- Email: [email protected]
Assessing Pain and Mood in a Poorly Resourced Country in a Post-Conflict Setting
Vol. 42 No. 2 August 2011
Journal of Pain and Symptom Management
301
Brief Methodological Report
Assessing Pain and Mood in a Poorly Resourced Country in a Post-Conflict Setting Amanda C. de C. Williams, BSc, MSc, PhD, Minha Rajput-Ray, BSc, Ost Med, MBChB, MICR, Xavier Lassalle, CRNA, Iain Crombie, PhD, CStat, FFPH, and Philippe Lacoux, MBBS, BSc, MD, FRCA, FFPMRCA Research Department of Clinical, Educational and Health Psychology (A.C.d.C.W.), University College London, London; Addenbrooke’s Hospital (M.R.-R.), Cambridge University Hospitals NHS Foundation Trust, Cambridge; Department of Public Health (I.C.), University of Dundee Ninewells Hospital and Medical School, Dundee, United Kingdom; and Department of Anesthesia (P.L.), Technical Advisor for Anaesthesia, MSF (X.L.), Paris, France
Abstract Context. Accurate pain assessment is important but presents problems in poorly resourced countries. A civil war in Sierra Leone resulted in civilian casualties with pain from deliberate amputation and other injuries. Objectives. To examine the quality of simple pain and mood scales. Methods. Within a pain treatment project, pain was assessed using numerical and verbal descriptor scales, and mood using visual analogue and verbal descriptor scales. The relationships between these scales (translated where necessary) were examined by comparison of pairs of measures. Results. The overwhelming majority (99%) used the scales consistently. The verbal pain scale showed substantial discrimination between the pain words mild/small, moderate/half and half, and severe/serious (F ¼ 41.80, P < 0.001). Numerical and verbal pain scales were related at a modest level (Kendall’s tau-b ¼ 0.39, P < 0.001, n ¼ 272) and not dependent on the level of education. A smaller sample (n ¼ 30) provided pain data across three assessment occasions, and both pain scales appeared sensitive to change. The 5-point verbal mood scale collapsed into three categories, with reasonable distinction between mood words (F ¼ 14.75, P < 0.001). The visual analogue scale proved difficult to explain to civilian casualties in this setting. Conclusion. Verbal pain and mood scales and numerical pain scales appeared to yield useful information in a post-conflict situation. This suggests that adapting these established rating scales for pain, and for mood, was useful to clinicians and acceptable to patients. J Pain Symptom Manage 2011;42:301e307. Ó 2011 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words Rating scales, pain, mood, underdeveloped country, post-conflict, education level
Address correspondence to: Amanda C. de C. Williams, BSc, MSc, PhD, Research Department of Clinical, Educational and Health Psychology, University College London, Gower Street, London WC1E Ó 2011 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved.
6BT, United Kingdom. E-mail: amanda.williams@ ucl.ac.uk Accepted for publication: November 6, 2010. 0885-3924/$ - see front matter doi:10.1016/j.jpainsymman.2010.11.012
302
Williams et al.
Introduction Pain rating methods are well established in the developed world for systematic pain assessment,1,2 but these methods cannot be assumed to apply in underdeveloped countries. Assessment is an essential feature of health care, and inadequate assessment, generally and in diverse ethnic groups, is associated with poorer treatment.3 Sierra Leone is a resource-poor West African country with a population of approximately five and a half million and predominantly small-scale agricultural economy. It achieved independence from British colonial rule in 1961. There is a diversity of languages, predominantly Krio (based on English). Civil War started in 1991 and lasted for over a decade.4 During the war, civilians suffered many injuries, including the organized amputation of limbs, most often of arms below the elbow. An initial assessment showed that pain problems were common,5 leading the nongovernmental organization M edecins Sans Fronti eres (MSF; www.msf.org.uk) to decide to treat chronic pain in this unstable setting. The challenge in rating pain in poorly resourced countries arises from differences in culture, education, and language. The simplest approach, translation of instruments developed in Western countries, makes the assumption that well-established measures will transfer to cultures other than that in which they were standardized.6 Others argue that more attention should be paid to investigating shared meaning before examining psychometric qualities7,8 and that cultural and educational differences require more fundamental examination of how measures are used. Differences in the use of a question across cultures can arise from differences in understanding, ease of retrieving the information required to answer, and differences in forming the response.7,8 This is particularly true of words and constructs for emotion because emphasis and differentiation can vary considerably,9 making it hard to find equivalent terms when translating a measurement instrument. There have been consistent reports of cultural differences in distress associated with pain, even where pain intensity shows no difference.10 Inevitably, responding to scales requires reference to implicit normsdof disability, for instancedthat vary between cultures.8 In addition, tendencies
Vol. 42 No. 2 August 2011
to use or avoid extremes of scales might vary systematically with cultural norms and with the dominance of collectivism or individualism.8 Educational level can affect use of scales.11 Of the two African studies identified, one reported no effect of education on the use of a Yoruba-translated visual analogue scale (VAS) and verbal rating scale (VRS) by Nigerian patients with musculoskeletal pain,12 whereas the other found that Zimbabwean maternity patients after Caesarean section only used a Shona-translated VAS and VRS consistently with one another when they had received more than seven years of education.13 This study in Sierra Leone investigated whether numerical and verbal pain rating scales and VAS and verbal mood rating scale, based closely on those widely used in the developed world,1 could be used to express pain by respondents, and interpreted meaningfully by clinicians, in this poorly resourced health care setting at a time of national chaos. That is, the study aimed to investigate shared meaning. It also assessed whether responses on these scales were related to extent of education.
