- Email: [email protected]
Can registry data be used as a proxy for perceived stress? A cross-sectional study
Annals of Epidemiology xxx (2016) 1e7
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Original article
Can registry data be used as a proxy for perceived stress? A cross-sectional study Laura Schärfe Jensen MD a, *, Charlotte Overgaard MSc, PhD b, Jens Peter Garne MD c, Kathrine Carlsen Cand.Scient., PhD d, Henrik Bøggild MD, PhD, Post.doc b, Kirsten Fonager MD, PhD a, b a
Department of Social Medicine, Aalborg University Hospital, Denmark Public Health and Epidemiology Group, Department of Health Science and Technology, Aalborg University, Denmark Department of Breast Surgery, Ringsted Hospital, Denmark d Research Centre for Prevention and Health, Glostrup University Hospital, Denmark b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 6 November 2015 Accepted 17 May 2016 Available online xxx
Purpose: This study explores the applicability of registry data as a proxy for perceived stress by examining the association between perceived stress measured in health surveys and registry data. Methods: Of 35,700 randomly invited participants from the 2010 Health Survey in the North Denmark Region (age 16e99 years), 21,842 answered 10 items from Cohen’s Perceived Stress Scale. Respondents were divided into quartiles based on their stress score. Survey information was individually linked to national registries containing information on prescribed psychiatric medication and consultations with psychologists or psychiatrists from 2009 to 2011. Results: The percentage of persons with prescriptions or consultations was higher (37.6%) in the highest stress score group, compared with the lowest stress score group (7.7%). Odds ratio (95% confidence interval) for the highest score compared with the lowest score was 7.3 (6.5e8.1). Different combinations of treatment showed low sensitivity (8.7%e37.6%), positive predictive value (49.4%e56.8%), and positive agreement (16.2%e42.7%) were found, whereas specificity (88.5%e98.0%) and negative agreement (85.5% e87.2%) were higher. Kappa measure showed slight to fair agreement (0.104e0.285). Conclusions: Participants reporting high perceived stress were more often prescribed medications and referred for consultations with psychologists or psychiatrists. However, due to low predictive values, registry data may not be suitable as a proxy for perceived stress. Ó 2016 Elsevier Inc. All rights reserved.
Keywords: Stress Psychological Data collection Methods
Introduction In several studies, self-reported perceived stress has been associated with the severity of disease and symptoms (e.g., coronary heart disease, asthma, and pain intensity), and the use of health care (e.g., after cancer treatment) [1e4]. One of the most widely used scales for measuring perceived stress is Cohen’s Perceived Stress Scale (PSS), which measures the degree to which respondents felt they were unable to control important things in their life, their confidence in their ability to handle personal problems, how often they felt they could not cope with all the things they need to do, and how often difficulties were overwhelming within the past month [5]. Data on perceived * Corresponding author. Department of Social Medicine, Aalborg University Hospital, Havrevangen 1, 2. Sal 9000, Aalborg, Denmark. Tel.: þ45 97 66 41 40. E-mail address: [email protected] (L.S. Jensen).
stress can be collected through study-specific questionnaires and/or health surveys [6] and are used in epidemiologic research [1,7,8]. The validity of studies using data from questionnaires or health surveys may be threatened by the risk of misclassification and/or a low response rate. Furthermore, designing and conducting surveys can be time consuming and costly [9]. Registry-based studies have large sample sizes and, thus, high precision and a reduced risk of misclassification and bias due to nonresponse. However, data have been [10] collected for administrative rather than research use; therefore, relevant information may be unavailable [11]. According to Cohen [12], stress is seen as the product of two constructs, impinging demands and compromised resources, which conjoin to produce somatic and mental changes that put people at risk for pathology. Cohen has not described further theoretical aspects in the model, and the questionnaire has been found to be unidimensional in factor analyses. Stress, in general, has been found
http://dx.doi.org/10.1016/j.annepidem.2016.05.008 1047-2797/Ó 2016 Elsevier Inc. All rights reserved.
Please cite this article in press as: Jensen LS, et al., Can registry data be used as a proxy for perceived stress? A cross-sectional study, Annals of Epidemiology (2016), http://dx.doi.org/10.1016/j.annepidem.2016.05.008
2
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
to be related to depression [13,14], anxiety [14], and presumably psychotic disorders [15]. These disorders can be treated with medication (i.e., antidepressants, antipsychotics, and anxiolytics) and/or consultations with psychologists or psychiatrists. Information about treatment can be found in the large Danish administrative registries and may work as a proxy for survey data on perceived stress. Therefore, the present study compared the information about perceived stress measured in health surveys and data with prescribed medications and consultations with psychologists or psychiatrists found in two different Danish national registries to evaluate whether registry data can be used as a proxy for survey data on perceived stress. Materials and methods Study design This cross-sectional study was based on data from a Danish health survey conducted in 2010 and two Danish national registries from 2009 to 2011. Using the unique civil registration number (CPR number) assigned to all residents of Denmark [16], information on perceived stress from the North Denmark Region Health Survey was linked to two different administrative data sources containing information on prescribed medications and consultations with psychologists and psychiatrists in the primary sector. In the North Denmark Region, a total of 35,700 randomly selected persons (age > 16 years) received a questionnaire in February 2010; 23,392 (65.5%) returned the questionnaire [6]. The questionnaire was part of a nationwide survey. Nonresponders received two reminders. The questionnaire included 10 items constituting Cohen’s PSS, among other questions. Each response to the PSS items was scored 0-4 and the scores summarized for a total score between 0 and 40. Higher scores denoted higher levels of perceived stress. Of the 23,392 persons who returned the questionnaire, 21,842 completed all the PSS items. The PSS score groups were based on quartiles in accordance with other studies [1,17]: PSS score < 9 (lowest PSS score group), 9e12 (second lowest PSS score group), 13e16 (second highest PSS score group), and >16 (highest PSS score group). Different group sizes occurred because of the use of integers in the PSS score. In addition to perceived stress, information about gender and age was drawn from the survey, and the respondents were divided into three age groups: <40 years, 40e65 years, and >65 years. Information from the survey was linked to two national registries covering all Danish citizens: (1) the Danish National Prescription Registry [18] and (2) the National Health Service Registry [19,20]. 1) The Danish National Prescription Registry contains information about the prescriptions redeemed in Denmark. The type of medicine was classified by the Anatomical Therapeutic Chemical Classification System, and the three types of medicine used in this study were: anxiolytics (N05B), antipsychotics (N05A), and antidepressants (N06A). The number of prescriptions redeemed from 2009 to 2011 (i.e., 1 year before and after the respondents answered the questionnaire) was taken from the registries. Two variables were created, one describing whether respondents received any prescriptions and other containing the number of prescriptions reimbursed by each respondent during the period. 2) The National Health Service Registry contained information on the activities of health professionals (i.e., psychologists and psychiatrists) contracted within the Danish tax-funded and public health care system, which includes a large part of the primary sector [19,20]. In the National Health Service Registry,
all persons, all providers, and select services were recorded through invoices [20]. For each respondent, the number of consultations with psychologists and psychiatrists was extracted for 2009e2011. Two variables were created, one describing whether respondents had consulted psychologists or psychiatrists and the other containing the number of consultations for each respondent.
