OHDSI : Design and implementation of a comparative cohort study in observational health care data 아주대학교 의료정보학과 유승찬

OHDSI (Observational Health Data Sciences and Informatics ) The odyssey to evidence generation

OHDSI (Observational Health Data Sciences and Informatics ) The odyssey to evidence generation Patient-level data in source system/schema

OHDSI (Observational Health Data Sciences and Informatics ) The odyssey to evidence generation Patient-level data in source system/schema Evidence

OHDSI : a global community adopting OMOP CDM OHDSI.org

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation Reproducibility Community Collaboration Openness Beneficence

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation : we encourage fresh methodological approaches from disruptive thinking Reproducibility Community Collaboration Openness Beneficence

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation Reproducibility : we sought reproducible and well–calibrated evidence avoiding bias Community Collaboration Openness Beneficence

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation Reproducibility Community : everyone is welcome to participate in OHDSI Collaboration Openness Beneficence

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation Reproducibility Community Collaboration : we work collectively Openness Beneficence

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation Reproducibility Community Collaboration Openness : all our community’s process open Beneficence

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation Reproducibility Community Collaboration Openness Beneficence : we seek to protect the rights of individuals and organizations

Why OHDSI? OHDSI projects was initiated to improve health, by empowering a community to collaboratively generate the evidence with adopting OMOP CDM to aggregate health data across world Objectives Innovation : we encourage fresh methodological approaches from disruptive thinking Reproducibility : we sought reproducible and well–calibrated evidence avoiding bias Community : everyone is welcome to participate in OHDSI Collaboration : we work collectively Openness : all our community’s process open Beneficence : we seek to protect the rights of individuals and organizations

Vision of OHDSI OHDSI collaborators access a network of 1,000,000,000 patients to generate evidence about all aspects of healthcare.  Patients and clinicians and other decision-makers around the world use OHDSI tools and evidence every day

What evidence does OHDSI seek to generate from observational data? Clinical characterization Natural history: Who are the patients who have diabetes? Among those patients, who takes metformin? Quality improvement: What proportion of patients with diabetes experience disease-related complications? Population-level estimation Safety surveillance: Does metformin cause lactic acidosis? Comparative effectiveness: Does metformin cause lactic acidosis more than glyburide? Patient-level prediction Precision medicine: Given everything you know about me and my medical history, if I start taking metformin, what is the chance that I am going to have lactic acidosis in the next year? Disease interception: Given everything you know about me, what is the chance I will develop diabetes?

A standardized process for evidence generation and dissemination

A standardized process for evidence generation and dissemination

Counterfactual reasoning for one person

Counterfactual reasoning for one person

Counterfactual reasoning for a population

Randomized treatment assignment to approximate counterfactual outcomes

An observational comparative cohort design to approximate counterfactual outcomes

Propensity score e(x) = Pr(Z=1|x) Z is treatment assignment x is a set of all covariates at the time of treatment assignment Propensity score = probability of belonging to the target cohort vs. the comparator cohort, given the baseline covariates Propensity score can be used as a ‘balancing score’: if the two cohorts have similar propensity score distribution, then the distribution of covariates should be similar (need to perform diagnostic to check) Rubin Biometrika 1983

Process flow for formally defining a cohort in ATLAS Cohort entry criteria Initial events Events are recorded time-stamped observations for the persons, such as drug exposures, conditions, procedures, measurements and visits All events have a start date and end date, though some events may have a start date and end date with the same value (such as procedures or measurements). Initial event inclusion criteria Additional qualifying inclusion criteria The qualifying cohort will be defined as all persons who have an initial event, satisfy the initial event inclusion criteria, and fulfill all additional qualifying inclusion criteria Each qualifying inclusion criteria will be evaluated to determine the impact of the criteria on the attrition of persons from the initial cohort Cohort exit criteria Initial cohort Qualifying cohort

A database is full of cohorts, some of which may represent valid comparisons

What are the key inputs to a comparative cohort design? Input parameter Design choice Target cohort (T) Comparator cohort (C) Outcome cohort (O) Time-at-risk Model specification

Cohort restriction in comparative cohort analyses

The choice of the outcome model defines your research question

Design an observational study like you would a randomized trial Input parameter Design choice Target cohort (T) Comparator cohort (C) Outcome cohort (O) Time-at-risk Model specification

Before starting Read CDM specification carefully ohdsi.org->wiki->documentation->CDM specification http://www.ohdsi.org/web/wiki/doku.php?id=documentation:cdm ohdsi.org->wiki->documentation->CDM specification

Creating cohort 예제) B-viral hepatitis 치료를 위하여 Tenofovir (vs. Entecavir) 초치료 받은 환자 군 -tenofovir 180일 이상의 연속 처방 필요 -나이 30~80세 추가 Inclusion criteria -HDV, HCV 등 기타 급성 간염 환자 제외 -HIV 진단 환자 제외 (전체 follow-up 기간 동안에도) -사망 환자 및 이전 간이식 환자 제외 -간암 환자 제외 -응급실 또는 입원 진단으로 골절/골다공증/신부전/암 진단 받은 환자 제외 ohdsi.org->wiki->documentation->CDM specification