Reference values for the PROMIS^® physical function item bank version 2.0 in the general population: a multinational comparison study (Korea, Netherlands, and US)

Abstract

Background

We aimed to obtain general population-based Korean reference values for the Patient-Reported Outcomes Measurement Information System (PROMIS) Physical Function (PF) item bank v2.0 and to compare them with US and Netherlands reference values.

Methods

In April 2021, we surveyed Korean representative participants (N = 2,124) using the PROMIS PF item bank. We compared the mean T-scores of the Netherlands (N = 1,310) and US populations (N = 1,646) using ANOVA by age group, sex, presence of comorbidities, and general health. We also performed Differential item functioning (DIF) analyses between the US and Korea populations.

Results

In Korean, the PF T-score was 55.3 (SD = 9.1), ranging from 28.7 to 77.6. This score was higher compared to the mean score of 50.0 in the US and 49.8 in the Netherlands general populations. Differences in PF were generally observed in all age and sex groups. In the PF item bank, 57 of 131 items were flagged for uniform or nonuniform DIF. Among the general population in Korea, older adults, women, those with lower levels of education, and those with comorbidities had lower levels of PF compared to their counterparts in the general Korean population.

Conclusions

Considering the cultural differences between Western and Asian countries, we recommend establishing country-specific reference values on a global scale.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-25051-7.

Background

Physical function (PF) is one of the primary health effects of many chronic diseases, including musculoskeletal disorders, cardiovascular diseases, chronic respiratory diseases, and cancer [1–3]. In addition, PF can affect other functions, leading to disability and a decline in overall health [4–7]. While there are several measurement methods to determine the PF range [8], patient-reported outcome measures (PROMs) are the most efficient and precise [9].

PROMs have been shown to have positive effects when used routinely in daily clinical care, but several implementation barriers exist because of their time-consuming and costly nature [10]. The Patient-Reported Outcomes Measurement Information System (PROMIS) funded by the US National Institutes of Health has been developed to overcome the limitations of traditional instruments. PF is one of the PROMIS domains and has applied item response theory (IRT) and computerized adaptive testing (CAT) methods. IRT allows more accurate and realistic estimates of reliability by allowing precision to vary across scores. The method can create shorter yet more reliable instruments, particularly when coupled with CAT [11]. Furthermore, PROMIS PF provides support for transforming raw scores to a mean of 50 and a standard deviation (SD) of 10 using the US general population as a reference [12], and the reference values help support the use of PROMIS in daily clinical practice and research [13].

However, it is necessary to validate whether the reference set is appropriate for different countries [14, 15], given that PROMIS measures have been translated into over 60 languages and are increasingly being used worldwide [13]. In fact, the PROMIS guidelines recommend differential item functioning (DIF) analyses to test whether people from different groups with the same construct respond differently to an item [16, 17]. Although several studies have calculated the DIF between the European (EU) population and the US population, few studies have conducted such a comparison for Asian populations, including that of Republic of Korea [14, 15]. Therefore, we aimed to obtain general population-based Korean reference values for the PROMIS PF item bank v2.0 and to compare them with US and Netherlands reference values.

Methods

Participants

We conducted a national survey of 2,189 participants aged 19–84 years using proportional stratified sampling across all 17 provinces in Republic of Korea in April 2021. For the analysis, we excluded participants who had reported physical disabilities (N = 65), resulting in 2,124 participants in the final sample. The study was approved by the Institutional Review Board of Samsung Medical Center (IRB number: 2021-03-005). Participants provided written informed consent.

In this study, we also used US and Netherlands databases. For the US representative population, we used PROMIS Wave 1 data obtained from the Health Measures Dataverse repository. Specifically, we used the data from the Forms C (n = 815) and G (n = 831) [18], which included PF.

Measurement

Although an online survey was performed, we also conducted a face-to-face survey for people older than 65 years due to concerns regarding digital health literacy. The survey consisted of the PROMIS PF full item bank v2.0. The average response time was 30 min.

