Validation of the PEG scale in Spanish (PEG-S) among adults receiving care for pain in US primary care

Abstract

This study sought to evaluate the psychometric properties of a Spanish version of the PEG scale (PEG-S, whose items assess Pain intensity and pain interference with Enjoyment of life and General activity) in a sample of Spanish-speaking adults receiving care for pain at primary care clinics in the Northwestern United States. We evaluated the PEG-S’s: (1) internal consistency, (2) convergent validity, and (3) discriminant validity. All participants (n=200, mean age=52 years [SD=15], 76% women, mean PEG-S score=5.7 [SD=2.5]) identified as having Hispanic or Latino ethnicity, and detailed ethnic origin was predominantly Mexican or Chicano (70%). The PEG-S’s internal consistency (Cronbach’s alpha, .82) was good. Correlations between the PEG-S scale scores and established measures of pain intensity and interference ranged from .68 to .79, supporting the measure’s convergent validity. The correlation between the PEG-S scale score and the PHQ-9 (r = .53) was weaker than those between the PEG-S scale and measures of pain intensity and interference, supporting the measure’s discriminant validity. The findings support reliability and validity of the PEG-S for assessing a composite score of pain intensity and interference among Spanish-speaking adults.

1. Introduction

Chronic pain is a leading cause of disability globally.¹¹ There are documented inequities in pain care for minoritized groups, including people who prefer to speak Spanish with their healthcare team in the United States.^2,20 Valid and reliable assessment of pain and its impact on function is critical for the development of an appropriate treatment plan, and for ongoing monitoring of treatment effects. Given the subjective nature and sociocultural variability in pain expression, communicating in Spanish can improve pain control and satisfaction with pain care among Spanish-speakers with limited English proficiency.¹⁴ However, validated Spanish-language pain assessment tools are often unavailable.¹³ Although multidimensional pain measures such as the Brief Pain Inventory (BPI)⁷ and the Graded Chronic Pain Scale (GCPS)²⁸ – widely used in specialty care and research – have been translated to Spanish,^4,9 time constraints make these measures impractical for use in primary care,¹⁵ where most of the pain care in the United States is delivered.⁵

The PEG scale, consisting of 3 items from the BPI, assesses (1) average Pain intensity, (2) pain interference with Enjoyment of life, and (3) pain interference with General activity in the past week.¹⁵ Respondents rate each of these items on 0-10 numerical rating scales where higher scores indicate worse intensity or interference. These ratings are then averaged into a composite score that reflects a combination of pain intensity and pain interference. Initial validation of the English version of the PEG scale among adult primary care populations and other ambulatory care patients provides strong evidence for its reliability and construct validity.¹⁵ Given this evidence, the use of the PEG has been widely recommended as a part of best clinical practice, including for use during baseline evaluation and goal setting, as well as for follow-up and determination of clinically meaningful improvements.^8,22,30 The PEG has also recently been included in the NIH’s Helping to End Addiction Long-term Initiative (NIH HEAL Initiative) Clinical Pain Management Common Data Element Program, required for use in all HEAL-funded clinical pain studies of adult chronic pain.²⁹ However, a Spanish version of the PEG has not been validated among primary care populations in the United States. Such a measure is needed to help ensure access to the best care for individuals who prefer Spanish to communicate with their healthcare team; this includes those who speak Spanish primarily or in addition to other languages. For instance, according to the American Community Survey, approximately 14% of the adult population (32 million people) spoke Spanish at home in the United States in 2020.²⁷

Given these considerations, this study aimed to evaluate the psychometric properties of a Spanish version of the PEG (PEG-S) in a sample of Spanish-speaking adults receiving care for pain at primary care clinics in the United States. Specifically, we aimed to evaluate the PEG-S scale’s: (1) internal consistency, (2) convergent validity, and (3) discriminant validity. Based on findings from the development and initial validation of the PEG scale in English,¹⁵ we hypothesized that internal consistency would be at least adequate (i.e., Cronbach’s alpha ≥ .70). With respect to convergent validity, we hypothesized that PEG-S scores would evidence at least moderate associations (i.e., Pearson’s r ≥ .50) with two other standardized measures of pain interference (i.e., the BPI Pain Interference scale and GCP Disability scale scores) and pain intensity (i.e., BPI Pain Severity scale and GCP Intensity scale scores), with higher coefficients for pain interference measures, because two of the three PEG-S items assess pain interference. Finally, with respect to discriminant validity, we hypothesized PEG-S scores would evidence stronger associations with the above pain measures than with a measure of depressive symptoms (i.e., PHQ-9 scale scores^16,24).

