cross-national

name: cross-national description: End-to-end cross-national comparison study using KNHANES + NHANES + CHNS (or other parallel surveys). Variable harmonization, parallel weighted analysis, and comparison tables. Supports 2-country (KR+US) and 3-country (KR+US+CN) designs. triggers: cross-national, 한미 비교, Korea US comparison, KNHANES NHANES, 양국 비교, binational, cross-country, 비교연구, 3국 비교, CHNS, 한미중 tools: Read, Write, Edit, Bash, Grep, Glob model: opus

Cross-National Comparison Study Skill

You are assisting a medical researcher in conducting a cross-national comparison study using parallel nationally representative surveys (e.g., KNHANES for Korea, NHANES for the US, CHNS for China).

When to Use

• Researcher has a clinical question to compare across two countries
• KNHANES + NHANES data available (or other parallel survey pairs)
• Goal: produce a complete analysis with country-stratified results + comparison table

Inputs

1. Research question: exposure → outcome association to compare across countries
2. Korean data path: KNHANES CSV file
3. US data path: NHANES CSV directory (multiple tables to merge)
4. Harmonization table (optional): CSV mapping variables across surveys
   • Default: replicate-study skill's harmonization_knhanes_nhanes.csv

Reference Files

• Harmonization table: medsci-skills/skills/replicate-study/references/harmonization_knhanes_nhanes.csv
• Upstream:
  • medsci-skills/skills/write-paper/references/paper_types/cross_national.md — writing template
  • medsci-skills/skills/analyze-stats/references/analysis_guides/survey_weighted.md

Workflow

Phase 1: Study Definition

1. Confirm research question: Exposure → Outcome
2. Define variable coding for both countries:
   • Exposure: PHQ-9, BMI category, smoking, etc.
   • Outcome: diabetes, hypertension, mortality, etc.
   • Covariates: age, sex, education, income, smoking, alcohol, obesity, CVD
3. Check harmonization table for variable availability
4. Output: study protocol summary for user approval

Phase 2: Data Preparation

KNHANES (single CSV):

1. Load CSV, filter age ≥20 (or per protocol)
2. Derive variables using KNHANES coding:
   • Smoking: BS3_1 (1,2=current, 3=former, 8=never)
   • Alcohol: BD1_11 (2-6=frequent, 1=occasional, 8=never)
   • Obesity: HE_obe (≥4=obesity for BMI≥25 Asian cutoff)
   • PHQ-9: BP_PHQ_1~9, sum score, ≥10=depression
   • Diabetes: HE_glu≥126 | HE_HbA1c≥6.5 | DE1_dg=1
   • CVD: DI4_dg=1 | DI5_dg=1 | DI6_dg=1
3. Set survey design: svydesign(id=psu, strata=kstrata, weights=~wt_itvex, nest=TRUE)

NHANES (multiple CSVs):

1. Load and merge tables by SEQN (DEMO_J, DPQ_J, GHB_J, BIOPRO_J, BMX_J, SMQ_J, ALQ_J, DIQ_J, MCQ_J, BPQ_J)
2. Derive variables using NHANES coding:
   • Smoking: SMQ020 + SMQ040 (100 cigs + now smoke)
   • Alcohol: ALQ121 (past 12 mo frequency → categories)
   • Obesity: BMXBMI ≥30 (WHO cutoff, NOT Asian)
   • PHQ-9: DPQ010~DPQ090, sum score, ≥10=depression
   • Diabetes: LBXSGL≥126 | LBXGH≥6.5 | DIQ010=="Yes" (CRITICAL: LBXSGL not LBXSGLU)
   • CVD: MCQ160B=="Yes" (CHF) | MCQ160C=="Yes" (CHD) | MCQ160D=="Yes" (angina) | MCQ160E=="Yes" (MI)
   • HTN: BPXOSY3≥140 | BPXODI3≥90 | BPQ020=="Yes"
3. Set survey design: svydesign(id=SDMVPSU, strata=SDMVSTRA, weights=~WTMECPRP, nest=TRUE)

Phase 3: Parallel Analysis

For EACH country independently:

1. Table 1: Baseline characteristics by exposure (weighted counts + percentages)
2. Main analysis: Sequential logistic regression models
   • Model 1 (unadjusted)
   • Model 2 (age + sex)
   • Model 3 (fully adjusted: + education, income, smoking, alcohol, obesity, CVD)
3. Subgroup analyses: By sex, age group, education, income, alcohol, smoking, CVD, obesity
4. Dose-response (if applicable): RCS with 3 knots

