The Relationship of PROMIS Pain Interference and Physical Function Scales

Abstract

Objectives

To examine the relationship between the Patient Reported Outcome Measurement Information System (PROMIS) Pain Interference (PI) and PROMIS Physical Function (PF) scales in patients with spinal pain at a university spine center.

Design

Retrospective analysis of prospectively collected patient-reported outcome data at a university spine clinic. Pearson correlation was done to examine the relationship of the PROMIS PF and PROMIS PI scores. Age, gender, and race were analyzed by subgroups on the PROMIS Physical Function and Pain Interference score. Linear regression analyzed predictive relationships. Statistical significance was set at P < 0.05.

Results

A total of 1,992 participants completed an assessment, with 1,923 completing the PF CAT and 1,927 the PI CAT. Participants’ mean age was 52.8 years (range = 18–94 years, SD = 6.5 years). Correlation analysis of the PROMIS PF with the PROMIS PI showed a Pearson correlation value of –0.717 (P < 0.05). There was a strong linear relationship with a high negative correlation between PF CAT and PI CAT. The PI CAT predicted PF CAT scores (β = –0.707, P < 0.001).

Conclusions

For patients with pain from spinal origin, there is a strong negative correlation between self-reported physical function and pain interference related to physical, social, and mental health. The predictive relationship of function from pain scores supports the PROMIS PI being used as an important adjunct measure of physical function in patients with spinal pain.

Pain and the perception of pain can have significant impacts on individuals with disorders of the spine. Pain has a major impact in limiting multiple domains of an individual’s life, including physical, cognitive, and emotional function, as well as social activities. There are multiple domains of pain that can be measured, such as intensity, interference with function, and behaviors related to pain. Pain intensity is the most common measure. Partly as a result of accrediting organization requirements, it is routinely monitored; however, intensity only explains some of the variance in impact on an individual’s function [1]. The measurement of pain and physical function in individuals with painful conditions is increasingly being urged in guiding optimal care that has value to the patient [2,3]. Patient-reported outcomes (PROs) are used clinically to provide insight on the effectiveness of interventions on patients’ subjective general health, daily function, pain, and quality of life. The most commonly used PROs in the spine disorder population include the Pain Disability Index, Brief Pain Inventory, Oswestry Disability Index (ODI), Roland Morris Disability Questionnaire (RMDQ), Neck Disability Index (NDI), and the visual analog scale (VAS). While these measures have been shown to be reliable and valid, some have suggested shortcomings, including being static in nature, such that all questions need to be completed in order to obtain a score [1,4], unacceptable floor and ceiling effects [5–7], patient and clinician burden in collection [5,8], and some being disease specific.

With the goal of improving PROs, the National Institutes of Health (NIH) began development of a Patient Reported Outcome Measurement Information System (PROMIS) in 2004 to measure important patient outcomes including physical, mental, and social health [9]. The goal was a unified measurement system that could be used across ailments and chronic conditions. Using item response theory (IRT), the PROMIS measures use large question banks with unidimensional items to produce a summary T-score, which allows a comparison with a group mean. This ability inherent in computer adaptive testing allows for a high level of accuracy with a minimum number of questions, reducing patient burden [10]. Additionally, this model produces valid, reliable results of patient function with minimal ceiling and floor effects [11]. The results are equivalent to legacy measures such as the ODI, NDI, and SF-36 Physical Function Domain (SF-36 PFD) [5], which had also been a criticism of the previous PRO models [12].

Thus, the PROMIS Physical Function CAT (PF CAT; v1.2) has been found to be an efficient, reliable, and precise tool in the assessment of physical function in a broad range of orthopedic conditions [5,11–14]. Yet pain measurement is an important aspect of physical function, and pain interference addresses the ways pain limits one’s physical, cognitive, emotional, social, and recreational activities. The PROMIS Pain Interference CAT (PI CAT; v1.1) measure is designed to understand the impact of pain on a patient’s physical, emotional, and social functioning [1].

In this study, our goal was to further understand the relationship of the PROMIS PF CAT and the PI CAT in order to understand if one measure would accurately assess the clinical implications of functioning on an individual, which could reduce both clinician and patient burden in assessing impacts of treatments in a population of individuals with spinal pain. We hypothesize that there will be an inverse correlation between physical function and pain interference, in that physical function decreases as pain interference increases.

Methods

Data Collection

Consecutive patients presenting to a university spine clinic completed the PROMIS Physical Function CAT v1.2 and Pain Interference CAT v1.1. All questionnaires were collected electronically on a tablet computer prior to seeing the physician. Responses were linked with clinical and demographic data and imported to the electronic medical record (EMR) system. The study protocol was reviewed and approved by the institution’s review board.