2. Methods

This study was reviewed and approved by Institutional Review Boards (IRB) at both Sea Mar Community Health Centers and University of Washington (STUDY00008435).

2.1. Participants

Clinic staff extracted contact information from electronic health records of primary care Community Health Centers located in the Western region of Washington State. Potential participants were at least 18 years of age, had Spanish as preferred language, and had been seen in the office or virtually for a non-cancer pain complaint between 03/01/2019 and 10/31/2020. ICD-10 codes used to identify pain-related visits are available in Supplemental Table 1.

Bilingual-bicultural study staff (i.e., native/proficient level in English and Spanish, and Hispanic/Latinx background) attempted to contact 1,019 potential participants using the phone numbers extracted from healthcare records. Three call attempts were made at different times of the day and days of the week (including weekends), leaving messages in Spanish with a callback number whenever voicemail was available. Study staff reached and screened 567 potential participants. Of these, 183 declined to participate, and 184 were not eligible (i.e., 183 due to not having had bothersome pain in the last month, and 1 due to Spanish not being their preferred language for care). This yielded a final sample of 200 participants who completed the verbal informed consent process and enrolled in the study (Figure 1). All eligible and consenting participants were included in the analysis.

Figure 1.

Flow diagram of sample selection.

2.2. Procedures

Potential participants were contacted via phone by the study staff, who briefly introduced themselves and described the study. For those who were interested, study staff evaluated eligibility and initiated the informed consent process if eligible. Staff explained the study procedures emphasizing the voluntary nature of participation, elicited clarifying questions, and used the teach-back method to ensure potential participants understood what was expected of them before obtaining verbal consent to participate. Those who were eligible and agreed to participate but were unable to be interviewed at the time of the initial phone call, were offered a callback at their preferred date and time. All phone calls were conducted in Spanish.

Study staff administered study questions in a phone interview format, during which they read questions and provided response options to participants when appropriate. Study staff recorded participants’ responses on a Research Electronic Data Capture (REDCap)¹² form hosted at the Institute of Translational Health Sciences. Some of the standardized questionnaires used for data collection were obtained from the REDCap Shared Library.²¹

We collected responses to the four study’s standardized questionnaires: PEG-S, PHQ-9, GCPS, and BPI (in this sequence for all participants), described in more detail in the Measures section. Additionally, we assessed sociodemographic characteristics (age, sex assigned at birth, self-identified race and ethnicity, and marital status), pain history (pain site[s] in the past month, most bothersome pain site[s] in the past month, and approximate time since pain onset). After completing the study, all participants were offered a ten-dollar Amazon gift code as a token of appreciation. Each phone interview took approximately 20 minutes.

2.3. Measures

2.3.1. PEG-S

As noted previously, the PEG (and the PEG-S) includes three items from the BPI⁷ assessing average pain intensity, pain interference with enjoyment of life, and pain interference with general activities during the past week. We used Spanish-language items obtained from the Pain Research Group at MD Anderson Cancer Center (Copyright 1991; Charles S. Cleeland, PhD), previously validated as part of the BPI in Spanish.^3,4 The original development of the Spanish version of the BPI included cultural adaptation, translation, and back-translation, prior to its psychometric validation.⁴ Wording in the present study mirrored items presented in the PEG in English, which was appropriate for an interview delivery mode (Figure 2). For example, as in Krebs et al. (2009),¹⁵ we used questions starting with “What number best describes… [¿Que número mejor describe.…]” as opposed to “Circle the one number that best describes… [Rodee con un círculo el número que mejor describa…].” The PEG-S summary score was calculated as the average of the three item responses, each rated on a 0-10 numerical rating scale, where higher scores indicated worse intensity or interference. In addition to the PEG-S items, participants also completed the Spanish versions of the BPI, the GCPS, and the PHQ-9, described below.

