はじめに
ACHILLES (大規模長期的証拠システムにおける医療情報の自動特性評価) は、CDM データベースの 記述統計 を作成するツールです。 ACHILLES の結果は ATLAS データ ソースに表示され、分析前にデータを理解するのに役立ちます。
Vai trò ACHILLES trong workflow:
ETL → CDM Database → ┬→ ACHILLES ─→ Data Sources (ATLAS)
│ "Dữ liệu trông thế nào?"
│
├→ DQD ──────→ Data Quality Report
│ "Dữ liệu có đúng không?"
│
└→ Analysis ─→ Cohort, Estimation, Prediction
"Dữ liệu nói gì?"
1. ACHILLESをインストールする
1.1 要件
- R ≥ 4.0
- Java Runtime (cho DatabaseConnector / JDBC)
- CDM database đã có data
- WebAPI đã cấu hình source
1.2 R パッケージのインストール
# Cài từ GitHub
install.packages("remotes")
remotes::install_github("OHDSI/Achilles")
# Dependencies sẽ tự cài:
# DatabaseConnector, SqlRender, ParallelLogger
1.3 接続の構成
library(Achilles)
library(DatabaseConnector)
connectionDetails <- createConnectionDetails(
dbms = "postgresql",
server = "localhost/ohdsi",
user = "ohdsi_app",
password = keyring::key_get("ohdsi_password"),
port = 5432
)
# Kiểm tra kết nối
conn <- connect(connectionDetails)
querySql(conn, "SELECT COUNT(*) FROM cdm.person")
disconnect(conn)
2. アキレスを実行します。
2.1 フルラン
achilles(
connectionDetails = connectionDetails,
cdmDatabaseSchema = "cdm",
resultsDatabaseSchema = "results",
vocabDatabaseSchema = "cdm",
sourceName = "Hospital_VN_2024",
cdmVersion = "5.4",
createTable = TRUE,
smallCellCount = 5, # ẩn cell < 5 (privacy)
numThreads = 4,
tempEmulationSchema = NULL
)
2.2 ACHILLES テーブルの作成
-- ACHILLES tạo 2 bảng chính trong results schema:
-- 1. achilles_results: aggregate statistics
SELECT * FROM results.achilles_results LIMIT 5;
-- analysis_id | stratum_1 | stratum_2 | count_value
-- 1 | NULL | NULL | 150000 ← tổng patients
-- 2 | NULL | NULL | 2500000 ← tổng visits
-- 101 | 8507 | NULL | 78000 ← male count
-- 101 | 8532 | NULL | 72000 ← female count
-- 200 | 201826 | NULL | 45000 ← Type 2 DM count
-- 2. achilles_results_dist: distribution statistics
SELECT * FROM results.achilles_results_dist LIMIT 3;
-- analysis_id | stratum_1 | min | p10 | p25 | median | p75 | p90 | max
-- 103 | NULL | 0 | 12 | 25 | 45 | 62 | 78 | 105 ← age dist
-- 203 | 201826 | 1 | 10 | 30 | 90 | 180 | 365 | 3650 ← DM duration
2.3 重要な ID を分析する
Analysis ID | Mô tả
─────────────┼─────────────────────────────────────────
1 | Number of persons
2 | Number of visits
101 | Gender distribution
103 | Age at first obs distribution
108 | Obs period length distribution
200 | Condition occurrence counts by concept
400 | Condition era counts
700 | Drug exposure counts by concept
800 | Observation counts
1800 | Measurement counts by concept
2100 | Procedure counts by concept
3. ATLAS の ACHILLES レポート
3.1 データソースのアクティブ化
Sau khi chạy ACHILLES:
1. WebAPI tự đọc bảng achilles_results
2. ATLAS → Data Sources → chọn source
Nếu chưa thấy:
WebAPI → source config → achillesResultsSchema = "results"
Refresh WebAPI cache
3.2 概要ダッシュボード
ATLAS → Data Sources → Hospital_VN_2024
┌─────────────────────────────────────────────────────────┐
│ Hospital_VN_2024 — Data Source Report │
│ │
│ Persons: 150,000 │
│ Records: 12,500,000 │
│ Obs Period: 2015-01-01 → 2024-12-31 │
│ │
│ Tabs: │
│ [Dashboard] [Conditions] [Drugs] [Procedures] │
│ [Measurements] [Observations] [Visits] [Death] │
│ │
│ Gender: │
│ ████████████████ Male (52%) │
│ ██████████████ Female (48%) │
│ │
│ Age Distribution: │
│ 0-9: ██ 8% │
│ 10-19: ███ 12% │
│ 20-29: █████ 15% │
│ 30-39: ██████ 18% │
│ 40-49: ████████ 20% │
│ 50-59: ██████ 16% │
│ 60-69: ███ 7% │
│ 70+: ██ 4% │
└─────────────────────────────────────────────────────────┘
3.3 状態レポート
ATLAS → Data Sources → Conditions
Top 20 Conditions:
┌────┬──────────────────────────────────┬─────────┬────────┐
│ # │ Condition │ Persons │ % │
├────┼──────────────────────────────────┼─────────┼────────┤
│ 1 │ Essential hypertension │ 45,000 │ 30.0% │
│ 2 │ Type 2 diabetes mellitus │ 35,000 │ 23.3% │
│ 3 │ Hyperlipidemia │ 28,000 │ 18.7% │
│ 4 │ Upper respiratory infection │ 25,000 │ 16.7% │
│ 5 │ Osteoarthritis │ 18,000 │ 12.0% │
│ 6 │ Chronic kidney disease │ 12,000 │ 8.0% │
│ 7 │ Ischemic heart disease │ 10,000 │ 6.7% │
