Sleep is an important foundation for good physical and mental health as it supports every system in the body including metabolic, immune, cognitive function and emotional regulation [1]. Sleep quality is a complex construct and includes quantitative characteristics such as sleep duration, number and length of awakenings, sleep regularity, as well as subjective aspects like perceived sleep depth and daytime functioning. Sleep health is multifaceted and influenced by biology behaviours as well as contextual and social factors including age, gender, social engagement, socio-economic and work status [1].

Within New Zealand (NZ) the COVID-19 pandemic and the mandatory lockdown that followed abruptly changed daily life for everyone, nationally as it was the case globally. In many countries, public health authorities enforced social restrictions such as closures of non-essential workplaces, school and day-cares as an epidemiological containment strategy impacting daily routines, physical activity, and other drivers of sleep. In pursuit of an elimination strategy, the early lockdowns in NZ were considered some of the strictest in the world whilst the infection rates remained low [2]. During the strictest level of lockdown (Level 4), only essential services were open, and the population (except essential workers) had to remain within their ‘house bubbles’ (close family members only), for 33 days. In the less rigorous lockdown level (Level 3) gatherings were restricted to up to 10 people [3]. Research suggests the stress associated with the COVID-19 lockdowns and associated social restrictions significantly affected sleep quality on a global scale [4,5,6]. A meta-analysis found the prevalence of sleep disturbances during the pandemic was higher in females (41%) compared to males (31%) [7]. Furthermore, females were identified as more likely to have taken on the task of home-schooling and full-time care for children whilst fulfilling work obligations from home [8]. Therefore, research focusing on the impact of the pandemic on sleep and wellbeing within stratified samples of working females are warranted.

A measure for sleep quality used extensively in sleep research is the Pittsburgh Sleep Quality Index (PSQI) [9] which is the focus of this study. The PSQI is a self-report questionnaire comprised of 19 items that assess subjective sleep quality during the previous month. It is one of the most widely used self-report measures for assessing subjective sleep quality and is well validated in clinical and community populations. Typically, all variables contribute to calculating one or more of the seven component scores which indicate various indicators of sleep health (or disturbance). They are then collapsed again to form a global score (between 0 and 21) making it an easy applicable instrument for research and clinical practice [10]. However, more recently, the performance of self-reported sleep scales has been debated. For example across time, clinical profiles, and age-groups or different cultures [11]. This calls to question how sleep quality scales such as the PSQI performed during the pandemic lockdown (a context which dramatically changed routines and behaviours). Few studies have assessed the PSQI structure validity during the COVID-19 pandemic [12] and in special populations such as working females.

To date, most researchers use the PSQI [9] with the global score (i.e. as a one-factor model) aiming to capture all attributes of subjective sleep quality. Prior the COVID-19 pandemic, some studies have shown that multifactorial models for PSQI can improve the probability of detailing the severity of sleep disturbance because components are represented and weighted across separate domains [13, 14]. Fabbri et al.’s recent systematic review of the PSQI psychometric properties reported good internal reliability and validity however, different factorial structures were noted; six papers reported a single dimension, six studies indicated a two-factor model and two papers a three-factor model [11]. Manzar et al. (2018) [15] conducted another meta-review of the PSQI factor structure, summarising 30 distinct PSQI models proposed in the literature. However, due to methodological discrepancies between the 45 studies included (for example, adequacy of sample evaluation, application of factor analysis, variation in software used, tests conducted, and outputs reported), the application of these findings is limited. To overcome these shortcomings, the authors proposed methodological guidelines for examining the structure validity of the PSQI in future studies. This is especially the case for habitual sleep quality under different environmental and social constraints. The internal factor validity may differ between circumstances that vary compared to those within which it was initally validated. Based on Manzar et al.’s (2018) [15] suggestions, Jia and colleagues [14] re-examined the PSQI structural validity in a large (N = 2189) non-clinical sample of Americans (64% female, mean age 35.9 years, SD = 12.2) by testing one, two, and three-factor models. Their results indicated that the two-factor model (which they named ‘sleep efficiency’ and ‘sleep latency’) and three-factor models (which they named ‘sleep efficiency’, ‘sleep latency’ and ‘sleep quality’) were statistically superior to the one-factor PSQI (i.e. the original global score). Because Jia et al.’s [14] models used a rigorous methodology to improve discrepancies in the validation literature, especially when the PSQI is used under different environmental and social constraints as described above, a similar approach was used in this study. Note, due to the naming of factors by Jia et al., the terminology around ‘sleep latency’ and ‘habitual sleep efficiency’ are used differently here compared to clinical definitions used elsewhere [16].

These works indicate the importance of considering the internal reliability of the PSQI items in various research contexts. The COVID-19 lockdowns created a situation where, due to the social restrictions, self-reported sleep status changed [4,5,6]. However, the interpretation and reliability of responses to items within surveys such as the PSQI may also have been affected. For example, the reliability of estimating bed and sleep times may have been hindered because, for many, the external drivers to physically attend work or educational facilities were dropped (and therefore regularity and remembrance of routines and use of aids such as alarm clocks reduced). Furthermore, it is anticipated that how participants interpret and estimate frequencies of issues such as having “trouble staying awake while driving, eating meals, or engaging in social activity?” may also be questionable during a period when confined to their homes with limited social engagement, despite having low infection rates (as was the case in NZ).

Given prior studies questioning the use and interpretation of the PSQI in various conditions [11, 14, 15], the current study aimed to assess the factorial validity of the PSQI amongst a unique population of working females during New Zealand’s first national lockdown (April–May 2020). This is novel as few studies have examined the structural validity of PSQI during the context of the COVID-19 pandemic restrictions. It is also the first study to assess factorial validity in this population who, as outlined above, have unique factors affecting their sleep. The structure validity of the original one-factor model (hypothesised to be superior due to its common-use, validity, and reliability across different populations and contexts prior to the pandemic) was evaluated against Jia et al.’s (2019) [14] two- and three- factor models to evaluate which has the best model data fit within this unique context and population.