Figure 2.

The PEG scale in Spanish (PEG-S).

Footnote: The PEG scale in English (Krebs et al., 2009) reads as follows: 1. What number best describes your pain on average in the past week [0 = No pain; 10 = Pain as bad as you can imagine]? 2. What number best describes how, during the past week, pain has interfered with your enjoyment of life [0 = Does not interfere; 10 = Completely interferes]? 3. What number best describes how, during the past week, pain has interfered with your general activity [0 = Does not interfere; 10 = Completely interferes]?

2.3.2. Brief Pain Inventory (BPI)

A Spanish version of the BPI^3,4 was administered separately from the PEG-S items and was used to compute measures of pain severity and pain interference. The BPI Pain Severity scale score was calculated as the average of four item responses assessing current pain intensity as well as worst, least, and average pain intensity during the past week, each rated on a 0-10 numerical rating scale (0 = “No pain [Ningún dolor]”, 10 = “Pain as bad as you can imagine [El peor dolor imaginable]”). The BPI Pain Interference scale score was calculated as the average of seven item responses assessing pain interference with general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life during the past week, each rated on a 0-10 numerical rating scale (0 = “Does not interfere [No me ha afectado]”, 10 = “Interferes completely [Me ha afectado por completo8]”). The BPI has been validated for telephone administration.³¹ The internal consistencies of the BPI Pain Severity and Pain Interference scales in the current sample were .90, and .93, respectively, indicating excellent reliability.

2.3.3. Graded Chronic Pain Scale (GCPS)

A Spanish version of the GCPS⁹ was administered to compute measures of pain intensity and pain disability. The GCP Intensity scale is computed using responses to three items assessing current pain, worst, and average pain intensity in the past three months, each rated on a 0-10 numerical rating scale (0 = “No pain [Ningún dolor]”, 10 = “Pain as bad as you can imagine [El peor dolor imaginable]”). Responses are averaged and multiplied by 10 to yield the GCP Intensity scale score, which can range from 0 to 100. The GCP Disability scale is computed from three items assessing pain interference with: (1) daily activities; (2) recreational, social, and family activities; and (3) ability to work, including housework. Each item is rated on a 0-10 numerical rating scale (0 = “No interference [Ningúna interferencia]”, 10 = “Unable to carry on any activity [Incapaz de realizar ningúna actividad]”). Responses to these items are averaged and multiplied by 10 to yield the GCP Disability scale score, which can range from 0 to 100. The GCPS has been validated for telephone administration.¹⁹ In the current sample, the internal consistencies of the GCP Intensity and Disability scales were .74 and .88, indicating adequate and good reliability, respectively.

2.3.4. Patient Health Questionnaire-9 (PHQ-9)

Depressive symptom severity was assessed using a Spanish version of the 9-tem PHQ-9 scale.²⁴ This measures was selected over other possible measures of depressive symptoms because it was already a part of the workflow of the clinics where participants were recruited. The items are designed to reflect the nine criteria for a major depressive disorder as listed in the DSM-V¹). Respondents are asked to rate the frequency with which they experienced each symptom in past two weeks on a 4-point Likert scale (0 = “Not at all [ningún día],” 1 = “Several days [varios días],” 2 = “More than half the days [más de la mitad de los días],” and 3 = “Nearly every day [casi todos los días].”). If a participant endorsed a response other than 0 for the item assessing “thoughts that you would be better off dead or of hurting yourself in some way,” study staff immediately notified the team’s Clinical Lead, who contacted the study participant’s primary care provider. This provider then initiated existing procedures to manage the potential risk for self-harm. These procedures include a brief mental health intervention and referral for further services if appropriate. The PHQ-9 total score is calculated as the sum of all responses and can range from 0 to 27. The PHQ-9 has been validated for telephone administration.^10,17,18,23 The internal consistency (Cronbach’s alpha) of the PHQ-9 scale in the current sample was .81, indicating good reliability.