Methods People with pain came from two temporary camps in the area around Freetown (capital of Sierra Leone) to a primary health care center in one of the camps. Most injuries had occurred in 1998 or 1999, up to four years before. Patients were assessed during the therapeutic consultation, usually in English but interpreted where required. Most consultations were with one of the authors (P. L.) and the remainder with a clinician supervised by him. Information also was collected by a trained local health care worker in English or in Krio (translated), using a questionnairebased interview. The questionnaire covered demographics, pain location, pain intensity, mood, function including prosthesis use, and education level. If the patient did not understand, a health worker experienced with the questionnaire translated the clinician’s questions to Krio or other dialects. Subsequent assessments reviewed the effects of treatment4 using the same pain and mood measures. The assessment occasions and numbers are shown in Fig. 1. Data in this study were
Vol. 42 No. 2 August 2011
Pain and Mood Assessment in a Refugee Camp
Fig. 1. Timeline of assessment and treatment of pain in Sierra Leone. A ¼ pilot assessment, n ¼ 40; B ¼ initial clinical assessment, n ¼ 304; C ¼ review on medication, n ¼ 86; D ¼ review at completion of medication, n ¼ 45; E ¼ review after completion of medication, n ¼ 122.
taken from therapeutic records with no patient identifiers, so confidentiality was not an issue. Related work used photographs and video, with verbal consent; patients were largely keen for their plight to be advertised as widely as possible. There was no national or institutional ethics committee from whom permission could be sought.
303
inspecting the NRS for normality and treating the verbal pain scale as ordinal. A one-way analysis of variance (ANOVA) with post hoc testing for significant differences between adjacent words was performed. Analyses were repeated, dichotomizing education, to examine whether education level made a difference. Sensitivity of the NRS to change with treatment was examined in a subsample of 30 patients who were assessed before, during, and after treatment (B, C, and E, respectively, in Fig. 1) by a parametric ANOVA, testing for linear (continuous change) and quadratic (discontinuous change) functions, and by a nonparametric ANOVA on the verbal scale. Discontinuous change would indicate a return toward baseline after initial improvement. Paired mood ratings (VAS and verbal) from each patient were compared using Pearson’s correlation and ANOVA with post hoc testing (Tukey’s honestly significant difference [HSD]).
Results Pain and Mood Rating Scales Translations to Krio were developed through liaison with bilingual MSF office workers in the locality and were piloted on the first few patients seen, before their regular use. Pain ratings were requested using a vertical 11-point numerical rating scale (NRS), where 0 ¼ no pain and 10 ¼ worst imaginable pain, and a verbal scale with the labels none, mild, moderate, and severe, both of which are widely used and have evidence of reliability and validity.1,2 The latter was translated to Krio as none, small, half and half, and serious. Where more than one pain was identified by the patient, only the first was used in this analysis. Mood ratings were collected using a horizontal 10-cm VAS, with the extremes labeled very happy and very sad. Bilateral upper limb amputees were unable to point but could indicate their mood on the VAS while the clinician ran a pointer along the line. Mood also was rated using a verbal scale in English with the terms very sad, sad, normal, happy, very happy; or in Krio, heart poil bad, heart poil, e nor bad, glady, glady too mos.
Data Analysis A comparison of the numerical and verbal pain scales was made by Kendall’s tau-b, after
Three hundred sixteen patients were assessed, of whom 303 provided pain measures: 223 men and 80 women, with a mean age of 39 years (range 12e80). For the 247 individuals for whom data were available, mean time since injury was 3.3 years, with a range of 10 to 48 months. One hundred three patients recorded only one pain location; the remainder had multiple pain sites, up to a maximum of six. Paired pain scores (numerical and verbal) were available from 275 of the 303 patients (91%). Amputees made up most of these patients (180/275, 65%), and the most common pain site was associated with the amputated limb. Of the 269 patients for whom data were available, 107 reported pain from arm, hand, or finger amputation and 48 from leg, foot, or toe amputation. Of the remainder, 41 reported pain in the shoulder, arm, or hand; 40 in the buttock, leg, or foot; 10 had back pain; 10 had pain in the head, face, or mouth; eight reported chest pain; six reported generalized body pain; and two each reported abdominal or pelvic pain and neck pain. As both amputees and nonamputees (including those with unsuccessful attempts at amputation) were heterogeneous groups with overlapping characteristics, it was not thought useful to compare them.
304
Williams et al.
Comparison of Verbal and Numerical Pain Scores Despite one outlier who scored worst pain as 1 out of 10 on the numerical scale and serious on the verbal scale, and two who scored small on the Krio verbal scale and nine out of 10 on the numerical scale, verbal and numerical scales showed a modest positive correlation, with a Kendall’s tau-b ¼ 0.431 (P < 0.001; n ¼ 275). Means for the NRS were calculated for the three pain intensity words (small, half and half, and serious) on the verbal scale. This revealed a progressive increase in the numerical score from small ¼ 3.38 (95% confidence interval [CI] 2.72e4.05) to half and half ¼ 4.40 (95% CI 3.95e4.85), to serious ¼ 6.21 (95% CI 5.92e6.49), with serious endorsed more than half and half. An ANOVA showed a between-groups F (degrees of freedom [df] ¼ 2) of 41.80 (P < 0.001) and post hoc testing (Tukey’s HSD) gave a statistically significant difference between serious and half and half (P < 0.001) and a just statistically significant difference between small and half and half (P ¼ 0.047). These findings suggest that the numerical and verbal scales were used reasonably consistently to make distinctions between levels of pain. This was particularly true at the upper end of the verbal scale, where almost all (46/49) of the pains rated 7e10 were also described as serious and the remaining three of 49 as half and half.
