Tools used

Two types of sleep diaries were used in this study. The USD (Fig 1A), used in common practice in a sleep department, was in the form of a grid. The hours of the day were in columns and the days of the month were in rows. The patient reported the time of day (with vertical arrows pointing up and down), the time of night (hatched rectangle), and the time of day (vertical arrows pointing down and up), respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Presentation of the sleep diaries. (A) The usual sleep diary (USD) – The downward arrow marks the lights-out time, the upward arrow indicates the wake-up time, and the hatch marks represent the sleep duration. (B) The light/sleep diary (LSD) – The downward arrow marks the lights-out time, the upward arrow indicates the wake-up time, and the hatch marks represent sleep duration. Unshaded rectangles indicate periods of outdoor daylight exposure.

https://doi.org/10.1371/journal.pone.0317788.g001

The light/sleep diary (LSD) (Fig 1B), an innovative tool, is the result of a test phase of seven successive versions of the diary, which initially included 5 light levels. Feedback from patients and medical teams has led to a simplified consensual version: it distinguishes between exposure to natural light outdoors, exposure to artificial light indoors, and exposure to darkness. The LSD was similar to the previous one, but additionally included periods of outdoor daylight exposure (unshaded rectangles) and hours spent in the dark (solid horizontal lines delimited by thick dots).

These two diaries have a precision of a quarter of an hour and are accompanied by an explanatory leaflet and a questionnaire to record certain variables of interest. Both diaries incorporate the elements from the consensus sleep diary[7], while maintaining the more visual graphic aspect for clinical practice.

Each diary was randomly assigned to either USD or LSD, with its own identification number. It was distributed either by mail by the department secretary or given by the nurse in the ambulatory sector along with explanations on how to complete it. A register of distributed diaries was established. They were asked to fill it out daily until their next consultation.

Data collection

The patients included were randomized in two parallel groups that were considered independent: distribution of an USD versus distribution of a LSD.

To be included, patients had to be of legal age, had a consultation or hospitalisation in the sleep pathology department of Strasbourg University Hospital, and provided written consent. Patients with physical or cognitive disabilities resulting in an inability to complete the diary were not included.

For each type of diary, the main outcome was the proportion of diaries that were returned and correctly completed relative to the total number distributed.

In the absence of a reference to consider whether an agenda was correctly completed, the research team established a consensus according to the following criteria: at least one line had to be dated and at least 7 consecutive days were completed. Each line corresponding to one day was to have a pair of arrows pointing in opposite directions. A hatched section was to have clear boundaries that would allow accuracy to within 15 minutes. Any other symbol or text not provided for in the instructions and which could interfere with interpretation should not be present on the grid. At the same time, the questionnaire attached to the diary was used to record the patient’s age, sex, socio-professional group (INSEE categories), level of diploma (INSEE categories), subjective difficulty in completing the diary using a 100 mm visual analogue scale (VAS), and history of sleep pathologies grouped according to the major categories of the ICSD-3.

Data analysis

For the analysis, patients who returned their diary were divided into two groups according to the distributed diary, namely USD or LSD.

For a first species alpha risk at 5%, a study power at 80%, a predicted correctly completed diary rate at 50% and a chosen non-inferiority threshold at 10%, the number of diaries to be distributed was estimated at 350 distributed diaries per group.

The characteristics of the two patient populations (USD versus LSD group) were first described. Quantitative variables were presented according to their means and standard deviations (if normally distributed), if not according to their median and interquartile range. Qualitative variables were presented by numbers (percentages) for each modality.

In order to verify the non-inferiority conditions, we presented the analysis as per protocol and then as intention to treat.

To meet the secondary objective, we first sought the links, through univariate analysis, between the proportion of correctly completed agendas (variable to be explained) and the risk factors studied (explanatory variables) using Mann–Whitney Wilcoxon tests, Chi2 tests or exact Fisher tests depending on the conditions of application of the present tests. A multivariate logistic regression model was then used, incorporating the risk factors from the univariate analysis for which the p-value was less than 0.15 according to a descending stepwise selection of variables. Finally, we looked for the distribution by pathology of patients consulting or hospitalized in the department in order to compare it with that of patients who returned the diary.