2.4. Data analysis

We used descriptive statistics to characterize the sample’s self-reported age, sex, race, ethnicity, marital status, pain locations, pain duration, as well as the scores for the PHQ-9, BPI scales, GCP scales, PEG-S scale, and each of the three PEG-S items. To address the study hypothesis about the PEG-S’s internal consistency, we computed the scale’s Cronbach’s alpha. To test the study hypotheses regarding convergent validity (i.e., the extent to which a construct measured in different ways yields similar results), we computed Pearson correlation coefficients (r) between the PEG-S score and the Spanish versions of the BPI and GCP scales. Finally, to test the study hypothesis regarding discriminant validity (i.e., the extent to which a measure is not simply a reflection of some other construct), we computed the Pearson correlation coefficient (r) between the PEG-S score and score of the Spanish version of the PHQ-9. We obtained bias-corrected 95% confidence intervals (95% CI) for alpha and r using bootstrapping with 10,000 replications. All analysis were performed using STATA (v.14) software (College Station, TX)²⁵.

3. Results

3.1. Participant sociodemographics

Two hundred participants met the eligibility criteria for the study and agreed to participate. Their average age was 52 years and 76% of participants were women. All participants identified as having Hispanic or Latino ethnicity. Detailed ethnic origin was predominantly Mexican or Chicano (n=139; 70%), and also included Salvadorean (n=17; 9%), Puerto Rican (n=10; 5%), Guatemalan (n=6; 3%), Venezuelan (n=3; 3%), Colombian (n=3; 2%), Peruvian (n=3; 2%), and other Hispanic/Latino ethnic origins (n=16; 8%). Fifty five percent of participants were married (Table 1).

Table 1.

Sample sociodemographic characteristics (n=200).

Categories	Mean (SD) or n (%)
Age, years
mean (SD)	52 (15)
18-24	7 (4%)
25-34	22 (11%)
35-44	35 (18%)
45-54	54 (27%)
55-64	37 (18%)
65-74	32 (16%)
75+	13 (6%)
Sex assigned at birth
Female	152 (76%)
Male	48 (24%)
Self-identified race
Indigenous only	84 (42%)
Indigenous AND White	71 (36%)
Hispanic, Latina or Mexican only	15 (8%)
White only	12 (6%)
Indigenous, Black AND White	8 (4%)
Mestiza, morena, or trigueña only	3 (2%)
Indigenous AND Asian/Pacific Islander	2 (1%)
Indigenous AND Black/African American	1 (<1%)
Latina AND White	1 (<1%)
Sonoran only	1 (<1%)
None/Chose not to say	2 (1%)
Self-identified Hispanic/Latino ethnic origin *
Mexican or Chicano	139 (70%)
Salvadorean	17 (8%)
Puerto Rican	10 (5%)
Guatemalan	6 (3%)
Venezuelan	6 (3%)
Colombian	3 (2%)
Peruvian	3 (2%)
Other Hispanic or Latino	16 (8%)
Marital status
Single, never married	40 (20%)
Married or living together	110 (55%)
Separated or divorced	38 (19%)
Widowed	12 (6%)

^*All study participants identified as of Hispanic/Latino ethnicity.

3.2. Clinical characteristics

Most participants (58%) described high average pain intensities in the past week on the PEG-S, with scores between 7-10, and a mean pain intensity score of 6.8/10 (SD = 2.3, mode = 8). They described slightly less severe mean pain interference with enjoyment of life (5.2/10, SD = 3.2, mode = 5) and mean pain interference with general activities in the last week (5.0/10, SD = 3.3, mode = 0). The mean PEG-S scale score was 5.7/10 (SD = 2.5). More details about the clinical characteristics of the sample are presented in Table 2.

Table 2.

Sample clinical characteristics (n=200).

