Author's response to reviews Title: Insomnia and its correlates in a representative sample of the Greek population Authors: Thomas Paparrigopoulos (tpaparrig@med.uoa.gr) Chara Tzavara (htzavara@med.uoa.gr) Christos Theleritis (chtheler@med.uoa.gr) Constantin Psarros (cpsarros@med.uoa.gr) Constantin Soldatos (egslelabath@hol.gr) Yiannis Tountas (chsr.med.uoa@gmail.com) Version: 2 Date: 19 July 2010 Author's response to reviews: see over
From: Thomas Paparrigopoulos, MD Assistant Professor of Psychiatry Athens University Medical School, 1 st Department of Psychiatry, Eginition Hospital 74, Vas. Sofias Ave. 115 28 Athens, Greece Phone/Fax: +30 (210) 72 89 324 E-mail: tpaparrig@med.uoa.gr To: Natalie Pafitis, MSc The BioMed Central Editorial Team, BMC Public Health Athens, July 19, 2010 Dear Mrs Pafitis, I am pleased to forward to you the revised version of our article Insomnia and its correlates in a representative sample of the Greek population (BMC Public Health, MS: 1331748248384198) authored by T. Paparrigopoulos, Ch. Tzavara, Ch. Theleritis, C. Psarros, C. Soldatos and Y. Tountas. In this revision we took into consideration the useful comments and suggestions of the referees; accordingly, we tried to ameliorate the manuscript by clarifying certain points, as well as to provide adequate answers to the issues raised by the referees. There follow the modifications of the manuscript point by point: Comment of the editor: Provide context information within the background section of your abstract, in addition to the aims of your study. This is now provided in our revision. It reads: Insomnia is a major public health concern affecting about 10% of the general population in its chronic form. Furthermore, epidemiological surveys demonstrate that poor sleep and sleep dissatisfaction are even more frequent problems (10-48%) in the community. This is the first report on the prevalence of
insomnia in Greece, a southeastern European country which differs in several socio-cultural and climatic aspects from the rest of European Community members. Referee 1: Major Compulsory Revisions In the logistic regression authors should carefully review the data within table 3. The investigation on mental health problems are not put into the table. Moreover the adjusted data show in fact a reduction in the odds of insomnia in people with mental health problems. In addition, which is the reference in which mental health is compared with? The mental health variable presented in the logistic regression results (Table 3) is the mental health summary score derived from the Short Form-36 self-administered questionnaire (SF-36, Greek standard version 1.0). According to the SF-36 scoring, higher scores reflect less mental health problems; consequently, an increase on the score of the mental health dimension leads to a reduction of the odds for insomnia. Also, the mental health summary score is a continuous variable and therefore there is no reference category. Table 3 and table 2 should be completed with data of mental health, number of hospitalizations included. The number of hospitalizations was reported in Table 1 of the original submission, but was omitted by mistake from Table 3 (logistic regression analysis). This information has been added in the present revision. Minor Essential Revisions 1-The authors must re-write the results section, discussion, and mainly conclusions in accordance with the interpretation of the data.
In the results section of the revised manuscript it has been added that chronic disease, the number of hospitalizations in the previous year and mental health significantly predicted insomnia in the multiple logistic regression analysis with odds ratios 1.58 (95% CI: 1.06-2.34), 1.90 (95% CI: 1.18-3.05) and 0.96 (95% CI: 0.95-0.97) respectively. In view that in overall no significant changes have resulted from the addition of the above parameters, we believe that the discussion and the conclusions stand as they are and do not need to be re-written. 2-Data comparison will be more interesting if author include low and middle income countries. We believe that pertinent comparisons are with high income countries, especially Southern European countries, instead of low and middle income countries as suggested, since Greece according to the World Bank data was included in the high income countries [GNI per capita: $ 28,400 (2008 OECD statistics)] at least at the time of the survey (2006). Such data from Portugal, Spain, Italy, France and other developed countries, as well as relevant references (9, 15, 16, 27, etc.) had been provided in our original submission. 3- Limitations- the manuscript must include information if patients in hospitals and sheltered homes were included. In addition, were homeless people investigated? As mentioned in the sample description, this was a national household survey (Hellas Health I) where the households interviewed were randomly selected by means of systematic sampling. Therefore, patients in hospitals, sheltered homes, and homeless people were not investigated. This is now mentioned in the revised manuscript. 4- Due to the sampling procedure adopted, the authors could choose a statistical procedure that allowed correction of confidence intervals for complex sample designs. I am sure the authors are perfectly aware of such information. However, if the same procedure is adopted they could include another limitation in the discussion section of the paper.
To respond to this comment we want to clarify the sampling procedure adopted in order to justify the statistical analysis that we used: Multistage Systematic PPS sampling was implemented for the selection of the sample. In this sampling method, the properties of systematic sampling and sampling proportional to size are combined into a single sampling scheme. The whole process is computerized via a specific SPPS programme code being developed for sampling purposes. More specifically, in the first stage PSUs (buildings blocks) are selected as follows: The sampling frame to be used is a detailed electronic data file of the latest population census of year 2001. Sampling frame covers all geographical regions and building blocks of the country providing detailed information about the population size of each building block, blocks codes which specify the exact location, urbanity and regional information (region, prefecture, town, etc). First, the sampling frame is sorted by the resident population size (in descending order of building blocks into every region and urbanity codes). Second, the sampling interval is calculated (q=tz/n) where Tz is the Greece resident population total. Similarly to the ordinary one-random-start systematic sampling, we select a random number from the closed interval [1, q]. Let it be Qo. The n selection numbers for inclusion in the sample are hence Qo, Qo+q, Qo+2q,,Qo+(n-1)q. The population element identified for the sample from each selection is the first unit in the list (buildings blocks) for which the cumulative size Gk is greater than or equal to the selection number. Given this method, the inclusion probability of the kth element in the sample is proportional to its size. In the second stage, a constant number of households is selected within PSU as follows: The upper left point of each sampling block is selected as the starting address In case the sampled address corresponds to a single house, this is the starting household. In case the sampled address corresponds to a block of flats starting dwelling is the block of flats. In this case, all households in the dwelling will be recorded and a starting household is selected via simple random method. Via systematic sampling a random route will be designed from the sampled starting household.