Statistical analysis The distribution of prescribed medications (anxiolytics, antipsychotics, and antidepressants) and consultations with psychologists or psychiatrists in the different PSS score groups were described using percentages and medians. The respondents were divided into three different age groups and the percentages of persons receiving prescriptions or consultations in each group calculated. To determine the association between PSS scores and prescriptions or consultations, we applied logistic regression models presented with odds ratios (ORs) and 95% confidence intervals (CIs), adjusting the models for gender and age. Registry data from each year were analyzed separately to examine whether the associations changed depending on whether register data were collected before or after perceived stress was measured. The agreement between the PSS score and prescribed medication and consultations with psychologists or psychiatrists was presented by: (1) sensitivity, specificity, and positive predictive value using PSS score as a gold standard, (2) positive and negative agreement, and (3) kappa measure. Analyses with different combinations of numbers of prescriptions and referrals were conducted to test whether some combinations were better predictors than others. All analyses were performed using STATA version 11 (StataCorp. 2009, Stata Statistical Software: Release 11; College Station, TX: StataCorp LP. v). Ethics The Danish Data Protection Agency approved this study (Ref. GEH-2014-014). All data were linked and stored in computers held by Statistics Denmark and made available with deidentified personal information to ensure that individuals could not be identified. In accordance with Danish legislation, only aggregated statistical analyses and results are published [21,22]. Registry-based studies do not require written informed consent and ethical approval in Denmark [21,22]. Results The completeness of answers to items on perceived stress among those who returned the questionnaire declined with age (<40 years, 95.9%; >65 years, 86.4%). The “missing group” represented 6.6% of the respondents. Women had a higher mean PSS score than men (12.9 vs. 11.5), and more women than men received prescriptions or consulted psychologists or psychiatrists (19.2% vs. 11.4%). The most frequently redeemed medication was antidepressants, and the least redeemed was anxiolytics. Overall, 0.6% of the respondents redeemed prescriptions for anxiolytics, 3.5% for antipsychotics, and 12.1% antidepressants. A higher percentage of respondents consulted psychologists (3.2%) than psychiatrists (1.0%). The overall percentage of respondents who were prescribed medication or consulted psychologists or psychiatrists was 15.4%. The percentage of respondents who received prescribed medication or consulted with psychologists or psychiatrists was higher (37.6%) in the highest PSS score group, compared with the lowest
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
3
Table 1 Characteristics of the different PSS score groups and the relationship with anxiolytics, antipsychotics, antidepressants, and referrals from 2009 to 2011 (n ¼ 23,392) PSS score <9 n ¼ 6578 Age 52 (39e63) Median (IQR)* Gender Women 2902 (54.1) Prescribed anxiolytics Yes 12 (0.2) Median (IQR)* 12.5 (9.0e22.5) Prescribed antipsychotics Yes 126 (1.9) 1.0 (1.0e3.0) Median (IQR)* Prescribed antidepressants Yes 312 (4.7) Median (IQR)* 9.0 (3.0e17.5) Prescribed 1, 2, or 3 Yes 450 (6.8) Median (IQR)* 5.0 (1.0e13.0) Consultations with psychologist Yes 81 (1.2) Median (IQR)* 5.0 (2.0e12.0) Consultations with psychiatrist Yes 12 (0.2) Median (IQR)* 6.5 (2.0e20.5) Consultations with 5 or 6 Yes 93 (1.4) Median (IQR)* 5.0 (2.0e12.0) Prescribed medicine or consultations Yes 505 (7.7) Prescribed medicine and consultations Yes 38 (0.6)
PSS score 9e12 n ¼ 5615
PSS score 13e16 n ¼ 4643
49 (36e62)
52 (37e64)
2825 (53.3) 18 (0.3) 1.5 (1.0e8.0)
2506 (54.0) 38 (0.8) 3.0 (1.0e15.0)
PSS score >16 n ¼ 5006 51 (36e66) 2959 (59.1)
No answer to PSS n ¼ 1550 51 (38e63) 909 (58.6)
66 (1.3) 11.0 (2.0e25.0)
14 (0.9) 9 (1.0e17.0)
294 (5.9) 2.0 (1.0e10.0)
97 (6.3) 2.0 (1.0e7.0)
152 (2.7) 1.0 (1.0e3.0)
149 (3.2) 2.0 (1.0e5.0)
387 (6.9) 8.0 (12.0e17.5)
501 (10.8) 10.0 (3.0e20.0)
1,372 (27.4) 12.0 (5.0e25.0)
246 (15.9) 12.0 (5.0e29.0)
557 (9.9) 5.5 (1.0e12.0)
688 (14.8) 7.0 (2.0e16.0)
1,732 (34.6) 11.0 (3.0e22.0)
357 (23.0) 10.0 (2.0e23.0)
123 (2.2) 6.0 (3.0e9.0)
149 (3.2) 6.0 (3.0e11.0)
362 (7.2) 7.0 (4.0e11.0)
35 (2.3) 6.0 (3.0e11.0)
29 (0.5) 10.0 (4.0e23.0)
41 (0.9) 7.0 (3.0e23.0)
140 (2.8) 11.5 (5.5e23.0)
15 (1.0) 9.0 (4.0e16.0)
152 (2.7) 6.0 (3.0e10.0)
190 (4.1) 6.0 (3.0e12.0)
502 (10.0) 8.0 (4.0e12.0)
50 (3.2) 7.0 (3.0e12.0)
639 (11.4)
786 (16.9)
1,884 (37.6)
373 (24.1)
70 (1.3)
92 (2.0)
350 (7.0)
34 (2.2)
Data are n (%) unless otherwise noted. * The median and interquartile range (IQR) of prescriptions and CI per person for those who received medicine or referrals in each PSS group.
PSS score group (7.7%) (Table 1). More in the older age groups received medication, and fewer in the older age groups consulted psychologists or psychiatrists compared with the younger age group (Fig. 1). A high PSS score was associated with increased risk of being prescribed medication or consulting psychologists or psychiatrists compared with a low PSS score (Table 2). The highest OR was found for consultations with psychiatrists (15.7, 95% CI 8.7e28.4). The probability of being prescribed medication or consulting psychologists or psychiatrists increased with PSS score using the lowest PSS score group as a reference (second lowest PSS score group: OR 1.5, 95% CI 1.4e1.7; second highest PSS score group: OR 2.5, 95% CI 2.2e2.8; highest PSS score group: OR 7.3, 95% CI 6.5e8.1). Adjusting for gender and age did not change the results. Subanalysis comparing survey data solely to registry data from each year (Fig. 2) showed the same pattern. Sensitivity (8.7%e37.6%), specificity (88.5%e98.0%), positive predictive value (49.4%e56.8%), positive (16.2%e42.7%) and
negative (85.5%e87.2%) agreement along with kappa coefficient (0.095e0.285) were estimated for different combinations of numbers of prescriptions and consultations with psychologists or psychiatrists (Table 3). Discussion Key findings The study demonstrated a significant association between perceived stress scores and redeemed prescriptions or consultations with psychologists or psychiatrists. Furthermore, low sensitivity, positive predictive value and positive agreement, but high specificity and negative agreement were found for different combinations of numbers of prescriptions and consultations with psychologists or psychiatrists. Kappa measure showed slight to fair agreement between registry data and perceived stress. Finally, the study showed that young respondents (age < 40 years) with high
Fig. 1. Percentage of persons receiving prescriptions or consulting psychologists or psychiatrists for the different age and stress score groups.
4
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
Table 2 Odds of being prescribed medicine or consulting a psychologist or psychiatrist in the different PSS score groups from 2009 to 2011 (n ¼ 21,842) The different stress score groups PSS score
Prescribed anxiolytics Prescribed antipsychotics Prescribed antidepressants 1, 2, or 3 Consultations with psychologist Consultations with psychiatrist Consultations with 5 or 6 Either medicine or consultation
Low <9
1 1 1 1 1 1 1 1
Second lowest 9e12
Second highest 13e16
Highest >16
OR (95% CI)
Adj. OR (95% CI)*
OR (95% CI)
Adj. OR (95% CI)*
OR (95% CI)
Adj. OR (95% CI)*
1.8 1.4 1.5 1.5 1.8 2.8 1.9 1.5
1.8 1.4 1.5 1.5 1.6 2.8 1.8 1.5
4.5 1.7 2.4 2.4 2.7 4.9 3.0 2.5
4.6 1.6 2.3 2.3 2.4 5.0 2.8 2.3
7.3 3.2 7.6 7.2 6.3 15.7 7.8 7.3
7.4 2.8 7.2 6.9 5.5 16.1 7.2 6.8
(0.8e3.7) (1.1e1.8) (1.3e1.7) (1.3e1.7) (1.4e2.4) (1.4e5.6) (1.5e2.5) (1.4e1.7)
(0.9e3.8) (1.1e1.8) (1.3e1.7) (1.3e1.7) (1.2e2.1) (1.4e5.4) (1.4e2.3) (1.3e1.7)
(2.4e8.6) (1.3e2.2) (2.1e2.8) (2.1e2.7) (2.0e3.5) (2.6e9.3) (2.3e3.8) (2.2e2.8)
(2.4e8.8) (1.2e2.0) (2.0e2.7) (2.0e2.6) (1.8e3.2) (2.6e9.4) (2.2e3.6) (2.1e2.6)
(3.9e13.7) (2.6e3.9) (6.7e8.3) (6.4e8.1) (4.9e8.0) (8.7e28.4) (6.2e9.7) (6.5e8.1)
(4.0e13.8) (2.2e3.4) (6.3e8.2) (6.1e7.7) (4.3e7.0) (8.9e29.2) (5.7e9.0) (6.1e7.6)
Adj. OR ¼ adjusted odds ratio. PSS < 9 ¼ reference group. * Adjusted for age and gender.
PSS scores were more likely to consult with psychologists or psychiatrists and less likely to receive prescriptions compared with older respondents (age > 65 years). Interpretation In line with the results of this study, a Swedish cross-sectional study found that high perceived stress was associated with psychotropic medicine (especially antidepressants). Instead of using Cohen’s PSS, that study used a 30-item stress scale (the Perceived Stress Questionnaire) (24), indicating that the association is not depending on a specific measurement tool. Despite finding an association between perceived stress and redeemed prescriptions or consultations with psychologists or psychiatrists, caution should be exercised in the use of the tested registries as a proxy for perceived stress due to low sensitivity and low positive predictive value. Sensitivity and specificity must be interpreted in relation to PSS, whereas kappa statistic, positive and negative agreement do not require a golden standard. The kappa coefficient was low, properly due to low prevalence, especially for consultations. Positive and
negative agreement is less affected of low prevalence, and still positive agreement values were less than 43% (21). Other studies have compared survey data and registry data, but the results have not been consistent [23,24], indicating that registry data can sometimes, but not always, be used as a proxy. The registry data on prescribed medication and consultations with psychologists or psychiatrists were able to correctly classify the persons with a low stress level (true negative) to a greater extent than the persons with a high stress level (true positive). If registries are to be used as a proxy for perceived stress, then they should correctly classify the true positives instead of the true negatives. The predictive value of a positive test describes the ability of the proxy measure to correctly identify the respondents with a high PSS score; in this study, it was 43%e55%, which is too low. Of the three types of medication examined in our study, antidepressants was most widely used, which is in line with antidepressants being used to treat different conditions (e.g., depression, insomnia, anxiety, and neuropathic pain) [25e27], whereas anxiolytics and antipsychotics are used for other indications and to a lesser extent. Our study found that 5% of the respondents in the
Fig. 2. The association between PSS and prescriptions or consultations with psychologists or psychiatrists in 2009e2011 and in subsets of data from each year (n ¼ 21,842).