Categories	Mean (SD) or n (%)
PEG-S scale
Mean (SD)	5.7 (2.5)
Average pain intensity, past week (0-10)
Mean (SD)	6.8 (2.3)
0-3	19 (10%)
4-6	66 (33%)
7-10	115 (57%)
Pain interference with enjoyment of life, past week (0-10)
Mean (SD)	5.2 (3.2)
0-3	66 (33%)
4-6	57 (28%)
7-10	77 (38%)
Pain interference with general activities, past week (0-10)
Mean (SD)	5.0 (3.3)
0-3	63 (32%)
4-6	59 (30%)
7-10	78 (39%)
Brief Pain Inventory (BPI) Severity scale (0-10)
Mean (SD)	5.4 (2.4)
Brief Pain Inventory (BPI) Interference scale (0-10)
Mean (SD)	3.8 (2.7)
Graded Chronic Pain (GCP) Intensity scale (0-100)
Mean (SD)	65.1 (19.7)
Graded Chronic Pain (GCP) Disability scale (0-100)
Mean (SD)	41.0 (29.1)
Number of pain sites
1	67 (34%)
2	71 (36%)
3	32 (16%)
4+	30 (15%)
Most bothersome pain site(s), past month *
Leg/foot	49 (24%)
Back	47 (24%)
Arm/Hand	34 (17%)
Abdomen	34 (17%)
Head	24 (12%)
Joints	19 (10%)
Chest	11 (6%)
Face/mouth	7 (4%)
Knee	7 (4%)
Pelvic/genital	7 (4%)
Hip	5 (2%)
Shoulder	3 (2%)
Ear	1 (<1%)
Bone	1 (<1%)
Pain duration
Mean (SD), years	3.6 (5.1)
≤3 months	63 (32%)
>3 months, ≤6 months	21 (11%)
>6 months, ≤1 year	20 (10%)
>1 year, ≤3 years	28 (14%)
>3 years, ≤5 years	23 (12%)
>5 years, ≤10 years	26 (13%)
>10 years	18 (9%)
Depression symptoms, past 2 weeks (PHQ-9, 0-27)
Mean (SD)	7.5 (5.4)
0-4, Minimal	71 (36%)
5-9, Mild	65 (32%)
10-14, Moderate	40 (20%)
15-19, Moderately severe	19 (10%)
20-27, Severe	5 (2%)

^*‘Knee’, ‘pelvic/genital’, ‘hip’, ‘shoulder’, ‘ear’, and ‘bone’ were entered as the description when participants selected ‘other’ most bothersome pain site(s); These categories may have received more endorsements had they been asked of all participants.

3.3. PEG-S psychometric properties

The study findings with respect to the psychometric properties of the PEG-S are presented in Table 3. The internal consistency of the 3-item scale was good, with a Cronbach’s alpha of .82 (95% CI: .77, .86). Regarding convergent validity, the Pearson’s correlation coefficients between the PEG-S scale scores and the measures of pain intensity and interference were moderate to strong, ranging from .68 to .79. Moreover, the correlation between the PEG-S scale score and the BPI measure of pain interference (r = .79, 95% CI: .71, .84) was stronger than between the PEG-S scale score and BPI measure of pain intensity (r = .68, 95% CI: .60, .64). However, the correlation between the PEG-S and GCP scales assessing pain intensity (r = .69, 95% CI: .61, .76) and interference (r = .69, 95% CI: .59, .77) were essentially the same. Regarding discriminant validity, the correlation between the PEG-S scale score and the PHQ-9 measure of depressive symptom severity (r = .53, 95% CI: .43, .62) was weaker than that between the PEG-S scale and measures of pain intensity and interference, described above.

Table 3.

Psychometric properties of the PEG-S among Spanish-speaking adults receiving care for pain in US primary care (n=200).

Internal consistency	alpha (95% CI)
PEG-S scale	.82 (.77, .86)
Convergent validity	r (95% CI)
Brief Pain Inventory (BPI) Severity scale	.68 (.60, 0.74)
Brief Pain Inventory (BPI) Interference scale	.79 (.71, 0.84)
Graded Chronic Pain (GCP) Intensity scale	.69 (.61, 0.76)
Graded Chronic Pain (GCP) Disability scale	.69 (.59, 0.77)
Discriminant validity	r (95% CI)
PHQ-9 scale	.53 (.43, .62)

r= Pearson correlation coefficient.

alpha= Cronbach’s alpha.