Education Data on education were available for 173 of 269 patients (64%). Ninety patients (52%) had no formal education, 18 (10.5%) had primary education, 32 (18.5%) had secondary education, and 33 (19%) had education beyond secondary level. (For reference, about half of adult men and about a quarter of adult women in Sierra Leone are literate [https://www.cia. gov/library/publications/the-world-factbook/ geos/sl.html]). Educational level did not influence the relationship between verbal and numerical scores. For those with no education or primary only (n ¼ 108), the tau-b for the relationship between NRS and VRS for pain was 0.468. For those with at least secondary education (n ¼ 65), tau-b ¼ 0.371. For those for whom education level was unknown (n ¼ 96), tau-b ¼ 0.380: all values P < 0.001.
Vol. 42 No. 2 August 2011
Change with Treatment
For 30 patients, NRS (n ¼ 28) and verbal (n ¼ 30) measures of pain in the same location were compared for three different occasions, initial assessment, during treatment, and after the treatment was stopped (B, C, and E in Fig. 1). The numerical scores satisfied assumptions for parametric testing and showed statistically significant change over time (F ¼ 8.45, df ¼ 2, P ¼ 0.001), with means of 5.46 (standard deviation [SD] 1.73) at initial assessment, 3.57 (SD 2.18) during treatment, and 4.29 (SD 2.07) after treatment stopped. There were significant linear (F ¼ 7.17, P ¼ 0.012) and quadratic functions (F ¼ 9.47, P ¼ 0.005) indicating, as expected, that pain was greatest at initial assessment before treatment, least during treatment, and recurred to some extent after cessation of treatment. The verbal scale did not satisfy requirements for parametric testing: a nonparametric Friedman test showed significant change across time with a median score of serious at initial assessment, small during treatment, and half and half after the end of treatment (c2 ¼ 13.98, df ¼ 2, P < 0.001).
Mood Fewer paired scores (VAS and verbal) were collected for mood (n ¼ 71) than for the paired pain scores (n ¼ 275). The mood VAS was discontinued because it took too long to explain reliably in the field. There were 53 men and 18 women respondents with a mean age of 41 years (range 16e70). Ten patients responded to the English language mood verbal descriptor and 61 to the Krio version. As there was no significant difference in variances (F ¼ 0.04), the two were combined for further analysis. Of these 71 respondents, two (2.8%) scored very sad, 11 (15.5%) scored sad, 26 (36.6%) scored normal, 30 (42.3%) scored happy, and two (2.8%) scored very happy. The mood VAS and verbal scales were correlated at a reasonable level (Pearson’s r ¼ 0.591, P < 0.001). Fig. 2 shows a progressive increase in mean VAS score across each point of the verbal mood scale. Because there were only two individuals in each of the very sad and very happy categories, these were collapsed into the adjacent sad and happy categories. A oneway ANOVA gave a between-groups F of 14.75 (P < 0.001). Post hoc testing showed
Vol. 42 No. 2 August 2011
Pain and Mood Assessment in a Refugee Camp
a progressive increase in the VAS means for the three verbal groups, sad: 43.8 (95% CI 31.9e55.6, n ¼ 13), normal: 59.3 (95% CI 52.4e66.2, n ¼ 26), and happy: 73.3 (95% CI 67.6e78.9, n ¼ 32).
Discussion Despite initial concern about comprehension of pain scales, all but a few of the patients interviewed were able to provide both numerical and translated verbal ratings for their pain. Only three of 275 patients provided inconsistent numerical and verbal ratings (indicating a failure to understand one or both). Scales appeared to be understood by the patients once explained and reliably used on more than one occasion. There was a good relationship between the two pain rating scales, as shown by the correlation coefficient, and satisfactory distinction between the verbal labels. Although the pain locations in this group were dominated by pain from amputations, there were a variety of other chronic pains, so the findings allow generalization beyond amputation pain alone. Level of education did not appear to affect the use of the scales. Although this is consistent with findings from Brazil14 and Nigeria,12 two other studies, from Zimbabwe13 and Mexico,15 suggest that six or more years of education is an aid to comprehension of pain scales. Our results did not indicate that the amount of formal education received is a prime determinant of comprehension of pain scales.
Fig. 2. Box and whisker plot of mood visual analogue score against verbal descriptor scale.
305
Translation may have posed fewer problems in Sierra Leone than it might in some other settings because the Krio language is English based. However, other unrelated languages were sometimes the first language of patients. Training of the lay translators was necessary, and ideally, periodic checks would be made on the consistency within and between these translators. Mood data are harder to interpret because of additional complexities of meaning. It is interesting to observe that most patients rated themselves happy on the verbal mood scale, despite their circumstances, and perhaps the comparison was with their lives at the time of injury or with having been spared worse injury or death. Further exploration of this would facilitate interpretation of scores. The use of a VAS was abandoned because of difficulty explaining it to patients. We have not examined whether a numerical scale, as used successfully for pain, would have been satisfactory. Alternatively, the problem may lie in the construct of mood, conceptualized and expressed differently across cultures, and perhaps an unexpected question in a medical context. There are several limitations to this work. There is no consensus on best practice for cross-cultural measurement,7 and best practice proposals are generally formulated in the context of a developed host country rather than by consultation with those who might use it in a less developed setting. Studies that include several language groups are inevitably dependent on translation and explanation, and as a result, findings can never be completely standardized across patients. The pain scores with verbal and numerical scales were lower after treatment, suggesting that sensitivity was satisfactory. However, this result was based on an analysis of mean scores, not change within individuals. Last, the issue of appropriate and informed ethical approval remains unresolved in such settings, and where suitable institutions do not exist in the country where the data are collected, institutional research ethics committees in the developed world need to clarify their responsibilities. The utility of the pain scale might have been improved by the addition of a term to express worse pain than serious on the pain VRS, the most heavily endorsed term. This would be consistent with the findings of
306
Williams et al.