The PROMIS PF item bank v2.0 included two subdomains and seven PROMIS PF short forms for mobility (44 items), upper extremities (46 items), and five generic short forms of different lengths: SF4a, SF6b, SF8b, SF10a, and SF20a, which had 4, 6, 8, 10, and 20 items, respectively. There were three 5-point Likert response scales, in which the lowest and highest scores indicated an inability to perform an activity or the ability to perform the physical activity without difficulty, respectively. No specific recall period was prescribed within the PROMIS PF item bank, with the intention that the responses should reflect the participant’s current status. We converted the raw scores into T-scores standardized for the general US population (mean (SD) 50 (10) using the conversion table provided at the Assessment Center (https://www.assessmentcenter.net; Northwestern University, Evanston, IL, USA) [18, 19]. Higher scores represented better PF. The mobility and upper extremity subscale scores were calculated individually. In addition, general health status was assessed using the Korean version of the SF-36 v2 physical functioning [20]. We also collected data on age, sex, region of residence, marital status, educational level, employment status, and comorbidities.

Statistical analyses

Descriptive statistics were used to summarize the participants’ sociodemographic characteristics and their responses to the K-PROMIS PF items. We compared the mean T-scores of the US populations and their subgroups. We did not have access to raw data for the PROMIS physical function in the Netherlands. However, we found a publication that reported PROMIS physical function scores and standard deviations for each characteristic. Using these published data, we divided the group into equal categories and compared Korea and the Netherlands indirectly using a formula to manually calculate the p-value of a t-test to compare the two countries.

In the absence of raw data for the Netherlands, the DIF analyses evaluated whether people from different populations with similar levels of PF responded similarly to the US vs. Korean items. The absence of DIF allowed valid comparisons of T-scores between the populations. DIF was evaluated using the R package Lordif (version 0.3–3.3), which employs an ordinal logistic regression framework. There are two types of DIF: uniform and nonuniform. A uniform DIF exists if its magnitude is consistent across the entire range of the assessed trait. A nonuniform DIF exists if the magnitude of the DIF varies across different levels of the trait. Because statistical power is dependent on sample size, the difference in population parameters was significant, given the sufficiently large sample. In response to this concern [21], we used McFadden’s pseudo R²-change between the two models of 0.02 as the critical value to flag for possible DIF [21]. This method of analysis involved three models to assess DIF.

Then, the T-scores of the full items (PF item bank), mobility, and upper extremity were compared by age group (18–34, 35–44, 45–54, 55–64, 65–74, and ≥ 75 years), gender, comorbidities, and general health within the Korean population using the t-test. Multivariate linear regression was performed to assess the factors associated with PF. Age, sex, marital status, education level, income level, region of residence, current employment status, number of chronic diseases, and disability were adjusted as covariates. In terms of comorbidities, patients with each condition were compared to patients without each condition (e.g., patient with cardiovascular disease vs. patient without cardiovascular disease, but could have other diseases).

Statistical significance was set at p < 0.05, and two-sided tests were used for all calculations. Statistical analyses were performed using R 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria).

Results

The mean age was 47.0 years (SD = 16.3), with a range of 18–84 years. Of the participants, 72% had more than a high school education, and 65.87% reported to have at least one chronic condition (Table 1). The PF T-score was 55.0 (SD = 9.3) with a range of 28.7–77.6. Compared to the mean score of 50.0 in the US and 49.8 in the Netherlands general populations, the PF T-score of the Korean population was higher (Fig. 1). Participants older than 75 years had a T-score < 50 (Table 2). People with cerebrovascular disease also had a PF score < 50 (Table 2). The majority of the Korean population maintained normal PF up to age 65, as illustrated in Fig. 1. This superior physical functioning, compared to the US population, was consistent across genders and comorbidities, as observed in several short-format questionnaires of PROMs. Furthermore, the mean scores and patterns derived from these abbreviated versions closely mirrored those found in the full item bank for the Korean cohort (Supplementary Table 1). The mobility, and upper extremity reference values were similar trends (Supplementary Table 2).

Table 1.