0.266 86.6 38.7 53.7 92.2 30.3
k ¼ kappa; NA ¼ negative agreement; PA ¼ positive agreement; PPV ¼ positive predictive value; Sens. ¼ sensitivity; Spec. ¼ specificity.
0.279 86.4 40.6 52.7 91.2 33.0 42.7 49.4 88.5
40.9
85.5 37.6 Prescriptions or consultations
0.285
8.7 98.0 55.1 15.0 87.2 0.095 At least three prescriptions and/or consultation (summarized) 10.0 97.4 53.6 16.9 86.9 0.104 At least one prescription or one consultation Only consultations
9.5 97.7 54.8 16.2 87.0 0.101 At least two prescriptions and/or consultations (summarized)
27.0 93.4 55.0 At least three consultations 0.262 86.7 38.2 29.5 92.5 54.0 At least two consultations 0.276 85.9 34.6 89.9 50.5 At least one consultation Only prescriptions
PPV Spec. Sens.
k NA PA PPV Spec. Sens.
At least two prescriptions
k NA PA PPV Spec.
At least one prescription
Sens.
Table 3 Agreement between a high PSS (>16) and number of prescriptions and consultations with psychologists or psychiatrists (all values are stated as percentages, except kappa)
At least three prescriptions
PA
36.2
NA
86.9
k
0.249
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
5
lowest PSS score group had received antidepressants. One explanation could be that the medication was given for another indication, such as neuropathic pain. However, a recent study reported that only 2% of antidepressants are given for this indication [25]. Another explanation could be that individuals in the lowest PSS score group had a higher PSS score before medical treatment and, therefore, could be considered well treated with a low PSS score. Although, Figure 2 shows the same association independent of treatment period. On the other hand, only 38% of the respondents in the highest PSS score group received either prescriptions or consulted psychologists or psychiatrists. One explanation could be that Cohen’s PSS does not measure the type of stress treated with psychiatric medicine or consultations with psychologists or psychiatrist, indicating that a high PSS does not correlate to any specific psychic problem, but psychic problem of various sorts. Furthermore, this study only included consultations with psychologists or psychiatrists in the primary sector; individuals in the highest PSS score group may have been treated to a greater extent in the secondary sector. More individuals were referred to psychologists than psychiatrists, which is consistent with the fact that psychologists more often treat conditions related to stress than psychiatrists in Denmark. Young respondents (age < 40 years) were less likely to receive prescriptions and more likely to consult with psychologists or psychiatrists compared with older respondents (age > 65 years), which could indicate that general practitioners use different approaches to treat younger and older patients. In Denmark, until 2011, there was an age limit for referrals to psychologists, which meant that older patients (>37 years old if the diagnosis was depression and >28 years old if the diagnosis was anxiety) could not get a subsidy [28]. The age limits were removed for depression and increased to 38 years old for anxiety [29]. An American study in 2015 reported that 13% of older individuals (age > 65 years) compared with 34% of younger individuals (age 25e64 years) diagnosed with depression were treated with psychotherapy [30]. The pattern is convergent with the results of the present study, although the percentages are smaller. In relation to medical treatment, the American study reported that 26% of the elderly adults and 34% of the younger adults received prescriptions for antidepressants, which contradicts the results of this study. Individuals participating in the survey received the questionnaire in February 2010, whereas the data from the registries were recorded from 2009 to 2011, which may affect the interpretation of the results. However, analyzing for years separately showed the same pattern as found for the entire period, indicating rather robust estimates. Of those, invited to participate, 61.2% returned the questionnaire, which is similar to or higher than other studies in the field [31,32] and acceptable taking into account the tendency to declining response rates [33]. Of those, 93.4% answered all 10 items related to Cohen’s PPS. Several types of nonresponses are generally found [23,34]. Nonresponse due to illness could underestimate the perceived stress level if the person has an illness that increases stress but not necessarily affect the association between perceived stress and the use of medication or consultations with psychologists or psychiatrists. On the other hand, the stress level of the study population could be overestimated if nonresponders were healthy individuals who did not have time to complete questionnaires. Strengths The two national registries are described as having high-quality data [18e20]. The two registries cover all Danish citizens with a civil registration number, meaning that for all individuals who
6
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
returned the questionnaire, a complete linkage was made to registry data regardless of whether they completed the PSS items. In the National Health Service Registry, all persons, all providers, and select services are recorded through invoices [20], which makes registration necessary for general practitioners, psychologists, and psychiatrists to receive payment for their services, encouraging accurate registration. Cohen’s PSS is one of the most used stress-related scales in the world and has been translated into different languages and validated in several countries [35e38], including Denmark [39]. Limitations Stress is a confused concept and might be understood as an individual reaction to demands [40], as a relationship between environment and individual taxed by the person as exceeding resources [41] or as a failed reciprocity between efforts and rewards [42] to name a few. The choice of methodological background is important and might have implications for the interpretation of our results. Although Cohen’s PSS is borrowing from Lazarus and Folkman, it is less theoretically founded and thus useful in epidemiologic surveys. Furthermore, the PSS score items do not allow for suggesting preventive actions in relation to use of prescribed medication or use of psychologists or psychiatrists. The Danish National Prescription Registry has two limitations. First, the registry lacks information on the indication for treatment. Second, the data provide information on prescribed medications but not on whether it was taken. However, it is more relevant if a general practitioner assessed the need for treatment than whether the treatment was carried out. In contrast, the National Health Service Registry contains information about consultations with psychologists or psychiatrists, meaning that it is known that the individuals referred by general practitioners received treatment. Not all psychologists in Denmark are contracted within the tax-funded public health care system, and contracted psychologists may be treating patients without reimbursement; thus, individuals in Denmark can be consulting psychologists without it being recorded in the national registry. The consequences for this study could be nondifferential and differential information bias, but as we used combined measures, it should be of lesser importance. What this study adds This study contributes new knowledge on whether registries can act as a proxy for survey data on perceived stress and identifies some of the problems associated with this connection. We found that, although the associations between perceived stress and the use of prescribed medications and consultations with psychologists and psychiatrists are high, the association itself is not enough to predict whether a person has a high level of perceived stress. Further research connecting survey data and registry data could be relevant and would create an opportunity to use registry data as proxy measures for, for example, psychological problems and allow research to be conducted without having to rely on surveys, which are expensive and have problems with low response rates. Conclusions This study demonstrated an association between perceived stress and registry data on prescribed anxiolytics, antipsychotics, and antidepressants or consultations with psychologists or psychiatrists. However, due to low predictive values, the registry data may not be suitable as a proxy for perceived stress.
Acknowledgments The North Denmark Region Health Survey 2010 was funded by the North Denmark Region.
