The psychometric properties of the Physical Resilience Instrument for Older Adults (PRIFOR): a Rasch analysis

Prior psychometric evidence of the Physical Resilience Instrument for Older Adults (PRIFOR) showed good criterion-related validity, concurrent validity, known-group validity, predictive validity, and internal consistency. However, it is unclear whether older patients with different treatment diagnoses interpret the PRIFOR similarly. This study aimed to test the psychometric properties of the PRIFOR scores among different treatment diagnoses of older patients. We recruited 413 hospitalized older patients with a medical diagnosis and 207 with a surgical diagnosis in a 1343-bed tertiary-care medical center in Taiwan. Data analyses included Rasch models, Principal Components Analysis (PCA), and Pearson correlations. The Rasch analyses showed that all PRIFOR items were embedded within their belonged constructs, reflecting good construct validity and unidimensionality. Person and item separation reliability support the internal consistency of the studied samples and PRIFOR items. However, six PRIFOR items were found to have meaningful differential item functioning (DIF) problems among treatment diagnoses. The PRIFOR is a solid measurement and can be used for monitoring the status of older adults’ physical resilience. However, because six items were found to have meaningful DIF among treatment diagnosis groups, future studies should consider designing specific items for different patient populations to assess their needs in physical resilience.


Background
The world has been experiencing unprecedented growth in the aging population.The population of people aged 80 years or older is expected to triple between 2020 and 2050 to reach 426 million [1].The World Health Organization (WHO) has described this demographic shift as a major societal challenge and proposed 2021-2030 the decade of healthy aging, aiming to maintain optimum functional ability for older people throughout their life course [2].However, maintaining one's functional ability is not enough; recovering from injury, illness, or other health stressors that inevitably occur in a lifetime is crucial [3].An emerging concept stemming from this work is "physical resilience", which is defined as the ability to recover or optimize function in the face of age-related losses or disease [4].Indeed, healthy aging requires a life course approach, and improving physical resilience may attenuate the cascade of adverse outcomes associated with aging [5].
A better understanding of physical resilience may contribute to healthy aging.The National Institute on Aging in the United States has prioritized the development of instruments to objectively measure an individual's level of physical resilience [3].However, the most straightforward way to measure physical resilience is to expose the individual to an experimental stressor, which may induce measurable changes in the individual's parameters of internal equilibrium determine the individual's capacity to recover.Unfortunately, exposing people to experimental stressors is not feasible or ethical, especially in older adult populations.Therefore, subjective measures such as patient-reported outcomes offer promising alternatives for the assessment of physical resilience in older adults.
With the use of rigorous methodological approaches, the authors have developed and validated the Physical Resilience Instrument for Older Adults (PRIFOR), a questionnaire for assessing physical resilience in older adults suffering from acute health stressors [6].Prior psychometric evidence of the PRIFOR in older patients supported a three-factor structure (Positive Thinking, Cope and Adjust Lifestyle, and Belief and Hopeful Mindset) and revealed good criterion-related validity, known-group validity, predictive validity, and internal consistency [6,7].Moreover, according to the results of a confirmatory factor analysis, the PRIFOR has good concurrent validity in older patients undergoing surgical treatment [8].Although the psychometric properties of PRIFOR have been examined for hospitalized older patients [6][7][8], whether older patients with different treatment diagnoses interpret the PRIFOR similarly has remained unknown.A sound instrument needs to be tested using different statistical methods across different populations.Therefore, the use of the Rasch model for psychometric assessment examination is necessary, as it has the following advantages: (a) the validity of the items can be individually analyzed to determine any redundancy, which may not be detected by classical test theory; (b) it separates the estimates of item difficulty and person ability; and (c) an ordinal-to-interval conversion table can be produced that can help healthcare professionals use the items to understand the latent traits of respondents [9,10].Thus, the purpose of this study was to test the psychometric properties of the PRIFOR scores among different treatment diagnoses in older patients.