Sociodemographic characteristics of study participants

Variables	Korean population (n = 2124)	US population (n = 1646)	Netherlands population (n = 1310)
Age, years	47.00 (16.30)	51.09 (18.23)
18–34	558 (26.27)	364 (22.11)	282 (21.53)
35–44	409 (19.26)	239 (14.52)	214 (16.34)
45–54	446 (21.00)	315 (19.14)	199 (15.19)
55–64	367 (17.28)	291 (17.68)	279 (21.30)
65–74	223 (10.50)	202 (12.27)	280 (21.37)
≥ 75	121 (5.70)	235 (14.28)	56 (4.27)
Gender
Male	1054 (49.62)	793 (48.18)	691 (50.07)
Female	1070 (50.38)	853 (51.82)	689 (49.93)
Highest education level
≤Middle school	257 (12.10)	28 (1.70)	393 (30.00)
High school	347 (16.34)	291 (17.68)	524 (40.00)
≥Colleges	1520 (71.56)	1325 (80.50)	393 (30.00)
Unknown	0 (0)	2 (0.12)
Marital status
Single	714 (33.62)	296 (17.98)
Married or living together	1266 (59.60)	1068 (64.88)
Divorced	56 (2.64)	186 (11.30)
Widowed	88 (4.14)	96 (5.83)
Household income (a year)
Less than $20,000	395 (18.60)	161 (9.78)
$20,000–$49,999	865 (40.73)	523 (31.77)
$50,000–$99,999	726 (34.18)	644 (39.13)
$100,000 or more	138 (6.50)	282 (17.13)
Unknown	0 (0)	36 (2.19)
Current employment status, employed	1372 (64.60)	843 (51.22)
Chronic disease
Cardiovascular disease	566 (26.65)	609 (37.00)
Endocrine diseases	245 (11.53)	148 (8.99)
Cerebrovascular disease	26 (1.22)	46 (2.79)
Musculoskeletal disease	371 (17.47)	349 (21.20)
Respiratory disease	373 (17.56)	269 (16.34)
Gastrointestinal disease	246 (11.58)	69 (4.19)
Cancer	54 (2.54)	123 (7.47)
Psychiatric disorders	142 (6.69)	457 (27.76)
No	725 (34.13)	439 (26.67)

Values are presented as n (%) or mean (SD)

Fig. 1.

Distributions of T-scores (based on full item bank) in the US and Korean populations

Table 2.

PROMIS PF full for the Korean general population and comparisons with the US and Netherlands

Categories	Korean population (n = 2124)	US participants (n = 1646)	Korean vs. US population P–value	Netherlands population (n = 1310)	Korean vs. Netherlands population P–value
	Mean (SD)	Mean (SD)		Mean (SD)
Overall	55.0 (9.3)	50.0 (10.0)	< 0.001	49.8 (10.8)	< 0.001
Age
18–34	58.3 (8.6)	55.9 (7.7)	< 0.001	55.2 (9.5)	< 0.001
35–44	56.9 (8.8)	53.6 (9.3)	< 0.001	52.8 (10.5)	< 0.001
45–54	55.9 (8.3)	51.0 (9.7)	< 0.001	50.0 (11.5)	< 0.001
55–64	54.3 (9.1)	49.6 (8.5)	< 0.001	46.4 (10.1)	< 0.001
65–74	50.8 (6.9)	47.8 (8.0)	< 0.001	46.7 (9.6)	< 0.001
≥ 75	44.9 (6.8)	44.4 (7.6)	0.578	42.6 (9.5)	0.068
Gender
Male	57.7 (8.9)	52.0 (9.4)	< 0.001	50.9 (11.2)	< 0.001
Female	52.9 (8.6)	49.8 (9.1)	< 0.001	48.8 (10.3)	< 0.001
Comorbidity
No	58.6 (8.9)	56.5 (7.3)	< 0.001
Cardiovascular disease	52.1 (9.0)	47.1 (9.0)	< 0.001
Endocrine diseases	51.0 (9.5)	44.4 (9.9)	< 0.001
Cerebrovascular disease	49.0 (7.7)	43.1 (7.9)	0.003
Musculoskeletal disease	50.0 (8.2)	43.6 (7.6)	< 0.001
Respiratory disease	54.9 (8.9)	47.0 (9.5)	< 0.001
Gastrointestinal disease	52.8 (8.5)	46.6 (9.9)	< 0.001
Cancer	52.1 (8.7)	45.0 (8.0)	< 0.001
Psychiatric disorders	51.9 (9.2)	47.8 (9.2)	< 0.001