All the analyses were carried out using R software version 3.1

Limitations of the study

The a priori calculation of the number of patients required involved 350 diaries per group, assuming that 50% of the diaries were returned and filled satisfactorily. In the absence of data on the subject in the literature, this value was proposed on the basis of our clinical judgement. This figure seems to have been accurately estimated since out of 295 diaries returned, 148 (50.2%) were completed according to the criteria. We considered a 9-month follow-up period adequate given the high activity at the consultation center. While this did not achieve the expected number of subjects, it came close (649 diaries were actually distributed versus the 700 expected). The return rate of 45.5% had not been anticipated and is responsible for the unmet requirement. The non-inferiority threshold for the main objective was –10%. This value was chosen arbitrarily in the absence of literature to guide the decision. It is a good compromise between a sufficiently small difference so that clinical comparability can be maintained between the two agendas and an acceptable number of patients to be included in order to conclude. A non-inferiority threshold of 5% would have required 350 to 1,000 patients per group, an even more difficult goal to achieve, given that it took more than six months to distribute these 649 diaries. In the case of a non-inferiority trial, the per protocol analysis (with exclusion of unreturned diaries) is theoretically to be preferred, because it allows us to place ourselves in the maximum hypothesis of showing a difference [17]. The intention-to-treat analysis (with inclusion of diaries not returned by patients as well as diaries not correctly filled in) is biased because it reduces the risk of showing a difference, and is therefore better suited to a superiority trial [18]. We chose to perform both analyses because the inclusion of unreturned diaries had an important practical significance: the diary being the basic tool for most sleep pathologies, the inclusion of patients who do not complete the diary is critical. It was not possible to conclude that the new diary was not inferior to the previous one, either in the intention-to-treat analysis or in the per-protocol analysis. This can be explained by the low proportion of returned agendas, which we had not anticipated and leads to a lack of power of our study. In spite of this, this result is not surprising. Indeed, the addition of light exposure data adds complexity to the diary, which can be a brake on its filling. This confirms the clinical impression of keeping a tool as simple as possible, and of targeting the use of a more complex diary in specific situations [19]. Although there is probably little difference between the two diaries in the proportion of correct filling, the burden of filling the diary should be reduced in order to achieve better compliance [20]. This could be achieved by limiting the time for filling the diaries, but it would also be possible to better target patients to whom the diaries should be offered by offering them to those who feel the need for them and do not perceive them as redundant. All this also requires oral explanations when the diary is handed over. That less than half of the agendas were returned was a real surprise and led us to question and hypothesize the causes of this result. Although a high rate of return of these diaries is ideal, in reality, adherence to medical instructions by patients is generally estimated to be around 50%, which significantly varies due to chronic illnesses. This study shows that the return rate of diaries is close to this average, with no significant difference between USD and LSD. Multiple factors contribute to this shortfall: the complexity of some diaries can make daily completion burdensome, and the requirement for regular updates might deter consistent use over several weeks. Certain requirements could also pose data privacy concerns. Instructions are also sometimes lacking, which could further hinder diary compliance and completeness after distribution. Clearer explanations or periodic reminders could motivate patients who might not fully understand the importance of diary compliance. Regular follow-up might encourage higher diary return rates and provide valuable data for further investigation. All these elements are important levers for improving diary return rates in future studies. The identification of the specific characteristics of those lost to sleep was not possible due to the data collection that took place when the diaries were returned. These data are not known for patients who did not return the distributed diary. Another hypothesis is that of targeting patients to whom it should be distributed: Patients who were more engaged with their treatment might have been more likely to return their diaries. Motivation appears to be a key factor in completing the sleep diary. Insomniac patients were proportionally twice as represented in the population, having returned the diaries compared to the population seen in the sleep department during this period. This reinforces our impression that this population, which is one of the main targets of our agenda, found the filling of this agenda useful[15,21]. We can also hypothesize that certain factors evaluated for the quality of agenda completion influenced this return rate.

We thus made the hypothesis that the factors associated with correct filling were not different from one schedule to another. The analysis of these factors was done by considering the overall number of diaries returned, regardless of the type of diary, in order to increase the power of the results. The univariate analysis showed that the quality of filling was better when the diary was given in person and with explanations, and when it was deemed easier to fill. In multivariate analysis, we suspected an interaction between the variables: distribution methods, explanations, subjective difficulty felt. Indeed, the nurses’ interventions accompanied by explanations with patients could have had an effect both on the quality of the information collected and on the impression of difficulty felt when filling in the diary. Effect not found when the diaries were distributed by the secretaries by post. During the construction of the model, no significant interaction model tested was found. This nevertheless led to a questioning of the way the service operates for the distribution of this tool: patients’ adherence to an assessment and support tool, whatever the pathology, depends on their understanding of it and the way they take it on board [22]. In the evaluation of the factors associated with a correct filling of the diary, we had hypothesized a better filling for patients with a higher level of education and related socio-professional categories. This was not the case. This result is interesting because it suggests that it is a tool that would not deepen social inequalities in health. It also raises the hypothesis that its completion is more related to patient motivation than to the understanding of the tool. In fact, when the diaries were correctly filled out, we found that they were often filled out over a longer period than recommended, several weeks. This element confirms the good appropriation of the tool by these patients, and shows the interest they can find in it.

In current practice, diaries are used to try to limit recall bias phenomena encountered when information on past events needs to be gathered retrospectively from simple patient-clinician interviews by recording the event at the time of its occurrence [23]. However, these agendas require repeated recording, which is the reason for the resulting compliance problem [20]. The problem of concealment of data must also be addressed. A study by Mazze et al. found that 2/3 of the patients had recorded false values to mask hypo- or hyperglycaemia in their blood glucose diary [24].

At present, electronic diaries are available for use with smartphones or tablets, and a sleep diary for patients, doctors and researchers is in the development phase [25]. They would have the advantage of time-stamping the data and thus limit the memory bias to improve compliance [26]. For the sleep diary, it would reduce reading ambiguities related to handwritten entries, limit interpretation time during consultations, and provide easier access to the diary [22]. Tonetti et al. compared an electronic tablet diary with a paper diary and actimetry in 15 healthy volunteers. The authors concluded that the two forms of the diary were equivalent and that there was a practical advantage for the electronic diary. Some studies also support the feasibility of estimating light exposure with a diary tool [27].

The use of the diary could lead to better compliance and greater ease of use and availability of the diary.

In conclusion, this study introduced a modified sleep diary to report usual lighting conditions in order to identify situations at risk of light exposure deficiency. The non-inferiority of this new diary compared to the traditional diary concerning the quality of filling could not be shown by a lack of power resulting from a large number of lost sight. Nevertheless, the targeting of patients, the provision of explanations, the limitation of information and the switch to electronic support offer interesting prospects for improving the return and correct filling rates of the distributed diaries. Future studies could be conducted on a larger scale, involving additional centers over a longer study period. Focusing on specific sleep troubles, especially chronic insomnia and circadian rhythm troubles should improve the patients’ motivation. Furthermore, follow-up campaigns, a hybrid format (electronic or paper), clear instructions at the time of distribution, and verification of understanding could help achieve a better return rate. Several additional studies could gradually validate the new LSD diary.