Several factors may affect the quality of measurements made of subjective complaints. Both the variation in complaints over time and the methods of measurement may invalidate the data. Health complaints vary over time due to variations in exposure to pathogenic factors, variations in underlying somatic condition, and possibly systematic seasonal changes. With subjective measures, additional sources of variation are changes in the perception of somatic and other sensations (Watson and Pennebaker, 1989; McDermid et al., 1996), processes that alter ways of coping with symptoms, and reporting behaviour (Macleod et al., 2002).
A short recall period should produce the most accurate data. However, the data merely represent a snapshot of the most recent situation of the respondent and may not be representative of a long-term condition. This may be a serious problem when a condition exhibits large fluctuations, as pain disorders do (Jensen and McFarland, 1993). Longer reporting periods may introduce recall bias, due to the factors mentioned above (Marshall et al., 1995).
In addition to the variation in the complaints themselves, the initial response to a questionnaire may be influenced by other sources of variation. The respondents receive attention during the investigation and are asked to direct their attention to symptoms and complaints that they may not have been consciously aware of before. Problems or issues receiving attention (e.g. by news coverage) at the time of the investigation may influence responses within the study population. These factors may render the single-sample reporting of complaints prone to errors and systematically confound associations between exposures or treatments and symptoms.
Psychological factors may contribute to the reporting of complaints. An important issue in assessing health complaints is the extent to which subjectively reported symptoms represent specific organ affliction rather than a generalized tendency to experience and report discomfort or negative sensations (Watson and Pennebaker, 1989; McDermid et al., 1996; Macleod et al., 2002). Several syndromes characterized by symptoms and functional impairment are not supported by physiological findings (see Barsky and Borus, 1999 for a review). The psychophysiological mechanisms contributing to these syndromes may play a role in the perception of any symptom, e.g. forearm pain (Hall and Morrow, 1988). On the other hand, psychological mechanisms may affect the reporting of symptoms in healthy subjects. Cognitive schemata may determine whether a subject recalls and decides to report a symptom that does not impair function or cause concern (Cope et al., 1994). Furthermore, depression, anxiety and distress may influence the subjective appraisal and reporting of symptoms (Halder et al., 2002). Hence, the association between psychological and somatic complaints and the possible generalized tendency to report complaints need to be elucidated in order to improve knowledge of subjective health complaints.
The objectives of the present study were (i) to assess variations in subjective health complaints over an extended period, (ii) to determine whether the single-sample method produce data that represent the respondents' subjective complaints over time, and (iii) to determine whether there are associations between the reporting of psychological and somatic complaints in a working population.
2. Subjects and methods
2.1. Study design and procedures
This study is a part of a prospective randomized intervention study (1997–2000) comprising a comprehensive questionnaire (conducted in February each of the three years), monthly health-complaint reports over 32–34 months (starting in March, April, or May 1997), two medical examinations (at one-year intervals), observation of the work routines of each participant in their workplaces (once or twice a year for 2 years), annual monitoring of electromyography, blood pressure and heart rate during a complete working day (over 2 years) and during standardized laboratory tests (over 3 years), and 12 week intervention programs (from September to December 1997).
Annual questionnaires and monthly reports were distributed at the workplaces during the study. They were sent by mail to those on long-term sick leave and to those no longer working at the post offices included in the study. Before filling in their initial monthly health-complaint report, all participants were given detailed instructions on items and scales.
The present data were obtained from the monthly health-complaint reports (1997–2000) and the comprehensive questionnaire (the socio-economic situation in 1997). The Regional Ethical Committee for Medical Research approved the study.
Participants were recruited from 11 post offices in Oslo. The inclusion criteria were: permanent employment at the post offices, working at least 19 h a week, counter service as their main work function, and age less than 61 years. The exclusion criteria were: any known systemic diseases, pregnancy, and participants' self-professed inability to maintain a seated position for 3 h.
Written, informed consent to participate was acquired from the 104 subjects (67 women) out of a total of 116 employees who met the inclusion criteria. Subjects were then excluded due to altered work functions (n=13), pregnancy (n=5), or long-term sick leave (n=6) before the first collection of health-complaint reports. One person dropped out. The study sample therefore consisted of 79 participants (51 women) for the initial report. In the pre-intervention period, further exclusion was due to altered work functions (n=15), pregnancy (n=2), and lower back pain (n=2). No participants were excluded after the interventions started, but 18 subjects dropped out during the intervention and post-intervention periods due to changes of work (n=4) and for unknown reasons.
