The impact of sampling and measurement on the prevalence of self-reported pain in Canada

1Department of Anesthesiology, Queen's University; 2Deparment of Community Health and Epidemiology, Queen's University; 3Clinical Research Unit, Kingston General Hospital; 4Deparment of Medicine (Rheumatology), Queen's University; 5Deparment of Medicine (Biochemistry), Queen's University; 6Deparment of Surgery, Queen's University, Kinston, Ontario; 7list appears at the end of the article Correspondence: Dr Elizabeth G Van Den Kerkhof, Deparment of Anesthesiology, Queen's University, Kingston General Hospital, 76 Stuart Street, Kingston, Ontario K7L 2V7. Telephone 613-549-6666x3964, fax 613-548-1375, e-mail ev5@post.queensu.ca EG Van Den Kerkhof, WM Hopman, TE Towheed, TP Anastassiades, DH Goldstein. The impact of sampling and measurement on the prevalence of self-reported pain in Canada. Pain Res Manage 2003;8(3):157-163.

chronic pain, the wording of specific questions and a lack of representativeness in studies whose results are being generalized to the general population.The purpose of this paper is to critically examine the potentially misleading use of overall prevalence estimates in the pain literature and to use two Canadian population-based crosssectional surveys to illustrate the impact of measurement (pain duration and questions asked) and sampling (eg, age and sex of the respondents) on prevalence estimates.To determine trends in the reporting of chronic pain in Canada, data from the 1996/97 NPHS were compared to data obtained from the 1994/95 NPHS.

METHODS AND MATERIALS
Two secondary data sets were used for this analysis the 1996 NPHS and the Canadian Multicentre Osteoporosis Study (CaMos) (3,4).The NPHS was conducted by Statistics Canada and was designed to collect information related to the health of the Canadian population.The survey has a crosssectional and a longitudinal component.The 1996/97 cross-sectional survey consisted of 210,377 general interviews and 81,804 in-depth interviews.The overall response rate at the household level was 82.6%.The 1996/97 longitudinal sample consisted of 16,168 respondents with an overall response rate of 93.6%.Additional details about the NPHS are available elsewhere (5).The data were weighted to reflect the sample design and the adjustments for nonresponse, and post-stratification (3) and estimates were calculated to represent the total Canadian population.The NPHS is based on randomly selected noninstitutional respondents of Canadian households.This paper includes NPHS respondents age 25 years and older who were interviewed in the 1996/97 cross-sectional study.
The CaMos is a prospective cohort study involving 9423 randomly selected Canadian men and women aged 25 years or older and living in the community.The sample was drawn from a 50 km radius of nine Canadian cities, including St John's, Newfoundland; Halifax, Nova Scotia; Quebec City, Quebec; Kingston, Toronto and Hamilton, Ontario; Saskatoon, Saskatchewan; Calgary, Alberta; and Vancouver, British Columbia.Households were randomly selected from telephone directories, mailed a letter of introduction, and subsequently telephoned to assess their interest in participation.If more than one eligible person resided in the home, a random number table was used to select the participant.Of the 80,163 households sampled, 59.0% were ineligible because the age, sex or calendar period stratum had already been filled.Another 7.8% were invalid numbers, and 5.2% were unreachable after 12 attempts.Of the remaining households, 28.4% declined to participate, 29.6% completed a short questionnaire that provided information about the age, sex distribution and fracture history of the residents, and 9423 (42.0%) went on to participate fully in the study.Additional details about the CaMos study are available elsewhere (4,6).Ethical approval for the study was obtained through the review boards of each participating centre, as well as through the coordinating centre in Montreal.
CaMos is designed to collect epidemiological data related to the incidence and prevalence of osteoporosis.As a result, the sampling framework is designed to include more women than men, and a higher number of older than younger Canadian residents, based on the current knowledge regarding osteoporosis risk factors (4).The study collects detailed demographic information, family history, medical history, medication use, activity, food consumption, health-related quality of life (HRQOL) and lifestyle variables.HRQOL was assessed by means of the Medical Outcomes Trust's 36-item short form health survey (SF-36) (7), and the Health Utilities Index (HUI) (8,9).
Demographic data and responses to the pain variables were obtained from the databases.This paper is based on the prevalence of reported pain over three time periods (usual pain, pain in the past four weeks and pain in the past week).The NPHS used three questions to reference 'usual pain' with no specified time frame.For the purpose of this paper, the NPHS definition will be referred to as 'chronic pain'.CaMos collected responses to the SF-36, which referenced pain in the past four weeks and the HUI, which referenced pain in the week before participation.Both the SF-36 and the HUI would therefore collect both short term or acute pain as well as chronic pain.However, they will be referred to collectively as short term pain and will be referred to individually as 'four-week pain' and 'one-week pain', respectively.Table 1 contains the content and response options for the questions used.Data are presented as frequencies and percentages overall and stratified by age and sex.The study is designed to describe the prevalence of self-reported pain on the basis of question asked, time frame used, age of respondent and sex of respondent.In addition, the overall prevalence estimates of pain reported in the CaMos (four-week pain and one-week pain) have been adjusted, using direct standardization, to the population distribution of the NPHS, thereby removing the effect of sampling method and allowing for assessment of the impact of measurement on overall rates.Direct standardization using the Canadian Census data was repeated and produced the same results (data not included).Bivariate analysis was not performed to assess for statistically significatnt differences in prevalence estimates by age or sex because statistically significant P values would have been generated for even the smallest differences in prevalence, due to large sample sizes.

