A total of 80 participants (71 females, 89%) completed the questionnaire and actigraphy measurement. The mean age of participants was 66.8 years (SD = 4.39). Nearly two thirds (65%) of participants were retired, 12.5% were in part-time employment, 8.8% in full-time employment, 7.5% self-employed, 2.5% were not employed and not currently looking for work, and 1.3% full-time carers, ‘not employed and looking for work’, and students, respectively. Only one participant was a smoker. Six participants (7.5%) consumed more than four standard drinks on any given drinking occasion.
Based on the classification of sleep groups, where participants were classed as poor sleepers if their sleep onset latency or WASO was ≥31min three times or more during the actigraphy recording period, and where participants were categorised as complaining sleepers if they had a self-reported sleep problem for at least six months, 22 participants were categorised as complaining-good (CG; 27.5%), 30 as complaining poor (CP; 37.5%), 14 as non-complaining good (NG; 17.5%) and 14 as non-complaining poor sleepers (NP; 17.5%). We used linear mixed modelling to conduct significance testing for the continuous outcome data: objective total sleep time (objTST), objective sleep onset latency (objSOL), objective wake after sleep onset (objWASO), and objective number of awakenings (obj#Awake), all of which were measured using actigraphy, and subjective total sleep time (subTST), subjective sleep onset latency (subSOL), subjective wake after sleep onset (subWASO), and subjective number of awakenings (sub#Awake), which were self-reported sleep measures we obtained from the participant sleep diaries, as well as scores on the DBAS16. We employed a Bonferroni correction to adjust for multiple comparisons. Restricted maximum likelihood was used to fit the model. We used ID as random effects, and as fixed effects included group, gender, alcohol consumption (more than four standard drinks per drinking session), smoking status, employment status, age, and caffeine consumption. An overview of the outcome data is shown in Table 1.
Actigraphy measured objective total sleep time (objTST) showed no significant main effect between the four groups NG, CG, NP, and CP sleepers F(3,76)=0.772, p=0.513.
For objective sleep onset latency (objSOL), linear mixed modelling revealed that there was a significant main effect for group F(3,63)=2.846, p=0.045, although this was a weak effect (partial η2 = 0.05). There was also an effect of smoking status F(1,63)=5.719, p=0.02. CP sleepers had significantly lower objective sleep onset latency compared with NP sleepers (between group mean difference =-2.969, p=0.034), but there was no statistically significant difference between the other groups.
There was a significant main effect for group for objective wake-after-sleep-onset (objWASO) F(3,63)=28.966, p=0.000; this was a large effect (partial η2=0.41). CG sleepers had significantly lower WASO than CP sleepers (between group mean difference=-37.06, p=0.000), and NP (between group mean difference =-23.55, p=0.000) but not NG sleepers. CP sleepers had significantly higher WASO than NG sleepers (between group mean difference=39.34, p=0.000), whereas NG sleepers had significantly lower WASO than NP sleepers (between group mean difference =-25.89, p=0.001). There was no statistically significant difference between CP and NP sleepers.
Furthermore, analyses showed a significant main effect for group in objective number of awakenings (obj#Awake) F(3,63)=20.73, p=0.000, and this was a large effect size (partial η2=0.33). There was also an effect of age F(1,63)=5.86, p=0.018. CG sleepers had a significantly lower number of objective awakenings than CP sleepers (between group mean difference=-6.32, p=0.000) and NP (between group mean difference=-4.47, p=0.009), but not NG sleepers (between group mean difference=2.87, p=0.317). CP sleepers’ (between group mean difference=9.19, p=0.000) and NP sleepers’ objective number of awakenings (between group mean difference=7.34, p=0.000) were significantly higher than those of NG sleepers, but there was no difference between CP and NP sleepers.
For subjective total sleep time (subTST), linear mixed modelling revealed that there was a significant main effect for group F(3,63)=2.827, p=0.046, and this was a medium effect size (partial η2=0.06). CG sleepers displayed significantly lower self-reported subjective total sleep time than NP sleepers (between group mean difference=-69.51, p=0.031), but not compared with CP or NG sleepers. There were no significant differences between any of the other groups.
There were no significant differences between the groups in subjective sleep onset latency (subSOL) F(3,64)=2.021, p=0.120.
Subjective WASO (subWASO) displayed a significant main effect for group F(3,62)=7.231, p=0.000, showing a medium effect size (partial η2=0.095), and an effect of age F(1,63)=4.663, p=0.035. CP displayed significantly higher self-reported subWASO than NG sleepers (between group mean difference=45.34, p=0.004) and NP sleepers (between group mean difference=48.02, p=0.001), but not CG sleepers. There were no significant differences between the other groups.
There were no significant differences between the groups in subjective number of awakenings (sub#Awake) F(3,64)=2.021, p=0.120.
Levels of dysfunctional sleep beliefs (measured by the DBAS16) displayed a significant main effect for group F(3,308)=10.740, p=0.000), showing a medium effect size (partial η2=0.095). CG displayed significantly higher levels of dysfunctional beliefs than NG sleepers (between group mean difference=0.94, p=0.002) and NP sleepers (between group mean difference=1.12, p=0.000), but not compared with CP sleepers. CP sleepers had significantly higher levels of dysfunctional beliefs than NG (between group mean difference=0.82, p=0.004) and NP (between group mean difference=1.00, p=0.000). There was no statistically significant difference between NG and NP sleepers.