The 60 subjects still participating at the start of the intervention programs were randomly assigned to two intervention groups and one control group. They were stratified according to their neck and shoulder complaints reported at the first medical examination (affliction: daily complaints or visual analogue scale >10 on a 100 mm line), sex, and their work place. The interventions, described by Eriksen et al. (2002), consisted of stress management training (SMT, n=19) and an integrated health practice program (IHP, n=22) and were carried out during the participants' working hours. Approximately half of the subjects participated in less than 50% of these sessions. According to the participants, non-compliance was mainly due to the personnel-situation at their work places during the intervention period.
The median age of participants at the time of the initial report was 35 years (range 24–60), median years of postal work 12 (range 3–29), median years of present work tasks five (range 1–5), and median days of sick leave during previous six months was six (range 1–135). The mean period of formal education was 12 years (SD 2) and mean number of paid working hours was 37.4 (SD 2.6). Forty-two percent of the workers were smokers.
2.3. Monthly health-complaint reports
At the end of each month, the subjects rated intensity and duration of 30 health complaints during the preceding 14 days. Each of the complaints was rated on a four-point scale for intensity (0=not troubled, 1=a little troubled, 2=quite troubled, 3=seriously troubled), and a three-point scale for duration (1=1–5 days, 2=6–10 days, 3=11–14 days). A complaint-severity score for each of the 30 health complaints was constructed by multiplying the intensity score by the duration score (score range 0–9). Six complaint-severity indices (index range 0–9) were computed as a mean of different complaint-severity scores (see Table 1).
2.4. Data analysis and statistics
The Statistical Package for Social Sciences (SPSS 9.0–11.5) and Confidence Interval Analysis (CIA 2.0) were used for the analyses. Unadjusted P-values are given and a 5% significance level adopted. As recommended by Perneger (1998) and Campbell et al. (2000), we chose not to adjust for multiple tests. Since the musculoskeletal, gastrointestinal, allergic, upper-airway-infection, and psychological complaint-severity indices are subsets of the total complaint-severity index, one may argue that the tests for the differences in the subset indices to some extent test the same hypothesis as when using the total complaint-severity index. Similar argumentation refers to the complaint-severity scores as subsets of the indices. Results including subsets of data should therefore be interpreted with caution.
Differences in complaints (initial report) between those subjects who were excluded or dropped out of the study and those who completed the whole study period were studied by Mann–Whitney U-tests. Missing data comprised 1.6% of the intensity and duration variables.
2.4.1. Variation in subjective complaints over an extended period
Individual level: To assess the proportion of variation in complaints due to individual variation (individual variation/(individual variation+error variation)), a variance-component analysis of the complaint-severity indices was performed with complaint as the dependent variable, participant as a random factor and time as covariate (last 24 months of the study period). Five complaint-severity indices (MSI, GI, AI, UAII, PI) and the complaint-severity scores of lower back pain and fatigue were analysed.
Complaint severity and variation:The individual subject's complaint-severity means (complaint-severity scores and complaint-severity indices) for the last 24 months of the study were plotted against their standard deviations to assess whether the variation was a function of complaint-severity (Bland and Altman, 1996, 1999). Similar plots were made for the intensity and duration scores. In addition, the mean month to month difference in complaint-severity between the last 24 consecutive reports were calculated for each individual and plotted in ascending order of complaint-severity.
Group level: The Milton–Friedman test for repeated measurements was used to determine if there were systematic changes over time in the complaint-severity scores or indices. To decrease the effect of short-time variability on the analysis of the slower temporal changes of complaints, the analysis comprised the means of the respective complaint-severity scores and indices over five six-month periods: July–December 1997, January–June 1998, July–December 1998, January–June 1999 and July–December 1999. In order to determine whether there were seasonal differences in complaint-severity, the Milton–Friedman tests for repeated measurements were additionally made on reports from five seasonal periods: winter (January–February), spring (March–April), summer (May–August), autumn (September–November) and Christmas (December) (all periods calculated as mean of data from 1998 and 1999).