T-score, higher scores represent better physical functioning

When we performed DIF analyses between US and Korean populations, 57 of 131 items were flagged for uniform or nonuniform DIF (Supplement Table 3). The Korean population had higher DIF scores at the same PF level as the US population. Specifically, PFA1 (“Does your health now limit you in vigorous activities?”), PFB5 (“Does your health now limit you in hiking a couple of miles?”), PFB51 (“Does your health now limit you in participating in active sports?”), PFB50 (“How much difficulty do you have during daily physical activities because of your health?”), PFA11(“Are you able to do chores such as vacuuming or yard work?”), PFA19 (“Are you able to run or jog for two miles?”), PFA31 (“Are you able to get up off the floor from lying on your back without help?”), PFA39 (“Are you able to run at a fast pace for two miles?”), PFB24 (“Are you able to run a short distance?”), PFC11 (“Does your health now limit you in doing yard work?”), PFC7 (“Are you able to run five miles?”), PFC32 (“Are you able to climb up 5 flights of stairs?”), PFC33 (“Are you able to run ten miles?”), PFC41 (“Are you able to sit down in and stand up from a low, soft couch?”), PFC42 (“Are you able to open a tight or new jar?”), and PFC48 (“Are you able to carry household items, such as heavy boxes or furniture, up a flight of stairs?”) had higher scores for the uniform or nonuniform DIF (Fig. 2).

Fig. 2.

Graphical display of the items that exhibited differential item functioning between the US and Korean

Among the general population in Korea, older adults (≥ 75 vs. 18–34: −7.92; 95% confidence interval [CI] = −10.49, −5.34), women (female vs. male: −4.55; 95% CI = −5.24, −3.85), and those with lower levels of education (low vs. high: −2.94; 95% CI = −4.86, −1.03) had lower levels of PF compared to those who were younger, male, or had a higher level of education, respectively (Table 3; Fig. 3). In terms of comorbidity, participants with cerebrovascular (−6.38; 95% CI = −9.51, −3.24), musculoskeletal (−5.34; 95% CI = −6.23, −4.45), gastrointestinal (−4.03; 95% CI = −5.11, −2.96), or psychiatric disorders (−5.27; 95% CI = −6.65, −3.89) were more likely to have lower PF than those without these diseases.

Table 3.

Factors associated with lower physical function among the general population

Variables	Crude (95% CI)	Adjusted* (95% CI)
Age, years
18–34	Reference	Reference
35–44	−1.34 (−2.41, −0.26)	−1.30 (−2.49, −0.10)
45–54	−2.43 (−3.48, −1.38)	−2.33 (−3.56, −1.10)
55–64	−3.98 (−5.10, −2.87)	−2.88 (−4.25, −1.51)
65–74	−7.47 (−8.78, −6.16)	−4.35 (−6.37, −2.32)
≥ 75	−13.39 (−15.05, −11.74)	−7.92 (−10.49, −5.34)
Gender
Male	Reference	Reference
Female	−4.88 (−5.63, −4.14)	−4.55 (−5.24, −3.85)
Education level
≤Middle school	−9.52 (−10.65, −8.39)	−2.94 (−4.86, −1.03)
High school	−2.17 (−3.17, −1.18)	−0.20 (−1.25, 0.84)
≥Colleges	Reference	Reference
Marital status
Single	Reference	Reference
Married or living together	−2.95 (−3.74, −2.15)	0.06 (−1.01, 1.12)
Divorce	−4.59 (−6.96, −2.22)	−1.44 (−3.71, 0.82)
Widow	−12.59 (−14.52, −10.66)	−2.40 (−4.57, −0.23)
Income level (a year)
Less than $20,000	−4.73 (−6.46, −3.00)	−0.46 (−2.15, 1.23)
$20,000–$49,999	0.15 (−1.48, 1.77)	−0.80 (−2.25, 0.64)
$50,000–$99,999	−1.17 (−2.77, 0.43)	−0.20 (−1.63, 1.24)
$100,000 or more	Reference	Reference
Current employment status
Unemployed	−1.72 (−2.52, −0.91)	−0.46 (−1.22, 0.30)
Employed	Reference	Reference
Chronic disease†
Cardiovascular disease	−4.37 (−5.23, −3.52)	−2.54 (−3.40, −1.68)
Endocrine diseases	−4.81 (−6.01, −3.62)	−2.62 (−3.72, −1.51)
Cerebrovascular disease	−6.33 (−9.83, −2.83)	−6.38 (−9.51, −3.24)
Musculoskeletal disease	−6.38 (−7.36, −5.40)	−5.34 (−6.23, −4.45)
Respiratory disease	−0.44 (−1.46, 0.57)	−1.99 (−2.92, −1.07)
Gastrointestinal disease	−2.77 (−3.97, −1.57)	−4.03 (−5.11, −2.96)
Cancer	−3.29 (−5.73, −0.84)	−1.99 (−4.20, 0.22)
Psychiatric disorders	−3.59 (−5.13, −2.05)	−5.27 (−6.65, −3.89)