Study design and participants
We conducted a cross-sectional study, and potential participants were identified from the adult wards of a 1343-bed tertiary-care medical center in Taiwan.Patients were eligible for inclusion if they were 65 years old or older and able to communicate independently with the researchers.Additionally, potential participants must have intact or mild cognitive impairment, which was measured by the Short Portable Mental Status Questionnaire (SPMSQ).The SPMSQ contains 10 questions with each wrong answer is scored as 1 point.Fewer than 2 points indicate intact cognitive function, 3 to 4 points indicate mild cognitive impairment, 5 to 7 points indicate moderate cognitive impairment, and more than 8 points indicate severe cognitive impairment.Also, scoring in SPMSQ is adjusted for educational level [11].A score less than 3 is identified as intact cognitive function if the participant's educational level is lower than elementary school.For subjects who have an educational level greater than high school, only one wrong answer is allowed.Exclusion criteria for this study were the following: admission due to severe acute illness (immediately requiring intensive care) or needing hospice care.The participants provided written informed consent, and the study was approved by the Institutional Review Board of the participating hospital (IRB No. B-ER-108-064).

Measures and procedures
Eligible patients were recruited within 48 h of admission.After obtaining consent, medical records were reviewed, and participating patients were interviewed face to face at the bedside by a trained research nurse.The medical records provided demographic data (i.e., age, gender, educational level, marital status, and treatment diagnosis).During the interviews, patients completed the PRIFOR, which consists of 16 items scored on a 5-point Likert response scale (1 = strongly disagree; 5 = strongly agree) and are distributed within three factors (i.e., Positive Thinking, Coping and Adjusting Lifestyle, and Belief and Hopeful Mindset).The items are totaled with a point given for each affirmative response, and higher scores reflect greater physical resilience [6].The internal consistency of the PRIFOR was determined by a Cronbach's alpha of 0.94, and the good validity, knowngroup validity, concurrent validity, and predictive validity have been also demonstrated [6][7][8].

Data analysis
Patients' demographic characteristics were analyzed with descriptive statistics, such as mean, standard deviation, frequency, and percentage.We used Facets Version 3.84.0 and Winsteps 5.3.1 to perform Rasch analysis to validate the psychometric properties of the PRIFOR.Given that the three domains of the PRIFOR were examined in separated analyses in a previous study [6,8], we also conducted analyses separately for each of the domains.The rating scale functioning was examined according to Linacre's criteria [12] that the rating scale should increase monotonically across the rating categories; the person separation criterion should be over 2; all items should contain at least 10 responses in each rating category; and each item's performance should fit the Rasch model's expectation of a mean square (MnSq) value less than 1.4 and a standardized mean square (Zstd) value less than 2.
Person reliability, separation, and strata were examined with Rasch analysis.Person strata should be over 2 [13] to demonstrate that the PRIFOR items were sensitive enough to differentiate enrolled patients into different physical resilience levels.Item reliability was examined to ensure the enrolled patients were sufficient to precisely locate the PRI-FOR items on the latent variables (i.e., physical resilience) [13].Principal Components Analysis (PCA) of the residuals [14] was used to further examine the underlying structure of PRIFOR; the criterion was set that the eigenvalue of the first contrast should be less than 3 [13].
A valid and reliable assessment tool should contain test items that would perform consistently across different subgroups.Differential item functioning (DIF) has been frequently used to detect the presence of potential item bias [15] as the bias may result in skew measurement outcomes; therefore, examining DIF has become necessary for every newly developed assessment tool [16].In the current study, we examined whether DIF existed across the enrolled subgroups defined by gender (male vs. female) and treatment diagnosis (medical vs. surgical).The Rasch-Welch t statistics were used to evaluate the significant DIF (p < 0.05); the meaningful DIF was identified when the absolute value of the DIF contrast across the subgroups was over 0.64 [17].

Participant characteristics
The mean age of the 620 patients was 75.3 years (SD = 6.5), and 50.3% were male.The majority of patients had elementary degrees (52.3%) and were married (80.2%).Of the 620 patients, 413 (66.6%) had a medical treatment diagnosis and 207 (33.4%) had a surgical treatment diagnosis (Table 1).

