Should people who have joint symptoms, but no diagnosis of arthritis from a doctor, be included in surveillance efforts?†
The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
In 2005, 27% of adults reported doctor-diagnosed arthritis, and 14% reported chronic joint symptoms but no doctor-diagnosed arthritis (i.e., possible arthritis). We evaluate the value of including persons classified as having possible arthritis in surveillance of arthritis.
In 2005, Kansas, Oklahoma, North Carolina, and Utah added extra questions to their Behavioral Risk Factor Surveillance System (BRFSS) telephone survey targeted to a subsample of those classified as having possible arthritis.
Persons classified as having possible arthritis (n = 2,884) were younger, more often male, and had less activity limitation than persons with doctor-diagnosed arthritis. Of those classified as having possible arthritis, half had seen a doctor for their symptoms, 12.5% reported arthritis, and 61.9% gave other causes. Of the half who had not seen a doctor, most reported mild symptoms (64.8%).
Only 6.3% of those classified as having possible arthritis had what we considered to be arthritis. Most who did not see a doctor reported mild symptoms and, therefore, would be unlikely to be amenable to medical and public health interventions for arthritis. Although including possible arthritis would slightly improve the sensitivity of detecting arthritis in the population, it would increase false-positives that would interfere with targeting state intervention efforts and burden estimates. The ability to add back questions to the BRFSS survey allows for the reintroduction of possible arthritis in case national surveillance indicates it necessary or if studies document an increased rate at which possible arthritis turns into arthritis. Currently, possible arthritis does not need to be included in state arthritis surveillance efforts, and limited question space on surveys is better spent on other arthritis issues.
Arthritis is the most common cause of disability (1) in the US, is an important contributor to health care costs (2), and is expected to increase in prevalence as the population ages (3). Arthritis surveillance, the ongoing and systematic collection, analysis, and interpretation of data (4), allows monitoring of the prevalence of arthritis. This monitoring is crucial to planning health services and programs to prevent arthritis-related disability, as well as for tracking progress toward national Healthy People objectives (5).
Identifying adults with arthritis by using self-reported information from telephone or in-person interviews is a practical method of public health surveillance that is intended to monitor disease burden and help target arthritis interventions. Questions to identify people who have arthritis or chronic joint symptoms have been on the National Health Interview Survey (NHIS) since 2002 and on the Behavioral Risk Factor Surveillance System (BRFSS) since 1997. Respondents are classified as having doctor-diagnosed arthritis if they report that a doctor had ever told them they have arthritis or another arthritis-related condition. People who report chronic joint symptoms (i.e., joint pain in the past 30 days that began more than 3 months prior), but who have not been told by a doctor that they have arthritis, are classified as having “possible arthritis.”
Given the limited space for arthritis-related questions on surveys like the BRFSS, whether to maintain surveillance of possible arthritis is an important consideration. The purpose of our study was to better describe the characteristics of people classified as having possible arthritis to help evaluate whether their continued surveillance was warranted.
Subjects and Methods
Data were collected as part of the 2005 BRFSS. The BRFSS survey is a state-based, random, and digit-dialed telephone survey of the noninstitutionalized, civilian US adult population ages ≥18 years and is conducted annually in all 50 states and the District of Columbia. Each state administers the core BRFSS survey and can add questions to the end of the survey. Kansas, North Carolina, Oklahoma, and Utah each added the same set of special survey questions for part of the year with the goal of accumulating a subsample of a minimum of 600 respondents classified as having possible arthritis in each state.
Respondents to the core survey were classified as having possible arthritis if they answered “yes” to the first 2 and “no” to the third of the following questions: 1) The next question refers to your joints. Please do not include the neck or back. During the past 30 days have you had any pain, aching, or stiffness in or around a joint?, 2) Did your joint symptoms first begin more than 3 months ago?, and 3) Have you ever been told by a doctor or other health professional that you have some form of arthritis, rheumatoid arthritis, gout, lupus, or fibromyalgia?
Those who answered yes to the first 2 questions were then asked, “Have you ever seen a doctor or other health professional for these joint symptoms?” Respondents were also asked, “Are you now limited in any way in any of your usual activities because of arthritis or joint symptoms?”
The subsample of respondents classified as having possible arthritis was asked the following special survey questions: 1) How many months or years ago did those joint symptoms first begin?, 2) Were the symptoms present on most days for at least 1 month?, 3) Over the past 7 days, how many days have you had pain, aching, or stiffness from arthritis or joint symptoms?, and 4) On a scale of 0 to 10, where 0 is no pain and 10 is pain or aching as bad as it could be, and over the past 7 days, how bad was your joint pain or aching on average?
Respondents who reported a pain score of ≥7 were classified as having severe pain. Respondents were also asked, “Have you taken any prescription or nonprescription medication for these joint symptoms in the past 7 days?”
