Introduction
PERSON is the central table of the entire OMOP CDM — every clinical table references PERSON person_id. This is where the patient's demographic information is stored.
Each line in PERSON = a unique patient (unique person).
1. PERSON table structure
1.1. Full column list
| Column | Type | Required | Description |
|---|---|---|---|
person_id | INTEGER | ✅ PK | Unique ID for each patient |
gender_concept_id | INTEGER | ✅ | Gender (Standard Concept) |
year_of_birth | INTEGER | ✅ | Year of birth |
month_of_birth | INTEGER | Month of birth | |
day_of_birth | INTEGER | Date of birth | |
birth_datetime | DATETIME | Full date and time of birth | |
race_concept_id | INTEGER | ✅ | Race (Standard Concept) |
ethnicity_concept_id | INTEGER | ✅ | Ethnicity (Standard Concept) |
location_id | INTEGER | FK | Address (reference LOCATION) |
provider_id | INTEGER | FK | Primary Physician (reference PROVIDER) |
care_site_id | INTEGER | FK | Medical facility (reference CARE_SITE) |
person_source_value | VARCHAR(50) | Original patient code from HIS | |
gender_source_value | VARCHAR(50) | Original gender (eg: "Nu", "F") | |
gender_source_concept_id | INTEGER | Original Gender ID Concept | |
race_source_value | VARCHAR(50) | Origin Race | |
race_source_concept_id | INTEGER | Origin Race ID Concept | |
ethnicity_source_value | VARCHAR(50) | Ethnic origin | |
ethnicity_source_concept_id | INTEGER | Concept ethnic ID |
1.2. Entity-Relationship
┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ LOCATION │←──────│ PERSON │──────→│ PROVIDER │
│ location_id │ │ person_id │ │ provider_id │
│ address_1 │ │ gender_* │ │ provider_name│
│ city │ │ birth_* │ │ specialty_* │
│ state │ │ race_* │ └──────────────┘
│ zip │ │ ethnicity_* │ ↑
│ country_* │ │ location_id │ │
└──────────────┘ │ provider_id │ ┌──────┴───────┐
│ care_site_id│──────→│ CARE_SITE │
└──────┬───────┘ │ care_site_id │
│ │ care_site_name│
┌───────────┼───────────┐ └──────────────┘
↓ ↓ ↓
VISIT_OCC. CONDITION DRUG_EXPOSURE
OBSERVATION MEASUREMENT ... (tất cả clinical)
2. Detailed important fields
2.1. person_id
- Type: INTEGER, Primary Key
- Rule: Unique, no change, no clinical significance
- Not the original patient code (original patient code stored in
person_source_value)
-- ĐÚNG: person_id là số tự tăng hoặc hash
person_id = 100001
person_source_value = 'BN-2024-00123' -- Mã gốc từ HIS
-- SAI: Không dùng mã gốc làm person_id
-- person_id = 'BN-2024-00123' ← SAI (phải là INTEGER)
2.2. gender_concept_id
| Concept ID | Concept Name | Description |
|---|---|---|
| 8507 | Male | Male |
| 8532 | Female | Female |
| 8551 | UNKNOWN | Unknown |
| 8521 | OTHER | Other |
-- Ví dụ ETL cho dữ liệu Việt Nam
CASE
WHEN gioi_tinh IN ('Nam', 'M', '1') THEN 8507 -- Male
WHEN gioi_tinh IN ('Nữ', 'Nu', 'F', '2') THEN 8532 -- Female
ELSE 8551 -- UNKNOWN
END AS gender_concept_id,
gioi_tinh AS gender_source_value
2.3. year_of_birth, month_of_birth, day_of_birth
year_of_birth: Required — if not present, do not load patientmonth_of_birth,day_of_birth: Optional — set NULL if not presentbirth_datetime: Optional — useful for pediatrics (accurate age calculation)
-- Ví dụ: BN sinh ngày 15/03/1980
year_of_birth = 1980
month_of_birth = 3
day_of_birth = 15
birth_datetime = '1980-03-15 00:00:00'
2.4. race_concept_id and ethnicity_concept_id
These are two schools according to US Census standards. For Vietnam data:
| School | Recommendations for Vietnam |
|---|---|
race_concept_id | 8515 (Asian) |
race_source_value | "Kinh", "Tay", "Muong"... |
ethnicity_concept_id | 0 (No matching concepts) |
ethnicity_source_value | Record ethnic origin if any |
Note:
raceandethnicityin OMOP according to American standards (OMB). When ETL VN data, we still have to set the value (use 0 if it cannot be mapped) but keep the original information in*_source_value.
