This is a cross-sectional survey of patients from a GP setting in Germany with at least one onset of neck pain between March 2005 and April 2006. Follow-up surveys of this cross-sectional cohort are under way. The study was approved by the local research ethics committee.
Recruitment of patients
As part of a project on the quality of medical care in general practice (MedViP), a network of 104 general practices has been established.  Fifteen of these within a radius of 30 km around Göttingen were selected for participation and provided anonymised electronic patient data (date of birth, sex, diagnosis). Patients were included in a list of potentially eligible persons if at least one consultation because of neck pain was documented in the electronic patient record during the period from March 2005 to April 2006. All GPs were asked to exclude patients from a list of potentially eligible persons, if they had their neck pain consultation because of a new trauma, were terminally ill, suffered from cancer, were in need of nursing care or had severe cognitive impairment. Additionally, patients seen by locums only, patients who had moved to a region outside of the study area or who were not able to speak German were excluded from the study.
Participants received a comprehensive self-administered questionnaire covering multiple domains such as socio-demographic information, anxiety, depression, social support and neck pain. Participants received the questionnaire from their primary care physicians together with written instructions on average 3 months after the consultation because of neck pain. Due to budgetary constraints no mail or telephone follow-up was done when persons did not or did incompletely return the questionnaire.
Neck and Pain Disability Scale (NPAD) [10, 11]
The NPAD is a 20 item measure specifically developed for patients with neck pain to assess neck pain and related disability (see Additional File 1). It measures the intensity of pain, its interference with vocational, recreational, social and functional aspects of living and the extent of associated emotional factors. Patients respond to each item by marking along a 10-cm visual analogue scale. Item scores range from 0 to 5, and the total score (possible range 0–100) is the sum of the item scores. A valid NPAD score can be generated if no more than 15% of the items are missing. The NPAD has been shown to have validity in comparison to other self-reported pain measures  as well as supporting constructs of mood and neuroticism.  Recently, a German version of the NPAD (NPAD-d) was developed and validated for the use in primary care settings. 
Age, gender, employment status, education, living with a partner and number of persons living in the same household were assessed by single item questions. Persons who were less than 10 years at school were considered to have only basic education. Depressive mood and anxiety were measured by the Hospital Anxiety and Depression Scale (HADS), [13–15] a widely used short self-assessment questionnaire mainly asking for psychological manifestations of (generalised) anxiety and depressive mood. It consists of two subscales with seven items each. Possible subscale scores range from 0 to 21. According to the German test manual,  patients with a depression score > 8 were considered depressive, patients with an anxiety score > 10 were considered anxious. Perceived social support was measured by the 14-item short form of the Social Support Questionnaire ("Fragebogen zur Sozialen Unterstützung"; F-SozU).  The items refer to different aspects of perceived social support (emotional support, instrumental support and social integration), resulting in a global scale with higher scores indicating better social support (five-point scale: from "relevant" to "not relevant"). Deficits in social support were defined as having 4 or less points on the F-SozU scale. Single item questions were used to ask for injuries of the cervical spine previous to completing the questionnaire and for exercise frequency per week. Additionally, three single item questions asked whether or not neck pain was present on the day of questionnaire completion, on more than 100 days in the last year and whether or not neck pain was constantly present during the last year.
First, summary statistics including simple counts and percents were computed to describe the baseline characteristics of the sample. Then NPAD-d total scores were calculated as described previously. Up to three missing item values were imputed by value substitution based on each subject's valid responses to NPAD-d items. Specifically, imputed values for missing NPAD items were calculated by dividing the sum of the non-missing NPAD-d items by the number of the non-missing items. We then analysed mean NPAD-d scores by baseline characteristics.
In a next step, we performed crude (bivariate) linear regression models to assess the association between baseline variables and neck pain (as measured by the continuous NPAD-d score). Baseline variables included were dichotomous socio-demographic characteristics (age 50 years or older, female, unemployed or retired, basic education, living without partner, living with 2 or less persons in the same household), psychometric characteristics (HADS depression and anxiety subscales, deficits in social support), one medicinal history characteristic (previous cervical spine injury), and one health-promoting lifestyle characteristic (exercise once or less per week). The depression and the anxiety subscale of the HADS were included as continuous variables in the regression analyses to increase power. The F-SozU scale for measurement of deficits in social support was dichotomised because of her skewed distribution. Then we calculated adjusted (multivariate) linear regression models including NPAD-d scores as dependent variable and the baseline characteristics described previously as independent variables. Regression coefficients > 0 denote higher NPAD-d scores (higher levels of neck pain) with increasing levels of the baseline characteristic, regression coefficients < 0 denote lower NPAD-d scores (lower levels of neck pain) with increasing levels of the baseline characteristic.
For sensitivity analyses, we recalculated linear regression models including only those participants with complete answers to all NPAD-d items (350 persons).
The adjusted linear regression analysis revealed that continuous independent variables (HADS depression and anxiety subscales) were significantly correlated with neck pain. As regression coefficients for continuous independent variables that range from 0 to 21 are difficult to interpret clinically, we used analysis of variance to investigate how those scales varied across patients with different levels of neck pain. Therefore, study participants were allocated to the following three groups: Those with NPAD-d values between the percentiles 0 and the 25 were assigned to the lowest quartile group, those with values between percentiles 25 and 75 were assigned to the middle quartiles group and those with values between the percentiles 75 and the 100 were assigned to the highest quartile group representing those with the highest levels of neck pain in this sample. Mean values and standard deviations of the two scales were derived to illustrate the crude extent of variation attributable to the level of neck pain.
All p values reported were two-sided and all analyses were performed using Stata 9.2 (Stata Corporation, College Station, Texas/USA).