Changes in self-efficacy and health status over 5 years: A longitudinal observational study of 306 patients with rheumatoid arthritis




To investigate changes in self-efficacy and health status over 5 years in patients with rheumatoid arthritis (RA), the relationships between these changes, and the influence of baseline values on subsequent changes.


306 adult patients with RA, born in 1926 or later, were examined by questionnaire in 1994 and again in 1999. We analyzed data regarding pain (visual analogue scale [VAS], Arthritis Impact Measurement Scale [AIMS2] symptom scale, Short Form-36 [SF-36] pain scale), fatigue (VAS, SF-36 vitality scale), mental distress (AIMS2 affect scale, SF-36 mental health scale) and self-efficacy (Arthritis Self-Efficacy Scales for pain and for other symptoms).


On group level, all health status measures were numerically somewhat improved, and self-efficacy slightly reduced. Changes in self-efficacy and in corresponding health status measures were significantly correlated. For patients with above average educational level self-efficacy for pain at baseline was positively correlated to improvement in pain measures. Good mental health at baseline was correlated to improvement in self-efficacy for other symptoms, but only for patients with below average educational level.


Baseline self-efficacy seems to influence future level of perceived pain and baseline mental health status seems to influence future self-efficacy. These associations seem to be affected by level of education.


It is widely recognized that psychological and sociological characteristics of patients influence the outcome of chronic diseases, including rheumatoid arthritis (RA) (1–4). However, the mechanisms and their relative contribution are poorly understood, despite the availability of several measures of socioeconomic and psychological status (5, 6). The concept of self-efficacy has provided a useful tool for investigation of some of these associations. The self-efficacy theory was developed within the framework of social-cognitive theory by Bandura (7). Perceived self-efficacy is defined as “Peoples' judgement of their capabilities to organize and execute courses of action required to attain designated types of performance” (7). It is concerned not with the skill one has, but with the judgement of what one can do with whatever skill one possesses. This judgement, whether accurate or faulty, is based on four major sources of information: mastery experiences, modeling, social persuasion, and psychological state. Enactive experience seems to be the most powerful determinant of self-efficacy level (8).

Although self-efficacy is related to other psychological concepts, such as self-esteem and locus of control, self-efficacy differs by being very specific to a given behavior (9). Thus, self-efficacy is measured as a task specific state, not as a generalized trait (10). For example, a patient may have high self-efficacy for using medication regularly, but low self-efficacy regarding carrying out prescribed exercise. Self-efficacy regarding a specific domain is little changed by transient day-to-day moods (8), and is thought to be fundamental to behavioral change, by governing the interrelationship between knowledge and action (7). It seems that individuals' ability to act in knowledge-consistent ways, for example in following the doctor's advice, depends on their level of self-efficacy (11). Lorig et al (10) developed and evaluated the Arthritis Self-Efficacy Scale (ASES) to measure perceived self-efficacy in people with arthritis. Several authors, mainly in cross-sectional studies, have shown associations between self-efficacy and certain aspects of RA (2, 4, 12–17), for example negative correlations with level of fatigue (12), daily pain (13), and patients' demand for informal and professional care (14), and positive association with compliance with treatment programs (4, 15, 16).

Self-efficacy may be changed by therapeutic interventions, like the Arthritis Self-Management Program (18), presenting disease-related information and problem solving techniques, and focusing on enhanced self-efficacy. As far as we have been able to assess through database searches, few studies have examined longitudinal change in self-efficacy unrelated to intervention programs aimed at enhancing it. Lefebvre et al (13) found that self-efficacy was related to perceived pain and mood in a period of 30 days prior to measurement of self-efficacy, but stated that their study could not determine the direction of the relationships. In a previous study we found that high self-efficacy scores were associated with favorable changes in health status over two years (19). In that study we did not include ASES in the second data collection, and thus were unable to investigate the directionality of the correlation between self-efficacy and health-status. We therefore decided to examine self-efficacy as well as health status again—this time five years after the initial survey, to possibly contribute some answers regarding the magnitude of change and the directionality of the relations. We had three hypotheses: first, that changes in self-efficacy and in corresponding health status measures were positively associated; second, that baseline self-efficacy level was positively associated with improved health status; third, that favorable health status at baseline was positively associated with increased self-efficacy. We wanted to examine the strength of each relationship, to try to assess the factors with greatest impact on the future health of the patient with RA.