2.4.2. Determining whether the single-sample approach represent the respondents' subjective health complaints over time
To estimate whether the initial sample varied from other samples, differences in complaints between (i) the first and second report and (ii) the first and 13th report were analysed by the Wilcoxon matched-pairs signed-rank-sum test on paired data.
To estimate the number of reports needed to obtain data representative of average complaint-severity over time, a complaint-severity reference for each individual was obtained from the mean of 12 consecutive reports that were chosen randomly from different starting points during the post-intervention period (the last 24 months of the study). The rationale for different starting points was to avoid possible effects of intervention, season, or alterations in work organization. We calculated the differences between complaint-severity scores and indices (computed by the increasing number of reports; one report, mean of two reports, up to mean of 11 reports) and the individual respondent's respective complaint-severity reference (mean of 12 reports). The number of participants showing a difference of at least ±1.0 from their reference was counted (Bland and Altman, 1999).
2.4.3. Associations between reported psychological and somatic complaints
The relationships between the different complaint-severity indices were determined by Spearman's rank correlations (rs). Both the means of the last 31 months of the study period, the means of a four-month period (second to fifth report), and reports from five seasonal periods (see above) were tested.
In order to determine whether the relations between the psychological and somatic complaint types (indices) only pertained to those subjects reporting more severe symptoms, the lower level half-sample and the higher level half-sample of psychological complaints (median split) were analysed separately.
The proposition that a generally high level of psychological complaint-severity (mean of the last 31 months of the study) determined a tendency to respond with high complaint levels on the initial report (difference between the complaint-severity of first report and the mean of complaint-severity of second to fifth report) was tested by Spearman's rank correlations.
We found strong associations between severity of musculoskeletal and psychological complaints. Hence, we added more in-depth analyses on these relations. Rank correlations were used to determine the stability of the relation between musculoskeletal and psychological complaint reporting; i.e. whether the same subjects report higher levels of both musculoskeletal and psychological complaints over time. The product of MSI and PI ((MSI+0.1)*(PI+0.1)) over four months in the early period (second–fifth) was correlated to the product of MSI and PI over four months one year later (13th–16th).
The data were averaged over four four-months periods: (i) prior to the intervention (May–August 1997), (ii) the intervention months (September–December 1997), (iii) the first 4 months following the intervention (January–April 1998), and (iv) the same season as the pre-intervention period (May–August 1998). A repeated measures analysis of variance (ANOVA), with the four time-periods as within-subject factors and the three groups as between-subject factors, was performed. For all within-subject effects, Greenhouse–Geisser corrections were used.
3.1. Prevalence and severity of reported complaints
The group that left the study (n=37; 19 subjects excluded, 18 dropouts) exhibited complaint-severity similar to the group completing the entire study (n=42;P>0.061).
The prevalence rates in the second report were lower than in the initial report (Table 2). In those who completed the whole study period, a 12-month prevalence (second to 13th report) was above 60% in all the different complaint types (Table 2), but the median values of the indices were less than 0.30 (scale 0–9).
Of the single items, headache, coughing, common cold, and fatigue exhibited a 12-month prevalence of 70% or more. Twenty-six percent of the workers reported fatigue and 24% reported lower back pain with a complaint-severity score ≥6 over a one-year period.
3.2. Variation in subjective complaints over an extended period
Individual level: In about one third of the participants the reported complaint-severity scores varied between zero and nine (scale 0–9). This means that in some of the reported consecutive 2-week periods the workers were free of complaints (i.e. lower back pain) and in others they were seriously troubled by that complaint on at least 11 of 14 days (Fig. 1). Sixteen percent of the participants reported that they were free of specific complaint during the last 14 days, while in the adjacent month reporting that they were seriously troubled by that complaint on at least 11 of 14 days. This pattern was seen for musculoskeletal and psychological complaints, but not for the other complaints studied.
For the last 24 months of the study period (post-intervention), the proportions of the complaint variation due to individual variation were 74% (MSI), 40% (GI), 18% (AI), 10% (UAII), 50% (PI), 61% (lower back pain) and 59% (fatigue).
