The present paper reports on a 3-year longitudinal study including one baseline survey (T0) and 10 equally spaced (i.e., at 3-month intervals) consecutive surveys (T1 to T10) as well as telephone interviews. The baseline survey was issued in spring 2012 (T0) and was returned by mail together with a completed informed consent form. The subsequent 10 surveys were completed online using the software Textalk Websurvey (www.textalk.se; Gothenburg, Sweden). The second survey (T1) was issued the first week of September 2012. The data were thereafter collected in December, March, June, and September until December 2014 (T10). Each websurvey had a response window of circa three weeks (or slightly longer in case of interfering public holidays), during which a maximum of three e-mail reminders were issued to non-responders. Telephone interviews served as a supplement to the surveys for participants with prospectively identified sustained stress or exhaustion warning in LUCIE (for details see section
In total, 1355 participants from southern Sweden were eligible to be prospectively followed over the 3-year study period. The recruitment process was described in detail by Persson et al. , and a brief summary follows. A total of 7799 persons, who had either been respondents to a population survey  or identified through the population registry of Skåne University Hospital in southern Sweden, were invited by letter to participate. The invitation letter stated that participants were expected to be gainfully employed at least 75 % of full-time, not to have had any period of long-term full-time sick leave for the past six months, nor to have had a chronic disease or been on daily medication. In total, 1400 persons (18 %) returned a completed baseline questionnaire together with a completed informed consent sheet and were included as potential participants. Data from the baseline questionnaire were used to refine the selection of participants. Exclusion criteria were self-reports of somatic disease, daily medication with psychotropic drugs, excessive alcohol consumption, part-time work below 75 % of full time (<30 h/week), and recent longer full-time sick leave. Forty-five persons failed to pass these criteria and the final cohort consisted of 1355 participants (57 % women). Their mean age at baseline was 41.1 years (SD 6.7 years; range 27–52). University education was reported by 67 %, secondary school by 32 % and elementary school by less than 1 %. Full-time occupational activity (40 h/week) was reported by 83 %. The vast majority were salaried employees, and 10 % were either self-employed or combined paid employment with self-employment .
Lund University Checklist for Incipient Exhaustion (LUCIE) consists of 28 items describing behaviors and feelings associated with the prodromal stages of exhaustion disorder. LUCIE is intended to be a tool for identifying the prodromal stages of work-stress-related exhaustion and is based on qualitative analyses of ED patients’ interviews/narratives concerning their earliest signs of ED. (See Persson et al.  for a detailed description of the basic development of LUCIE and item contents). In LUCIE, the instruction to the respondent is: “For the past month, to what extent have you felt or observed the following?” The response to each item is made on a 4-point scale: 1 = not at all, 2 = somewhat, 3 = quite a bit, and 4 = very much. The LUCIE items cover six domains: (a) sleep and recovery, (b) separation between work and spare time, (c) sense of community and support in the workplace, (d) managing work duties and personal capabilities, (e) private life and spare time activities, and (f) health complaints.
The detection of incipient exhaustion in LUCIE builds on two algorithms comprising two separate, supplementary scales: the Stress Warning Scale (SWS), which is sensitive to milder signs of incipient exhaustion, and the Exhaustion Warning Scale (EWS), which is intended to reflect more severe signs of exhaustion. The general difference between the SWS and the EWS algorithms concerns the intensity of the replies, as the EWS score is based mainly on replies at the highest level (“very much”), while replies on the next lower level (“quite a bit”) are also included in computation of the SWS. A wide range of stress signs on the next lower level is thus only reflected in a high SWS score, while the extent of replies on the highest level are recorded on the EWS scale. The purpose with this division on two scales is to enable the clinician to easily assess whether the LUCIE result indicates a slight to moderate state of stress (high SWS and low EWS) or if signs are so intense that ED might be suspected (high EWS) . The SWS and EWS computation algorithms are presented in detail in Additional file 1.
The scores on both scales range from 0 to 100. A low SWS score (≤ 17.00; ‘the green zone’) is intended to indicate normal/negligible long-term stress symptoms. A slightly higher SWS score (between 17.01 and 38.50; ‘the yellow zone’) suggests possible slight stress symptoms. A rather high SWS score (≥38.51; ‘the red zone’) indicates mild to moderate stress symptoms. When the SWS score reaches the red zone, it is recommended to start checking the EWS score for more severe symptoms of stress, possibly indicating exhaustion. A low EWS score (≤ 21.50; ‘the EWS green zone’) indicates that signs of exhaustion are mostly absent or mild, while a higher score (> 21.50; ‘the EWS red zone’) suggests severe symptoms that might indicate exhaustion disorder, in that case overriding any SWS score. In practice, the combined scores on the SWS and the EWS provide a 4-step severity ladder of stress symptomatology:
Step 1-GG (SWS green zone and EWS green zone) = no or negligible lasting stress symptoms
Step 2-YG (SWS yellow zone and EWS green zone) = possible slight lasting stress symptoms
Step 3-RG (SWS red zone and EWS green zone) = mild to moderate lasting stress symptoms, but less severe than ED
Step 4-RR (SWS red zone and EWS red zone) = lasting stress symptoms of a severity indicating possible ED.