Results of Rasch analysis
We conducted three separate analyses for each of the PRI-FOR domains.According to the Rasch analysis results, the rating scale was monotonically increased in all rating categories (i.e., the thresholds between rating categories were properly ordered), except one item (i.e., "When I need to, I can find someone to help") in the Cope and Adjust Lifestyle domain.The rating category 1 (i.e., strongly disagree) was the least endorsed category with all items having less than 10 responses.Further details of the rating category average measures can be found in Table 2.
Table 3 shows the item calibrations, standard errors, and fit statistics.The item calibrations can be used to determine the item difficulty hierarchy along the physical resilience trait.The most difficult item in each domain was: "I am able to recover from illness or injury in the expected duration" (Domain: Positive Thinking); "I believe I can recover to do my daily activities after illness or injury" (Domain: Cope and Adjust Lifestyle); and "I feel I can handle my life" (Domain: Belief and Hopeful Mindset).The easiest item in each domain was: "I focus on my remaining abilities, not on what I cannot do" (Domain: Positive Thinking); "When I am ill or injured, I accept help from my families and friends" (Domain: Cope and Adjust Lifestyle); and "I am a strong person when I am facing illness or injury" (Domain: Belief and Hopeful Mindset).All the items in each of the three domains had infit statistics less than 1.4, indicating good fit to the Rasch model's expectation.The first principal component analysis with the unexplained variance in the first contrast had eigenvalues in the three domains of 1.86, 2.34, and 1.71, respectively.These results further supported the construct validity and confirmed the unidimensionality of the three domains in PRIFOR.
The person reliability in the three domains was 0.85, 0.87, and 0.90, respectively.Additionally, the person separation in the three domains was 2.38, 2.56, and 3.08, which resulted in the person strata of 3.5, 3.74, and 7.49, respectively.These values met the minimum criterion of person separation over 2 and indicated that the PRIFOR items could effectively distinguish 4 to 8 strata of enrolled patients into different physical resilience levels.The Cronbach's alpha of the internal consistency of the three domains were 0.87, 0.98, and 0.97, respectively.It indicated that the enrolled patients were sufficient and were able to respond reliably and consistently on the PRIFOR items.The mean person measures of the three domains were 2.88, 2.48, and 2.02, respectively, which was about 2 standard deviations greater than the mean of the item calibrations.In this case, it showed that the enrolled patients possessed higher levels of physical resilience than the PRI-FOR item set targeted.However, there were only 6 (1.0%), 19 (3.1%), and 7 (1.1%)patients who achieved the maximum measure scores in each of three domains, which can be considered negligible ceiling effects.No patients achieved the minimum scores; therefore, no floor effects were found.Table 4 shows the DIF items for the subgroups of gender and treatment diagnosis.For gender, 75%, 67%, and 50% of items at the three domains were easier for female patients compared to male patients.However, no items demonstrated significant DIF (p > 0.05) between male and female subgroups; the absolute values of DIF contrasts range from 0.05 to 0.30, which can be considered negligible.For patients' treatment diagnosis, 25%, 67%, and 17% of items at the three domains were easier for patients with a medical treatment diagnosis compared to those with a surgical treatment diagnosis, respectively.Additionally, ten items showed significant DIF (p < 0.05).Among them, six items demonstrated meaningful DIF (i.e., DIF contrast of > │0.64│).One item in Positive Thinking domain (i.e., "I believe I can recover from every illness or injury"; DIF contrast = − 0.76) and one item in Cope and Adjust Lifestyle domain (i.e., "I believe I can recover to do my daily activities after illness or injury"; DIF contrast = − 1.07) were easier for patients with a surgical treatment diagnosis.Alternatively, one item in Positive Thinking domain (i.e., "I focus on my remaining abilities, not on what I cannot do"; DIF contrast = 1.51), two items in Cope and Adjust Lifestyle domain (i.e., "When I am ill or injured, I accept help from my families and friends"; DIF contrast = 0.69 and "When I need to, I can find someone to help"; DIF contrast = 0.80), and one item in Belief and Hopeful Mindset domain(i.e., "I am a strong person when I am facing illness or injury"; DIF contrast = 1.16) were easier for patients with a medical treatment diagnosis.More details regarding each item and the DIF results can be found in Table 4.