Complaint severity and variation: The amount of variation over an extended period (24 months) was related to complaint-severity (examples given in Fig. 2). Those with more complaints exhibited larger variation. The results were similar when the intensity and duration variables were plotted separately.
The month to month difference in complaint-severity the last 24 months of the study was also generally larger for those with more complaints (Fig. 3).
Group level: When blocks of data from six months were pooled, all the indices and scores exhibited stability during the last 30 months of the study period (complaint-severity indices P>0.08,χ2<9.0, df 4; complaint-severity scores P>0.06,χ2<8.9, df 4). The data did not reveal seasonal effects on the complaint-severity over the five different season periods (P>0.082,χ2<8.3, df 4).
Based on inspection of graphs and the analyses of group and individual variation, one may conclude that there were no seasonal patterns or time trends in the complaint-severity during the last 30 months of the study period.
3.3. Determining whether the single-sample approach represent the respondents' subjective health complaints over time
Higher complaint-severity indices of gastrointestinal (P=0.019, median differences 0.17), psychological (P=0.019, median differences 0.20) and total health complaints (P=0.012, median differences 0.05) were reported in the initial report compared with the second report and the 13th report one year later (n=77; other indices p>0.317).
Twenty of the 30 complaint-severity scores showed larger individual deviations from the 12-sample severity reference of consecutive reports (taken from different starting points during the last 24 months of the study) in the first report of the respective reference period compared with a two-report sample (the mean of the first and second report of the respective reference period; example of one complaint-severity score given in Fig. 4). Eight complaint-severity scores showed similar deviations from the reference when using one-report or two-report samples.
With an increasing number of reports, the number of participants providing data that deviated by ±1.0 or more from their 12-month reference declined in all complaint-severity scores and indices (Fig. 5). A mean of four reports of PI was needed for every subject to attain this criterion (Fig. 5A); up to 11 reports for pain in the lower back (Fig. 5B and C). Complaints with the largest number of participants deviating from the criterion were lower back pain and fatigue. These complaints had the highest one-year prevalence of severity score of >6 (see above).
Up to eight samples were needed for complaint-severity scores included in the MSI for 90% of the participant's reports to deviate less than ±1.0 from their reference scores (Fig. 5B). Up to four samples were needed for the complaint-severity scores included in the PI to reach the same criterion, and one sample for the GI, AI and UAII.
Four consecutive samples were needed to obtain agreement with the reference criterion (deviation of less than ±1.0) for 95% of the data (30 complaint-severity scores*42 participants).
3.4. Associations between reported psychological and somatic complaints
There was a strong correlation between the MSI and PI, both based on the means of the last 31 months of the study period (n=42; see Table 3) and based on a four-month sample (means of second to fifth report, Symbol CI 0.51 to 0.78). Correlations between the somatic indices were generally weak or moderate (n=42, see Table 3; n=69,rs<0.301). Similar relations were observed between the indices for the five different seasonal periods.
There was no association between psychological complaint-severity (the last 31 months of the study) and a tendency to respond with higher complaint-severity levels (PI, MSI, GI, AI, UAII) on the initial report (difference between the initial report and the following four-month period; rs<0.171).
The relation between the product of musculoskeletal and psychological complaints (MSI*PI) was stable over time (rs=0.763, CI 0.60 to 0.87).
3.5. Effects of interventions
There were no statistically significant main effects between the three groups on any of the indices, neither when all the subjects that participated in any of the intervention sessions were included in the analysis (p>0.234; mean differences <0.97, median differences <1.0, SMT n=19; IHP n=22, control group n=19) nor when only those who participated in at least half of the sessions were included (subgroup of completers; p>0.161; SMT n=10, IHP n=12). Median UAII increased 0.25 from the pre-intervention period to the intervention period Symbol Because of the small number of participants in the intervention part of the study and the low adherence in the intervention programs, we do not have enough statistical power to conclude about the effectiveness of the intervention programs.
Healthy service-sector workers exhibited large intra-individual complaint variability in monthly reports over about three years. At a group level, however, the complaint-severity was stable during this period. No trend or seasonal variation was found in health-complaint reporting among those that participated for the entire study period, neither at a group nor at an individual level. The single-sample approach does not produce data that adequately represent average complaints over time. There was a strong association between the reporting of psychological and the reporting of musculoskeletal complaints.