In the third stage, one person falling in target population based on the closest birthday method is selected. Based on the above methodology, no stratification is used and in each stage up to the respondent s level a fair probability of selection of each population element in the sample has been kept. On this rationale, sampling design weights are assumed to be equal to one. Therefore, according to the methodology no changes in terms of weights and complex design should be made. Referee 2: 1. Why was it hypothesized that data from Greece might differ from data previous population based studies. As mentioned in our original manuscript, Greece differs in several socio-cultural and climatic aspects from the rest of European Community members and this may influence sleep habits and consequently the prevalence of insomnia symptoms. Thus, Greece has a warmer climate than the rest of European Community members and daylight hours are relatively extended compared to other latitudes, which gives more opportunities for social evening activities and leads to delayed bedtime hours. Furthermore, in the Greek culture daytime napping of about one hour duration remains a socially acceptable behavior, even in large cities; although a continuous working schedule has been operating in most cases during the last two decades, napping behavior still appears to be relatively prevalent in Greece. Finally, because strong emotional and financial bonds still exist within both the core and extended Greek family, housing conditions of the family may differ from those in the other European countries. These details have been included in the background section of the revised manuscript. 2. The cutoff for case selection of insomnia requires that the AIS score of 6 or greater. However, an individual can achieve this score with simply severe nocturnal symptoms without any daytime symptoms. However both the ICD 10 and the DSM IV require daytime impairment or distress for the diagnosis of insomnia.
Indeed, a score of 6 on the AIS has been found to be the best cut-off for diagnosing insomnia (based on the balance between sensitivity and specificity). Thus, when diagnosing individuals with a score of 6 or higher as insomniacs, the scale presents with 93% sensitivity and 85% specificity (90% overall correct case identification) (Soldatos CR, Dikeos DG, Paparrigopoulos TJ: The diagnostic validity of the Athens Insomnia Scale. J Psychosom Res 2003, 55:263-267). Although, as the referee points out, the possibility that an individual can achieve this score with simply severe nocturnal symptoms without any daytime symptoms theoretically exists, this is exceptionally rare and does not challenge the diagnostic validity of the scale. Regarding the ICD-10 and the DSM-IV requirement of daytime impairment or distress for the diagnosis of insomnia, this is true for the DSM-IV criteria, but not strictly for the ICD criteria, which mostly focus on subjective distress. [ICD-10 Criteria for Nonorganic Insomnia: A condition of unsatisfactory quantity and/or quality of sleep, which persists for a considerable period of time, including difficulty falling asleep, difficulty staying asleep, or early final wakening. Insomnia is a common symptom of many mental and physical disorders, and should be classified here in addition to the basic disorder only if it dominates the clinical picture. DSM-IV-TR Criteria for Primary Insomnia: 1. The predominant complaint is difficulty initiating or maintaining sleep, or nonrestorative sleep, for at least 1 month. 2. The sleep disturbance (or associated daytime fatigue) causes clinically significant distress or impairment in social, occupational, or other important areas of functioning ]. 3. The response rate of 44.5% seems very low The response rate may seem low but it is fairly good for Greek standards. This has been added to our revision. 4. How long was the interview? The duration of the interview was 35-40 minutes. This information is now provided in the revision.
5. What is meant by the demographic residence? In the revised manuscript it is clarified that residence is a dichotomous variable, i.e. urban (2000 or more inhabitants) vs. rural (fewer than 2000 inhabitants). 6. Data on work shift would be helpful; was it collected? Unfortunately, information on shift work was not collected. This is an obvious limitation of the study, although shift work is not as frequent in Greece as in other developed countries, and consequently lack of data is not expected to have influenced our results in a considerable way. This had already been highlighted in our original submission (in the conclusion) where it was mentioned that circadian sleep-wake parameters were not assessed. This is pointed out in our revision by mentioning that circadian sleep-wake parameters, such as shift work, were not assessed. 7. What was the reliability of the data from the interview? The authors indicate that a pilot study was done but no results are presented. We would like to clarify that the self-administered questionnaire was pre-tested only in terms of the comprehension and the order of the questions included. The reliability of the data from the interview need not to be assessed because it comprised of questionnaires (e.g. Athens Insomnia Scale, SF-36, Greek standard version 1.0) previously standardized and checked for their reliability in the Greek population. This is now clearly stated in our revised manuscript, which reads: the questionnaire had been pre-tested in terms of the comprehension and the order of the questions included. The term pilot study has been removed from the text to avoid misunderstandings. 8. One cigarette per day seems like a very strict cutoff for smoking. Were any other cut points used?
Several cut-offs for smoking were checked in the analysis but no interesting results relevant to insomnia were found. The cut-off of one cigarette per day was chosen because it is frequently applied in epidemiological studies. 9. I prefer the term insomnia symptoms rather than sleep problems for example see AIS criteria section. Not all the questions in the AIS, nor the diagnostic criteria are all sleep problems. We agree with the referee s suggestion and therefore the term sleep problems has been replaced with the term insomnia symptoms throughout the text. Following the above modifications, we hope that this revision will be acceptable for publication in BMC Public Health. Sincerely, Thomas Paparrigopoulos, MD