*Adjusted for age, sex, marital status, education level, income level (years), current employment status, and presence of chronic disease

†The physical functioning of patients with chronic conditions was compared to that of patients without those comorbidities

Fig. 3.

Proportions of normal, mild, moderate, and severe physical function, as defined by the Health Measures

Discussion

In this study, the Korean general population had higher PF score than 50 on average. The score was higher than the Netherlands and the US populations regardless of age and sex. In the PF item bank, 40% of items were flagged as DIF. Among the general population in Korea, older adults, women, those with lower levels of education, and those with cerebrovascular, musculoskeletal, gastrointestinal, or psychiatric disorders had lower levels of PF.

The T-scores for PF were higher among the Korean population than in the Netherlands and US populations. Comparative studies focusing on the general population in the UK, France, and Germany previously documented an average T-score range of approximately 51–53 [22, 23]. These figures were relatively lower than those recorded for the Korean population. These differences could be influenced by several cultural and societal factors. Emerging evidence indicates that East Asian populations, including Korea, consistently report better self-perceived health status and health-related quality of life (HRQoL) than Western European populations. According to the multinational study, East Asian populations showing a tendency to report fewer health problems across domains compared to their Western European counterparts [24]. Similarly, a systematic review found that East and South-East Asian populations reported more favorable outcomes on generic preference-based measures of health [25]. One key factor is the cultural context of self-reporting in East Asia, where respondents may be more inclined to report higher satisfaction or fewer difficulties due to social desirability or normative expectations about health. In addition, the reference values for the US PROMIS PF domain were established using data collected in the year 2000 [26]. Considering the passage of time and significant demographic changes within the US population since then, a score of 50 on the US metric may no longer accurately reflect the current mean score for populations both within and outside the US [11]. Following this, the survey conducted in our study in 2021 revealed results that might be different from those observed in 2000.

In our study, we observed a DIF in 40% of the PF item bank items when comparing the US and Korean populations. Despite rigorous efforts to ensure semantic and conceptual equivalence in the Korean linguistic validation study, linguistic and cultural discrepancies persisted in 14.1% of the PF items [27]. Even though common cultural issues highlighted in previous PROMIS linguistic validation studies were carefully addressed during the translation process, age-related and healthcare system variances contributed to differing interpretations of PF items, underscoring the need for nuanced, context-aware assessment tools. For instance, some items in the PF item bank that pertain to everyday activities, such as “yard work” or “carrying a laundry basket,” may not align with the experiences or typical activities of individuals in Korea, especially those living in urban environments where such tasks are less common. Similarly, items referencing tools such as “utensils” or “can openers” posed unique challenges because their design and functionality can vary significantly across cultures. These differences in tool design affect the amount of physical effort required to use them, particularly in terms of grip strength or hand dexterity, leading to potential discrepancies in how respondents from different cultural backgrounds evaluate their physical capabilities when responding to these items. Additionally, the use of metrics (e.g., km and kg) posed difficulties for participants, highlighting a disconnect with their everyday practices [27]. In addition, differences in healthcare systems and generational perspectives further accounted for the DIF observed between the US and Korean populations. The distinct healthcare landscapes, encompassing preventive care, public health policies, and access to services, differently shaped the individuals’ experiences and expectations of physical health in each country. Moreover, generational influences, reflecting diverse historical, technological, and social contexts, could alter perceptions and reporting of PF. These findings underscored the critical need for separate psychometric evaluations for each language version, particularly in Asia, to account for unique cultural and linguistic context. Thus, while psychometric evaluation is a critical step, cross-cultural adaptation is equally important for ensuring the PROMIS PF tool’s utility and validity across diverse populations. Both approaches together can provide a comprehensive framework for developing robust and universally applicable assessment tools. However, if the differences was too larger than adapting, the use of a standardized core set of items that did not show significant DIF could be considered to maintain comparability across countries. Such a core set would allow researchers to assess physical function consistently across diverse populations while minimizing the cultural and contextual biases associated with specific items.