Of 116 eligible workers, 90% were willing to participate in the present study. This group is likely to represent the working population in the service sector in Norway. Almost 60% of the participants were later excluded from the study, mostly due to organizational changes resulting in being mover or having altered work functions during the research period. Importantly, those who completed the entire study reported complaint-severity similar to the group that left the study.
The large intra-individual variability in the complaint reporting seems to reflect the fluctuation in pain and other complaints over time (Von Korff and Saunders, 1996). The reporting may, however, be influenced by factors other than the complaints per se. The participants' mood while filling in the self-reports and while experiencing the complaints may affect their recollection (Bower, 1981; Singer and Salovey, 1988; Eich and Maculay, 2000). The pain level at the time of reporting may influence both the reported pain intensity and duration (Salovey et al., 1993).
The initial report produced higher complaint-severity compared with the second and later reports. Factors related to the first encounter with a questionnaire, such as bringing attention to symptoms and difficulties in understanding the questions, may have influenced results. Another contribution to the higher complaint level reported in the initial report of the present study may have been through differences in way the reports were administered. The initial report was administered immediately after participants received detailed instructions on its completion and had the opportunity to ask questions. The remaining reports were either distributed in the workplaces or sent by mail. However, during the first encounter participants were strongly advised to contact the researchers if they had any questions regarding filling in the later reports. Several participants contacted us for more information, but at different stages in the study. Additional reasons for higher values in the initial report may be that the participants became bored with filling in the same questionnaire repeatedly over an extended period and therefore became inaccurate. However, it is not likely that the participants had already become bored when completing the second report. Conversely, later reports may be more accurate than the initial reports since the participants know about the forthcoming reporting and become more aware of their complaint intensity and complaint duration. In addition, one would expect that practice would alleviate problems in filling in the questionnaires, resulting in more valid data.
The intra-individual deviation from the 12-month severity reference (from the last 24 months of the study period) was largest when data were collected with only one sample, and it decreased as the number of consecutive reports increased. Larger deviations from the reference were seen in complaints with a higher severity score. This raises the question of whether a higher number of reports are needed in populations with severe complaints. With symptoms naturally fluctuating, there is no absolute acceptance criterion for the amount of deviation from the reference. Four reports or more were needed to obtain agreement for 95% of data when the criterion of accepted deviation from the 12-month reference was set to ±1.0 (on a 0–9 scale of intensity×duration). Since most of the participants exhibited low complaint scores, this criterion represents a deviation of at least 70% from the group reference of the most severe complaints (including only participants reporting complaints during the reference period). Therefore, it seems that the common procedure of collecting subjective complaint data at only one time-point does not produce adequate representation of the complaint-severity over time. This is of particular importance when designing intervention studies. Due to the large degree of intra-individual variability, at least two pre-intervention samples and at least two post-intervention samples should be gathered in the intervention group and control group to reveal effects of interventions on health. One must be aware of the possibility that the initial report may reveal higher complaint levels than subsequent reports. It may be reasonable to omit the initial report from the analyses. Thus, in intervention studies, it may be argued that at least three pre-intervention samples (omitting the initial report) and at least two post-intervention samples should be gathered to reveal the effects of intervention on health. When interpreting effects of interventions on study groups with different complaint levels, one should also take into consideration that the amount of variation might be related to complaint-severity, as seen in the present data and also in Bland and Altman, (1996) and (1999).
The lack of synchronized seasonal differences in complaint-severity is consistent with self-reported seasonal influences on prevalence rates of neck and back pain (Hildebrandt et al., 2002) and variation of symptoms in chronic fatigue patients (Zubieta et al., 1994; Garcia-Borreguero et al., 1998). The literature shows inconsistent findings on the effect of seasons on psychological complaints (Zubieta et al., 1994; Garcia-Borreguero et al., 1998; Hardt and Gerbershagen, 1999; Stiles et al., 1993; Harmatz et al., 2000; Nayyar and Cochrane, 1996). A possible relation between positive and negative affect and season of the year was therefore, not found in the individual psychological or somatic complaint data.
