Development and Validation of a questionnaire to describe behaviours and attitudes of physicians towards their own process of falling ill.

Background: Physicians have not learned their role as patients. Medical associations focused mainly on mental health programs for doctors in different countries. Nevertheless, the anomalous behaviours of ill doctors exist independently of their health problem. We described the development and validation of a questionnaire to describe Behaviours and Attitudes of Doctors towards their Own Illness (BADOI). Methods: Development of the questionnaire and appearance and content validation (ACV) by experts’ debriengs; construct validation by exploratory (EFA) and conrmatory factor analysis (CFA); reliability by Cronbach's alpha index (CAI) and feasibility by completion time (CT), accessibility to questionnaire platform and % of not answer items. Results: A Questionnaire with 56 items was developed with 3 sections: 1) socio-demographic, professional and work-related data. 2) Activities for self-care and work impact (sick leaves and presenteeism) and physicians’ healthcare pathways (Self-management and use of healthcare services). 3) Doctors of ill doctors and the role of medical associations. The validation sample size was of 4308 physicians’ answers. The factorial model presented 5 factors explaining 78.08% variance, a high value. Factor 1 explained 26.75% of the model describing the relationship between illness and work; Factor 2 explained 16.68%, with items regarding the negative impact of mental issues, toxic habits and the use of healthcare services; Factor 3 explained 15.06%, with items considering presenteeism and sick leaves; Factor 4 explained 9.95% regarding the handling of an ill colleague and medical associations. Finally, Factor 5 explained 9.64%, with items describing healthcare given to ill physicians and the revalidation of medical profession. Discussion: The factors grouped together items relating to self-care, impact on work and physicians’ healthcare pathways (F1-F2-F3) and the role of doctors of ill doctors and of the professional associations (F4-F5), being consistent with the 2 main perspectives: personal experience and caring for ill colleagues. Conclusion: A reliable Multidimensional Questionnaire on the process of physicians becoming ill has been validated in a large sample of registered doctors. It will help to identify and quantify physicians’ healthcare pathways allowing healthcare managers to set evidence-based interventions and formulate policies and

(F1-F2-F3) and the role of doctors of ill doctors and of the professional associations (F4-F5), being consistent with the 2 main perspectives: personal experience and caring for ill colleagues.
Conclusion: A reliable Multidimensional Questionnaire on the process of physicians becoming ill has been validated in a large sample of registered doctors. It will help to identify and quantify physicians' healthcare pathways allowing healthcare managers to set evidence-based interventions and formulate policies and ethical recommendations regarding the management of sick doctors.

Background
The culture of self-sacri ce has been promoted in the medical profession and health organizations prioritising the needs of patients above all else, even above doctors' own health needs (1). Moreover, when doctors fall ill, they become atypical patients (2) and have di culties in taking on the role of patient (3). Consequently, behaviours of self-diagnosis (4), selftreatment (5) and the use of alternative healthcare pathways (6) are frequent. Doctors who treat ill colleagues very often do not act as they would with a standard patient and may feel pressured, tense and uncertain (7). Therefore, the care received by ill doctors may be of poorer quality than standard. In the meanwhile, working while ill, physicians might pose risks to their patients (8), (9).
It is di cult to nd epidemiological information regarding ill doctors in Spain and other European countries, except for those being treated under Mental Health Programs as the Spanish Integrated Care Programme for Ill Doctors (PAIME) (10), which is devoted to mental disorders and addiction. Nonetheless, the impact of doctors' anomalous behaviour affects the way they cope with any illness.
The subject is never approached during the years of undergraduate study, and once they are practising, doctors will have internalised the notion that patients are others (11). A con ict then emerges between the roles of professional and patient.
This con ict negatively impacts the process of providing care and professional competence (12) and leads to high rates of presenteeism (13).
The World Medical Association expressed its concern in this respect and issued a declaration to this end (14). The declaration incorporated a key reference to doctors' own health into its Physician's Pledge: 'I will attend to my own health, well-being, and abilities in order to provide care of the highest standard' (15).
We started a project: 'Getting Sick is Human: When the Doctor is the Patient' to provide an understanding of the process of doctors falling ill through a qualitative approach based on semi-structured interviews with ill doctors and focus groups, together with a descriptive approach involving a self-reported survey for the purpose of describing the perceptions of doctors facing their own illness and to identify improvements for the provision of a better healthcare quality in the health system throughout every stage of the medical profession, from medical students to retired doctors.
The present study aims to present the development and validation of the Behaviours and Attitudes of Doctors facing their Own Illness (BADOI) questionnaire.