a European working group on cancer pain,16 which recommended that verbal scales should have more than four points (including no pain as the lowest point). The chaotic national context and limited size of the clinic team in this study required the compromise of using unidimensional pain rating scales, where ideally multidimensional pain rating scales would be used;10,17,18 the same compromise also is made in many better resourced settings. Despite the above limitations, our investigations gave us confidence in shared meaning of the numerical and verbal pain scales, not dependent on education. It subsequently proved possible to assess pain well enough to evaluate pain interventions,4 identifying for treatment those with the most severe pain. Using only the limited resources of clinician, translators, and pain scales, both the restricted verbal pain scale and numerical scale presented visually met the needs of the treatment service. The verbal mood scale also functioned adequately. There is interest in developing effective assessment tools for such settings and a pressing need for such tools,19 but the necessary conditions for development of culturally embedded scales8 are unlikely ever to be achieved in resource-poor and unstable settings. Nevertheless, this study demonstrates a serious attempt to ensure accurate interpretation of responses to unfamiliar measurement instruments, using a method that could be replicated in similar settings. Supplementing such studies with qualitative examination of response selection, and testing reliability across a clinically realistic time period without intervention, would add further confidence in the use of these scales. The World Health Organization and other agencies are supportive of the case for analgesia as part of health care strategies. Effective treatment of pain depends on adequate pain assessment, and this study suggests that some of the same scoring approaches for pain and mood used in the developed world can work in resource-poor settings such as this.8 It is reasonable to expect that these findings should generalize further than Sierra Leone, and further similar studies will enable us to select with confidence the best ways to assess people’s pain in under-resourced and conflict settings.
Vol. 42 No. 2 August 2011
Disclosures and Acknowledgments The residents of Murray Town Amputees Camp and Grafton War Wounded Camp, Freetown, Sierra Leone, were generous of their time. We are grateful to the late Ruth Miller for data entry and analysis, to Pauline McGoldrick for advice on questionnaire design, and to Bill Macrae for project support.
References 1. Jensen MP, Karoly P. Self-report scales and procedures for assessing pain in adults. In: Turk DC, Melzack R, eds. Handbook of pain assessment, 2nd ed. New York: Guilford Press, 2001: 15e34. 2. Turk DC, Okifuji A. Assessment of patients’ reporting of pain: an integrated perspective. Lancet 1999;353:1784e1788. 3. Green CR, Anderson KO, Baker TA, et al. The unequal burden of pain: confronting racial and ethnic disparities in pain. Pain Med 2003;4:277e294. 4. Lacoux PA, Ford N. Treatment of neuropathic pain in Sierra Leone. Lancet Neurol 2002;1: 190e195. 5. Lacoux PA, Crombie IK, Macrae WA. Pain in traumatic upper limb amputees in Sierra Leone. Pain 2002;99:309e312. 6. Napoles-Springer AM, Stewart AL. Overview of qualitative methods in research with diverse populations. Making research reflect the population. Med Care 2006;44:S5eS9. 7. Johnson TP. Methods and frameworks for crosscultural measurement. Med Care 2006;44:S17eS20. 8. McHorney CA, Fleishman JA. Assessing and understanding measurement equivalence in health outcome measures. Issues for further quantitative and qualitative inquiry. Med Care 2006;44: S205eS210. 9. Elfenbein HA, Ambady N. On the universality and cultural specificity of emotion recognition: a meta-analysis. Psychol Bull 2002;128:202e235. 10. Edwards CL, Fillingim RB, Keefe F. Race, ethnicity, and pain. Pain 2001;94:133e137. 11. Edwards RR, Moric M, Husfeldt B, Buvanendran A, Ivankovich O. Ethnic similarities and differences in the chronic pain experience: a comparison of African-American, Hispanic, and White patients. Pain Med 2005;6:88e98. 12. Soyannwo OA, Amanor-Boadu SD, Sanya AO, Gureje O. Pain assessment in NigeriansdVisual Analogue Scale and Verbal Rating Scale compared. West Afr J Med 2000;19:242e245. 13. Jelsma JM, Machiri G, Madzivire DM. The use of pain measurement scales in the Zimbabwean context. Cent Afr J Med 1997;43:256e259.
Vol. 42 No. 2 August 2011
Pain and Mood Assessment in a Refugee Camp
307
Quarasma MR, Aquino LRL, of pain scales in the assessment illiterate patients with rheumaJ Rheumatol 1990;17:1022e
Palliative Care. J Pain Symptom Manage 2002;23: 239e255.
15. Clark P, Lavielle P, Martınez H. Learning from pain scales: patient perspective. J Rheumatol 2003; 30:1584e1588.
18. Clark WC, Yang JC, Tsui S-L, Ng K-F, Clark SB. Unidimensional pain rating scales: a multidimensional affect and pain survey (MAPS) analysis of what they really measure. Pain 2002;98:241e247.