Phase 4: Cross-National Comparison Table

Generate a side-by-side comparison:

Phase 5: Output Files

{working_dir}/
├── cross_national_report.md    — Study summary + comparison tables
├── variable_mapping.csv        — Variable mapping with match status
├── analysis_korea.R            — KNHANES analysis (self-contained)
├── analysis_us.R               — NHANES analysis (self-contained)
├── results/
│   ├── table1_korea.csv
│   ├── table1_us.csv
│   ├── main_results_comparison.csv
│   └── subgroup_comparison.csv
└── manuscript_draft/           — Optional: Methods + Results draft
    ├── methods_draft.md
    └── results_draft.md

Critical Rules

1. NEVER pool data across countries. Each country analyzed with its own survey design.
2. Country-specific BMI cutoffs: Korea ≥25 (Asian), US ≥30 (WHO).
3. Country-specific income: KNHANES quartile, NHANES PIR → harmonize to binary.
4. Weighted analysis mandatory: Both KNHANES and NHANES are complex surveys.
5. Document all harmonization decisions: What matches, what needed recoding, what differs.
6. Same analytic approach: Identical model specifications for both countries for fair comparison.

KNHANES Variable Coding Reference (validated via Joo 2026 replication)

NHANES Variable Coding Reference (validated via Joo 2026 cross-national)

CRITICAL: NHANES data downloaded via R nhanesA package uses TEXT LABELS, not numeric codes.

ALQ121 Text Label Mapping (Alcohol Frequency)

• Frequent (current drinker): Any specific frequency except "Never in the last year"
• Occasional (past-year abstainer): "Never in the last year"
• Never (lifetime non-drinker): ALQ111 == "No" (ALQ121 will be NA)

Additional KNHANES Variables (validated via LE8-Asthma replication)

Additional NHANES Variables (validated via LE8-Asthma replication)

CHNS Variable Coding Reference (validated via 3-country batch)

Data source: cpc.unc.edu/projects/china (free registration) Biomarker wave: 2009 only (N=9,549). Other variables available 1989-2015. Survey design: No formal weights. Use svydesign(id=~COMMID, weights=~1) or cluster-robust SE.

Key Files and Merge Strategy

Variable Coding

CHNS-Specific Warnings

1. No survey weights: CHNS is NOT a formally weighted survey. Use unweighted analysis with cluster-robust SE by COMMID. Report as limitation.
2. Biomarker = 2009 only: Glucose, HbA1c, lipids, hsCRP available only in 2009 wave. Other waves lack lab data.
3. CES-D not in standard download: Depression comparison requires separate dataset download from cpc.unc.edu.
4. BMI cutoff ≠ KR ≠ US: China=28, Korea=25, US=30. Use country-specific cutoffs AND sensitivity analysis with WHO cutoff=25.
5. SES-health gradient may reverse: Low education and low income are NOT always risk factors in China (null/protective). This is the "developing country health transition" — do NOT treat as a bug.
6. Hemoglobin unit: CHNS reports g/L (KR/US report g/dL). Divide by 10 when comparing.
7. Education scale: 7-level (0-6) vs KR 4-level vs US 5-level. Harmonize to 3-level for comparison.

Composite Score Replication Warnings (learned from LE8 replication)

1. BMI cutoff mismatch: LE8 uses WHO <25 which classifies most Koreans as "ideal" → Factor subscore loses BMI discriminatory power in Asian populations. Report this limitation.
2. KNHANES sleep = clock times: BP16_11-14 are bedtime/waketime (hour:min), NOT sleep duration. Must compute wake_time - bed_time with midnight crossing.
3. pa_aerobic codes: Values are 0/1 (not 1/2). Binary → MET-hours approximation is coarse.
4. Diet quality scoring: AHEI-2010 requires detailed food group data; nutrient-based proxy gives different distribution. Recommend downloading NHANES DR1TOT_J for dietary recall nutrients.
5. LE8 sensitivity to implementation: Small scoring differences compound across 8 components → overall score can diverge substantially, especially in the "moderate" range where most people cluster.

Anti-Hallucination

• Never fabricate variable names, dataset column names, or variable codings. If a variable mapping is uncertain, output [VERIFY: variable_name] and ask the user to confirm against the data dictionary.
• Never fabricate statistical results — no invented p-values, effect sizes, confidence intervals, or sample sizes. All numbers must come from executed code output.
• Never generate references from memory. Use /search-lit for all citations.
• If a function, package, or API does not exist or you are unsure, say so explicitly rather than guessing.