location_id / OMOP Person / A foreign key to the place of residency for the person in the location table, where the detailed address information is stored.
provider_id / OMOP Person / A foreign key to the primary care provider the person is seeing in the provider table.
care_site_id / OMOP Person / A foreign key to the primary care site in the care site table, where the details of the care site are stored.
LOW_INCOME_SUBSIDY_CD / Source data (Rx, demographic)
GENERIC INDICATOR / Source data (Rx) / Identifies if a prescription is a brand or generic drug
GENERIC AVAIL CD / Source data (Rx) / Indicates the generic availability
RX ORIG CD / Source data (Rx) / Prescription origin code - not specified, written, telephone, electronic
PHARMACY ID / Source data (Rx) / Novel field which identifies each individual pharmacy.
TIER CODE / Source data (Rx) / Defines and classifies the Rx tier
PRIMARY DIAG POA CD / Source Data (Medical) / Represents presence of diagnosis code at time of admission for inpatient services or acute care hospitals.
SECONDARY DIAG POA CD / Source Data (Medical) / Represents presence of diagnosis code at time of admission for inpatient services or acute care hospitals.
ADMISSION DIAG CODE / Source Data (Medical) / Member's admitting ICD-9 diagnostic code for admission to an inpatient facility
ADMISSION TYPE / Source Data (Medical) / Indicates level of service at admission of the patient (emergency, trauma, elective, etc.)
READMIT IND / Source Data (Medical) / Indicates whether the claim is considered a readmit of a previous claim (for inpatient admissions)
READMIT DRG MATCH IND / Source Data (Medical) / Indicates whether the DRG between the originator and readmit is the same
CLAIM TYPE / Source Data (Medical) / Indicates whether the claim is a facility or professional claim

Table 3. Mean and median exposure time for drug of interest and comparator drugs (in days)

Pair / HOI / DOI / Mean / Median / Comparator Drug / Mean / Median
1 / AMI / Indomethacin / 35 / 16 / Loratadine / 73 / 31
2 / GI bleeding / Ketorolac / 9 / 1 / Loratadine / 73 / 31
3 / Hip Fracture / Benzodiazepines / 131 / 31 / Lisinopril / 355 / 194
4 / Acute Liver Injury / Valproic acid / 230 / 96 / Fluticasone / 76 / 31
5 / Acute Liver Injury / Carbamazepine / 283 / 95 / Fluticasone / 76 / 31
6 / Anaphylactic Shock / Amoxicillin / 12 / 11 / Levothyroxine / 471 / 302

Table 4. High Dimensional Propensity Score based analytic procedure run time in hours

OMOP HDPS / MS HDPS
DOI-HOI / Risk period: / Risk period: / Risk period: / Risk period:
exposure +30 days / fixed 30 days / exposure + 30 days / fixed 30 days
Indomethacin – AMI / 1.1 / 1.2 / 2.3 / 2.2
Ketorolac – GI Bleeding / 1.0 / 0.9 / 1.8 / 1.7
Benzodiazepine – Hip Fracture† / 1.3 / 0.8 / 6.4 / 11.9
Valproic acid – Acute Liver Injury / 0.9 / 0.9 / 41.5 / 40.7
Carbamazepine – Acute Liver Injury† / 0.9 / 0.9 / 10.4 / 14.2
Amoxicillin – Anaphylactic Shock† / 1.4 / 0.9 / 6.2 / 5.7

Program run time determined from SAS log files.

† Both OMOP and MS used subset data

Table 5. Observational Medical Outcomes Project High Dimensional Propensity Score based analytic procedure results – propensity score adjusted vs. propensity score stratified odds ratios (fixed risk period*)
Pair / HOI / DOI / Comparator Drug / Adjusted OR** / Stratified OR***
1 / AMI / Indomethacin / Loratadine / 1.57 (1.14, 2.15) / 1.54 (1.13, 2.11)
2 / GI Bleeding / Ketorolac / Loratadine / 2.49 (1.68, 3.68) / 2.47 (1.67, 3.65)
3 / Hip Fracture / Benzodiazepines / Lisinopril / 1.47 (1.18, 1.84) / 1.45 (1.17, 1.80)
4 / Acute Liver Injury / Valproic acid / Fluticasone / 1.67 (0.72, 3.87) / 1.63 (0.72, 3.73)
5 / Acute Liver Injury / Carbamazepine / Fluticasone / 1.49 (0.20, 11.24) / 1.47 (0.20, 10.74)
6 / Anaphylactic Shock / Amoxicillin / Levothyroxine / 1.68 (0.81, 3.47) / 1.63 (0.80, 3.34)

*Fixed risk period 30 days starting from index date

** OR based on PS 10 group adjustment using logistic regression model

*** OR based on PS decile stratification using Mantel-Haenszel method

Table 6. Observational Medical Outcomes Project High Dimensional Propensity Score based analytic procedure results – propensity score adjusted vs. propensity score stratified odds ratios (variable risk period*)
Pair / HOI / DOI / Comparator Drug / Adjusted OR** / Stratified OR***
1 / AMI / Indomethacin / Loratadine / 0.77 (0.63, 0.95) / 0.77 (0.63, 0.95)
2 / GI Bleeding / Ketorolac / Loratadine / 0.93 (0.72, 1.21) / 0.93 (0.72, 1.21)
3 / Hip Fracture / Benzodiazepine / Lisinopril / 0.76 (0.68, 0.85) / 0.77 (0.69, 0.86)
4 / Acute Liver Injury / Valproic acid / Fluticasone / 1.82 (1.08, 3.05) / 1.76 (1.07, 2.91)
5 / Acute Liver Injury / Carbamazepine / Fluticasone / 1.72 (0.62, 4.79) / 1.69 (0.62, 4.62)
6 / Anaphylactic Shock / Amoxicillin / Levothyroxine / 0.26 (0.17, 0.41) / 0.27 (0.17, 0.43)

*Variable risk period =Drug exposure time plus 30 days from the exposure end date

** OR based on PS 10 group adjustment using logistic regression model

*** OR based on PS decile stratification using Mantel-Haenszel method