3. Real-life example
3.1. Vietnamese patient
INSERT INTO person VALUES (
100001, -- person_id
8532, -- gender_concept_id (Female)
1980, -- year_of_birth
3, -- month_of_birth
15, -- day_of_birth
'1980-03-15 00:00:00', -- birth_datetime
8515, -- race_concept_id (Asian)
0, -- ethnicity_concept_id (N/A)
1001, -- location_id (→ LOCATION table)
5001, -- provider_id (→ PROVIDER table)
2001, -- care_site_id (→ CARE_SITE table)
'BN-2024-00123', -- person_source_value
'Nữ', -- gender_source_value
0, -- gender_source_concept_id
'Kinh', -- race_source_value
0, -- race_source_concept_id
NULL, -- ethnicity_source_value
0 -- ethnicity_source_concept_id
);
3.2. Basic SQL queries
-- Đếm bệnh nhân theo giới tính
SELECT
c.concept_name AS gender,
COUNT(*) AS patient_count
FROM person p
JOIN concept c ON p.gender_concept_id = c.concept_id
GROUP BY c.concept_name;
-- Phân bố tuổi
SELECT
EXTRACT(YEAR FROM CURRENT_DATE) - year_of_birth AS age,
COUNT(*) AS count
FROM person
GROUP BY 1
ORDER BY 1;
-- Tìm bệnh nhân có dữ liệu gốc
SELECT
person_id,
person_source_value AS ma_bn_goc,
gender_source_value AS gioi_tinh_goc,
race_source_value AS dan_toc
FROM person
WHERE person_source_value IS NOT NULL
LIMIT 10;
4. ETL Conventions
4.1. Important rule
| Rules | Details |
|---|---|
| 1 person = 1 record | No duplication, need deduplicate |
| year_of_birth required | Ignore patient if there is no year of birth |
| gender_concept_id required | Put 8551 (UNKNOWN) if unknown |
| person_id has no meaning | Do not use original patient code or ID card/CCCD |
| Does not store PII directly | There are no columns for name, ID card, or phone number in PERSON |
4.2. De-identification
OMOP CDM does not have columns for name, ID/CCCD number, phone number. This is intentional design:
HIS gốc (có PII): OMOP CDM (de-identified):
┌─────────────────────────┐ ┌──────────────────────────┐
│ ma_bn: BN-2024-00123 │ → │ person_id: 100001 │
│ ho_ten: Nguyễn Thị Lan │ → │ (không có cột tên!) │
│ cmnd: 079123456789 │ → │ (không có cột CMND!) │
│ sdt: 0901234567 │ → │ (không có cột SĐT!) │
│ ngay_sinh: 15/03/1980 │ → │ year_of_birth: 1980 │
│ gioi_tinh: Nữ │ → │ gender_concept_id: 8532 │
└─────────────────────────┘ └──────────────────────────┘
Note:
person_source_valuemay contain the original patient code (for tracing). Depending on the organization, this value can be hashed or encrypted.
4.3. Handling duplicates
When a patient has ≥2 codes in multiple systems:
HIS BV Chợ Rẫy: BN-CR-001 ┐
HIS BV Bạch Mai: BN-BM-555 ├──→ person_id = 100001
BHXH: DN-7900123456789 ┘ (1 person duy nhất)
ETL needs to perform Patient Matching (patient matching) before loading into PERSON.
5. Relationships with other tables
-- Tất cả dữ liệu lâm sàng của 1 bệnh nhân
SELECT 'Visits' AS data_type, COUNT(*) AS count
FROM visit_occurrence WHERE person_id = 100001
UNION ALL
SELECT 'Conditions', COUNT(*)
FROM condition_occurrence WHERE person_id = 100001
UNION ALL
SELECT 'Drugs', COUNT(*)
FROM drug_exposure WHERE person_id = 100001
UNION ALL
SELECT 'Measurements', COUNT(*)
FROM measurement WHERE person_id = 100001
UNION ALL
SELECT 'Observations', COUNT(*)
FROM observation WHERE person_id = 100001;
6. Common ETL errors
| Error | Consequences | How to fix |
|---|---|---|
| Duplicate person_id | Overwrite data | Check for unique constraints |
| year_of_birth = NULL | Violation of NOT NULL | Ignore or impute |
| gender_concept_id is wrong | Misgender analysis | Accurate Mapping |
| Put PII in source_value | Violation of de-identification | Hash or remove |
| Do not duplicate | 1 patient becomes many people | Patient matching before ETL |
Summary
- PERSON is the central table — all clinical tables referenced
person_id - Required fields: person_id, gender_concept_id, year_of_birth, race_concept_id, ethnicity_concept_id
- Does not contain PII (name, ID card, phone number) — de-identified design
- Link: LOCATION (address), CARE_SITE (facility), PROVIDER (main doctor)
- ETL VN: gender mapping, race = Asian (8515), ethnicity = 0
Next article: OBSERVATION_PERIOD — why it's important to know the "monitoring period" and how it affects every analysis.