References [1] Redmond N, Richman J, Gamboa CM, Albert MA, Sims M, Durant RW, et al. Perceived stress is associated with incident coronary heart disease and all-cause mortality in low- but not high-income participants in the Reasons for Geographic And Racial Differences in Stroke study. J Am Heart Assoc 2013;2(6):e000447. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid¼ 3886761&tool¼pmcentrez&rendertype¼abstract [accessed 04.08.2015]. [2] Rod NH, Kristensen TS, Lange P, Prescott E, Diderichsen F. Perceived stress and risk of adult-onset asthma and other atopic disorders: a longitudinal cohort study. Allergy 2012;67(11):1408e14. Available at: http://www.ncbi.nlm.nih. gov/pubmed/22943607 [accessed 02.09.2015]. [3] White RS, Jiang J, Hall CB, Katz MJ, Zimmerman ME, Sliwinski M, et al. Higher Perceived Stress Scale scores are associated with higher pain intensity and pain interference levels in older adults. J Am Geriatr Soc 2014;62(12):2350e6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25516031 [accessed 13.05.2015]. [4] Han X, Lin CC, Li C, de Moor JS, Rodriguez JL, Kent EE, et al. Association between serious psychological distress and health care use and expenditures by cancer history. Cancer 2015;121(4):614e22. Available at: http://www.ncbi. nlm.nih.gov/pubmed/25345778 [accessed 17.03.2015]. [5] Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav 1983;24(4):385e96. Available at: http://www.ncbi.nlm. nih.gov/pubmed/6668417 [accessed 08.08.2014]. [6] The National Health Profile. Available at: http://www.sundhedsprofil2010.dk/ [accessed 21.01.2015]. [7] Roepke-Buehler SK, Dong X. Perceived stress and elder abuse: a populationbased study of adult protective services cases in Chicago. J Am Geriatr Soc 2015;63(9):1820e8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 26280151 [accessed 22.10.2015]. [8] Treharne GJ, Lyons AC, Booth DA, Kitas GD. Psychological well-being across 1 year with rheumatoid arthritis: coping resources as buffers of perceived stress. Br J Health Psychol 2007;12(3):323e45. Available at: http://www.ncbi.nlm. nih.gov/pubmed/17640450 [accessed 22.10.2015]. [9] Cook JV, Dickinson HO, Eccles MP. Response rates in postal surveys of healthcare professionals between 1996 and 2005: an observational study. BMC Health Serv Res 2009;9:160. Available at: http://www.pubmedcentral. nih.gov/articlerender.fcgi?artid¼2758861&tool¼pmcentrez&rendertype¼ abstract [accessed 13.08.2014]. [10] Sorensen HT, Sabroe S, Olsen J. A framework for evaluation of secondary data sources for epidemiological research. Int J Epidemiol 1996;25(2):435e42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9119571 [accessed 18.11.2014]. [11] Thygesen LC, Ersbøll AK. When the entire population is the sample: strengths and limitations in register-based epidemiology. Eur J Epidemiol 2014;29(8):551e8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 24407880 [accessed 29.11.2014]. [12] Cohen S, K R, G L. Measuring stress: a guide for health and social scientists. Oxford University Press, Inc; 1998. p. 3e28. [13] Gherardi-Donato EC, Cardoso L, Teixeira CAB, Pereira Sde S, Reisdorfer E. Association between depression and work stress in nursing professionals with technical education level. Rev Lat Am Enfermagem 2015;23(4):733e40. Available at: http://www.ncbi.nlm.nih.gov/pubmed/26444177 [accessed 22.10.2015]. [14] Fan LB, Blumenthal JA, Watkins LL, Sherwood A. Work and home stress: associations with anxiety and depression symptoms. Occup Med (Lond) 2015;65(2):110e6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 25589707 [accessed 22.10.2015]. [15] Lataster T, Valmaggia L, Lardinois M, van Os J, Myin-Germeys I. Increased stress reactivity: a mechanism specifically associated with the positive symptoms of psychotic disorder. Psychol Med 2013;43(7):1389e400. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/23111055 [accessed 21.09.2015]. [16] Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011;39(7 Suppl):22e5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 21775345 [accessed 13.02.2015]. [17] Tavolacci MP, Ladner J, Grigioni S, Richard L, Villet H, Dechelotte P. Prevalence and association of perceived stress, substance use and behavioral addictions: a cross-sectional study among university students in France, 2009-2011. BMC Public Health 2013;13:724. Available at: http://www.pubmedcentral.nih.gov/ articlerender.fcgi?artid¼3750571&tool¼pmcentrez&rendertype¼abstract [accessed 30.07.2015]. [18] Kildemoes HW, Sørensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health 2011;39(7 Suppl):38e41. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/21775349 [accessed 05.02.2015]. [19] Andersen JS, Olivarius ND, Krasnik A. The Danish National Health Service Register. Scand J Public Health 2011;39(7 Suppl):34e7. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/21775348 [accessed 05.02.2015]. [20] Olivarius NF, Hollnagel H, Krasnik A, Pedersen PA, Thorsen H. The Danish National Health Service Register. A tool for primary health care research. Dan
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
Med Bull 1997;44(4):449e53. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/9377908 [accessed 12.08.2014]. Smith B, Chu LK, Smith TC, Amoroso PJ, Boyko EJ, Hooper TI, et al. Challenges of self-reported medical conditions and electronic medical records among members of a large military cohort. BMC Med Res Methodol 2008;8:37. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid¼2447848&tool¼ pmcentrez&rendertype¼abstract [accessed 14.01.2016]. Thygesen LC, Daasnes C, Thaulow I, Brønnum-Hansen H. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scand J Public Health 2011;39(7 Suppl):12e6. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/21898916 [accessed 28.09.2015]. The Danish Data Protection Agency. Datatilsynet. Available at: https://www. datatilsynet.dk/english/the-act-on-processing-of-personal-data/read-the-act-onprocessing-of-personal-data/compiled-version-of-the-act-on-processing-of-pers onal-data. [accessed 28.09.2015]. Bergdahl J, Bergdahl M. Perceived stress in adults: prevalence and association of depression, anxiety and medication in a Swedish population. Stress Heal 2002;18(5):235e41. Available at: http://doi.wiley.com/10.1002/smi.946 [accessed 10.02.2016]. Gundgaard J, Ekholm O, Hansen EH, Rasmussen NK. The effect of nonresponse on estimates of health care utilisation: linking health surveys and registers. Eur J Public Health 2008;18(2):189e94. Available at: http://www. ncbi.nlm.nih.gov/pubmed/17984128 [accessed 25.06.2015]. Chini F, Pezzotti P, Orzella L, Borgia P, Guasticchi G. Can we use the pharmacy data to estimate the prevalence of chronic conditions? A comparison of multiple data sources. BMC Public Health 2011;11:688. Available at: http://www. pubmedcentral.nih.gov/articlerender.fcgi?artid¼3223740&tool¼pmcentrez& rendertype¼abstract [accessed 25.03.2015]. Bourgeois J, Elseviers MM, Van Bortel L, Petrovic M, Vander Stichele RH. The use of antidepressants in Belgian nursing homes. Drugs Aging 2012;29(9):759e69. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 23018612 [accessed 17.04.2015]. Schofield MJ, Khan A. Predictors of prescribed medication use for depression, anxiety, stress, and sleep problems in mid-aged Australian women. Soc Psychiatry Psychiatr Epidemiol 2014;49(11):1835e47. Available at: http://www. ncbi.nlm.nih.gov/pubmed/24907046 [accessed 06.02.2015]. Dharmshaktu P, Tayal V, Kalra BS. Efficacy of antidepressants as analgesics: a review. J Clin Pharmacol 2012;52(1):6e17. Available at: http://www.ncbi.nlm. nih.gov/pubmed/21415285 [accessed 03.01.2015]. Bekendtgørelse om tilskud til psykologbehandling i praksissektoren for særligt udsatte persongrupper - retsinformation.dk. Available at: https://www. retsinformation.dk/forms/R0710.aspx?id¼131868. [accessed 22.10.2015]. Tilskud til psykologbehandling - Bekendtgørelse om tilskud til psykologbehandling i praksissektoren for særligt udsatte persongrupper -
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40] [41]
[42]
7
retsinformation.dk. Available at: https://www.retsinformation.dk/forms/ R0710.aspx?id¼131868. [accessed 22.10.2015]. Sanglier T, Saragoussi D, Milea D, Tournier M. Depressed older adults may be less cared for than depressed younger ones. Psychiatry Res 2015;229(3):905e12. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 26233825 [accessed 28.09.2015]. Kramer M, Schmalenberg C, Brewer BB, Verran JA, Keller-Unger J. Accurate assessment of clinical nurses’ work environments: response rate needed. Res Nurs Health 2009;32(2):229e40. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/19152405 [accessed 13.08.2014]. Grava-Gubins I, Scott S. Effects of various methodologic strategies: survey response rates among Canadian physicians and physicians-in-training. Can Fam Physician 2008;54(10):1424e30. Available at: http://www.pubmedcentral.nih. gov/articlerender.fcgi?artid¼2567275&tool¼pmcentrez&rendertype¼abstract [accessed 13.08.2014]. Hazell ML, Morris JA, Linehan MF, Frank PI, Frank TL. Factors influencing the response to postal questionnaire surveys about respiratory symptoms. Prim Care Respir J 2009;18(3):165e70. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/19104738 [accessed 13.08.2014]. Christensen AI, Ekholm O, Gray L, Glümer C, Juel K. What is wrong with nonrespondents? Alcohol-, drug- and smoking-related mortality and morbidity in a 12-year follow-up study of respondents and non-respondents in the Danish Health and Morbidity Survey. Addict 2015;110(9):1505e12. Available at: http:// www.pubmedcentral.nih.gov/articlerender.fcgi?artid¼4538793&tool¼pmcentrez &rendertype¼abstract [accessed 05.10.2015]. Mimura C, Griffiths P. A Japanese version of the perceived stress scale: translation and preliminary test. Int J Nurs Stud 2004;41(4):379e85. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15050849 [accessed 11.06.2014]. Remor E. Psychometric properties of a European Spanish version of the Perceived Stress Scale (PSS). Span J Psychol 2006;9(1):86e93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16673626 [accessed 06.08.2014]. Chaaya M, Osman H, Naassan G, Mahfoud Z. Validation of the Arabic version of the Cohen Perceived Stress Scale (PSS-10) among pregnant and postpartum women. BMC Psychiatry 2010;10:111. Available at: http://www.pubmedcentral.nih.gov/ articlerender.fcgi?artid¼3016315&tool¼pmcentrez&rendertype¼abstract [accessed 11.06.2014]. Seley H. Stress of life. New York: New York: McGraw-Hill; 1956. Lazarus RS, Folkman S. Stress, appraisal, and coping. Springer Publishing Company; 1984. p. 445. Available at: https://books.google.com/books?id¼i -ySQQuUpr8C&pgis¼1 [accessed 12.02.2016]. Siegrist J. Adverse health effects of high-effort/low-reward conditions. J Occup Health Psychol 1996;1(1):27e41. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/9547031 [accessed 24.02.2016].