Respondents who reported that they had never seen a doctor for their joint symptoms were asked, “Earlier you said you had never seen a doctor about your joint symptoms. Please tell me why you did not see a doctor about your joint symptoms?” Respondents who reported they had seen a doctor for their joint symptoms were asked, “Earlier you said you had seen a doctor for your joint symptoms. What did the doctor say was the cause of your joint symptoms?” The last 2 questions had defined response categories, but interviewers were given the option to enter a verbatim response if they were not sure how to categorize a response. Three coauthors (JB, CGH, and JJS) each reviewed all of the text responses and classified them independently. Disagreements in coding were resolved by consensus before final codes were entered.
The analyses used information from the core BRFSS survey and the subsample. A statistical software package (Sudann; RTI International) was used to calculate point estimates and 95% confidence intervals to account for the complex survey design of the BRFSS. Data were weighted to reflect each state's most recent adult population estimates. Unweighted sample sizes and weighted percentages were reported. We used the full BRFSS survey to compare the demographic characteristics and activity limitations of the possible arthritis group with the doctor-diagnosed arthritis group. Then, we used the subsample with possible arthritis to compare the responses of those who had seen a doctor for their joint symptoms with the responses of those who had not seen a doctor for their symptoms.
In 2005, the overall BRFSS response rates (i.e., the percentage of persons who completed interviews among all eligible persons, including those who were not successfully contacted) were 51.8% for Kansas, 44.6% for North Carolina, 53.6% for Oklahoma, and 59.4% for Utah. The median cooperation rates (i.e., the percentage of persons who completed interviews among all eligible persons contacted) were 81.1% for Kansas, 77.4% for North Carolina, 81.4% for Oklahoma, and 81.5% for Utah (6).
For those 4 states combined, the total 2005 BRFSS sample was 44,326. Approximately 27.2% of adults (n = 15,786) were classified as having doctor-diagnosed arthritis and 14% as having possible arthritis (n = 6,064). Persons classified as having possible arthritis had a different demographic profile than those with doctor-diagnosed arthritis; they were significantly younger (i.e., more likely to be ages <45 years [54.5% versus 22.5%; P < 0.001]), more often male (53.3% versus 41.3%; P < 0.001), and had less activity limitation (18.8% versus 39.3%; P < 0.001) than persons with doctor-diagnosed arthritis (Table 1). The demographic characteristics of all respondents classified as having possible arthritis were similar to the subsample of respondents classified as having possible arthritis, who answered the additional special study questions.
Table 1. Characteristics of respondents with doctor-diagnosed arthritis, possible arthritis, and possible arthritis respondents who answered additional questions*
|Age, years|| || || || || || |
|Race/ethnicity|| || || || || || |
| African American||10.3||9.7–11.0||8.7||7.8–9.8||5.1||4.1–6.4|
|Activity limitation, yes||39.3||38.3–40.4||18.8||17.5–20.3||18.1||16.2–20.1|
Among those classified as having possible arthritis in the subsample (n = 2,884), there were 2 primary subgroups of interest: those who had seen a doctor for their symptoms (n = 1,388) and those who had not seen a doctor for their symptoms (n = 1,496). Compared with the adults who did not see a doctor for their symptoms, those who saw a doctor had a significantly higher (P < 0.001) prevalence of arthritis-attributable activity limitation (24.0% versus 12.2%), more frequent symptoms (joint symptoms on most days of the past month [75.3% versus 54.1%; P < 0.001]), severe pain (16.4% versus 9.7%; P < 0.001), and use of pain medication (40.3% versus 30.6%; P < 0.001) (data not shown).
Among those adults classified as having possible arthritis and who saw a doctor for their symptoms (n = 1,388), only 12.5% reported a specific diagnosis consistent with arthritis as defined by the National Arthritis Data Workgroup (7). In addition, 28.1% reported injury- or work-related joint problems, 23.3% reported musculoskeletal conditions not related to arthritis, 7.1% had joint problems caused by other nonmusculoskeletal diseases, and 25.6% did not know what the doctor told them or reported they did not know what the doctor said (Table 2).
Table 2. Doctor diagnosis of chronic joint symptoms among respondents with possible arthritis by demographic characteristics*
|Age, years|| || || || || || || || |
|Race/ethnicity|| || || || || || || || |
| African American||17.0¶||29.4¶||7.6¶||26.9¶||9.7¶||0.7¶||8.8¶||69|
Among adults classified as having possible arthritis who did not see a doctor (n = 1,496), 64.8% were not motivated to seek medical treatment (i.e., their symptoms were not that bad, they did not want to go, or said they were going soon) (Table 3). An additional 18.2% said they did not go to a doctor because of the cost, 7.9% reported that they were able to successfully manage their symptoms on their own, 1.2% had actually seen a doctor, and 7.9% did not know why they had not seen a doctor.
