Low-Risk Sample Dataset (저위험 표본데이터셋)

The Low-Risk Sample Dataset (저위험 표본데이터셋) is a heavily aggregated sample built to widen access to NHIS data for private-sector research and development. It follows 1,000,000 randomly sampled people who used medical care between 2014 and 2023, with one record per person, year, and main diagnosis group, together with their health screening results. The NHISS portal lists it as the low-risk sample cohort (저위험 표본코호트DB) in its menu for applying for sample research databases (표본연구DB신청). The portal's pages describing each database do not yet include it, and it is absent from the NHIS guide to new applications of September 2024.

Source

This page follows the NHIS user manual for the dataset (저위험 표본데이터셋 사용자 매뉴얼), which is undated and covers data through 2023.

Sample design

The sample is drawn from people who used medical care in the ten-year window, with the design below.

• Population (모집단). Everyone with at least one medical claim (의과 요양급여청구명세서) between 2014 and 2023, each counted once.
• Sample (표본). 1,000,000 people drawn at random, with RN_INDI (개인일련번호, person identifier) assigned at random from 1 to 1,000,000.
• Years (기간). Care from 2014 to 2023 and deaths from 2014 to 2024.

Because the sample is drawn from people who used care, people with no medical claims in the ten years are absent, which matters for any estimate of prevalence or utilization rates in the population. The manual reports that the sample resembles the population in its sex ratio, regional distribution, and mean age.

Tables

The dataset has two tables, plus two lookup files for diagnosis names (SMPL_MC_SICK_NM) and the list of masked diagnoses (SMPL_MC_SICK_MSK).

Table	한글	Unit of observation	Records	People	Variables
SMPL_MC	의료이용, medical use	Person, year, and main diagnosis group	42,814,1551	1,000,000	12
SMPL_HC	건강검진, health screening	Person and screening year	2,773,017	678,120	25

The medical use table has the following variables.

Variable	한글	English	Notes
RN_INDI	개인일련번호	Person identifier	1 to 1,000,000
GENDER	성별	Sex	1 male, 2 female
BTH_YYYY	출생연도	Birth year	1929LE for people born in 1929 or earlier
DTH_YYYY	사망연도	Year of death	2014 to 2024
MDCARE_YYYY	수진연도	Year of care	2014 to 2023
SICK_SYM	주상병	Main diagnosis group	KCD block, chapter, or three-character code
SICK_SYM_NM	주상병 설명	Name of the main diagnosis group	Provided in SMPL_MC_SICK_NM
SIDO	거주지(시도)	Province of residence	17 provinces
INSUR_TYPE	건강보험자격	Insurance type	1 self-employed, 2 employee, 3 Medical Aid
CTRB_VTILE	건강보험료 5분위	Premium quintile
DD_ADM	입원일수	Inpatient days	Annual total for the diagnosis group
DD_OPD	외래내원일수	Outpatient visit days	Annual total for the diagnosis group

The screening table holds 23 measurements and questionnaire items, including height, weight, waist circumference, blood pressure, blood and urine tests, estimated kidney function, smoking, drinking, and physical activity.

Known issues

The dataset trades detail for ease of access, and the following features limit what it can answer.

• Diagnoses are coarse. Only the main diagnosis is recorded, at the level of a KCD block (중분류) or, for sensitive diagnoses, a KCD chapter (대분류). Eight major cancers are kept at a finer level, namely stomach (C16), colorectal (C18-C20), liver (C22), gallbladder and biliary (C23-C24), pancreas (C25), lung (C33-C34), kidney (C64), and thyroid (C73), while breast and cervical cancer appear only at the chapter level because they are sensitive.
• Only medical claims are included. Dental, Korean medicine, and pharmacy claims are excluded, and the data contain no costs, dates of care, procedures, drugs, or providers.
• Days are annual totals. Inpatient and outpatient days are summed by year and diagnosis group and capped at the number of days in the year, so individual visits and admissions cannot be identified.
• Screening values are rounded and top-coded. Height, weight, and waist circumference are top-coded at 190 cm, 126 kg, and 121 cm, and several measurements are rounded to whole numbers.
• Questionnaire items are harmonized. Responses were recoded to be comparable across screening reforms, with drinking frequency per month converted to a weekly rate. Walking days exist only through 2017, and muscle-strengthening days only from 2018.
• Birth year is top-coded. BTH_YYYY (출생연도, birth year) is recorded as 1929LE for people aged 85 or older in 2014.
• Triglyceride entry in the manual. The manual's description of G1E_TG (트리글리세라이드, triglycerides) gives the formula for calculated LDL cholesterol, which appears to belong to G1E_LDL (LDL-콜레스테롤, LDL cholesterol).

---

1. The manual's text gives 45,814,155 medical use records, but its table gives 42,814,155, and only the second figure is consistent with its reported total of 45,587,172 records across both tables. ↩