Contents lists available at ScienceDirect
Annals of Epidemiology journal homepage: www.annalsofepidemiology.org
Original article
Can registry data be used as a proxy for perceived stress? A cross-sectional study Laura Schärfe Jensen MD a, *, Charlotte Overgaard MSc, PhD b, Jens Peter Garne MD c, Kathrine Carlsen Cand.Scient., PhD d, Henrik Bøggild MD, PhD, Post.doc b, Kirsten Fonager MD, PhD a, b a
Department of Social Medicine, Aalborg University Hospital, Denmark Public Health and Epidemiology Group, Department of Health Science and Technology, Aalborg University, Denmark Department of Breast Surgery, Ringsted Hospital, Denmark d Research Centre for Prevention and Health, Glostrup University Hospital, Denmark b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 6 November 2015 Accepted 17 May 2016 Available online xxx
Purpose: This study explores the applicability of registry data as a proxy for perceived stress by examining the association between perceived stress measured in health surveys and registry data. Methods: Of 35,700 randomly invited participants from the 2010 Health Survey in the North Denmark Region (age 16e99 years), 21,842 answered 10 items from Cohen’s Perceived Stress Scale. Respondents were divided into quartiles based on their stress score. Survey information was individually linked to national registries containing information on prescribed psychiatric medication and consultations with psychologists or psychiatrists from 2009 to 2011. Results: The percentage of persons with prescriptions or consultations was higher (37.6%) in the highest stress score group, compared with the lowest stress score group (7.7%). Odds ratio (95% confidence interval) for the highest score compared with the lowest score was 7.3 (6.5e8.1). Different combinations of treatment showed low sensitivity (8.7%e37.6%), positive predictive value (49.4%e56.8%), and positive agreement (16.2%e42.7%) were found, whereas specificity (88.5%e98.0%) and negative agreement (85.5% e87.2%) were higher. Kappa measure showed slight to fair agreement (0.104e0.285). Conclusions: Participants reporting high perceived stress were more often prescribed medications and referred for consultations with psychologists or psychiatrists. However, due to low predictive values, registry data may not be suitable as a proxy for perceived stress. Ó 2016 Elsevier Inc. All rights reserved.
Keywords: Stress Psychological Data collection Methods
Introduction In several studies, self-reported perceived stress has been associated with the severity of disease and symptoms (e.g., coronary heart disease, asthma, and pain intensity), and the use of health care (e.g., after cancer treatment) [1e4]. One of the most widely used scales for measuring perceived stress is Cohen’s Perceived Stress Scale (PSS), which measures the degree to which respondents felt they were unable to control important things in their life, their confidence in their ability to handle personal problems, how often they felt they could not cope with all the things they need to do, and how often difficulties were overwhelming within the past month [5]. Data on perceived * Corresponding author. Department of Social Medicine, Aalborg University Hospital, Havrevangen 1, 2. Sal 9000, Aalborg, Denmark. Tel.: þ45 97 66 41 40. E-mail address: [email protected] (L.S. Jensen).
stress can be collected through study-specific questionnaires and/or health surveys [6] and are used in epidemiologic research [1,7,8]. The validity of studies using data from questionnaires or health surveys may be threatened by the risk of misclassification and/or a low response rate. Furthermore, designing and conducting surveys can be time consuming and costly [9]. Registry-based studies have large sample sizes and, thus, high precision and a reduced risk of misclassification and bias due to nonresponse. However, data have been [10] collected for administrative rather than research use; therefore, relevant information may be unavailable [11]. According to Cohen [12], stress is seen as the product of two constructs, impinging demands and compromised resources, which conjoin to produce somatic and mental changes that put people at risk for pathology. Cohen has not described further theoretical aspects in the model, and the questionnaire has been found to be unidimensional in factor analyses. Stress, in general, has been found
http://dx.doi.org/10.1016/j.annepidem.2016.05.008 1047-2797/Ó 2016 Elsevier Inc. All rights reserved.
Please cite this article in press as: Jensen LS, et al., Can registry data be used as a proxy for perceived stress? A cross-sectional study, Annals of Epidemiology (2016), http://dx.doi.org/10.1016/j.annepidem.2016.05.008
2
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
to be related to depression [13,14], anxiety [14], and presumably psychotic disorders [15]. These disorders can be treated with medication (i.e., antidepressants, antipsychotics, and anxiolytics) and/or consultations with psychologists or psychiatrists. Information about treatment can be found in the large Danish administrative registries and may work as a proxy for survey data on perceived stress. Therefore, the present study compared the information about perceived stress measured in health surveys and data with prescribed medications and consultations with psychologists or psychiatrists found in two different Danish national registries to evaluate whether registry data can be used as a proxy for survey data on perceived stress. Materials and methods Study design This cross-sectional study was based on data from a Danish health survey conducted in 2010 and two Danish national registries from 2009 to 2011. Using the unique civil registration number (CPR number) assigned to all residents of Denmark [16], information on perceived stress from the North Denmark Region Health Survey was linked to two different administrative data sources containing information on prescribed medications and consultations with psychologists and psychiatrists in the primary sector. In the North Denmark Region, a total of 35,700 randomly selected persons (age > 16 years) received a questionnaire in February 2010; 23,392 (65.5%) returned the questionnaire [6]. The questionnaire was part of a nationwide survey. Nonresponders received two reminders. The questionnaire included 10 items constituting Cohen’s PSS, among other questions. Each response to the PSS items was scored 0-4 and the scores summarized for a total score between 0 and 40. Higher scores denoted higher levels of perceived stress. Of the 23,392 persons who returned the questionnaire, 21,842 completed all the PSS items. The PSS score groups were based on quartiles in accordance with other studies [1,17]: PSS score < 9 (lowest PSS score group), 9e12 (second lowest PSS score group), 13e16 (second highest PSS score group), and >16 (highest PSS score group). Different group sizes occurred because of the use of integers in the PSS score. In addition to perceived stress, information about gender and age was drawn from the survey, and the respondents were divided into three age groups: <40 years, 40e65 years, and >65 years. Information from the survey was linked to two national registries covering all Danish citizens: (1) the Danish National Prescription Registry [18] and (2) the National Health Service Registry [19,20]. 1) The Danish National Prescription Registry contains information about the prescriptions redeemed in Denmark. The type of medicine was classified by the Anatomical Therapeutic Chemical Classification System, and the three types of medicine used in this study were: anxiolytics (N05B), antipsychotics (N05A), and antidepressants (N06A). The number of prescriptions redeemed from 2009 to 2011 (i.e., 1 year before and after the respondents answered the questionnaire) was taken from the registries. Two variables were created, one describing whether respondents received any prescriptions and other containing the number of prescriptions reimbursed by each respondent during the period. 2) The National Health Service Registry contained information on the activities of health professionals (i.e., psychologists and psychiatrists) contracted within the Danish tax-funded and public health care system, which includes a large part of the primary sector [19,20]. In the National Health Service Registry,
all persons, all providers, and select services were recorded through invoices [20]. For each respondent, the number of consultations with psychologists and psychiatrists was extracted for 2009e2011. Two variables were created, one describing whether respondents had consulted psychologists or psychiatrists and the other containing the number of consultations for each respondent.