The Oslo RA register.

The basis of the study was the permanent Oslo RA register (20), where inclusion criteria are a diagnosis of RA (21) and a residential address in Oslo. Patients with juvenile arthritis, i.e., disease onset before the age of 16, are excluded. Inclusion of patients began in 1991 and a variety of procedures were performed to identify all patients with RA living in Oslo. The register is continuously updated, and now includes about 1,600 living patients. The completeness of the register is estimated to be 85% (20, 22).

Data collection and study sample.

The two waves of data collection were done in a slightly different way. In 1994, a 30-page questionnaire comprising health status measures and self-efficacy was mailed to the 1,542 patients in the Oslo RA register, of which 1,024 (66.4%) responded. As part of a longitudinal observational study, patients born in 1926 or later were invited to a followup examination in 1999, which also included completion of some of the questionnaires that were mailed in 1994. The total number of patients in the Oslo RA-register possibly eligible for being incorporated into the five-year cohort was 659 (born in 1926 or later, being in the Oslo RA-register in 1994, and alive and still living in Oslo in 1999). A total of 306 (46.4%) patients have paired data from 1994 and 1996 and constitute the sample for the present analyses.

Measures on self-efficacy and health status.

The ASES is a valid and reliable instrument developed to measure self-efficacy in patients with rheumatic diseases (10). The questionnaire designed by Lorig et al contains 5 items concerning coping with pain, 6 items concerning other symptoms related to RA (fatigue, mental distress), and 9 items concerning function. Each item presents a statement where the patient may agree or disagree. The scores are expressed with values between 10 and 100, with 10 the lowest possible self-efficacy level. Because of the size of the questionnaire used in 1994 (30 pages) we had to omit some instruments and questions, among them the self-efficacy function scale. We thus express self-efficacy as 2 scores: 1 regarding pain and 1 regarding other symptoms.

In addition, the 2 waves of data collection comprised several instruments examining self-reported health status: the Arthritis Impact Measurement Scales (AIMS2) (23), the Short Form-36 (SF-36) (24) and a 100 mm visual analogue scale (VAS) for assessment of joint pain and fatigue. The AIMS2 comprise 78 items, and health status may be expressed by 5 sub-scales (physical, symptom [= pain], role, social interaction, affect) with scores in the 0–10 range (0 = the best score). The SF-36 is a 36-item generic health status measure providing information on eight different aspects of health status (physical functioning, physical role, mental health, emotional role, pain, social functioning, general health, and vitality). It has been used in a variety of conditions and has been adapted to and validated in various populations, including the Norwegian, according to internationally accepted guidelines (25). SF-36 is expressed with values between 0 and 100 (0 = the poorest score).

The questionnaire scales had previously been translated into Norwegian according to published guidelines and examined for their performance in patients with RA (25).

Statistical analyses.

In this study we exclusively present health status data from the sub-scales giving information on pain (AIMS2 symptom, SF-36 bodily pain, VAS), fatigue (SF-36 vitality scale, VAS), and mental distress (AIMS2 affect, SF-36 mental health). We carried out analyzes of the correlation between measures on self-efficacy and health-status within similar dimensions: ASES pain was related to perceived intensity of pain and ASES symptom to level of fatigue and mental distress.

Statistical analyses were made using the Statistical Packages for the Social Sciences version 9.0 (SPSS, Chicago, IL) (SPSS for Windows version 9.0). Pearson's correlation coefficients were computed to examine the cross-sectional bivariate relationships between study variables related to pain, between variables related to other symptoms, in 1994 as well as in 1999, and between changes in self-efficacy and in related health status measures over the 5-year period. Correlations with coefficients of 0.30 or below were regarded as weak, between 0.30 and 0.70 as moderate to substantial, and with coefficients of 0.70 or above as high.

