Norwegian reference data on the Fatigue Questionnaire and the Patient Health Questionnaire-9 and their interrelationship

The FQ contains 11 fatigue-related symptoms experienced during the last month compared with how the subject felt when he/she last was feeling well. Four items concern mental and seven physical fatigue (Table 1). Each item is rated from 0 (‘less’/’better than usual’) to 3 (‘much worse than usual’). The mental fatigue sum score ranges from 0 to 12 and the physical fatigue sum score from 0 to 21, and higher scores mean more fatigue on both subscales. The total fatigue sum score represents the sum of all 11 items and ranges from 0 to 33. An additional FQ item covers the duration of the fatigue experience with one response alternative being “6 months or more”.

To identify cases with chronic fatigue (CF), a dichotomized score for each response alternative (0 = 0, 1 = 0, 2 = 1, 3 = 1) was used with a range of 0 to 11, and CF was defined as a dichotomized sum score of ≥ 4 with a duration of ≥ 6 months [23].

The PHQ-9 covers depressive symptoms experienced during the last 2 weeks, and each item is scored from 0 (‘not at all’) to 3 (‘nearly every day’), providing a 0–27 severity score.[16]. MDE is defined either by a sum score ≥ 10 [17], or by an algorithm as in the DSM system. We used the algorithm recommended by the Spitzer group in which an item is counted as a positive criterion if scored above threshold. Items #1–8 were scored as ‘bothersome’ for at least ‘more than half the days’ (score 2), or ‘on some days’ (score 1) on item #9 (thoughts about suicide or self-harm), to fulfill the algorithm. At least five criteria must be present among which either item #1 (anhedonia) or item #2 (depressed mood) is obligatory [17, 24, 25].

Descriptive statistics were applied (concerning) of means and proportions, which presented with their 95% confidence intervals (95%CI). Non-overlapping 95%CI represent a p value of < 0.05 [26]. The associations between the FQ and the PHQ-9 items and sum scores were examined with Spearman’s correlation coefficient r. Correlations were also reported as shared variance defined as the correlation coefficient squared in percent. Internal consistencies of the instruments were tested with Cronbach’s coefficient alpha. The p value was set at < 0.05, and all tests were two-sided. The statistical software applied was IBM SPSS version 25 for PC (IBM Corporation, Armonk, USA).

Attrition analysis of non-responders versus responders of 2015 sample has already been published [22]. Significantly more women than men responded to the invitation, and the responders were significantly older than the non-responders. The 2015 sample consisted of 923 (45%) males and 1,118 (55%) females who completed the FQ, and of 943 (45%) men and 1,143 (55%) women who completed the PHQ-9, respectively.

Internal consistencies of the FQ measured by Cronbach’s coefficient alphas were 0.88 for physical, 0.74 for mental, and 0.88 for total fatigue in the 2015 sample. Concerning the FQ samples of 1996 and 2015, several significant differences were observed on socio-demographic variables (Table 2).

The prevalence rates of CF according to sex and age groups are displayed in Table 3. No significant differences of prevalence rates were observed for males in any age groups. For females, however, the prevalence rates in 2015 are significantly higher in the 18–29 years age group and significantly lower in the ≥ 60 years age group.

As to the FQ mean scores, the mental and total scores for males ≥ 60 years were significantly lower in 2015 compared to 1996. The significant age group differences observed in 1996 were not the same in 2015. The only significant mean difference was a lower mean score for total fatigue in the 30–39 years versus ≥ 60 years age group (Table 4). In females, the age group of ≥ 60 years reported significantly lower mean scores for all FQ dimensions in 2015. The significant age group differences observed for physical and total fatigue in 1996 were not found in 2015.

The Cronbach’s coefficient alpha was 0.87 for the PHQ-9. The mean PHQ-9 scores according to the cut-off score and the algorithm score are displayed in Table 5 divided by sex and age groups. For males, no significant differences in mean scores were observed between the 10-year age groups. In contrast, among females the mean PHQ-9 scores in the 18–29 years age group were significantly higher relative to the age groups ≥ 50 years and to the overall mean score.

Based on the PHQ-9 sum score ≥ 10 definition of MDE, the prevalence was 8.1% (95%CI 6.9—9.2%) in the total sample, and among males 5.9% (95%CI 4.4–7.4%) and females 9.8% (95%CI 8.2–11.7%). Sex and age group distributions are shown in Table 5, and significant differences were observed between the 18–29 years versus the 70–79 years age groups for females, while no significant age differences were observed for males.