Statistical analysis The distribution of prescribed medications (anxiolytics, antipsychotics, and antidepressants) and consultations with psychologists or psychiatrists in the different PSS score groups were described using percentages and medians. The respondents were divided into three different age groups and the percentages of persons receiving prescriptions or consultations in each group calculated. To determine the association between PSS scores and prescriptions or consultations, we applied logistic regression models presented with odds ratios (ORs) and 95% confidence intervals (CIs), adjusting the models for gender and age. Registry data from each year were analyzed separately to examine whether the associations changed depending on whether register data were collected before or after perceived stress was measured. The agreement between the PSS score and prescribed medication and consultations with psychologists or psychiatrists was presented by: (1) sensitivity, specificity, and positive predictive value using PSS score as a gold standard, (2) positive and negative agreement, and (3) kappa measure. Analyses with different combinations of numbers of prescriptions and referrals were conducted to test whether some combinations were better predictors than others. All analyses were performed using STATA version 11 (StataCorp. 2009, Stata Statistical Software: Release 11; College Station, TX: StataCorp LP. v). Ethics The Danish Data Protection Agency approved this study (Ref. GEH-2014-014). All data were linked and stored in computers held by Statistics Denmark and made available with deidentified personal information to ensure that individuals could not be identified. In accordance with Danish legislation, only aggregated statistical analyses and results are published [21,22]. Registry-based studies do not require written informed consent and ethical approval in Denmark [21,22]. Results The completeness of answers to items on perceived stress among those who returned the questionnaire declined with age (<40 years, 95.9%; >65 years, 86.4%). The “missing group” represented 6.6% of the respondents. Women had a higher mean PSS score than men (12.9 vs. 11.5), and more women than men received prescriptions or consulted psychologists or psychiatrists (19.2% vs. 11.4%). The most frequently redeemed medication was antidepressants, and the least redeemed was anxiolytics. Overall, 0.6% of the respondents redeemed prescriptions for anxiolytics, 3.5% for antipsychotics, and 12.1% antidepressants. A higher percentage of respondents consulted psychologists (3.2%) than psychiatrists (1.0%). The overall percentage of respondents who were prescribed medication or consulted psychologists or psychiatrists was 15.4%. The percentage of respondents who received prescribed medication or consulted with psychologists or psychiatrists was higher (37.6%) in the highest PSS score group, compared with the lowest
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
3
Table 1 Characteristics of the different PSS score groups and the relationship with anxiolytics, antipsychotics, antidepressants, and referrals from 2009 to 2011 (n ¼ 23,392) PSS score <9 n ¼ 6578 Age 52 (39e63) Median (IQR)* Gender Women 2902 (54.1) Prescribed anxiolytics Yes 12 (0.2) Median (IQR)* 12.5 (9.0e22.5) Prescribed antipsychotics Yes 126 (1.9) 1.0 (1.0e3.0) Median (IQR)* Prescribed antidepressants Yes 312 (4.7) Median (IQR)* 9.0 (3.0e17.5) Prescribed 1, 2, or 3 Yes 450 (6.8) Median (IQR)* 5.0 (1.0e13.0) Consultations with psychologist Yes 81 (1.2) Median (IQR)* 5.0 (2.0e12.0) Consultations with psychiatrist Yes 12 (0.2) Median (IQR)* 6.5 (2.0e20.5) Consultations with 5 or 6 Yes 93 (1.4) Median (IQR)* 5.0 (2.0e12.0) Prescribed medicine or consultations Yes 505 (7.7) Prescribed medicine and consultations Yes 38 (0.6)
PSS score 9e12 n ¼ 5615
PSS score 13e16 n ¼ 4643
49 (36e62)
52 (37e64)
2825 (53.3) 18 (0.3) 1.5 (1.0e8.0)
2506 (54.0) 38 (0.8) 3.0 (1.0e15.0)
PSS score >16 n ¼ 5006 51 (36e66) 2959 (59.1)
No answer to PSS n ¼ 1550 51 (38e63) 909 (58.6)
66 (1.3) 11.0 (2.0e25.0)
14 (0.9) 9 (1.0e17.0)
294 (5.9) 2.0 (1.0e10.0)
97 (6.3) 2.0 (1.0e7.0)
152 (2.7) 1.0 (1.0e3.0)
149 (3.2) 2.0 (1.0e5.0)
387 (6.9) 8.0 (12.0e17.5)
501 (10.8) 10.0 (3.0e20.0)
1,372 (27.4) 12.0 (5.0e25.0)
246 (15.9) 12.0 (5.0e29.0)
557 (9.9) 5.5 (1.0e12.0)
688 (14.8) 7.0 (2.0e16.0)
1,732 (34.6) 11.0 (3.0e22.0)
357 (23.0) 10.0 (2.0e23.0)
123 (2.2) 6.0 (3.0e9.0)
149 (3.2) 6.0 (3.0e11.0)
362 (7.2) 7.0 (4.0e11.0)
35 (2.3) 6.0 (3.0e11.0)
29 (0.5) 10.0 (4.0e23.0)
41 (0.9) 7.0 (3.0e23.0)
140 (2.8) 11.5 (5.5e23.0)
15 (1.0) 9.0 (4.0e16.0)
152 (2.7) 6.0 (3.0e10.0)
190 (4.1) 6.0 (3.0e12.0)
502 (10.0) 8.0 (4.0e12.0)
50 (3.2) 7.0 (3.0e12.0)
639 (11.4)
786 (16.9)
1,884 (37.6)
373 (24.1)
70 (1.3)
92 (2.0)
350 (7.0)
34 (2.2)
Data are n (%) unless otherwise noted. * The median and interquartile range (IQR) of prescriptions and CI per person for those who received medicine or referrals in each PSS group.
PSS score group (7.7%) (Table 1). More in the older age groups received medication, and fewer in the older age groups consulted psychologists or psychiatrists compared with the younger age group (Fig. 1). A high PSS score was associated with increased risk of being prescribed medication or consulting psychologists or psychiatrists compared with a low PSS score (Table 2). The highest OR was found for consultations with psychiatrists (15.7, 95% CI 8.7e28.4). The probability of being prescribed medication or consulting psychologists or psychiatrists increased with PSS score using the lowest PSS score group as a reference (second lowest PSS score group: OR 1.5, 95% CI 1.4e1.7; second highest PSS score group: OR 2.5, 95% CI 2.2e2.8; highest PSS score group: OR 7.3, 95% CI 6.5e8.1). Adjusting for gender and age did not change the results. Subanalysis comparing survey data solely to registry data from each year (Fig. 2) showed the same pattern. Sensitivity (8.7%e37.6%), specificity (88.5%e98.0%), positive predictive value (49.4%e56.8%), positive (16.2%e42.7%) and
negative (85.5%e87.2%) agreement along with kappa coefficient (0.095e0.285) were estimated for different combinations of numbers of prescriptions and consultations with psychologists or psychiatrists (Table 3). Discussion Key findings The study demonstrated a significant association between perceived stress scores and redeemed prescriptions or consultations with psychologists or psychiatrists. Furthermore, low sensitivity, positive predictive value and positive agreement, but high specificity and negative agreement were found for different combinations of numbers of prescriptions and consultations with psychologists or psychiatrists. Kappa measure showed slight to fair agreement between registry data and perceived stress. Finally, the study showed that young respondents (age < 40 years) with high
Fig. 1. Percentage of persons receiving prescriptions or consulting psychologists or psychiatrists for the different age and stress score groups.
4
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
Table 2 Odds of being prescribed medicine or consulting a psychologist or psychiatrist in the different PSS score groups from 2009 to 2011 (n ¼ 21,842) The different stress score groups PSS score
Prescribed anxiolytics Prescribed antipsychotics Prescribed antidepressants 1, 2, or 3 Consultations with psychologist Consultations with psychiatrist Consultations with 5 or 6 Either medicine or consultation
Low <9
1 1 1 1 1 1 1 1
Second lowest 9e12
Second highest 13e16
Highest >16
OR (95% CI)
Adj. OR (95% CI)*
OR (95% CI)
Adj. OR (95% CI)*
OR (95% CI)
Adj. OR (95% CI)*
1.8 1.4 1.5 1.5 1.8 2.8 1.9 1.5
1.8 1.4 1.5 1.5 1.6 2.8 1.8 1.5
4.5 1.7 2.4 2.4 2.7 4.9 3.0 2.5
4.6 1.6 2.3 2.3 2.4 5.0 2.8 2.3
7.3 3.2 7.6 7.2 6.3 15.7 7.8 7.3
7.4 2.8 7.2 6.9 5.5 16.1 7.2 6.8
(0.8e3.7) (1.1e1.8) (1.3e1.7) (1.3e1.7) (1.4e2.4) (1.4e5.6) (1.5e2.5) (1.4e1.7)
(0.9e3.8) (1.1e1.8) (1.3e1.7) (1.3e1.7) (1.2e2.1) (1.4e5.4) (1.4e2.3) (1.3e1.7)
(2.4e8.6) (1.3e2.2) (2.1e2.8) (2.1e2.7) (2.0e3.5) (2.6e9.3) (2.3e3.8) (2.2e2.8)
(2.4e8.8) (1.2e2.0) (2.0e2.7) (2.0e2.6) (1.8e3.2) (2.6e9.4) (2.2e3.6) (2.1e2.6)
(3.9e13.7) (2.6e3.9) (6.7e8.3) (6.4e8.1) (4.9e8.0) (8.7e28.4) (6.2e9.7) (6.5e8.1)
(4.0e13.8) (2.2e3.4) (6.3e8.2) (6.1e7.7) (4.3e7.0) (8.9e29.2) (5.7e9.0) (6.1e7.6)
Adj. OR ¼ adjusted odds ratio. PSS < 9 ¼ reference group. * Adjusted for age and gender.
PSS scores were more likely to consult with psychologists or psychiatrists and less likely to receive prescriptions compared with older respondents (age > 65 years). Interpretation In line with the results of this study, a Swedish cross-sectional study found that high perceived stress was associated with psychotropic medicine (especially antidepressants). Instead of using Cohen’s PSS, that study used a 30-item stress scale (the Perceived Stress Questionnaire) (24), indicating that the association is not depending on a specific measurement tool. Despite finding an association between perceived stress and redeemed prescriptions or consultations with psychologists or psychiatrists, caution should be exercised in the use of the tested registries as a proxy for perceived stress due to low sensitivity and low positive predictive value. Sensitivity and specificity must be interpreted in relation to PSS, whereas kappa statistic, positive and negative agreement do not require a golden standard. The kappa coefficient was low, properly due to low prevalence, especially for consultations. Positive and
negative agreement is less affected of low prevalence, and still positive agreement values were less than 43% (21). Other studies have compared survey data and registry data, but the results have not been consistent [23,24], indicating that registry data can sometimes, but not always, be used as a proxy. The registry data on prescribed medication and consultations with psychologists or psychiatrists were able to correctly classify the persons with a low stress level (true negative) to a greater extent than the persons with a high stress level (true positive). If registries are to be used as a proxy for perceived stress, then they should correctly classify the true positives instead of the true negatives. The predictive value of a positive test describes the ability of the proxy measure to correctly identify the respondents with a high PSS score; in this study, it was 43%e55%, which is too low. Of the three types of medication examined in our study, antidepressants was most widely used, which is in line with antidepressants being used to treat different conditions (e.g., depression, insomnia, anxiety, and neuropathic pain) [25e27], whereas anxiolytics and antipsychotics are used for other indications and to a lesser extent. Our study found that 5% of the respondents in the
Fig. 2. The association between PSS and prescriptions or consultations with psychologists or psychiatrists in 2009e2011 and in subsets of data from each year (n ¼ 21,842).