Multiple linear regression analyses were performed to examine the relationships between baseline level of self-efficacy and related health status measures after 5 years, adjusted for the baseline score of the health status measure, and the relationships between baseline level of the health status measures and self-efficacy after 5 years, adjusted for the baseline level of self-efficacy. Subgroup analyses were performed on patients with low education (7–11 years) and high education (12–18 years). Level of statistical significance was chosen at 5% (P < 0.05). Bonferroni adjustment was carried out to exclude chance significance, i.e., the P values were multiplied by the number of analyses made.


The respondents to both questionnaires (n = 306, which were 46.4% of the potentially eligible patients in the Oslo RA register) had a mean age (in 1994) of 53.3 ± 11.5 years, a mean disease duration of 11.0 ± 9.3 years, 49.7% had a positive test result for rheumatoid factor (RF), 80.4% were females, 59.8% were living with a spouse, and mean educational length was 11.9 ± 3.42 years. The non-respondents (n = 353) comprised less females (71.7%), but otherwise did not differ significantly (mean age 52.8 ± 12.1 years), mean disease duration 11.8 ± 8.8 years, mean educational length 11.9 ± 3.38 years, and 50.4% had a positive test result for RF).

The disease course over the 5-year study period was rather stable on group level (Table 1), both for ASES and health status measures. However, there were major interindividual variations as indicated by the 10th and 90th percentiles for change. The pain scores (VAS and AIMS2) as well as the SF-36 vitality score were slightly improved on a group level (Table 1). The disability score (Modified Health Assessment Questionnaire, range 1–4) of the cohort was 1.57 at baseline and increased by 0.05 over the period (data not shown).

Table 1. Baseline scores of arthritis self-efficacy scales and health status measures in 306 rheumatoid arthritis patients*
Measures (score)1994 Mean ± SDChanges 1994–1999
Mean (95% CI)10th percentile90th percentile
  • *

    VAS = visual analog scale; AIMS2 = Arthritis Impact Measurement Scales; SF-36 = Short-form 36.

  • Low score = poor health.

  • Low score = good health.

  • §

    Statistically significant.

Self-efficacy pain (10–100)54.6 (18.4)−0.9 (−3.2, 1.4)−24.024.0
Self-efficacy symptoms (10–100)65.8 (17.0)−1.5 (−3.5, 0.5)−20.018.3
Pain VAS (0–100)44.8 (21.9)−7.7 (−10.3, −5.2)§−37.217.6
AIMS2 symptom (0–10)5.25 (2.55)−0.27 (−0.53, −0.01)§−3.303.00
SF-36 pain (0–100)43.1 (19.9)0.3 (−2.1, 2.7)−22.027.6
Fatigue VAS (0–100)50.2 (25.5)−4.4 (−7.6, 1.2)−40.231.0
SF-36 vitality (0–100)40.5 (21.1)4.4 (2.1, 6.6)§−15.030.0
AIMS2 affect (0–10)3.25 (1.67)−0.11 (−0.25, 0.34)−1.751.25
SF-36 mental health (0–100)71.6 (18.3)1.5 (−0.4, 3.5)−18.820.0

The bivariate cross-sectional correlations between the different study variables in 1994 and in 1999 are shown in Table 2 (variables regarding pain) and Table 3 (variables regarding fatigue and mental distress). Correlations were strong between measures on similar dimension of health within the same year, and otherwise moderate to substantial. The correlation between self-efficacy pain and each of the 3 pain measures was almost identical in 1994 and in 1999, even if self-efficacy pain in 1994 and in 1999 correlated only 0.5 (Table 2). For the measures on fatigue and mental health the patterns were similar, but less obvious (Table 3). ASES pain in 1994 was more strongly associated with pain level 5 years later than was pain level in 1994 with subsequent ASES (Table 2).

Table 2. Pearson correlation coefficient matrix for study variables related to pain (1994 and 1999)*
  • *

    All P values <0.001. See Table 1 for definitions.