│ 8 │ Depression │ 8,500 │ 5.7% │
└────┴──────────────────────────────────┴─────────┴────────┘
Drill-down per condition:
- Prevalence by month (trend over time)
- Age at first diagnosis distribution
- Type (inpatient / outpatient / ER)
- Frequency distribution per person
3.4 薬物報告
ATLAS → Data Sources → Drugs
Drug Exposure by Class (ATC):
┌──────────────────────────────────────────────────┐
│ A10 — Drugs for diabetes │
│ ├── Metformin: 25,000 persons │
│ ├── Glimepiride: 8,000 persons │
│ ├── Insulin Glargine: 5,000 persons │
│ └── Empagliflozin: 3,000 persons │
│ │
│ C09 — ACE Inhibitors & ARBs │
│ ├── Enalapril: 15,000 persons │
│ ├── Losartan: 12,000 persons │
│ └── Valsartan: 8,000 persons │
│ │
│ C10 — Lipid Lowering Agents │
│ ├── Atorvastatin: 20,000 persons │
│ ├── Rosuvastatin: 10,000 persons │
│ └── Simvastatin: 5,000 persons │
└──────────────────────────────────────────────────┘
Duration distribution, frequency, dosage patterns
##4.アキレス腱
4.1 データ品質に関する警告
# ACHILLES Heel chạy tự động cùng achilles()
# Kết quả lưu trong bảng achilles_heel_results
# Xem warnings:
conn <- connect(connectionDetails)
heel <- querySql(conn, "
SELECT *
FROM results.achilles_heel_results
ORDER BY record_count DESC
LIMIT 20
")
disconnect(conn)
4.2 ヒール警告の例
┌─────┬──────────────────────────────────────────────┬──────────┐
│ ID │ Warning Message │ Severity │
├─────┼──────────────────────────────────────────────┼──────────┤
│ 1 │ 5,200 records with date before 1900 │ ERROR │
│ 2 │ 800 persons with age > 150 years │ ERROR │
│ 3 │ Condition_occurrence has 12% unmapped codes │ WARNING │
│ 4 │ Drug_exposure end_date < start_date (350) │ ERROR │
│ 5 │ 15% persons with observation < 30 days │ WARNING │
│ 6 │ Measurement has no unit_concept_id (25%) │ WARNING │
│ 7 │ Visit_occurrence has 0 ER visits │ NOTICE │
│ 8 │ Death records only from 2020+ (COVID bias?) │ NOTICE │
└─────┴──────────────────────────────────────────────┴──────────┘
Severity levels:
ERROR: Lỗi nghiêm trọng cần fix trước khi phân tích
WARNING: Vấn đề nên xử lý nhưng không block
NOTICE: Thông tin cần kiểm tra
4.3 かかとの警告への対処
-- Fix ERROR #1: Records with date before 1900
UPDATE cdm.condition_occurrence
SET condition_start_date = NULL
WHERE condition_start_date < '1900-01-01';
-- Fix ERROR #2: Age > 150
DELETE FROM cdm.person
WHERE EXTRACT(YEAR FROM CURRENT_DATE) - year_of_birth > 150;
-- Fix ERROR #4: end_date < start_date
UPDATE cdm.drug_exposure
SET drug_exposure_end_date = drug_exposure_start_date
WHERE drug_exposure_end_date < drug_exposure_start_date;
-- Sau khi fix: chạy lại ACHILLES
5. インクリメンタルモード
5.1 ACHILLES をより速く実行する
# Lần đầu: full run (~2-4 giờ cho 1M patients)
# Lần sau: incremental (chỉ update thay đổi)
achilles(
connectionDetails = connectionDetails,
cdmDatabaseSchema = "cdm",
resultsDatabaseSchema = "results",
sourceName = "Hospital_VN_2024",
cdmVersion = "5.4",
createTable = FALSE, # không tạo lại bảng
updateGivenAnalysesOnly = TRUE,
analysisIds = c(1, 101, 200, 700, 1800) # chỉ update analyses cần
)
5.2 cron による自動化
#!/bin/bash
# /opt/ohdsi/scripts/run_achilles.sh
cd /opt/ohdsi/achilles
Rscript -e '
library(Achilles)
library(DatabaseConnector)
connectionDetails <- createConnectionDetails(
dbms = "postgresql",
server = Sys.getenv("CDM_DB_SERVER"),
user = Sys.getenv("CDM_DB_USER"),
password = Sys.getenv("CDM_DB_PASS"),
port = 5432
)
achilles(
connectionDetails = connectionDetails,
cdmDatabaseSchema = "cdm",
resultsDatabaseSchema = "results",
sourceName = "Hospital_VN_2024",
cdmVersion = "5.4",
numThreads = 4
)
cat("ACHILLES completed:", format(Sys.time()), "\n")
'
# Chạy ACHILLES hàng tuần (Chủ nhật 2:00 AM)
0 2 * * 0 /opt/ohdsi/scripts/run_achilles.sh >> /var/log/achilles.log 2>&1
概要
| 成分 | 機能 |
|---|---|
| アキレスコア | CDM データベースの集計統計を計算する |
| アキレス結果 | 表には、カウント、有病率、 |
| アキレス結果距離 | 分布を含む表 (p10、p25、中央値...) |
| アキレスヒール | データ品質の問題の検出 (エラー/警告/通知) |
| ATLAS データ ソース | ACHILLESの結果をダッシュボード形式で表示 |
| インクリメンタルモード | 高速に実行 - 変更のみを更新 |
次の記事: データ品質ダッシュボード — CDM データ品質の評価