Table 3. Respondent reasons for not visiting a doctor for chronic joint symptoms by demographic characteristics*
|Age, years|| || || || || || |
|Race/ethnicity|| || || || || || |
| African American||44.7‡||20.9‡||6.0‡||2.1‡||26.2‡||52|
There were 156 respondents (6.3% of the total subsample and 12.5% of persons who had seen a doctor about their symptoms) who were misclassified as having possible arthritis on the basis of their initial response to the arthritis case-finding question (Table 2). Although they first reported they did not have doctor-diagnosed arthritis, they reported diagnoses that were classified as arthritis when asked for more details regarding what the doctor had told them about their conditions. In general, this group had mild symptoms and was more similar to those with possible arthritis than those with doctor-diagnosed arthritis when we compared the prevalence of activity limitation for the 3 groups (15.9% for the misclassified group, 18.8% for possible arthritis, and 39.3% for doctor-diagnosed arthritis; P < 0.01).
Arthritis surveillance is intended to monitor disease burden and help target arthritis interventions. The different demographic profiles (i.e., younger, more often male, and less activity limitation) of those classified as having possible arthritis suggest that they are different from those who have doctor-diagnosed arthritis. The analysis shows that only 6.3% of those with possible arthritis reported a specific arthritis-related condition (12.5% of the one-half of possible arthritis respondents who saw a doctor for their arthritis symptoms) and therefore would be considered as a case of arthritis. However, among those who saw a doctor for their symptoms, 25.6% of the respondents did not know what the doctor told them or reported the doctor did not give them a diagnosis. If we assume a similar distributional response as the 74.4% of those who reported a diagnosis, then the percentage of all those with possible arthritis who had an arthritis-related condition could rise to 16.8%. Even at this level, it appears that the majority of those with possible arthritis do not have arthritis, and including them in burden counts of arthritis would add many false-positives. Unfortunately, the rate at which this group will develop arthritis is unknown and merits further study.
There is a tradeoff between detecting cases of arthritis (sensitivity) and properly identifying those without it (specificity). When failure to detect is costly (e.g., missing a case of syphilis can result in further spread), sensitivity is valued over specificity in public health surveillance. When surveillance is primarily used to monitor disease burden and target interventions, high specificity is preferable to high sensitivity for the best use of limited resources. Because only 6.3% of persons (perhaps as many as 8.4%) classified as having possible arthritis appear to have arthritis, including possible arthritis as cases of arthritis in surveillance would lower specificity by adding large numbers of false-positives (i.e., people with other conditions) and misdirect arthritis programs and resources to persons not in particular need of interventions.
Of those classified as having possible arthritis who did not see a doctor for their symptoms, the majority were not motivated to do so, and most had mild symptoms that did not interfere with daily activities. Current medical and public health arthritis-related interventions focus on helping persons with arthritis with some degree of activity limitation or pain become more active or manage their arthritis symptoms (8). It appears unlikely that people classified as having possible arthritis, who did not see a doctor about their symptoms, would be motivated to participate in such interventions.
This study has strengths and limitations. The questions were asked of a large sample of adults in 4 geographically and demographically diverse states. Verbatim responses yielded additional details about people classified as having possible arthritis, the severity and cause of their symptoms, and why many of them choose not to seek medical care. One limitation is that joint symptoms are self-reported in the BRFSS survey and have not been confirmed by a health care provider or objective monitoring. However, such self-reports have been shown to be valid for surveillance purposes (9, 10). The BRFSS is a telephone survey and did not include persons without landline telephones, persons in the military, or those residing in institutions at the time of this survey. The BRFSS response rates were low, and although the BRFSS weighting procedures partially correct for nonresponse, the effect of low response rates is uncertain.
The Centers for Disease Control and Prevention (CDC) Arthritis Program concluded that only a small percentage of those people classified as having possible arthritis appeared to have arthritis, and that those classified as having possible arthritis would not likely be amenable to currently available public health programmatic state-based interventions for arthritis because of their low level of activity limitation. Increasing limitations in space in the BRFSS and other information gaps led to difficult choices about surveillance coverage. Accordingly, in 2009, the CDC Arthritis Program replaced questions about possible arthritis on the BRFSS with questions relevant to other important arthritis public health issues. The questions about chronic joint symptoms that support identification of those classified as having possible arthritis remain on the NHIS to allow tracking of the Healthy People 2010 goal (5) to increase the proportion of adults with chronic joint symptoms who have seen a health care provider for their symptoms. The ability to add back questions at cyclic intervals to the BRFSS survey allows for the reintroduction of possible arthritis questions in case national surveillance indicates it necessary or if studies document an increased rate at which possible arthritis turns into arthritis.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Bolen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Bolen, Helmick, Sacks.
Acquisition of data. Bolen, Gizlice, Potter.
Analysis and interpretation of data. Bolen, Helmick, Sacks, Gizlice, Potter.
The authors thank Gary Langmaid of Northrop Grumman for computer programming services, Kathryn Marti, RN, MPH, at the Utah Department of Health for her assistance with data collection, and Derek Pate, MPH, and Joyce L. Kirksey at the Oklahoma State Department of Health for their assistance in pilot testing the special survey questions and data collection.