Because the other theoretically plausible combinations of scores (i.e., SWS green zone or SWS yellow zone in combination with EWS red zone score) are extremely rare in clinical settings and in population samples , the four ranking steps above are practical simplifications. In cases where a high SWS score is observed, the supplementary EWS measure will indicate whether the stress symptomatology is of an intensity indicative of ED (Step 4-RR), or is more benign in nature (Step 3-RG).
Changes in the situation at work and in private life were assessed using two newly constructed items that addressed perceived positive or negative changes in (a) the work situation or (b) private life. The items read: “Has your situation at work changed in a positive or negative direction during the past couple of months?” and “Has the situation in your private life changed in a positive or negative direction during the past couple of months?” Responses to both items were made on a 5-point scale: 1 = Yes, in a highly positive direction, 2 = Yes, to some extent in a positive direction, 3 = No, no significant change, 4 = Yes, to some extent in a negative direction, 5 = Yes, in a highly negative direction. As a supplement, participants were also encouraged to fill in an optional text field (480 signs) with free-text commentaries.
The Karolinska Exhaustion Disorder Scale (KEDS) was used to validate that a prospective elevation of LUCIE scores (see
Identification of cases
below) reflected genuine signs of exhaustion. KEDS is a recently developed tool for screening for the presence of ED; it contains nine items selected to correspond with the ED criteria specified by the NBHW in 2003 . The item contents are: (1) ability to concentrate, (2) memory, (3) physical stamina, (4) mental stamina, (5) recovery, (6) sleep, (7) hypersensitivity to sensory impressions, (8) experience of demands, and (9) irritation and anger. Each item has seven response alternatives, ranging from 0 to 6, with higher values reflecting more severe symptoms. The sum of item scores constitutes the outcome (range 0 – 54). A sum of item scores ≥ 19 is the recommended cutoff criterion for ED, which was shown to optimize both sensitivity and specificity .
Personality traits were assessed at baseline (T0) with a Swedish version of the 44-item Big Five Inventory (BFI), which includes the dimensions: Neuroticism, Extraversion, Openness, Agreeableness and Conscientiousness [14, 15]. The items of the BFI are short and easily understandable phrases, and each BFI item is rated on a 5-point scale with verbal labels ranging from “Disagree strongly,” (score 1), “Disagree a little,” “Neither agree nor disagree,” “Agree a little,” and “Agree strongly” (score 5). Each of the Big Five personality dimensions was calculated as the mean score of the 8–10 items covering the dimension.
Telephone interviews were carried out to collect data on the substance of LUCIE indications and the perceived sources of stress. Specifically, participants showing a prospective elevation of LUCIE scores (see
Identification of cases
below) were informed by letter that unspecified changes in their questionnaire replies had been observed and told we would contact them by telephone for a brief interview. The interviews were carried out by an experienced clinical psychologist and psychotherapist (N.V.). When we reached the participant by phone, he/she was informed that we had seen, during the past two quarters, increased ratings on a set of questions that might be related to work stress. The interviewer then asked the participant the following two questions, one regarding work stressors and the other concerning stressors outside work (private life). The first question read “In your opinion, have you experienced more stress at work lately (during the past six months) than previously and, if so, to what extent?” The second question read “Do you think there are other reasons (outside work) for the changes in your questionnaire replies and, if so, to what extent?” In practice, these two questions were not given in verbatim, and were often conveniently combined into one question; “Have you experienced more stress at work during the past quarters, or do you think there are other reasons (outside work; e.g., private stress) for the changes in your questionnaire replies?” The reply obtained within each area was scored by the interviewer as: 3 = a substantial increase; 2 = a moderate increase; 1 = a slight increase; 0 = no increase or a decrease. After the interview, the balance between reported work and private life stressors was scored as: 1 = work stressors only; 2 = predominantly work stressors but also some private life stressors; 3 = roughly equal shares of work stressors and private life stressors; 4 = predominantly private life stressors but also some work stressors; 5 = private life stressors only. Of the 56 participants we were able to reach by phone, 26 also accepted a clinical consultation with the psychotherapist, which provided a richer background which, however, only in a few cases led to minor adjustments in the interviewer ratings.