Self-efficacy pain 1994 (X1)1.00       
Pain VAS 1994 (X2)−0.341.00      
AIMS2 symptom 1994 (X3)−0.360.711.00     
SF-36 pain 1994 (X4)0.35−0.73−0.771.00    
Self-efficacy pain 1999 (X5)0.50−0.18−   
Pain VAS 1999 (X6)−0.340.490.48−0.45−0.401.00  
AIMS2 symptom 1999 (X7)−0.350.520.58−0.52−0.380.791.00 
SF-36 pain 1999 (X8)0.29−0.43−0.450.460.44−0.73−0.761.00
Table 3. Pearson correlation coefficient matrix for study variables related to fatigue and mental distress (1994 and 1999)*
  • *

    P values <0.001. See Table 1 for definitions.

Self-efficacy symptom 1994 (X1)1.00         
Fatigue VAS 1994 (X2)−0.381.00        
SF-36 vitality 1994 (X3)0.44−0.711.00       
AIMS2 affect 1994 (X4)−0.550.45−0.601.00      
SF-36 mental health 1994 (X5)0.52−0.420.59−0.831.00     
Self-efficacy symptom 1999 (X6)0.50−0.250.32−0.400.371.00    
Fatigue VAS 1999 (X7)−0.220.46−0.450.35−0.32−0.311.00   
SF-36 vitality 1999 (X8)0.32−0.380.54−0.400.350.48−0.581.00  
AIMS2 affect 1999 (X9)−0.460.29−0.460.70−0.64−0.590.37−0.531.00 
SF-36 mental health 1999 (X10)0.37−0.240.41−0.570.560.55−0.360.56−0.831.00

The bivariate correlations between changes in ASES and changes in measures within the same dimension of health over the 5-year period are shown in Table 4. All correlations were statistically significant, but moderate to weak.

Table 4. Correlation between changes in self-efficacy and changes in health status measures over five years*
Health status measuresPearson's correlation coefficient
  • *

    All P values <0.001. See Table 1 for definitions.

Self-efficacy pain, Pain VAS−0.24
Self-efficacy pain, AIMS2 symptom−0.24
Self-efficacy pain, SF-36 pain0.29
Self-efficacy symptoms, fatigue VAS−0.22
Self-efficacy symptoms, SF-36 vitality0.25
Self-efficacy symptoms, AIMS2 affect−0.38
Self-efficacy symptoms, SF-36 mental health0.34

The correlations between baseline level of self-efficacy and the 5-year changes in related health status measures are shown in Table 5, and between baseline level of health status measures and the 5-year changes in the related ASES measure in Table 6. A high score on ASES pain at baseline was significantly correlated to improved pain after 5 years (Table 5). Subgroup analyses revealed that this finding was confined to patients with high education. Good mental health scores at baseline were related to improved level of ASES symptoms, but only in the subgroup of patients with low education (Table 6).

Table 5. Correlation between 1994 level of self-efficacy and 1999 level of health status measures*
MeasuresAll patients (n = 306)Low education (n = 167)High education (n = 136)
ConstantUnst BPConstantUnst BPConstantUnst BP
  • *

    Values are the mean ± SD. Adjusted for 1994 level of health status measure. Low education = patients with 7–11 years of education; High education = patients with 12–18 years of education; Unst B = unstandardized beta; SE = self-efficacy; VAS = visual analog scale; AIMS2 = Arthritis Impact Measurement Scales; SF-36 = Short-form 36; sympt = symptoms.

  • Statistically significant (P < 0.05) after Bonferroni adjustment (Each P value has been multiplied by the number of separate analyses, n = 7).