Different types of reporting bias, such as systematic over- or under-reporting (Black and Cole, 2001) can influence results from subjective measures. In accordance with previous studies (Linton, 2000; Russel, 2001), we found strong correlations between musculoskeletal and psychological complaints. In general, only weak or moderate correlations were found between the different somatic complaint indices. If the reporting of complaints were strongly biased, one would expect all the different complaints to be highly correlated. Hence, reporting bias does not seem to explain the association between psychological and musculoskeletal complaints.
The strong relation between psychological and musculoskeletal pain suggests either a causal link between psychological and musculoskeletal complaints or vice versa, or shared causal factors. The association may also result from the persistent experience of pain (Chapman and Cavrin, 1999), so that chronic pain may result in symptoms of fatigue. However, the magnitude of the relationship was comparable in those with less and more severe complaints, indicating an overlap between musculoskeletal and psychological complaint reporting even in the absence of constant or debilitating pain. It should be noted that, of the five items included in the psychological complaint-severity index, fatigue exhibited the highest severity score.
Complaints may be grouped according to organ systems (anatomy) and common beliefs regarding causes, or according to factor-analysis of the results. The present instrument was constructed by listing subjective complaints belonging to common clinical groups of diagnoses. Hence, there were five compound indices in addition to the total complaint-severity index. Some complaints may be grouped in different indices, since they are symptoms that may signify several different problems. Chest pain may be a symptom of angina pectoris, a result of an asthmatic attack, or a primary musculoskeletal problem. Fatigue and sleep disturbances may be results of somatic symptoms or psychological phenomena. Some headaches may have a vascular origin like migraine, and pain can be referred from internal organs. Furthermore, the order of questions asked may be of importance to how complaints are reported. In our working population (median age 35 years), headache and chest pain were classified as musculoskeletal complaints and grouped accordingly in the questionnaire. Migraine attacks were not included in the musculoskeletal index.
During almost the entire study period, comprehensive organizationsational changes of the Norway Post took place. The extensive changes of the organization may modify possible effects of the interventions that were administered in 1997 and profoundly influence the compliance of participation. The Norway Post had planned to provide substitution during the intervention periods, but training of the assistants took longer time than expected and the substitution was therefore unsatisfactory. Several participants complained that the workload did not permit leaving the workplace.
High intra-individual complaint variability was revealed in a working population, but no trend or seasonal effects were demonstrated at a group or an individual level. Measuring subjective health with single-sample approaches does not produce data representative of average complaints over a period. Higher complaint levels were found in the initial report compared with subsequent reports. Our data indicate that more than two samples are required when collecting subjective data on health complaints, both when the aim is to reveal changes in health, as in interventions studies, and when the aim is to obtain representative data for the averaging of complaints over time. A strong association was found between reported psychological and musculoskeletal complaints.
We gratefully acknowledge the assistance and collaboration of Eva S. Bakke, Astrid Bolling, Øyvind Hesselberg, Ada Ingvaldsen, Christine Mohn, Birgitta Jarmark-Robertsson, Laila Rosenborg, Janne Schiøll, Bjørg Ingrid Selberg, Jorid T. Stuenæs and Morten Wærsted. Further, we thank Thore Egeland for statistical advice and Eva S. Bakke and Dagfinn Matre for comments on the manuscript. We also thank the Norway Post and the workers that participated in the project. The Research Council of Norway supported the study (Grant 109176/330).
Barsky AJ, Borus JF. Functional somatic syndromes. Ann Intern Med. 1999;130:910-921.
Black AE, Cole TJ. Biased over- or under-reporting is characteristic of individuals whether over time or by different assessment methods. J Am Diet Assoc. 2001;101:70-80.
Bland JM, Altman DG. Measurement error proportional to the mean. Br Med J. 1996;313:106.
Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999;8:135-160.
Bower GH. Mood and memory. Am Psychol. 1981;36:129-148.
Campbell MJ, Daly LE, Machin D. Special topics. Altman DG, Machin D, Bryant TN, Gardner MJ, editors. Statistics with confidence. 2nd ed. Bristol: BMJ; 2000: pp. 153-167.
Chapman CR, Gavrin J. Suffering: the contributions of persistent pain. Review. Lancet. 1999;353:2233-2237.