RESULTS
The demographic characteristics of the NPHS and CaMos study participants are presented in Table 2. Due to the different sampling frames, CaMos study participants were more likely to be female and older compared with the NPHS participants (Table 2).Table 3 presents unadjusted and adjusted overall prevalence estimates.The CaMos overall unadjusted pain prevalence estimates were higher than the estimates adjusted to the NPHS data.Overall adjusted prevalence of one-week pain was 39% and four-week pain was 66%.The prevalence of pain interfering with work in the past four weeks was 38%.Fifteen per cent of respondents reported chronic pain.Tables 4-7 are based on unadjusted CaMos figures.

One-week pain (CaMos -HUI)
The prevalence and impact of pain in the past week on activities are presented in Table 4. Forty-six per cent of the sample reported pain.Women were more likely to report pain than men (48.4% versus 40.5%).Men were more likely to report mild to moderate pain preventing no activities (49.5% versus 43.1%).The degree to which moderate to severe pain prevented some or most activities generally increased with increased age for both sexes.
The prevalence, medication use and disruption of normal activities due to pain in the past week are presented in Table 5. Forty-eight per cent of the sample reported pain.Women were more likely to report pain than men (50.7% versus 41.3%).Of those who reported pain, approximately two-thirds reported occasional pain that was relieved by nonprescription drug use.Women were somewhat more likely than men to report pain that was not relieved by any drugs (2.6% versus 1.9%).

Four-week pain (CaMos -SF-36)
The prevalence and severity of bodily pain in the past four weeks reported in the CaMos survey are presented in Table 6.Sixty-nine per cent of the sample reported pain.Half reported   very mild or mild pain, while 9.8% reported severe or very severe pain.Men were more likely to report very mild pain, while women were more likely than men to report moderate to very severe pain.The highest prevalence of very severe pain was reported in women 25 to 34 years of age (2.3%).When severe and very severe pain were combined, the prevalence of pain tended to increase with increasing age.The extent to which bodily pain interfered with normal work in the four weeks before the study is presented in Table 7.
Forty-two per cent of the sample reported pain interfering with work.Women were more likely than men to report pain interfering with work (45.4% versus 35.3%).Fifty-five per cent who reported that pain interfered with work reported that it interfered a little bit, 28.8% reported moderate interference and 16.1% reported that it interfered quite a bit or extremely.Men were more likely than women to report that pain interfered with work a little bit.Younger men were more likely than older men to report that pain interfered extremely with work, while older women were more likely than younger women to report the same.

Chronic pain (NPHS -1996/97)
The prevalence and severity of chronic pain based on the 1996/97 NPHS data are presented in Table 8.Just over 15% (15.1%) reported chronic pain.Prevalence among women was 16.4% and prevalence among men was 13.7%.In women, prevalence increased from 9.8% in the 25 to 34 age group to 28.7% for ages 75 and older.The prevalence in men for the same age groups increased from 8.1% to 27.1%.Prevalence increased at a relatively constant rate in women with the largest increase seen in the 45 to 64 year age range, however, in men, the largest increase occurred in ages 75 and older when it increased to 27.1%.The impact of chronic pain on activities is presented in Table 9.Of those who reported pain, 25.0% reported no impact on activities and 17.8% reported that pain prevented most activities.Men were more likely than women to report no activity limitation (28.9% versus 21.9%), but were also more likely to report that pain prevented most activities (19.7% versus 16.3%).Women were more likely than men to report that pain prevented few (36.0%versus 31.0%) or some (25.8% versus 20.5%) activities.Activity limitation increased with age for both sexes, but it increased more dramatically for women than men.
Comparison of 1994/95 and 1996/97 NPHS Results from the previous (1994-95) NPHS ( 1) revealed an overall prevalence of chronic pain of 17%, while the current (1996-97) NPHS data revealed a prevalence of 15.1%.Although there has been a reduction in the overall prevalence of chronic pain in Canada over the two-year sampling period between the two NPHS surveys, the distribution of pain severity has not appreciably changed.Women are still more likely to report chronic pain and increasing age is strongly associated with greater prevalence estimates of chronic pain in both sexes.