SE pain 1994—Pain VAS 199930.17−0.24 ± 0.07<0.00132.98−0.21 ± 0.110.0524.00−0.24 ± 0.080.006
SE 1994—AIMS2 sympt 19993.60−0.02 ± 0.0070.0013.05−0.013 ±−0.03 ± 0.010.005
SE pain 1994—SF-36 pain 199916.080.18 ± 0.060.00522.100.11 ± ± 0.080.006
SE sympt 1994—fatigue VAS 199928.75−0.10 ± 0.090.2931.55−0.08 ± 0.120.5119.39−0.05 ± 0.140.75
SE sympt 1994—SF-36 vitality 199919.080.10 ± 0.060.1120.430.06 ± 0.090.5020.190.11 ± 0.100.24
SE sympt 1994—AIMS2 affect 19991.60−0.08 ± 0.0050.110.14−0.07 ± 0.0060.251.44−0.03 ± 0.0080.68
SE sympt 1994—SF-36 mental health 199930.230.13 ± 0.060.0427.900.07 ± 0.090.3933.590.13 ± 0.090.13
Table 6. Correlation between 1994 level of health status measures and 1999 level of self-efficacy*
MeasuresAll patients (n = 306)Low education (n = 167)High education (n = 136)
ConstantUnst BPConstantUnst BPConstantUnst BP
  • *

    Values are the mean ± SD. Adjusted for 1994 level of health status measure. Low education = patients with 7–11 years of education; High education = patients with 12–18 years of education; Unst B = unstandardized beta; VAS = visual analog scale; SE = self-efficacy; AIMS2 = Arthritis Impact Measurement Scales; SF-36 = Short-form 36; sympt = symptoms.

  • No correlation reached level of statistical significance (P < 0.05) after Bonferroni adjustment (Each P value has been multiplied by the number of separate analyses, n = 7).

Pain VAS 1994—SE pain 199926.95−0.05 ± 0.050.3422.83−0.01 ± 0.070.8429.50−0.10 ± 0.080.20
AIMS-2 sympt 1994—SE pain 199928.18−0.55 ± 0.440.0223.520.15 ± 0.650.8132.33−1.23 ± 0.610.05
SF-36 pain 1994—SE pain 199921.99−0.08 ± 0.060.1625.51−0.04 ± 0.090.6417.990.20 ± 0.080.01
Fatigue VAS 1994—SE sympt 199935.74−0.05 ± 0.040.1837.94−0.07 ± 0.060.2437.86−0.05 ± 0.050.32
SF-vitality 1994—SE sympt 199930.850.11 ± 0.050.0231.720.15 ± 0.070.0332.970.08 ± 0.070.20
AIMS-2 affect 1994—SE sympt 199943.16−0.16 ± 0.660.0147.04−0.22 ± 0.860.0139.50−0.73 ± 1.070.50
SF-36 mental health 1994—SE sympt 199927.500.12 ± 0.060.0427.150.14 ± 0.080.0631.260.07 ± 0.090.42


This study focused on 5-year changes in and interrelationships between self-efficacy and selected health status measures in RA patients managed in a regular clinical setting. The patients did not undergo specific interventions aimed at enhancing self-efficacy. Probably, self-efficacy regarding the handling of pain and other symptoms related to a chronic disease may be influenced indirectly by several interventions, such as contact with health professionals, written information, cognitive therapy etc., but this has not been completely evaluated (26). The 306 patients in the study sample were all treated at the Oslo City Department of Rheumatology, Oslo, Norway. According to guidelines, disease-modifying antirheumatic drugs were used more aggressively in 1999 compared with 1994 (27% of patients used methotrexate in 1999 versus 16% in 1994), but as a whole, the therapeutic strategy, including a multidisciplinary bio-psychosocial approach, did not change over the study period. After 1985 patients with a recent diagnosis of RA had been offered an educational program without specific focus on enhancing self-efficacy.

Our results add evidence to the view that changes in self-efficacy, regarding pain as well as other symptoms, are positively related to changes in perceived health-status among patients with RA (Table 4). We chose to include more than one measure on each aspect (3 regarding pain, 2 regarding fatigue, and 2 regarding mental distress), in order to increase the strength of the results, and the findings were consistent for measures on the same dimension.