Despite variations in scores between Korea, the Netherlands, and the US, older adults and women reported lower PF compared to younger individuals and men across the board. Additionally, individuals suffering from cerebrovascular, musculoskeletal, gastrointestinal, or psychiatric disorders demonstrated reduced PF relative to those without the diseases. On the other hand, cancer patients reported the higher mobility scores, while low upper extremity subdomain. It could be that different types of cancer had different physical demands and health outcomes. For instance, breast cancer survivors often experience upper extremity dysfunction due to surgeries like mastectomy or procedures such as lymph node dissection, which can lead to lymphedema or restricted shoulder mobility [28]. In contrast, cancers such as colon or lung cancer may primarily impact systemic energy levels and endurance, influencing general physical function more than specific domains like upper extremity strength or dexterity [29, 30]. We assessed cancer history but did not collect information on specific cancer types. Future studies should consider characterizing cancer types to better understand functional implications. Although the full bank or short-form may be lack of discriminative ability, a CAT that takes into account the difficulty and discriminative ability of the different items in the K-PROMIS is expected to maximize its usefulness.

This study has several limitations. First, the data were collected in 2021 during the COVID-19 pandemic. Consequently, the current population’s level of PF and their participation rates may differ. Ideally, reference values should be updated periodically to accurately reflect the latest population status. Second, our comparison with data from the US and the Netherlands was indirect. Further studies are necessary to collect data concurrently, using consistent methods and protocols to ensure more accurate comparisons. Third, while efforts were made to ensure the sample was representative of the general Korean population, there may be limitations in representativeness and generalizability, especially concerning age, gender, and regional distribution.

Conclusions

In this study, we identified the need for distinct reference values and separate psychometric evaluations for the Korean general population, especially in the Asian context. The detection of DIF across 40% of the items indicated potential variability in item interpretation across cultures. Then we calculated the average T-scores segmented by age, gender, and presence of comorbidities within the Korean general population. Considering that there are several DIF observed between US and Korean populations, our data are valuable references for researchers to compare their study groups with the general Korean population.

Abbreviations

PROMIS

Patient-Reported Outcome Measurement Information System

PF

Physical Function

DIF

Differential Item Functioning

SD

Standard Deviation

Authors’ contributions

(I) Conception and design: Jiseon Lee, Danbee Kang, Juhee Cho(II) Administrative support: (III) Provision of study materials or patients: Jiseon Lee, Danbee Kang, Yeonjung Lim(IV) Collection and assembly of data: Jiseon Lee, Danbee Kang, Yeonjung Lim(V) Data analysis and interpretation: Jiseon Lee, Danbee Kang, Yeonjung Lim, Benjamin D. Schalet, Felix Fischer, Matthias Rose, Dong Gi Seo, Minji K. Lee, Juhee Cho(VI) Manuscript writing: Jiseon Lee and Danbee Kang (VII) Final approval of manuscript: All authors.

Funding

This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2020R1I1A2074210) and Bio&Medical Technology Development Program of the National Research Foundation (NRF) funded by the Korean government (MSIT) (No. RS-2024-00440881).

Data availability

All datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Institutional Review Board of Samsung Medical Center (IRB number: 2021-03-005). Informed consent was obtained from all individual participants included in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.