Methods
Questionnaire development (Fig. 1): A literature search was conducted on the process of doctors falling ill. No MeSH term were identi ed that could suitably re ect the concept of 'ill/sick doctor/physician'. The information search strategy, including documents published by medical associations, was structured according to the dimensions identi ed for qualitative research. At the same time, semi-structured interviews were conducted with ill doctors at three professional levels: junior doctors, practising doctors and retired doctors. Interviews were conducted with a total of 26 ill doctors. Focus groups were developed comprising doctors specialising in occupational health, clinical and educational supervisors, and representatives of Spanish provincial medical associations.
An initial questionnaire was designed with 26 items sourced from the literature and the map of conceptual codes which had been de ned for the qualitative analysis of interviews and focus groups. The questionnaire was piloted between 2014 and 2016 with debrie ng by experts. Modi cations were incorporated from the debrie ng, together with questions derived from the focus groups and interviews. At the suggestion of experts, questions from the 2017 Spanish National Health Survey (ENSE17) (16) were incorporated in order to allow comparison of certain lifestyle habits with the population with the same level of education and age. Questions on certain ethics-related aspects taken from the Ethics in the Coordination Between Levels of Care (ERNA) survey (17) were also incorporated.
The nal questionnaire items were grouped into the following three sections: 1) content related to socio-demographic aspects, speciality, employment status, public/private/combined practice and professional satisfaction; 2) content related to health-related aspects (chronic illness, u vaccination, smoking, alcohol use, relationship with general practitioner) and impact on work (sick leave, presenteeism, time in sick leave) and related to falling ill (self-diagnosis, selftreatment, alternative pathways to diagnosis and treatment, patient care and con dentiality). 3) Content related to being consulted by ill colleagues, being doctor of a doctor, medical associations, ethical considerations regarding periodic revalidation and aspects that could be improved with regard to provision of healthcare to ill doctors (Appendix 1: Final questionnaire).
Target population: The target population was the 270,000 Spanish registered physicians. Random sampling was conducted with a (1-α):99% con dence interval, 3% precision and sample proportion of p = 0.5 that would maximise sample size, with an expected attrition rate of 50%. The total sample consisted of 3,662 physicians. The questionnaire was designed with a self-report format to be lled online in the Organisation of Spanish Medical Associations (OMC) website http://www.cgcom.es/. All physicians registered in any of the 50 Provincial Medical Associations in Spain had accessed to the online questionnaire.
The questionnaire was available from June 2017 until June 2018. The study was approved by the Research Ethical Committee of Aragon in 2016.

Questionnaire validation:
The validation process was divided into two stages(18): 1) assessment of face validity, content validity and feasibility and 2) assessment of construct validity and reliability. 1) Process for assessing face validity, content validity and feasibility: The validation process consisted of assessing face validity and content validity through three debrie ngs with experts, in medical ethics and, experts from PAIME and Fundació Galatea (a foundation created to protect the health and well-being of all health professionals) (19) and the research team.
In terms of questionnaire feasibility, the evaluation took into consideration: 1) Percentage of unanswered questions, 2) the access to the questionnaire platform, available only to Spanish registered physicians.
2) Process for assessing construct validity and reliability (20): Construct validity was assessed by exploratory factor analysis (EFA). Prior to this, the suitability of the factor analysis was evaluated by the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy. Adequacy was also evaluated through a correlation matrix by Bartlett's Test of Sphericity. In our case, the p-value associated with Bartlett's test was < 0.001, con rming the use of this technique. Then a con rmatory factor analysis (CFA) was performed to corroborate the dimensions of the questionnaire and to analyse the distribution of the items within each dimension. The following goodness-of-t statistics were obtained: 1) Normed t index (NFI), which took values between 0 and 1. NFI values closer to 1 indicated a better t. 2) The root mean square error of approximation (RMSEA) index which was a subjective magnitude. A value less than 0.10 was usually considered an indicator of good model t. The comparative t index (CFI) measures model tted as well. In order to interpret the factors resulting from the CFA, the items whose correlations with the factors were highest in absolute value were identi ed rst, given that they could be directly or indirectly related.
The reliability was measured through the internal consistency of the questionnaire by correlation coe cients and Cronbach's alpha.
The development and validation stages are summarised in Fig. 1.