This study also revealed that self-efficacy at baseline was positively related to improved pain perception after 5 years (Table 5), and that mental distress at baseline was related to reduced self-efficacy after 5 years (Table 6). The results indicate that a 1 unit increase in baseline self-efficacy for pain (score range 10–100) predicts a 0.24 decrease in 100 mm pain VAS, and an increase in SF-36 pain score (range 0–100) of 0.23 5 years later in the subgroup of patients with high education. Associations of this magnitude, although weak in statistical terms, must be considered to be clinically interesting. To control pain is a main goal in the management of patients with arthritis, and assessment of the patients' self-efficacy regarding pain will probably be useful in planning and guiding their pain treatment (27). This measure (ASES pain) is easy to assess, and should increasingly be incorporated into clinical practice.

Studies have shown that mental distress is common in arthritis patients (28, 29), a fact that can easily be ignored by professionals expecting to deal with a strict somatic disorder. As several authors have pointed out, there is correlation between psychosocial factors and level of self-reported pain and disability among patients with RA (30, 31). We found that the patients' baseline level of mental distress was related to their subsequent level of ASES symptom. All these data taken together indicate the importance of being aware of depression and anxiety and offering proper treatment.

Surprisingly, our main findings regarding the associations between baseline self-efficacy and change in pain level as well as between baseline mental health and change in self-efficacy, varied according to the patients' educational level. The correlations between changes in self-efficacy and changes in health status measures did not differ across educational groups (data not shown). These findings may be statistical “artifacts” or only indirectly related to education. When we carried out Bonferroni adjustment (each P value is multiplied by the number of analyses made) the correlation between baseline self-efficacy and changes in perceived pain were still significant (Table 5), while the correlation between baseline mental health and changes in self-efficacy lost their significance level (Table 6). Because low educational level is associated with low self-efficacy (2, 3), we also performed the analyses separately for patients with below or above median self-efficacy level, to see if this would mimic the results of the educational groups, but it did not (data not shown). Because educational level is associated with age (older people tend to have less years of education), we carried out the regression analyses in Tables 5 and 6 adjusted for age. This did not change the results (data not shown). When we analyzed the data in a model to determine the relative contribution of self-efficacy and education to health status, we found that self-efficacy had a 2-fold impact on pain compared to educational length (data not shown).

The association between low levels of formal education and increased morbidity and mortality is well established for patients with RA (3, 32–34) and has been shown in most chronic diseases (5, 35). Education is unlikely to be the cause of poor outcomes, but may rather be seen as a surrogate marker for many variables that make a contribution, such as work strain, housing conditions, health habits, lifestyle, compliance with medical care, and coping strategies. Katz examined self-care activities performed by patients with RA with different levels of education (36). She found that persons with 13 or more years of education were significantly more likely to perform specific self-care activities, and suggested that this may be part of the explanation for the association between education and outcome in RA. Self-efficacy is claimed to be a predictor of disease management behavior (9) and to be closing the knowledge-behavior gap in health promotion (11). Several authors have shown that enhancement of self-efficacy through the Arthritis Self-Efficacy program is associated with positive changes in depression as well as in the perception of pain (37–39). Improvements are found to be commensurate with those obtained by nonsteroidal antiinflammatory drugs, and are additive to those obtained by other interventions (26). The self-management program is without side-effects, and reduces the utilization of and dependency upon health care services (11), and it is recommended to incorporate such programs in the care of patients with RA (40).

As Hawley points out in her survey on psychoeducational interventions in the treatment of arthritis (26), data suggest that the self-management programs were over-represented by highly educated individuals. In the mid 1980s Goeppinger et al implemented a self-care intervention called “Bone Up on Arthritis,” which was adapted to low literacy individuals (41). The 46 subjects in their study group with less than 12 years of schooling reported more depression and pain and less knowledge at baseline. However, level of education did not alter the effects of the intervention on knowledge, self-care behavior, perceived helplessness, and pain. As far as we have been able to find out, no study apart from this (41) has investigated the effect of education on the outcome of arthritis self-help programs. Our results indicate that this should be done, as the programs may not be equally successful for patients with low educational level.