Discussion
Prior psychometric evidence of the PRIFOR in older patients was supported to have good criterion-related validity, known-group validity, predictive validity, concurrent validity, and internal consistency [6][7][8].In general, our present findings agree with the previous studies that the PRIFOR is a valid instrument using different psychometric methods.In prior studies [6,8], we mainly used classical test theory to show the robust psychometric properties of the PRIFOR, while in the present study, we used a novel psychometric testing method (i.e., Rasch analysis) to demonstrate the validity of the PRIFOR.The Rasch analyses showed that all PRIFOR items were embedded within their belonged constructs, reflecting good construct validity and unidimensionality.Moreover, the appropriateness of the rating scales (i.e., the five-point Likert scale) was supported by the monotonically increased difficulties among the five options in the rating scales.Person and item separation reliability supported the internal consistency of the studied samples and PRIFOR items; person and item separation index indicated that the studied samples and PRIFOR items were across a variety of physical resilience abilities and difficulties.However, six PRIFOR items (two in Positive Thinking domain; three in Cope and Adjust Lifestyle domain; and one in Belief and Hopeful Mindset domain) were found to have DIF problems among treatment diagnoses.
Regarding the DIF items among treatment diagnoses, older patients with surgical diagnosis were more likely to express beliefs of recovery (Positive Thinking domain "I believe I can recover from every illness or injury"; Cope and Adjust Lifestyle domain "I believe I can recover to do my daily activities after illness or injury") than those with medical diagnosis.It is possible that because the nature of surgery involves the removal of diseased tissues or organs, older patients may tend to believe they can get back to normal after surgery.A systematic review of qualitative studies also reported patients with a surgical diagnosis often expect rapid recovery after surgery [18].In contrast, older patients with a medical diagnosis felt that recovery was more difficult, and they may choose adaptation to the chronic conditions rather than recovery (Positive Thinking domain "I focus on my remaining abilities, not on what I cannot do"; Belief and Hopeful Mindset domain "I am a strong person when I am facing illness or injury").A qualitative study explored older medical patients' experiences and management strategies for recovery.The results pointed out aging and disease aggravated symptoms; the patients expressed a strong sense of decline in physiological function (i.e., poor vitality, fatigue, and weakness), and they were more likely to feel unrecovered.Therefore, older medical patients adapted to chronic conditions through acceptance without misgivings, made positive attempts to boost their own morale, and changed their mindset to fight against chronic conditions [19].
Regarding the DIF items in Cope and Adjust Lifestyle domain among treatment diagnoses, older patients with a medical diagnosis felt that support was more available than surgical patients did (Item "When I am ill or injured, I accept help from my families and friends" and "When I need to, I can find someone to help").Older medical patients may have gradually experienced discomfort or life changes brought on by chronic conditions; therefore, they may have been more likely to seek help from external assistance (e.g., professionals, families, or friends) or use assistive devices to maintain their daily routines.This is consistent with the findings of Yueh et al. study, who reported that older medical patients usually connect to support systems to facilitate recovery from frailty after hospitalization [19].Moreover, previous research has shown that individuals with chronic conditions demonstrated a greater desire for their children to care for them [20].Many older patients believe that taking care of aging parents is their children's responsibility and it is an important filial piety of Chinese traditional culture.In Western cultures, Judaism and Christianity also stress on the importance of honoring and respecting their parents [21].Teng noted that older patients who received care from families developed positive attitudes to deal with adversity, which was consistent with the in Chinese society; that is, adequate physical and mental familial support can assist older patients in coping effectively with chronic conditions [20].However, with the advent of an industrialized society, Chinese family structure has undergone fundamental changes.There is an increasing trend of families hiring formal caregiver to take care for older patients rather than themselves.This has led to conflict between traditional filial piety principles and the new alternatives [22].Further studies could seek to discover how filial piety plays a role and older patients' expectation of new alternatives.
There are several limitations in this study.First, the patients were from the same tertiary-care medical center; our results may therefore have limited generalizability.Second, because the PRIFOR is a generic questionnaire, which does not contain any surgery-specific items, especially surgical complications, it cannot reflect the detailed status in physical resilience for older patients with a surgical treatment diagnosis.However, a generic questionnaire with good psychometric properties may still be useful in detecting physical resilience status for acute health stressors.Third, all the participants had the ability to complete the PRIFOR; that is, none of the participants had moderate or severe cognitive impairment.Our results are not applicable to older patients with moderate or severe cognitive impairment.The Physical Resilience Scale (PRS) developed by Resnick et al. and is reliable and valid when completed by proxy reports [23].Therefore, PRS is more suitable for measuring physical resilience for older adults living with moderate or severe cognitive impairment and needing proxy responses.Finally, the enrolled patients with graduate educational levels is limited and future studies should further explore whether the diverse educational levels may impact patients' physical resilience.

Conclusion
Our results demonstrate that the PRIFOR is a sound instrument to measure physical resilience for older patients.However, substantial DIF was found for six items, and we suggest using the non-DIF items when making comparisons among two treatment diagnosis groups of older patients.
Furthermore, specific items need to be designed for different patient populations to assess their needs in physical resilience.

Table 2
Category average measure of the physical resilience instrument for older adults (PRIFOR) + Indicates its rating category has less than 10 responses *Indicates average perception does not ascend with category calibration

Table 3
Calibration and fit statistics of the physical resilience instrument for older adults (PRIFOR)

Table 4
DIF in physical resilience instrument for older adults (PRIFOR)