DISCUSSION
Pain has varying levels of prevalence in the general population and its epidemiology is not well understood.Estimates of chronic pain in the general population range from 2% to 46% (2).The principle sources of heterogeneity in pain prevalence estimates are in the measurement domain (measurement differences with respect to pain duration and specific question wording) and in the sampling domain (lack of representativeness of the general population) as illustrated by the prevalence estimates reported in the NPHS and the CaMos.Adjustment for sampling bias through direct standardization of the CaMos with the NPHS marginally decreased the heterogeneity between prevalence estimates; however, wide variation in prevalence estimates remained.The remaining variation is largely due to measurement bias (ie, definition used).
International and national studies display similar variations in prevalence estimates.A random sample of patients from general practices in the United Kingdom, reported the prevalence of chronic pain at 46.5% (2), while a population-based telephone survey in Australia reported a prevalence of 17% in men and 20% in women (10).The prevalence of chronic pain in an Israeli population was estimated at 10% (11), while Croft et al (12) found a prevalence of 11% in a postal survey of adults in England.Prevalence estimates in Canadian studies ranged from 11% to 44% (13,14).These studies, drawn from the chronic pain literature, illustrate that the primary factors influencing the reported level of chronic pain in the general population are related to the selection of the subsample and the definition used.Definitions in the literature ranged from general definitions, such as, 'usually have pain or discomfort', to more specific definitions, such as pain defined by a validated pain questionnaire.In Canadian studies, the highest levels of pain reported used either a validated questionnaire (2,15), or were based on a pain definition of a relatively lengthy duration (13), as illustrated by the studies by Birse (13) and Moulin (16).Both cross sectional telephone surveys defined pain as lasting at least six months.In 1998, Birse reported a chronic pain prevalence of 44% (men respondents=34.5%,women respondents=65.5%) in the general population in Edmonton (13).In 2002 Moulin reported a prevalence of chronic pain of 29% in a stratified random sample of Canadians (16) , however, the estimate may have been affected by sampling bias given the response rate of 19.1%.The lowest level of reported pain (11%) was reported by Crook in 1984 (14).A two-part definition of pain was used to capture persistent pain ("Do you usually or always have pain?";"Has this pain been present in the past two weeks?").Those not reporting pain in the past two weeks were excluded from the persistent pain sample.Finally, a change in awareness and attitude towards pain over the past 18 years may have also had an impact on the disparity in prevalence estimates in the abovementioned studies.
Contrary to the stated literature, we found that use of a validated question and a relatively lengthy duration resulted in the lowest levels of reported pain (NPHS).The high prevalence estimates of short term pain were based on standardized estimates of HRQOL that included questions regarding pain and were targeted towards individuals more likely to experience higher levels of pain, with greater representation of older age strata (CaMos).Yet, standardization of the estimates only resulted in slight decreases in the overall estimate of pain, providing further evidence for the importance of measurement (ie, definition) in comparing prevalence estimates of pain.
The strengths of the CaMos and the NPHS relate to the fact that they are both large, randomly conducted and rigorously sampled, population-based Canadian studies with adequate response rates, making them likely to be representative of the Canadian noninstitutionalized population.A comparison of the results of each of the different questions provides valuable insights not only into the prevalence of self-reported pain in the Canadian population, but also the significant impact of question asked, the time frame used, and the age and sex of the respondent on the responses obtained.
In terms of the prevalence of pain in Canada, the results of the SF-36 and HUI are more similar to one another than they are to the NPHS.The NPHS consistently reports lower percentages of people affected with chronic pain, probably  because the questions designed to measure the prevalence of pain were quite different between the two studies.The SF-36 asked about the prevalence of any bodily pain within the past four weeks, while the HUI asked if the subject was free of pain and discomfort in the past week.Both the SF-36 and HUI will likely capture those with short term acute pain (eg, headaches, migraines, dental pain or menstrual cramps), as well as chronic pain.The NPHS, in contrast, asked the subjects if they were usually free of pain and discomfort, with no time frame attached.The NPHS question, therefore, is more likely to identify those with chronic and long lasting pain complaints, rather than minor or transient bodily pain.
There are a number of limitations with the NPHS and CaMos data.First, the analyses do not cover residents of institutions, and thereby, likely exclude a substantial number of people who have pain.Second, the data are self-reported, and have not been independently validated.Third, like all other available prevalence estimates, these data are cross-sectional, which limits the conclusions that can be drawn.
However, this study yields useful information about the selfreported responses to a variety of questions assessing pain in the general population.It is particularly useful in illustrating the impact of the question asked, the age of the respondent and the sex of the respondent on the estimates obtained.These results highlight the need for extremely careful definitions of the variable of interest and the time frame selected, as these have a significant impact on the responses elicited.In addition, any study attempting to estimate the prevalence of pain in the population must ensure representativeness of the general population through both adequate sampling and satisfactory response rates.
Responses to the different questions likely represent different categories of short term and chronic pain, which in turn may have different epidemiological risk factors and profiles.Longitudinal studies of the epidemiology, predictors, and natural history of chronic pain are urgently needed in the Canadian population.Such studies may be useful in identifying those whose acute pain syndromes are more likely to become chronic and persistent (17)(18)(19).Future research with the CaMos and NPHS survey data will be directed at identifying the correlates of those reporting chronic pain in terms of demographic variables, and the concomitant presence of other diseases and chronic conditions, as well as assessing long term changes in the population.