14. Ferraz MB, et al. Reliability of literate and toid arthritis. 1024.
16. Caraceni A, Cherny N, Fainsinger R, et al. Pain measurement tools and methods in clinical research in palliative care: recommendations of an Expert Working Group of the European Association of
17. Price DD. Psychological mechanisms of pain and analgesia. Seattle, WA: IASP Press, 1999.
19. De Jong JT, Komproe IH, Van Ommeren M. Common mental disorders in postconflict settings. Lancet 2003;361:2128e2130.
Journal of Pain and Symptom Management
301
Brief Methodological Report
Assessing Pain and Mood in a Poorly Resourced Country in a Post-Conflict Setting Amanda C. de C. Williams, BSc, MSc, PhD, Minha Rajput-Ray, BSc, Ost Med, MBChB, MICR, Xavier Lassalle, CRNA, Iain Crombie, PhD, CStat, FFPH, and Philippe Lacoux, MBBS, BSc, MD, FRCA, FFPMRCA Research Department of Clinical, Educational and Health Psychology (A.C.d.C.W.), University College London, London; Addenbrooke’s Hospital (M.R.-R.), Cambridge University Hospitals NHS Foundation Trust, Cambridge; Department of Public Health (I.C.), University of Dundee Ninewells Hospital and Medical School, Dundee, United Kingdom; and Department of Anesthesia (P.L.), Technical Advisor for Anaesthesia, MSF (X.L.), Paris, France
Abstract Context. Accurate pain assessment is important but presents problems in poorly resourced countries. A civil war in Sierra Leone resulted in civilian casualties with pain from deliberate amputation and other injuries. Objectives. To examine the quality of simple pain and mood scales. Methods. Within a pain treatment project, pain was assessed using numerical and verbal descriptor scales, and mood using visual analogue and verbal descriptor scales. The relationships between these scales (translated where necessary) were examined by comparison of pairs of measures. Results. The overwhelming majority (99%) used the scales consistently. The verbal pain scale showed substantial discrimination between the pain words mild/small, moderate/half and half, and severe/serious (F ¼ 41.80, P < 0.001). Numerical and verbal pain scales were related at a modest level (Kendall’s tau-b ¼ 0.39, P < 0.001, n ¼ 272) and not dependent on the level of education. A smaller sample (n ¼ 30) provided pain data across three assessment occasions, and both pain scales appeared sensitive to change. The 5-point verbal mood scale collapsed into three categories, with reasonable distinction between mood words (F ¼ 14.75, P < 0.001). The visual analogue scale proved difficult to explain to civilian casualties in this setting. Conclusion. Verbal pain and mood scales and numerical pain scales appeared to yield useful information in a post-conflict situation. This suggests that adapting these established rating scales for pain, and for mood, was useful to clinicians and acceptable to patients. J Pain Symptom Manage 2011;42:301e307. Ó 2011 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved. Key Words Rating scales, pain, mood, underdeveloped country, post-conflict, education level
Address correspondence to: Amanda C. de C. Williams, BSc, MSc, PhD, Research Department of Clinical, Educational and Health Psychology, University College London, Gower Street, London WC1E Ó 2011 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved.
6BT, United Kingdom. E-mail: amanda.williams@ ucl.ac.uk Accepted for publication: November 6, 2010. 0885-3924/$ - see front matter doi:10.1016/j.jpainsymman.2010.11.012
302
Williams et al.
Introduction Pain rating methods are well established in the developed world for systematic pain assessment,1,2 but these methods cannot be assumed to apply in underdeveloped countries. Assessment is an essential feature of health care, and inadequate assessment, generally and in diverse ethnic groups, is associated with poorer treatment.3 Sierra Leone is a resource-poor West African country with a population of approximately five and a half million and predominantly small-scale agricultural economy. It achieved independence from British colonial rule in 1961. There is a diversity of languages, predominantly Krio (based on English). Civil War started in 1991 and lasted for over a decade.4 During the war, civilians suffered many injuries, including the organized amputation of limbs, most often of arms below the elbow. An initial assessment showed that pain problems were common,5 leading the nongovernmental organization M edecins Sans Fronti eres (MSF; www.msf.org.uk) to decide to treat chronic pain in this unstable setting. The challenge in rating pain in poorly resourced countries arises from differences in culture, education, and language. The simplest approach, translation of instruments developed in Western countries, makes the assumption that well-established measures will transfer to cultures other than that in which they were standardized.6 Others argue that more attention should be paid to investigating shared meaning before examining psychometric qualities7,8 and that cultural and educational differences require more fundamental examination of how measures are used. Differences in the use of a question across cultures can arise from differences in understanding, ease of retrieving the information required to answer, and differences in forming the response.7,8 This is particularly true of words and constructs for emotion because emphasis and differentiation can vary considerably,9 making it hard to find equivalent terms when translating a measurement instrument. There have been consistent reports of cultural differences in distress associated with pain, even where pain intensity shows no difference.10 Inevitably, responding to scales requires reference to implicit normsdof disability, for instancedthat vary between cultures.8 In addition, tendencies
Vol. 42 No. 2 August 2011
to use or avoid extremes of scales might vary systematically with cultural norms and with the dominance of collectivism or individualism.8 Educational level can affect use of scales.11 Of the two African studies identified, one reported no effect of education on the use of a Yoruba-translated visual analogue scale (VAS) and verbal rating scale (VRS) by Nigerian patients with musculoskeletal pain,12 whereas the other found that Zimbabwean maternity patients after Caesarean section only used a Shona-translated VAS and VRS consistently with one another when they had received more than seven years of education.13 This study in Sierra Leone investigated whether numerical and verbal pain rating scales and VAS and verbal mood rating scale, based closely on those widely used in the developed world,1 could be used to express pain by respondents, and interpreted meaningfully by clinicians, in this poorly resourced health care setting at a time of national chaos. That is, the study aimed to investigate shared meaning. It also assessed whether responses on these scales were related to extent of education.