Cope H, David A, Pelosi A, Mann A. Predictors of chronic ‘postviral’ fatigue. Lancet. 1994;344:864-868.
Eich E, Maculay D. Are real moods required to reveal mood-congruent and mood-dependent memory? Psychol Sci. 2000;11(3):244-248.
Eriksen HR, Ihlebæk C, Mikkelsen A, Grønningsæter H, Sandal G, Ursin H. Improving subjective health at the work site: a randomized controlled trial of stress management training, physical exercise, and an integrated health program. Occup Med. 2002;383-391.
Garcia-Borreguero D, Dale JK, Rosenthal NE, Chiara A, O'Fallon A, Bartko JJ, Straus SE, et al. Lack of seasonal variation of symptoms in patients with chronic fatigue syndrome. Psychiatr Res. 1998;77:71-77.
Halder SL, McBeth I, Silman AJ, Thompson DG, Macfarlane GJ. Psychosocial risk factors for the onset of abdominal pain. Results from a large prospective population-based study. Int J Epidemiol. 2002;31:1219-1225.
Hall W, Morrow L. Repetition strain injury: an Australia epidemic of upper limb pain. Soc Sci Med. 1988;27:645-649.
Hardt J, Gerbershagen HU. No changes in mood with the seasons: observations in 3000 chronic pain patients. Acta Psychiatr Scand. 1999;100:288-294.
Harmatz MG, Well AD, Overtree CE, Kawamura KY, Rosal M, Ockene IS. Seasonal variation of depression and other moods: a longitudinal approach. J Biol Rhythms. 2000;15:344-350.
Hildebrandt VH, Bongers PM, van Dijk FJ, Kemper HC, Dul J. The influence of climatic factors on non-specific back and neck-shoulder disease. Ergonomics. 2002;45:32-48.
Jensen MP, McFarland CA. Increasing the reliability and validity of pain intensity measurement in chronic pain patients. Pain. 1993;55:195-203.
Linton SJ., 2000. Psychological risk factors for neck and back pain. In: Nachemson A, Jonson E, editors., Neck and back pain: the scientific evidence of causes. Lippincott Williams & Wilkins, Philadelphia, pp. 57-77.
Macleod J, Davey Smith G, Heslop P, Metcalfe C, Carroll D, Hart C. Psychological stress and cardiovascular disease: empirical demonstration of bias in a prospective observational study of Scottish men. Br Med J. 2002;324:1247-1251.
Marshall PD, O'Connor M, Hodgkinson JP. The perceived relationship between neck symptoms and precedent injury. Injury. 1995;26:17-19.
McDermid AJ, Rollman GB, McCain GA. Generalized hypervigilance in fibromyalgia: evidence of perceptual amplification. Pain. 1996;66:133-144.
Nayyar K, Cochrane R. Seasonal changes in affective state measured prospectively and retrospectively. Br J Psychiatry. 1996;168:627-632.
Perneger TV. What's wrong with Bonferroni adjustments. Br Med J. 1998;316:1236-1238.
Russel J., 2001. Fibromyalgia Syndrome. In: Mense S, Simons DG, editors., Muscle pain: understanding its nature, diagnosis, and treatment. Lippincott Williams and Wilkins, Philadelphia, pp. 289-337.
Salovey P, Smith AF, Turk DC, Jobe JB, Willis GB. The accuracy of memory for pain. Not so bad most of the time. Am Phys J. 1993;2:184-191.
Singer JA, Salovey P. Mood and memory: evaluating the network theory of affect. Clin Psychol Rev. 1988;8:211-251.
Stiles WB, Barkham M, Shapiro DA. Lack of synchronized seasonal variation in the intensity of psychological problems. J Abnorm Psychol. 1993;102:388-394.
Von Korff M, Saunders K. The course of back pain in primary care. Spine. 1996;21:2833-2837.
Watson D, Pennebaker JW. Health complaints, stress, and distress: exploring the central role of negative affectivity. Psychol Rev. 1989;96:234-254.
Zubieta JK, Engleberg NC, Yargic LI, Pande AC, Demitrack MA. Seasonal symptom variation in patients with chronic fatigue: comparison with major mood disorders. J Psychiatr Res. 1994;28:13-22.