Description of Instruments

The PROMIS PF CAT v1.2 consists of 121 items individually validated and calibrated using item response theory. The algorithm for the CAT, which assigns the next item to be answered by the patient based on the previous answers, was provided through an application program interface (API) connected to the PROMIS Assessment Center (Assessment Center, PROMIS Group, Chicago, IL, USA). Item category responses range from 1 to 5. The scores for the PF CAT were recorded in T-scores, derived from sampling weighted to reflect the US population, with a mean score of 50 and standard deviation of 10 points. Low scores in the PF CAT represent low physical function, while high scores represent high physical function.

The PROMIS PI CAT v1.1 consists of 40 items individually validated and calibrated using item response theory. The algorithm for the CAT was provided through an API connected to the PROMIS Assessment Center (Assessment Center, PROMIS Group, Chicago, IL, USA). Item category responses range from 1 to 5, and T-scores had a mean score of 50 and standard deviation of 10 points. Low scores in the PI CAT represent less pain interference with physical function, while high scores represent greater interference with physical function.

Statistics

Descriptive statistic including mean, standard deviation, range, and histograms were used to describe patient characteristics. For comparison of race, we used one-way analysis of variance to investigate differences. For gender differences between PI CAT and PF CAT, we used independent sample t tests. Pearson correlation was performed to investigate the relationship between the PROMIS PI CAT and PF CAT. A correlation value >0.4 was considered satisfactory, and >0.6 was considered very good. Subgroup analyses on the PROMIS PF and PI scores were assessed using race and gender groupings. Linear regression analyzed predictive relationships, controlling for effects of gender and ethnicity. Scatter plots of the PROMIS Pain Interference and the PROMIS Physical Function as well as the PROMIS Pain Interference and age were generated. Statistical significance was set at P < 0.05.

Results

Demographic characteristics of the sample are shown in Table 1. There were 1,992 participants who completed the questionnaires; 1,923 completed the PF CAT, and 1,927 completed the PI CAT. Participant mean age was 52.8 +/− 6.5 years (range = 18–94 years). Race was heavily Caucasian (85.2%) with 1,702 white participants. Participants of other races included 27 (1.4%) African Americans, 36 (1.8%) Asians, 23 (1.2%) American Indians, 6 (0.3%) Pacific Islanders, 34 (6.7%) classified as other, and 64 (3.2%) as not specified. Participant gender was 56.3% female. The mean PF CAT T-score was 38.2 +/− 8.5 (range = 15.4–73.3). The mean PI CAT T-score was 63.2 +/− 7.3 (range = 38.6– 83.8). Pearson correlation analysis of the PF CAT to the PI CAT resulted in a value of r = –0.717. There was a two-tailed significance of P = 0.000 for both, and the correlation was significant at the level of P < 0.01 (Table 2). This suggests a strong linear relationship with a high negative correlation for the PF CAT and PI CAT (Figure 1). Linear multivariate regression modeling found that the PROMIS PI CAT predicts PROMIS PF scores (β = –0.707, P < 0.001), where higher pain interference is associated with lower function. Additionally, there is a statistically significant effect of gender on PI, but not race (Table 3). However, despite statistical significance, the scores likely do represent a clinically significant difference as they are within the minimal clinically important difference (MCID) of 4.2 for PF CAT and 3.7 for PI CAT. There was a significant correlation between age and the PI CAT scores (P = 0.001).

Table 1.

Demographic characteristics of patients (N = 1,992)

Variable	Mean (SD)	No. (%)	Range
Age, y	52.19 (16.56)		18–94
Gender
Male		871 (43.7)
Female		1,121 (56.3)
Race
Native Hawaiian and  other Pacific Islander		6 (0.3)
White or Caucasian		1,702 (85.4)
Asian		36 (1.8)
American Indian and  Alaska Native		23 (1.2)
Black or African  American		27 (1.4)
Other		134 (6.7)
Missing		64 (3.2)
Primary reason for visit
Arm		35 (1.8)
Back		1,200 (60.2)
Hip/buttock		133 (6.7)
Leg		88 (4.4)
Neck		325 (16.3)
Shoulder		45 (2.3)
Missing		166 (8.3)

Table 2.

The PROMIS Physical Function and Pain Interference scores across gender and race

	Physical Function		Pain Interference
	Mean Score (SD)	P Value	Mean Score (SD)	P Value
Gender		0.000		0.000
Male	39.53 (8.86)		62.54 (7.57)
Female	37.47 (8.26)		63.52 (7.12)
Race		0.006		0.003
Native Hawaiian and other Pacific Islander	39.65 (10.35)		64.72 (9.72)
White or Caucasian	38.42 (8.43)		62.97 (7.32)
Asian	41.02 (10.05)		62.08 (8.39)
American Indian and Alaska Native	37.20 (9.14)		64.52 (6.54)
Black or African American	37.51 (9.25)		64.55 (7.63)
Other	37.52 (9.67)		64.68 (7.27)

PROMIS = Patient Reported Outcome Measurement Information System.

Figure 1.

Patient Reported Outcome Measurement Information System (PROMIS) Pain Interference CAT vs PROMIS Physical Functioning CAT scatterplot.