Methods People with pain came from two temporary camps in the area around Freetown (capital of Sierra Leone) to a primary health care center in one of the camps. Most injuries had occurred in 1998 or 1999, up to four years before. Patients were assessed during the therapeutic consultation, usually in English but interpreted where required. Most consultations were with one of the authors (P. L.) and the remainder with a clinician supervised by him. Information also was collected by a trained local health care worker in English or in Krio (translated), using a questionnairebased interview. The questionnaire covered demographics, pain location, pain intensity, mood, function including prosthesis use, and education level. If the patient did not understand, a health worker experienced with the questionnaire translated the clinician’s questions to Krio or other dialects. Subsequent assessments reviewed the effects of treatment4 using the same pain and mood measures. The assessment occasions and numbers are shown in Fig. 1. Data in this study were
Vol. 42 No. 2 August 2011
Pain and Mood Assessment in a Refugee Camp
Fig. 1. Timeline of assessment and treatment of pain in Sierra Leone. A ¼ pilot assessment, n ¼ 40; B ¼ initial clinical assessment, n ¼ 304; C ¼ review on medication, n ¼ 86; D ¼ review at completion of medication, n ¼ 45; E ¼ review after completion of medication, n ¼ 122.
taken from therapeutic records with no patient identifiers, so confidentiality was not an issue. Related work used photographs and video, with verbal consent; patients were largely keen for their plight to be advertised as widely as possible. There was no national or institutional ethics committee from whom permission could be sought.
303
inspecting the NRS for normality and treating the verbal pain scale as ordinal. A one-way analysis of variance (ANOVA) with post hoc testing for significant differences between adjacent words was performed. Analyses were repeated, dichotomizing education, to examine whether education level made a difference. Sensitivity of the NRS to change with treatment was examined in a subsample of 30 patients who were assessed before, during, and after treatment (B, C, and E, respectively, in Fig. 1) by a parametric ANOVA, testing for linear (continuous change) and quadratic (discontinuous change) functions, and by a nonparametric ANOVA on the verbal scale. Discontinuous change would indicate a return toward baseline after initial improvement. Paired mood ratings (VAS and verbal) from each patient were compared using Pearson’s correlation and ANOVA with post hoc testing (Tukey’s honestly significant difference [HSD]).
Results Pain and Mood Rating Scales Translations to Krio were developed through liaison with bilingual MSF office workers in the locality and were piloted on the first few patients seen, before their regular use. Pain ratings were requested using a vertical 11-point numerical rating scale (NRS), where 0 ¼ no pain and 10 ¼ worst imaginable pain, and a verbal scale with the labels none, mild, moderate, and severe, both of which are widely used and have evidence of reliability and validity.1,2 The latter was translated to Krio as none, small, half and half, and serious. Where more than one pain was identified by the patient, only the first was used in this analysis. Mood ratings were collected using a horizontal 10-cm VAS, with the extremes labeled very happy and very sad. Bilateral upper limb amputees were unable to point but could indicate their mood on the VAS while the clinician ran a pointer along the line. Mood also was rated using a verbal scale in English with the terms very sad, sad, normal, happy, very happy; or in Krio, heart poil bad, heart poil, e nor bad, glady, glady too mos.
Data Analysis A comparison of the numerical and verbal pain scales was made by Kendall’s tau-b, after
Three hundred sixteen patients were assessed, of whom 303 provided pain measures: 223 men and 80 women, with a mean age of 39 years (range 12e80). For the 247 individuals for whom data were available, mean time since injury was 3.3 years, with a range of 10 to 48 months. One hundred three patients recorded only one pain location; the remainder had multiple pain sites, up to a maximum of six. Paired pain scores (numerical and verbal) were available from 275 of the 303 patients (91%). Amputees made up most of these patients (180/275, 65%), and the most common pain site was associated with the amputated limb. Of the 269 patients for whom data were available, 107 reported pain from arm, hand, or finger amputation and 48 from leg, foot, or toe amputation. Of the remainder, 41 reported pain in the shoulder, arm, or hand; 40 in the buttock, leg, or foot; 10 had back pain; 10 had pain in the head, face, or mouth; eight reported chest pain; six reported generalized body pain; and two each reported abdominal or pelvic pain and neck pain. As both amputees and nonamputees (including those with unsuccessful attempts at amputation) were heterogeneous groups with overlapping characteristics, it was not thought useful to compare them.
304
Williams et al.
Comparison of Verbal and Numerical Pain Scores Despite one outlier who scored worst pain as 1 out of 10 on the numerical scale and serious on the verbal scale, and two who scored small on the Krio verbal scale and nine out of 10 on the numerical scale, verbal and numerical scales showed a modest positive correlation, with a Kendall’s tau-b ¼ 0.431 (P < 0.001; n ¼ 275). Means for the NRS were calculated for the three pain intensity words (small, half and half, and serious) on the verbal scale. This revealed a progressive increase in the numerical score from small ¼ 3.38 (95% confidence interval [CI] 2.72e4.05) to half and half ¼ 4.40 (95% CI 3.95e4.85), to serious ¼ 6.21 (95% CI 5.92e6.49), with serious endorsed more than half and half. An ANOVA showed a between-groups F (degrees of freedom [df] ¼ 2) of 41.80 (P < 0.001) and post hoc testing (Tukey’s HSD) gave a statistically significant difference between serious and half and half (P < 0.001) and a just statistically significant difference between small and half and half (P ¼ 0.047). These findings suggest that the numerical and verbal scales were used reasonably consistently to make distinctions between levels of pain. This was particularly true at the upper end of the verbal scale, where almost all (46/49) of the pains rated 7e10 were also described as serious and the remaining three of 49 as half and half.