According to the Spitzer group’s algorithm, cases of MDE were 3.2% (95%CI 2.4–4.0%) in the total sample, and among males 2.6% (95%CI 1.7–3.8%) and females 3.8% (95%CI 2.8–5.0%). No significant age group differences were observed for males, but significant differences were observed between the 18–29 years and the 70–79 years age groups for females (Table 5).

Sixty-seven persons (3.2% of the sample) were defined as depressed by both definitions, while 101 (4.8%) fulfilled the dimensional definition only, and no person fulfilled the algorithm definition only. The prevalence of MDE defined by the PHQ-9 sum score was significantly higher for all age groups and for both sexes compared to MDE defined by the algorithm. The differences between the MDE definitions were significant for both males’ and females’ total prevalence, and for the prevalence rates of females age 40–49, 50–59, and 70–79 years.

For the PHQ-9 mean scores, no significant differences were observed between the age groups in males, while among females the mean score of the 18–29 years age group was significantly higher than for the age groups of 50–59 years and older (Table 5).

According to Table 1, the FQ has two items that correspond with three PHQ-9 items: FQ item #4 covers PHQ-9 items #1 and #5, FQ item #7 and PHQ-9 item #8 seem equal, while FQ item #3 and PHQ-9 item #3 are somewhat different concerning sleep problems.

The correlation matrix (Table 6) showed that most correlation coefficients were < 0.50, implying less than 25% shared variance. Only PHQ-9 item #4 (feeling of tiredness) showed four correlations between 0.52 and 0.61 with the FQ items #1 (tiredness), #2 (need more rest), #3 (feels sleepy and drowsy), and #5 (lacking energy) as well as the total fatigue score. Maximum shared variance between the item sets was 37% for FQ item #5 and PHQ item #4. The FQ items #1, 2, 4 and 5 showed correlations of 0.50 to 0.61 with the PHQ-9 sum score. The FQ and the PHQ-9 sum scores showed a correlation of r = 0.59, implying 35% shared variance.

Males and females ≥ 60 years had significantly lower mean scores of mental and total fatigue in 2015 compared to 1996, and for the females that was the case also for physical fatigue. No other significant differences between age groups on FQ mean scores were observed over time. Concerning prevalence of CF, only females ≥ 60 years showed a reduction from 1996 to 2015. The increase in CF among females aged 18–29 we consider an artifact (see below). The same was relevant for the high PHQ-9 mean score and prevalence of MDE in that group. The prevalence rates of MDE based on PHQ-9 sum scores were higher than those based on the algorithm, but no significant differences were found between age groups in males, or between the age groups over 30 years in females. The shared variance between the FQ and the PHQ scales was 35%.

The randomly drawn national 2015 sample was representative of the Norwegian population concerning age, sex and place of dwelling. However, the response rate was only 36%. The previous attrition analysis showed that significantly more women than men responded, and the participants were significantly older than the non-participants [22]. About 21% of the population was between 18 and 29 years, but this age group only represented 5% of our sample. Hence, the number of participants in this age group was very low; only 36 males responding to the FQ and the PHQ-9, and 65 and 68 females, respectively. The significantly higher rates of males and females with CF in 2015 in this age group, indicate a bias towards more ill responders. Thus, we suggest that our FQ and PHQ-9 results for the 18–29 age group, should not be used as reference data.

In contrast, individuals aged 67 years or above were over-represented since they constituted 27% of the sample, while only representing 18% of the population. The high proportion of respondents over 70 years may reflect the increased life expectancy in Norway, and active and engaged persons constitute a larger proportion of this age group. Since fatigue and depression are characterized by lack of energy and initiative, which increases the risk of non-participation, our sample also carries a risk of under-reporting for this reason. We are not able to estimate of that factor on our findings.

The decline in response rate from 67% in 1996 to 36% in 2015 is in accordance with both national [27] and international trends [28] concerning mailed questionnaires. Several factors may be operating, but in general, the population has become more restrictive in sharing personal information. Another factor is that the general population nowadays is exposed to a higher rate of mailed and electronic questionnaires than previously, not only from the field of medicine.

In 2018, 96% of all Norwegian households with at least one person < 75 years had access to the Internet, and an alternative had been to make a digital rather than postal survey. However, that approach could trigger still more fear of spreading of sensitive information, and the response rate is not necessarily higher than in paper-based surveys [29]. Web-based studies, therefore, do not solve the problem of sinking response rate to population-based health studies in Norway.