Table 3.

Regression of PROMIS Physical Function by pain interference, patient gender, and race

Multivariate Model	B	Standard Error	Β	T Value	P Value
PROMIS Pain Interference	−0.833	0.014	−0.707	−59.32	0.000
Gender	−1.395	0.207	−0.080	−6.748	0.000
Race	−0.558	0.292	00.023	−1.912	0.056

PROMIS = Patient Reported Outcome Measurement Information System.

Discussion

Patient-reported outcome measures continue to be encouraged by many to understand the impact of a disease on an individual. However, this also can add to the burden of questions both on a provider and patient to understand the impact across a number of conditions. Our goal was to evaluate the relationship between the PROMIS PF CAT and PI CAT in patients with spinal pain to evaluate if there is a correlation of patient-reported pain interference with physical function. The results of this study indicate that the PROMIS-PI CAT has an excellent negative correlation to the PROMIS-PF CAT in individuals with spinal pain conditions. Intuitively, we would expect a strong negative correlation, as pain usually will limit an individual’s physical abilities. However, while this does suggest that physical function is a dominant factor in assessing the impact of pain on an individual’s reported function, the correlation is not perfect, suggesting that social and emotional factors have an impact on the total score for PI CAT as well. The PROMIS-PI item bank measures the impact of pain on functioning in several domains, including physical, emotional, and social. This effect of pain on physical function is similar to findings in a population of individuals with foot and ankle pain [15] and in individuals with a broad spectrum of pain disorders [1], where the correlation was r = –0.55.

The PROMIS set of outcome measures is an important NIH initiative that has produced an efficient and valid set of outcomes for physical function and pain interference that can be applied in a wide range of diseases and conditions. In comparison with legacy measures such as the BPI, ODI, NDI, or SF-36 PFD, use of the PI CAT or PF CAT can give a more accurate representation of physical function with less patient burden [5,8,16]. In the study by Papuga et al., patients on average needed only 35 seconds to answer an average of four to five questions to complete the PF CAT vs 188 seconds to complete the ODI. Despite the strong negative correlation of the PF CAT and PI CAT, it would be expected that only 49% of the variance is explained by this factor. This would suggest that physical function and pain interference, though inter-related, are truly separate domains in a patient with spinal pain.

The significant differences in PI CAT scores based on race and gender are a concept that we believe needs to be investigated further. Despite statistical significance, the scores likely do not represent a clinically significant difference as they are within the MCID of 3.5–5.5 for the PI CAT [17]. Even though the difference in our study is not clinically significant, as with other studies on the PF CAT [8,11,12,14] and PI CAT [1,14], there is an overwhelming Caucasian population. Further work needs to be done on the gender and race differences in this measure to determine if this is truly epidemiologically significant. This has not been reported previously; however, the differences are very small and within the calculated MCID using the distribution method [18,19].

One limitation to this study is we have not analyzed the impact of pain intensity on the overall score of PF CAT or PI CAT. Prior evaluation of the pain intensity and PI item bank in a mixed population of subjects has shown 25% shared variance in the numerical pain score and PI item bank, suggesting the measures are related but different domains [1]. Additionally, the impact of multiple sites of pain or multiple conditions other than a spinal disorder on the score of the PF CAT and PI CAT has not been evaluated, but has been shown to be important on total score for the PF CAT [12] and the PI CAT [1].

Another limitation of this study is the limited representation of races within this study population, which was almost entirely Caucasian. This could impair the generalizability to other populations; however, we are not aware of any published studies on PI CAT in specific minority populations. Additionally, the narrow range of conditions in this study population provides only information about spine clinic patients and is not necessarily generalizable to other conditions. The main focus was on the outpatient spinal population, and how well that translates to other diseases or conditions is difficult to interpret.

Further study on the responsiveness to change following treatment for spinal disorders is still needed for both the PI CAT and PF CAT. As well further evaluation of the PI CAT with legacy measures such as the BPI interference subscale, ODI and NDI, which are currently used to measure function, should be completed prior to further recommendation on inclusion in clinical decision-making.

Conclusion

The PROMIS PF CAT and PI CAT showed a high level of association to each other. Our study provides validation of support in the spine population, showing that as pain levels increase, functionality decreases. The PROMIS PI CAT therefore allows for direct comparison of a patient’s subjective experiences of pain on physical function, allowing for a focused, time-efficient measure in the spine population. The continued application of the PROMIS data bank and various CATs has the potential to increase visit efficiency, decrease patient burden, allow for a variety of disease comparisons, and reduce reliance on static standard legacy measures of patient-reported outcomes.

Funding sources: This research was funded by the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under award number U01AR067138. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. We thank the Quality Outcomes and Assessment (http://QualityOutcomesResearch) for supporting this study.

Prior presentations: Portions of this work were presented at the Association of Academic Physiatrists Annual Meeting 2017 in Las Vegas, Nevada.