Education Data on education were available for 173 of 269 patients (64%). Ninety patients (52%) had no formal education, 18 (10.5%) had primary education, 32 (18.5%) had secondary education, and 33 (19%) had education beyond secondary level. (For reference, about half of adult men and about a quarter of adult women in Sierra Leone are literate [https://www.cia. gov/library/publications/the-world-factbook/ geos/sl.html]). Educational level did not influence the relationship between verbal and numerical scores. For those with no education or primary only (n ¼ 108), the tau-b for the relationship between NRS and VRS for pain was 0.468. For those with at least secondary education (n ¼ 65), tau-b ¼ 0.371. For those for whom education level was unknown (n ¼ 96), tau-b ¼ 0.380: all values P < 0.001.
Vol. 42 No. 2 August 2011
Change with Treatment
For 30 patients, NRS (n ¼ 28) and verbal (n ¼ 30) measures of pain in the same location were compared for three different occasions, initial assessment, during treatment, and after the treatment was stopped (B, C, and E in Fig. 1). The numerical scores satisfied assumptions for parametric testing and showed statistically significant change over time (F ¼ 8.45, df ¼ 2, P ¼ 0.001), with means of 5.46 (standard deviation [SD] 1.73) at initial assessment, 3.57 (SD 2.18) during treatment, and 4.29 (SD 2.07) after treatment stopped. There were significant linear (F ¼ 7.17, P ¼ 0.012) and quadratic functions (F ¼ 9.47, P ¼ 0.005) indicating, as expected, that pain was greatest at initial assessment before treatment, least during treatment, and recurred to some extent after cessation of treatment. The verbal scale did not satisfy requirements for parametric testing: a nonparametric Friedman test showed significant change across time with a median score of serious at initial assessment, small during treatment, and half and half after the end of treatment (c2 ¼ 13.98, df ¼ 2, P < 0.001).
Mood Fewer paired scores (VAS and verbal) were collected for mood (n ¼ 71) than for the paired pain scores (n ¼ 275). The mood VAS was discontinued because it took too long to explain reliably in the field. There were 53 men and 18 women respondents with a mean age of 41 years (range 16e70). Ten patients responded to the English language mood verbal descriptor and 61 to the Krio version. As there was no significant difference in variances (F ¼ 0.04), the two were combined for further analysis. Of these 71 respondents, two (2.8%) scored very sad, 11 (15.5%) scored sad, 26 (36.6%) scored normal, 30 (42.3%) scored happy, and two (2.8%) scored very happy. The mood VAS and verbal scales were correlated at a reasonable level (Pearson’s r ¼ 0.591, P < 0.001). Fig. 2 shows a progressive increase in mean VAS score across each point of the verbal mood scale. Because there were only two individuals in each of the very sad and very happy categories, these were collapsed into the adjacent sad and happy categories. A oneway ANOVA gave a between-groups F of 14.75 (P < 0.001). Post hoc testing showed
Vol. 42 No. 2 August 2011
Pain and Mood Assessment in a Refugee Camp
a progressive increase in the VAS means for the three verbal groups, sad: 43.8 (95% CI 31.9e55.6, n ¼ 13), normal: 59.3 (95% CI 52.4e66.2, n ¼ 26), and happy: 73.3 (95% CI 67.6e78.9, n ¼ 32).
Discussion Despite initial concern about comprehension of pain scales, all but a few of the patients interviewed were able to provide both numerical and translated verbal ratings for their pain. Only three of 275 patients provided inconsistent numerical and verbal ratings (indicating a failure to understand one or both). Scales appeared to be understood by the patients once explained and reliably used on more than one occasion. There was a good relationship between the two pain rating scales, as shown by the correlation coefficient, and satisfactory distinction between the verbal labels. Although the pain locations in this group were dominated by pain from amputations, there were a variety of other chronic pains, so the findings allow generalization beyond amputation pain alone. Level of education did not appear to affect the use of the scales. Although this is consistent with findings from Brazil14 and Nigeria,12 two other studies, from Zimbabwe13 and Mexico,15 suggest that six or more years of education is an aid to comprehension of pain scales. Our results did not indicate that the amount of formal education received is a prime determinant of comprehension of pain scales.
Fig. 2. Box and whisker plot of mood visual analogue score against verbal descriptor scale.
305
Translation may have posed fewer problems in Sierra Leone than it might in some other settings because the Krio language is English based. However, other unrelated languages were sometimes the first language of patients. Training of the lay translators was necessary, and ideally, periodic checks would be made on the consistency within and between these translators. Mood data are harder to interpret because of additional complexities of meaning. It is interesting to observe that most patients rated themselves happy on the verbal mood scale, despite their circumstances, and perhaps the comparison was with their lives at the time of injury or with having been spared worse injury or death. Further exploration of this would facilitate interpretation of scores. The use of a VAS was abandoned because of difficulty explaining it to patients. We have not examined whether a numerical scale, as used successfully for pain, would have been satisfactory. Alternatively, the problem may lie in the construct of mood, conceptualized and expressed differently across cultures, and perhaps an unexpected question in a medical context. There are several limitations to this work. There is no consensus on best practice for cross-cultural measurement,7 and best practice proposals are generally formulated in the context of a developed host country rather than by consultation with those who might use it in a less developed setting. Studies that include several language groups are inevitably dependent on translation and explanation, and as a result, findings can never be completely standardized across patients. The pain scores with verbal and numerical scales were lower after treatment, suggesting that sensitivity was satisfactory. However, this result was based on an analysis of mean scores, not change within individuals. Last, the issue of appropriate and informed ethical approval remains unresolved in such settings, and where suitable institutions do not exist in the country where the data are collected, institutional research ethics committees in the developed world need to clarify their responsibilities. The utility of the pain scale might have been improved by the addition of a term to express worse pain than serious on the pain VRS, the most heavily endorsed term. This would be consistent with the findings of
306
Williams et al.