2 .
How would you describe the usual intensity of your pain or discomfort?Mild/Moderate/Severe 3. How many activities does your pain or discomfort prevent?None/Few/Some/Most CaMos -SF-36 1.How much bodily pain have you had during the past four weeks?None/Very mild/Mild/Moderate/Severe/Very severe 2. During the past four weeks, how much did pain interfere with your normal work (including both work outside the home and housework)?Not a bit/A little bit/Moderately/Quite a bit/Extremely CaMos -HUI 1. Are you free of pain and discomfort?(refers to 'over the past week') Yes/No 2. If not, which one of the following best describes your level of pain?Do you have: a) mild to moderate pain that prevents no activities, b) moderate pain that prevents a few activities, c) moderate to severe pain that prevents some activities, d) severe pain that prevents most activities.3. Are you free of pain and discomfort?(refers to 'over the past week' and is asked at the end of the questionnaire for a second time) Yes/No 4. If not, which one of the following best describes your usual level of pain?a) Occasional pain.Discomfort relieved by nonprescription drugs or self-control activity without disruption of normal activities b) Frequent pain.Discomfort relieved by oral medicines with occasional disruption of normal activities c) Frequent pain.Frequent disruption of normal activities.Discomfort requires prescription narcotics for relief d) Severe pain.Pain not relieved by drugs and constantly disrupts normal activities.Data from references 3 and 4. HUI Health Utilities Index; SF-36 Medical Outcomes Trust 36-item short form health survey

TABLE 1
Pain questions asked in the National Population and Health Survey (NPHS) and the Canadian Multicentre Osteoporosis Study (CaMos) NPHS 1. Are you usually free of pain or discomfort?

TABLE 2
Overall prevalence of pain and demographic characteristics of participants in the 1996-97 National Population and Health Survey (NPHS) and the Canadian Multicentre Osteoporosis Study (CaMos) *May not add up to 100% due to rounding and missing or not stated responses; † Data not available.

TABLE 5
Prevalence of pain and impact on medication use (%) in the one week before participation in Canadian Multicentre Osteoporosis Study by sex and age group, 1995/96

TABLE 6
Prevalence and severity of pain (%) in the four weeks before participation in Canadian Multicentre Osteoporosis Study by sex and age group, 1995/96

TABLE 7
Extent to which pain Interfered with normal work (%) in the four weeks before participation in Canadian Multicentre Osteoporosis Study by sex and age group, 1995/96

TABLE 9
Prevalence and impact on activities of chronic pain in Canada by sex and age group based on the National Population and Health Survey 1996/97

TABLE 8
Prevalence and severity of chronic pain in Canada by sex and age group based on the National Population and Health Survey 1996/97 Data from references 3 and 4. Responses to the question on the severity of pain were not provided by 36,490 individuals