The first Norwegian reference data on the PHQ-9 showed lower mean PHQ-9 scores with increasing age for both sexes, and consistently higher mean scores in females for all age groups. The sex differences of the mean scores of the PHQ-9 were also observed in a German population based study [30], but for both sexes, they observed an increase in mean PHQ-9 scores with age. The reason for the differences may be that their study had better representativity with 63% response rate.

In our study, closely the same pattern was observed for MDE defined by cut-off ≥ 10 with reduced prevalence with increasing age and female sex, but we did not find any other papers with such PHQ-9 population data.

MDE defined by the algorithm score showed lower prevalence rates than defined by cut-off score for all age groups and both sexes. The discrepancy of these prevalence rates of MDE presents a problem for the use of the PHQ-9 as a clinical screening instrument for depression. Especially since the discrepancy of prevalence rates between the dimensional and algorithmic definitions of MDE also was observed in studies from other countries [20213031] [20, 21, 30, 31].

For PHQ-9 score ≥ 10 definitions of MDE, we observed no significant prevalence differences in males between our study and the others; while for females, our study had significantly higher prevalence than reported by Kocalevent et al. [31]. Concerning total prevalence rates, our study found significantly higher rates than both Shin et al. [21] and Kocalevent et al. [31]. Based on the MDE algorithm, our prevalence rates did not differ significantly from the other studies concerning sex-based and total prevalence rates.

Several demographic and lifestyle-related changes have occurred in Norway during the last decades. Mean life expectancy has increased for both sexes to 81 years for males and 84 years for females in 2016. Self-reported health and activities of daily living have improved in persons above 70 years over the last two decades [32]. Of today’s adult Norwegian population, 13% are daily smokers, obesity (BMI ≥ 30) 12, 32% consume alcohol once a week or more frequently, and 71% are physically active once a week or more [33]. Fourteen per cent of the population are immigrants, making the Norwegian population more diverse than before. In addition, there are significant differences concerning socio-demographic variables in our 1996 and 2015 samples as shown in Table 2. In spite of these social changes and sample differences, the levels of physical, mental and total fatigue show only significant changes in the ≥ 60 year age group for both sexes. We consider the reduction of fatigue among elderly persons to be a consequence of their improved health status [32].

The consistent increase in physical fatigue with age in males observed in 1996 [7] was not observed in 2015. It is tempting to speculate that this finding could in part be explained by improved health in middle-aged men [32].

Concerning the prevalence of CF, no significant changes were observed for males, but we observed a significant reduction among women ≥ 60 years, probably due to improved health in this group of women. The significant increase in the prevalence of CF among women aged 18–29 years, we consider due to sampling bias, and we remind about our concern to not use data from that age group as reference data for either sex.

As indicated in Table 1, three items concerning energy, tiredness, and concentration are common to both instruments. The correlation matrix between the FQ and the PHQ-9 (Table 6) mostly showed coefficients below 0.50, except for the PHQ-9 items #4 ‘Feeling tired or having little energy’ that had coefficients above 0.50 with four physical fatigue items (#1, 2, 3, and 5) related to tiredness, lack of energy, and need for rest. The 35% shared variance indicate that these PROMs could be used together, since 65% of their variance is separate.

Our findings have confirmed the close relationship between fatigue and depression. Several explanations for this fact have been proposed. One of them is heredity, as documented in the study of a twin sample. Multivariate twin modeling estimated a common additive genetic component which explained 25%, and 20% of the variance in depression and fatigue, respectively. For depression, environmental factors explained 28% and for fatigue 54% of the variance [34].

Another explanation concerns malfunctioning neural circuits and neurotransmitters [10]. They build on the observation that treatment with antidepressants often is effective on the affective component of MDE, but less so on the lack of energy (fatigue) component. Therefore, antidepressants are considered to have different effects on the depression and fatigue neural circuits that may be due to different neurotransmitter profiles.

These and other potential explanations need further research.

We have produced Norwegian normative data for the FQ and the PHQ-9 scales. The FQ results show reduced fatigue in persons ≥ 60 years from 1996 to 2015 possibly reflecting improved general health in Norwegian during that period. The PHQ-9 data are new, but in line with the findings of other national samples [20, 21, 30, 31].

Limitations are the response rate of 36% leading to low numbers of responders in the 18–29 years age group; thus, the FQ and PHQ-9 data reported by that age group should be considered critically.