a European working group on cancer pain,16 which recommended that verbal scales should have more than four points (including no pain as the lowest point). The chaotic national context and limited size of the clinic team in this study required the compromise of using unidimensional pain rating scales, where ideally multidimensional pain rating scales would be used;10,17,18 the same compromise also is made in many better resourced settings. Despite the above limitations, our investigations gave us confidence in shared meaning of the numerical and verbal pain scales, not dependent on education. It subsequently proved possible to assess pain well enough to evaluate pain interventions,4 identifying for treatment those with the most severe pain. Using only the limited resources of clinician, translators, and pain scales, both the restricted verbal pain scale and numerical scale presented visually met the needs of the treatment service. The verbal mood scale also functioned adequately. There is interest in developing effective assessment tools for such settings and a pressing need for such tools,19 but the necessary conditions for development of culturally embedded scales8 are unlikely ever to be achieved in resource-poor and unstable settings. Nevertheless, this study demonstrates a serious attempt to ensure accurate interpretation of responses to unfamiliar measurement instruments, using a method that could be replicated in similar settings. Supplementing such studies with qualitative examination of response selection, and testing reliability across a clinically realistic time period without intervention, would add further confidence in the use of these scales. The World Health Organization and other agencies are supportive of the case for analgesia as part of health care strategies. Effective treatment of pain depends on adequate pain assessment, and this study suggests that some of the same scoring approaches for pain and mood used in the developed world can work in resource-poor settings such as this.8 It is reasonable to expect that these findings should generalize further than Sierra Leone, and further similar studies will enable us to select with confidence the best ways to assess people’s pain in under-resourced and conflict settings.
Vol. 42 No. 2 August 2011
Disclosures and Acknowledgments The residents of Murray Town Amputees Camp and Grafton War Wounded Camp, Freetown, Sierra Leone, were generous of their time. We are grateful to the late Ruth Miller for data entry and analysis, to Pauline McGoldrick for advice on questionnaire design, and to Bill Macrae for project support.
References 1. Jensen MP, Karoly P. Self-report scales and procedures for assessing pain in adults. In: Turk DC, Melzack R, eds. Handbook of pain assessment, 2nd ed. New York: Guilford Press, 2001: 15e34. 2. Turk DC, Okifuji A. Assessment of patients’ reporting of pain: an integrated perspective. Lancet 1999;353:1784e1788. 3. Green CR, Anderson KO, Baker TA, et al. The unequal burden of pain: confronting racial and ethnic disparities in pain. Pain Med 2003;4:277e294. 4. Lacoux PA, Ford N. Treatment of neuropathic pain in Sierra Leone. Lancet Neurol 2002;1: 190e195. 5. Lacoux PA, Crombie IK, Macrae WA. Pain in traumatic upper limb amputees in Sierra Leone. Pain 2002;99:309e312. 6. Napoles-Springer AM, Stewart AL. Overview of qualitative methods in research with diverse populations. Making research reflect the population. Med Care 2006;44:S5eS9. 7. Johnson TP. Methods and frameworks for crosscultural measurement. Med Care 2006;44:S17eS20. 8. McHorney CA, Fleishman JA. Assessing and understanding measurement equivalence in health outcome measures. Issues for further quantitative and qualitative inquiry. Med Care 2006;44: S205eS210. 9. Elfenbein HA, Ambady N. On the universality and cultural specificity of emotion recognition: a meta-analysis. Psychol Bull 2002;128:202e235. 10. Edwards CL, Fillingim RB, Keefe F. Race, ethnicity, and pain. Pain 2001;94:133e137. 11. Edwards RR, Moric M, Husfeldt B, Buvanendran A, Ivankovich O. Ethnic similarities and differences in the chronic pain experience: a comparison of African-American, Hispanic, and White patients. Pain Med 2005;6:88e98. 12. Soyannwo OA, Amanor-Boadu SD, Sanya AO, Gureje O. Pain assessment in NigeriansdVisual Analogue Scale and Verbal Rating Scale compared. West Afr J Med 2000;19:242e245. 13. Jelsma JM, Machiri G, Madzivire DM. The use of pain measurement scales in the Zimbabwean context. Cent Afr J Med 1997;43:256e259.
Vol. 42 No. 2 August 2011
Pain and Mood Assessment in a Refugee Camp
307
Quarasma MR, Aquino LRL, of pain scales in the assessment illiterate patients with rheumaJ Rheumatol 1990;17:1022e
Palliative Care. J Pain Symptom Manage 2002;23: 239e255.
15. Clark P, Lavielle P, Martınez H. Learning from pain scales: patient perspective. J Rheumatol 2003; 30:1584e1588.
18. Clark WC, Yang JC, Tsui S-L, Ng K-F, Clark SB. Unidimensional pain rating scales: a multidimensional affect and pain survey (MAPS) analysis of what they really measure. Pain 2002;98:241e247.
14. Ferraz MB, et al. Reliability of literate and toid arthritis. 1024.
16. Caraceni A, Cherny N, Fainsinger R, et al. Pain measurement tools and methods in clinical research in palliative care: recommendations of an Expert Working Group of the European Association of
17. Price DD. Psychological mechanisms of pain and analgesia. Seattle, WA: IASP Press, 1999.
19. De Jong JT, Komproe IH, Van Ommeren M. Common mental disorders in postconflict settings. Lancet 2003;361:2128e2130.