0.266 86.6 38.7 53.7 92.2 30.3
k ¼ kappa; NA ¼ negative agreement; PA ¼ positive agreement; PPV ¼ positive predictive value; Sens. ¼ sensitivity; Spec. ¼ specificity.
0.279 86.4 40.6 52.7 91.2 33.0 42.7 49.4 88.5
40.9
85.5 37.6 Prescriptions or consultations
0.285
8.7 98.0 55.1 15.0 87.2 0.095 At least three prescriptions and/or consultation (summarized) 10.0 97.4 53.6 16.9 86.9 0.104 At least one prescription or one consultation Only consultations
9.5 97.7 54.8 16.2 87.0 0.101 At least two prescriptions and/or consultations (summarized)
27.0 93.4 55.0 At least three consultations 0.262 86.7 38.2 29.5 92.5 54.0 At least two consultations 0.276 85.9 34.6 89.9 50.5 At least one consultation Only prescriptions
PPV Spec. Sens.
k NA PA PPV Spec. Sens.
At least two prescriptions
k NA PA PPV Spec.
At least one prescription
Sens.
Table 3 Agreement between a high PSS (>16) and number of prescriptions and consultations with psychologists or psychiatrists (all values are stated as percentages, except kappa)
At least three prescriptions
PA
36.2
NA
86.9
k
0.249
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
5
lowest PSS score group had received antidepressants. One explanation could be that the medication was given for another indication, such as neuropathic pain. However, a recent study reported that only 2% of antidepressants are given for this indication [25]. Another explanation could be that individuals in the lowest PSS score group had a higher PSS score before medical treatment and, therefore, could be considered well treated with a low PSS score. Although, Figure 2 shows the same association independent of treatment period. On the other hand, only 38% of the respondents in the highest PSS score group received either prescriptions or consulted psychologists or psychiatrists. One explanation could be that Cohen’s PSS does not measure the type of stress treated with psychiatric medicine or consultations with psychologists or psychiatrist, indicating that a high PSS does not correlate to any specific psychic problem, but psychic problem of various sorts. Furthermore, this study only included consultations with psychologists or psychiatrists in the primary sector; individuals in the highest PSS score group may have been treated to a greater extent in the secondary sector. More individuals were referred to psychologists than psychiatrists, which is consistent with the fact that psychologists more often treat conditions related to stress than psychiatrists in Denmark. Young respondents (age < 40 years) were less likely to receive prescriptions and more likely to consult with psychologists or psychiatrists compared with older respondents (age > 65 years), which could indicate that general practitioners use different approaches to treat younger and older patients. In Denmark, until 2011, there was an age limit for referrals to psychologists, which meant that older patients (>37 years old if the diagnosis was depression and >28 years old if the diagnosis was anxiety) could not get a subsidy [28]. The age limits were removed for depression and increased to 38 years old for anxiety [29]. An American study in 2015 reported that 13% of older individuals (age > 65 years) compared with 34% of younger individuals (age 25e64 years) diagnosed with depression were treated with psychotherapy [30]. The pattern is convergent with the results of the present study, although the percentages are smaller. In relation to medical treatment, the American study reported that 26% of the elderly adults and 34% of the younger adults received prescriptions for antidepressants, which contradicts the results of this study. Individuals participating in the survey received the questionnaire in February 2010, whereas the data from the registries were recorded from 2009 to 2011, which may affect the interpretation of the results. However, analyzing for years separately showed the same pattern as found for the entire period, indicating rather robust estimates. Of those, invited to participate, 61.2% returned the questionnaire, which is similar to or higher than other studies in the field [31,32] and acceptable taking into account the tendency to declining response rates [33]. Of those, 93.4% answered all 10 items related to Cohen’s PPS. Several types of nonresponses are generally found [23,34]. Nonresponse due to illness could underestimate the perceived stress level if the person has an illness that increases stress but not necessarily affect the association between perceived stress and the use of medication or consultations with psychologists or psychiatrists. On the other hand, the stress level of the study population could be overestimated if nonresponders were healthy individuals who did not have time to complete questionnaires. Strengths The two national registries are described as having high-quality data [18e20]. The two registries cover all Danish citizens with a civil registration number, meaning that for all individuals who
6
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
returned the questionnaire, a complete linkage was made to registry data regardless of whether they completed the PSS items. In the National Health Service Registry, all persons, all providers, and select services are recorded through invoices [20], which makes registration necessary for general practitioners, psychologists, and psychiatrists to receive payment for their services, encouraging accurate registration. Cohen’s PSS is one of the most used stress-related scales in the world and has been translated into different languages and validated in several countries [35e38], including Denmark [39]. Limitations Stress is a confused concept and might be understood as an individual reaction to demands [40], as a relationship between environment and individual taxed by the person as exceeding resources [41] or as a failed reciprocity between efforts and rewards [42] to name a few. The choice of methodological background is important and might have implications for the interpretation of our results. Although Cohen’s PSS is borrowing from Lazarus and Folkman, it is less theoretically founded and thus useful in epidemiologic surveys. Furthermore, the PSS score items do not allow for suggesting preventive actions in relation to use of prescribed medication or use of psychologists or psychiatrists. The Danish National Prescription Registry has two limitations. First, the registry lacks information on the indication for treatment. Second, the data provide information on prescribed medications but not on whether it was taken. However, it is more relevant if a general practitioner assessed the need for treatment than whether the treatment was carried out. In contrast, the National Health Service Registry contains information about consultations with psychologists or psychiatrists, meaning that it is known that the individuals referred by general practitioners received treatment. Not all psychologists in Denmark are contracted within the tax-funded public health care system, and contracted psychologists may be treating patients without reimbursement; thus, individuals in Denmark can be consulting psychologists without it being recorded in the national registry. The consequences for this study could be nondifferential and differential information bias, but as we used combined measures, it should be of lesser importance. What this study adds This study contributes new knowledge on whether registries can act as a proxy for survey data on perceived stress and identifies some of the problems associated with this connection. We found that, although the associations between perceived stress and the use of prescribed medications and consultations with psychologists and psychiatrists are high, the association itself is not enough to predict whether a person has a high level of perceived stress. Further research connecting survey data and registry data could be relevant and would create an opportunity to use registry data as proxy measures for, for example, psychological problems and allow research to be conducted without having to rely on surveys, which are expensive and have problems with low response rates. Conclusions This study demonstrated an association between perceived stress and registry data on prescribed anxiolytics, antipsychotics, and antidepressants or consultations with psychologists or psychiatrists. However, due to low predictive values, the registry data may not be suitable as a proxy for perceived stress.
Acknowledgments The North Denmark Region Health Survey 2010 was funded by the North Denmark Region.
