Of the 167 raw data files, 157 had valid data for at least one weekday and one weekend day. Once combined with the questionnaire and body composition data, 145 had complete data. Participants averaged 7.54 (0.61) days of valid data, with 94.3% (7.6) wear time and 5.7% (7.6) of non-wear time. The average number of week days and weekend days with valid data was 5.9 (0.5) and 2.0 (0.2) days, respectively. Except for self-reported PA (p = 0.002), there were no significant differences in socio-demographic, behavioural or biological variables between the study participants and the full DHDSS AWI-Gen sample .
A total of 1148 sleep periods were recorded, of which 13 (1.1%) fell completely within 06h00-18h00. Four of these sleep periods were recorded by HIV+ participants, of which three periods (309–340 minutes, 5–14 awakenings, WASO 28–57 minutes) were recorded by one HIV+ male (56 years, 19.2 kg/m2). Only WASO (55 vs. 49 minutes) and NOA (16 vs. 14 awakenings) were significantly higher in HIV+ compared with HIV− participants (p ≤ 0.0404).
Females were significantly higher in adiposity, prevalence of hypertension, consumed more sugar sweetened beverages (p ≤ 0.0375), and had significantly higher levels of objectively-measured PA (p = 0.0007) (Table 1). No males were obese (≥ 30 kg/m2) and significantly more males compared with females, reported current alcohol use (p < 0.0001) (Table 1).
[Insert Table 1 here]
Sleep quantity (nocturnal) was significantly higher in females (p ≤ 0.0464), while sleep quality indices were significantly poorer in females (p ≤ 0.0290) (Table 2). Males slept more during the day (p = 0.0290), were in bed more than 90 minutes earlier than females (p = 0.0016) and had significantly more sleep periods starting between 00:01–05:59 (p = 0.0084) (Table 2). There was no association between sufficient sleep categories for TST or nocturnal sleep time, and gender or day of the week (p ≥ 0.4904) (Table 2).
Sleep quality (SFI) was significantly poorer on weekdays (p = 0.048) and significantly more sleep periods starting between 00:01–05:59 (p = 0.0443) occurred on weekdays (Table 2).
There was no significant association between self-reported TST (Table 1) and objectively-measured TST (Table 2) (p ≥ 0.0568).
[Insert Table 2 here]
Significantly more nocturnal sleep time was accrued on Tuesday and Wednesday compared with the weekend (p ≤ 0.0409) (Fig. 1A), and a similar pattern was found for TST (Fig. 1B). Less than 40% of the participants slept 7–9 hours (Fig. 1B). Generally, the NOA and WASO were significantly higher on Tuesday-Wednesday, compared with the Saturday-Monday (p ≤ 0.0451) (Fig. 1C-D). Activity counts during sleep were significantly higher on Wednesday compared with Saturday (p = 0.0076) (Fig. 1E) and 60–70% of participants achieved ≥ 85% SE (Fig. 1F). There was no significant difference across days for diurnal sleep time (p = 0.3463), any fragmentation index (p ≥ 0.2828), or length of awakenings (p = 0.3463).
[Insert Fig. 1 here]
Except for TST (forced WC model: β=-0.033, p = 0.0453), sleep indices were not related adiposity measures in multi-variate analyses (Table S1-S2, see Additional file 1). In contrast, current morbidity status, socio-demographic indices and lifestyle factors were significantly associated with adiposity indices (Table S1-S2, see Additional file 1). In female models, parity was not related to adiposity level (p > 0.05).
PA was consistently and significantly, inversely associated with TST and SE (p ≤ 0.0077), while socio-economic and -demographic indices were significantly associated with sleep indices (p ≤ 0.0389) (Table S3-S4, see Additional file 1). In female models, parity was significantly and directly associated with TST and SE (p ≤ 0.0430) (Table S4, see Additional file 1).
PA was significantly and directly related to WASO and SFI in forced and backward-selection sleep models (p ≤ 0.0360).