CI= confidence interval; bias-corrected, obtained using bootstrapping with 10,000 replications.

4. Discussion

Overall, we found the PEG-S scale to have good psychometric properties among a sample of adults receiving care for pain in primary care clinics across the Western region of Washington State, in the Pacific Northwest of the United States, and who preferred Spanish as the language for care. Internal consistency of the PEG-S was good, and correlations with BPI, CGP, and PHQ-9 scales supported both convergent and discriminant validity.

For instance, internal consistency was higher for the PEG-S in the present study than in the initial development and validation of the original PEG in English¹⁵ (Krebs et al., 2009 Cronbach’s alpha = .73). In addition, correlations assessing convergent validity with GCPS scores were slightly stronger for the PEG-S in the present study than for the original PEG¹⁵ (Krebs et al., 2009 CGP Intensity r = .64; GCP Disability r = .67). Correlation coefficients assessing convergent validity with BPI scores were moderate to strong in the present study, and stronger for the BPI Interference scale. These correlations were slightly weaker than for the original PEG¹⁵ (Krebs et al., 2009 BPI Pain Severity: Study 1 r = .69, Study 2 r = .84; BPI Interference: Study 1 r = .89, Study 2 r = .95), but still within an adequate range. The mean PEG-S score (5.7/10) for the current sample was similar to that reported for Study 1 in the original PEG validation study (i.e., 6.1/10) study.¹⁵ However, the participants in the current study had, on average, higher pain intensity scores and lower pain interference scores, whereas the average pain intensity and interference scores were similar to each other in the original PEG validation samples.

Although the original PEG scale development did not report on discriminant validity, the authors reported having intentionally selected the item “Pain interference with enjoyment of life” instead of the item “Pain interference with mood” since the former produced better discrimination from depression.¹⁵ Indeed, the correlation of PEG-S scale scores with PHQ-9 scores in the present study was substantially weaker than with all pain measures tested, in support of the PEG-S’s discriminant validity.

A limitation of the study is that participant health literacy and numeracy were not assessed, although it could be relevant to characterize the generalizability of our findings to other Spanish-speaking populations. For context, at least two thirds of individuals served by the organization from where our study participants were recruited (Sea Mar Community Health Centers) have a family income at or below the Federal Poverty Level. Although numerical rating scales (NRS) may not be as intuitive for individuals with lower numeracy, none of the participants demonstrated hesitation in selecting their responses to the 0-10 NRS in the present study. This may be due, at least in part, to the ubiquitous use of 0-10 pain ratings in healthcare settings throughout the United States, which may have helped participants use this type of scale without major issues. Additionally, the BPI user guide recommends an interview administration for individuals with lower literacy.⁶ Thus, the interview administration may have also helped the participants to use this scale.

The overall composition of Hispanic/Latino ethnic origin in the present study was similar to estimates for Washington State. However, the proportion of participants in the current sample who self-identified their detailed ethnic origins as Puerto Rican, Cuban, and Dominican was lower than among Hispanic or Latinos in the United States as a whole.²⁶ Additional research to further evaluate the psychometric properties of the PEG-S in other settings – including among individuals that may differ in self-identified ethnic origins – would be useful.

Given evidence of the PEG-S scale’s internal consistency, convergent, and discriminant validity, this 3-item composite measure of pain intensity and pain interference would help clinicians and researchers assess pain among adults whose preferred language for care is Spanish. By evaluating the psychometric properties of this ultra-brief tool, this study provides findings needed to support greater equity and inclusion in pain care and research.

Perspective:

We present evidence supporting the reliability and validity of the PEG scale in Spanish (PEG-S) in a sample of adults receiving pain care at primary care clinics in the Northwestern US. This 3-item composite measure of pain intensity and interference can help clinicians and researchers assess pain among Spanish-speaking adults.