Identification of cases with sustained stress or sustained exhaustion warning
Each participant was prospectively followed throughout the study period with the intent to identify the first onset of an episode exhibiting a sustained stress warning or a sustained exhaustion warning in LUCIE. A LUCIE indication case was defined by an episode of two consecutive scores in the red zone on the SWS or the EWS scale, preceded by more or less low/normal scores for the two previous quarters (except in the first sampling round at T2, see below). Once identified, the interview protocol was commenced (see above).
Algorithms for identification of sustained stress warning cases
Abrupt onset to sustained stress warning (SWS-AS): This algorithm identified cases that had two consecutive LUCIE SWS scores in the green zone (score ≤ 17.00) followed by two consecutive scores in the red zone (score > 38.50).
Gradual onset to sustained stress warning (SWS-GS): This algorithm identified cases that started with SWS scores in the green or yellow zone (≤ 38.50), followed by a score in the SWS yellow zone (>17.00 and ≤ 38.50), and then two consecutive scores in the SWS red zone (score > 38.50). To avoid targeting individuals with very minor increases in SWS around the cutoff border (e.g., an increase from SWS = 37 to SWS = 39), an additional criterion was that the SWS red zone scores should be at least 20 points higher than observed in the preceding ‘yellow zone’ phase.
Algorithms for identification of sustained exhaustion warning cases
Onset to sustained exhaustion warning (EWS-S): Irrespective of SWS scores, this algorithm identified cases that had two consecutive EWS scores in the green zone (≤ 21.50) followed by two consecutive EWS scores in the red zone (> 21.50).
Modified algorithms in the first sampling round
In order to maximize the number of participants with a LUCIE indication, slightly simplified algorithms were used in the initial sampling round at T2. We targeted individuals already at T2 if their single baseline (T0) LUCIE score was normal, followed by elevated scores both at T1 and T2. The modified algorithms were as follows:
Abrupt onset to sustained stress warning (SWS-AS): A LUCIE SWS score in the green zone (score ≤ 17.00) followed by two consecutive scores in the red zone (score > 38.50).
Gradual onset to sustained stress warning (SWS-GS): A SWS score in the SWS yellow zone (>17.00 and ≤ 38.50) followed by two consecutive scores in the SWS red zone (score > 38.50), both being ≥ 20 points higher than the preceding yellow zone score.
Onset to sustained exhaustion warning (EWS-S): An EWS score in the green zone (≤ 21.50) followed by two consecutive EWS scores in the red zone (> 21.50).
Participants fulfilling any LUCIE algorithm criteria are henceforth referred to as ‘LUCIE indication cases’, and the phrase ‘after onset of stress/exhaustion indication’ refers to the latter two quarters with elevated LUCIE scores.
When four consecutive LUCIE scores, based on the algorithms given above, confirmed the onset of a sustained stress or exhaustion warning, the time for the fourth quarter was set to Q 0 (Q zero) and the preceding three quarters were defined as: Q minus 3, Q minus 2 and Q minus 1 (Q -3, Q -2, and Q -1). Thus, taking an event-based approach, we corralled all identified events into a new dataset consisting of four quarters (Q-3 to Q 0). Participants with no indication of a sustained stress or exhaustion warning during the study period (n = 1259), and with complete sets of data during the final four assessment waves (T7 to T10), constituted the control group (n = 745). For controls, the final websurvey (T10) was thus defined as Q 0. Of the 745 final controls, 82 % had responded to all 11 quarterly surveys, 17 % failed to reply on 1–3 quarters, and less than 1 % on ≥ 4 quarters. The subgroup of potential controls that did not have complete data during the final four quarters (n = 514) did not differ from the final control group concerning demographic data or self-reports concerning, for example, age, education, personality scores, or LUCIE scores (data not shown).
Statistical analysis and analysis of free-text answers
P-values ≤ 0.05 (two-tailed) were considered statistically significant. Visual inspection of the data revealed that the LUCIE scores were positively skewed. The data from the single-item questions on changes in the situation at work and in private life were also considered unsuitable for parametrical analysis. However, the BFI data were approximately normally distributed. Thus, all data except BFI data were analyzed using non-parametrical tests. Between-groups analyses entailing categorical data were conducted using Pearson’s χ2-tests. Only BFI data were analyzed using one-way ANOVA with Post Hoc LSD tests, and effect sizes are given as partial eta square (ηp2). All tests were made using SPSS/IBM software, version 22.
Thematic analyses of the free-text commentaries were made through an iterative process. To begin with, changes in the work/private situation were scrutinized to establish empirically common and broad general themes. Then a more fine-grained reading of each commentary sorted them into preliminary thematic categorizes. Next, changes in the work situation themes were, if possible, broadly categorized according to themes found in extant mainstream models of work stressors, e.g., the demand-control-support model, the effort-reward model and the organizational injustice model. Finally, an iterated check of the free-text commentaries was carried out to confirm or disconfirm the previous rounds of classification.