References [1] Redmond N, Richman J, Gamboa CM, Albert MA, Sims M, Durant RW, et al. Perceived stress is associated with incident coronary heart disease and all-cause mortality in low- but not high-income participants in the Reasons for Geographic And Racial Differences in Stroke study. J Am Heart Assoc 2013;2(6):e000447. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid¼ 3886761&tool¼pmcentrez&rendertype¼abstract [accessed 04.08.2015]. [2] Rod NH, Kristensen TS, Lange P, Prescott E, Diderichsen F. Perceived stress and risk of adult-onset asthma and other atopic disorders: a longitudinal cohort study. Allergy 2012;67(11):1408e14. Available at: http://www.ncbi.nlm.nih. gov/pubmed/22943607 [accessed 02.09.2015]. [3] White RS, Jiang J, Hall CB, Katz MJ, Zimmerman ME, Sliwinski M, et al. Higher Perceived Stress Scale scores are associated with higher pain intensity and pain interference levels in older adults. J Am Geriatr Soc 2014;62(12):2350e6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/25516031 [accessed 13.05.2015]. [4] Han X, Lin CC, Li C, de Moor JS, Rodriguez JL, Kent EE, et al. Association between serious psychological distress and health care use and expenditures by cancer history. Cancer 2015;121(4):614e22. Available at: http://www.ncbi. nlm.nih.gov/pubmed/25345778 [accessed 17.03.2015]. [5] Cohen S, Kamarck T, Mermelstein R. A global measure of perceived stress. J Health Soc Behav 1983;24(4):385e96. Available at: http://www.ncbi.nlm. nih.gov/pubmed/6668417 [accessed 08.08.2014]. [6] The National Health Profile. Available at: http://www.sundhedsprofil2010.dk/ [accessed 21.01.2015]. [7] Roepke-Buehler SK, Dong X. Perceived stress and elder abuse: a populationbased study of adult protective services cases in Chicago. J Am Geriatr Soc 2015;63(9):1820e8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 26280151 [accessed 22.10.2015]. [8] Treharne GJ, Lyons AC, Booth DA, Kitas GD. Psychological well-being across 1 year with rheumatoid arthritis: coping resources as buffers of perceived stress. Br J Health Psychol 2007;12(3):323e45. Available at: http://www.ncbi.nlm. nih.gov/pubmed/17640450 [accessed 22.10.2015]. [9] Cook JV, Dickinson HO, Eccles MP. Response rates in postal surveys of healthcare professionals between 1996 and 2005: an observational study. BMC Health Serv Res 2009;9:160. Available at: http://www.pubmedcentral. nih.gov/articlerender.fcgi?artid¼2758861&tool¼pmcentrez&rendertype¼ abstract [accessed 13.08.2014]. [10] Sorensen HT, Sabroe S, Olsen J. A framework for evaluation of secondary data sources for epidemiological research. Int J Epidemiol 1996;25(2):435e42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9119571 [accessed 18.11.2014]. [11] Thygesen LC, Ersbøll AK. When the entire population is the sample: strengths and limitations in register-based epidemiology. Eur J Epidemiol 2014;29(8):551e8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 24407880 [accessed 29.11.2014]. [12] Cohen S, K R, G L. Measuring stress: a guide for health and social scientists. Oxford University Press, Inc; 1998. p. 3e28. [13] Gherardi-Donato EC, Cardoso L, Teixeira CAB, Pereira Sde S, Reisdorfer E. Association between depression and work stress in nursing professionals with technical education level. Rev Lat Am Enfermagem 2015;23(4):733e40. Available at: http://www.ncbi.nlm.nih.gov/pubmed/26444177 [accessed 22.10.2015]. [14] Fan LB, Blumenthal JA, Watkins LL, Sherwood A. Work and home stress: associations with anxiety and depression symptoms. Occup Med (Lond) 2015;65(2):110e6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 25589707 [accessed 22.10.2015]. [15] Lataster T, Valmaggia L, Lardinois M, van Os J, Myin-Germeys I. Increased stress reactivity: a mechanism specifically associated with the positive symptoms of psychotic disorder. Psychol Med 2013;43(7):1389e400. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/23111055 [accessed 21.09.2015]. [16] Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011;39(7 Suppl):22e5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 21775345 [accessed 13.02.2015]. [17] Tavolacci MP, Ladner J, Grigioni S, Richard L, Villet H, Dechelotte P. Prevalence and association of perceived stress, substance use and behavioral addictions: a cross-sectional study among university students in France, 2009-2011. BMC Public Health 2013;13:724. Available at: http://www.pubmedcentral.nih.gov/ articlerender.fcgi?artid¼3750571&tool¼pmcentrez&rendertype¼abstract [accessed 30.07.2015]. [18] Kildemoes HW, Sørensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health 2011;39(7 Suppl):38e41. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/21775349 [accessed 05.02.2015]. [19] Andersen JS, Olivarius ND, Krasnik A. The Danish National Health Service Register. Scand J Public Health 2011;39(7 Suppl):34e7. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/21775348 [accessed 05.02.2015]. [20] Olivarius NF, Hollnagel H, Krasnik A, Pedersen PA, Thorsen H. The Danish National Health Service Register. A tool for primary health care research. Dan
L.S. Jensen et al. / Annals of Epidemiology xxx (2016) 1e7
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
Med Bull 1997;44(4):449e53. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/9377908 [accessed 12.08.2014]. Smith B, Chu LK, Smith TC, Amoroso PJ, Boyko EJ, Hooper TI, et al. Challenges of self-reported medical conditions and electronic medical records among members of a large military cohort. BMC Med Res Methodol 2008;8:37. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid¼2447848&tool¼ pmcentrez&rendertype¼abstract [accessed 14.01.2016]. Thygesen LC, Daasnes C, Thaulow I, Brønnum-Hansen H. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scand J Public Health 2011;39(7 Suppl):12e6. Available at: http:// www.ncbi.nlm.nih.gov/pubmed/21898916 [accessed 28.09.2015]. The Danish Data Protection Agency. Datatilsynet. Available at: https://www. datatilsynet.dk/english/the-act-on-processing-of-personal-data/read-the-act-onprocessing-of-personal-data/compiled-version-of-the-act-on-processing-of-pers onal-data. [accessed 28.09.2015]. Bergdahl J, Bergdahl M. Perceived stress in adults: prevalence and association of depression, anxiety and medication in a Swedish population. Stress Heal 2002;18(5):235e41. Available at: http://doi.wiley.com/10.1002/smi.946 [accessed 10.02.2016]. Gundgaard J, Ekholm O, Hansen EH, Rasmussen NK. The effect of nonresponse on estimates of health care utilisation: linking health surveys and registers. Eur J Public Health 2008;18(2):189e94. Available at: http://www. ncbi.nlm.nih.gov/pubmed/17984128 [accessed 25.06.2015]. Chini F, Pezzotti P, Orzella L, Borgia P, Guasticchi G. Can we use the pharmacy data to estimate the prevalence of chronic conditions? A comparison of multiple data sources. BMC Public Health 2011;11:688. Available at: http://www. pubmedcentral.nih.gov/articlerender.fcgi?artid¼3223740&tool¼pmcentrez& rendertype¼abstract [accessed 25.03.2015]. Bourgeois J, Elseviers MM, Van Bortel L, Petrovic M, Vander Stichele RH. The use of antidepressants in Belgian nursing homes. Drugs Aging 2012;29(9):759e69. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 23018612 [accessed 17.04.2015]. Schofield MJ, Khan A. Predictors of prescribed medication use for depression, anxiety, stress, and sleep problems in mid-aged Australian women. Soc Psychiatry Psychiatr Epidemiol 2014;49(11):1835e47. Available at: http://www. ncbi.nlm.nih.gov/pubmed/24907046 [accessed 06.02.2015]. Dharmshaktu P, Tayal V, Kalra BS. Efficacy of antidepressants as analgesics: a review. J Clin Pharmacol 2012;52(1):6e17. Available at: http://www.ncbi.nlm. nih.gov/pubmed/21415285 [accessed 03.01.2015]. Bekendtgørelse om tilskud til psykologbehandling i praksissektoren for særligt udsatte persongrupper - retsinformation.dk. Available at: https://www. retsinformation.dk/forms/R0710.aspx?id¼131868. [accessed 22.10.2015]. Tilskud til psykologbehandling - Bekendtgørelse om tilskud til psykologbehandling i praksissektoren for særligt udsatte persongrupper -
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40] [41]
[42]
7
retsinformation.dk. Available at: https://www.retsinformation.dk/forms/ R0710.aspx?id¼131868. [accessed 22.10.2015]. Sanglier T, Saragoussi D, Milea D, Tournier M. Depressed older adults may be less cared for than depressed younger ones. Psychiatry Res 2015;229(3):905e12. Available at: http://www.ncbi.nlm.nih.gov/pubmed/ 26233825 [accessed 28.09.2015]. Kramer M, Schmalenberg C, Brewer BB, Verran JA, Keller-Unger J. Accurate assessment of clinical nurses’ work environments: response rate needed. Res Nurs Health 2009;32(2):229e40. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/19152405 [accessed 13.08.2014]. Grava-Gubins I, Scott S. Effects of various methodologic strategies: survey response rates among Canadian physicians and physicians-in-training. Can Fam Physician 2008;54(10):1424e30. Available at: http://www.pubmedcentral.nih. gov/articlerender.fcgi?artid¼2567275&tool¼pmcentrez&rendertype¼abstract [accessed 13.08.2014]. Hazell ML, Morris JA, Linehan MF, Frank PI, Frank TL. Factors influencing the response to postal questionnaire surveys about respiratory symptoms. Prim Care Respir J 2009;18(3):165e70. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/19104738 [accessed 13.08.2014]. Christensen AI, Ekholm O, Gray L, Glümer C, Juel K. What is wrong with nonrespondents? Alcohol-, drug- and smoking-related mortality and morbidity in a 12-year follow-up study of respondents and non-respondents in the Danish Health and Morbidity Survey. Addict 2015;110(9):1505e12. Available at: http:// www.pubmedcentral.nih.gov/articlerender.fcgi?artid¼4538793&tool¼pmcentrez &rendertype¼abstract [accessed 05.10.2015]. Mimura C, Griffiths P. A Japanese version of the perceived stress scale: translation and preliminary test. Int J Nurs Stud 2004;41(4):379e85. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15050849 [accessed 11.06.2014]. Remor E. Psychometric properties of a European Spanish version of the Perceived Stress Scale (PSS). Span J Psychol 2006;9(1):86e93. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16673626 [accessed 06.08.2014]. Chaaya M, Osman H, Naassan G, Mahfoud Z. Validation of the Arabic version of the Cohen Perceived Stress Scale (PSS-10) among pregnant and postpartum women. BMC Psychiatry 2010;10:111. Available at: http://www.pubmedcentral.nih.gov/ articlerender.fcgi?artid¼3016315&tool¼pmcentrez&rendertype¼abstract [accessed 11.06.2014]. Seley H. Stress of life. New York: New York: McGraw-Hill; 1956. Lazarus RS, Folkman S. Stress, appraisal, and coping. Springer Publishing Company; 1984. p. 445. Available at: https://books.google.com/books?id¼i -ySQQuUpr8C&pgis¼1 [accessed 12.02.2016]. Siegrist J. Adverse health effects of high-effort/low-reward conditions. J Occup Health Psychol 1996;1(1):27e41. Available at: http://www.ncbi.nlm.nih.gov/ pubmed/9547031 [accessed 24.02.2016].