Results
The nal questionnaire consists of three sections with 55 closed questions and one open-ended question. (Appendix 1 contains the nal questionnaire in Spanish and English).
Content and face validation and feasibility were conducted through the debrie ng sessions held with the medical associations which led to the incorporation of questions related to the role of these associations in the process of doctors falling ill and on the periodic revalidation. The debrie ng sessions held with experts from PAIME led to questions on the availability of services to manage ill doctors. Furthermore, in order to facilitate comparison with general population of same age and level of education, the experts suggested the incorporation of 11 questions from the ESNE17.
The median completion time recorded by the online platform was 10-12 minutes, which was quite an optimal period.
With regard to the form, method and time required to classify the scale scores, the survey was very easy to ll as the items were categorically arranged and many of them were also rated by a Likert-like scale. Finally, in terms of the consideration of the person who lled the survey, the questions were highly focused in the eld of healthcare and physicians' experiences, and the survey was only available through the websites of medical associations, so there was a need to be a registered doctor to access and answer.
The questionnaire was completed by a total of 4,308 physicians, of whom 1,858 were men (43.13%) and 2,450 women (56.87%). Considering professional stages, more female junior doctors (73.06%) than male (26.94%) took part. With regard to medical specialists, female representation was also higher (59.37%) than men (40.63%). Finally, the retired doctors group comprised a majority of men (73.08%). The mean age of the male participants (54.78 ± 11.87 years) was higher than that of the female participants (45.95 ± 11.76) in all professional stages, with statistically signi cance among medical specialists and retired doctors (Table-1). The fact that our initial estimated sample was 3,662 medical practitioners and the current sample is slightly bigger reinforced the feasibility of the questionnaire. Only three questions were not answer by all participants (Table-2). Five dimensions were selected as a solution to the CFA, ful lling the Kaiser rule, selecting all factors whose eigenvalues were greater than one. The factor model with the ve factors explained 78.08% of model variance, a high value. Table-3 shows the variance that explains each of the factors, and their aggregate value. questions concerning the relationship between illness and work (p18, p20 and p21); Factor 2 explained 16.68% of the model. It was formed by mental health issues, impact of work on health and the weighting of toxic habits (consumption of toxic substances). Factor 3, explaining 15.06%, was made of sick leave and presenteeism; Factor 4, which explained 9.95%, included items related to the handling of an ill colleague and medical associations. Finally, Factor 5, which explained 9.64% of the model, contained questions related to the healthcare given to ill physicians and the revalidation of medical profession. To conclude, the internal consistency and homogeneity of the instrument was analysed by Crombach's Alpha Index for each dimension from the CFA. The nal column of Table-3 shows the values obtained, which are high and close to one for each factor. Each of the factors is consistent in the sense that the items that it comprises are stable in the dimension.

Discussion
A questionnaire with 56 items directed at compiling behaviours and attitudes in the medical profession concerning the process of doctors falling ill was developed and properly assessed the psychometric soundness of the questionnaire throughout appearance, content and construct validity and feasibility with an important sample size of 4,308 answers.
Sex distribution of participants clearly showed the feminisation of the current medical population (23)(24)(25). It may have been necessary to incorporate a gender perspective, as it has been explicitly done in other studies on doctors' health (26).
We focused on the impact of illness on the medical profession regardless gender issues. So certain variables were not directly taken into consideration, such as responsibility for children (27) and the care of elderly persons (28), compatibility with household chores and the profession (29), which most likely have an impact on health and illness in women and which would in uence decisions on the careers of future women doctors (30,31), particularly in surgical elds (32).
Therefore, we considered the gender perspective to be a very important area of research to develop in the immediate future by the research group.
With regard to the construct validity process, factor analysis allowed us to group the questions in factors that connected items with resemblance of the main sections: self-care, illness and health work impact and healthcare pathways which are mainly represented by factors 1, 2 and 3. Then Factors 4 and 5 are more related to the role of doctors who care for ill doctors and the role of medical associations and revalidation of the medical profession. All the factors included questions on the process of physicians' health-illness, but each one provided a nuance to the analysis. Both the factors themselves and the entire set of factors explain a model with statistical signi cance. With regard to internal validity, Cronbach's alpha provided an excellent result. All of this means that the questionnaire has been properly validated. The research team is assessing the incorporation of the gender perspective in the questionnaire analysis and approaching the limitations found with the questionnaire with new research strategies.
One of the main limitations for contrasting the information of BADOI questionnaire is the lack of previous research studies on the subject in Spain, excepting one regarding health habits by Fundació Galatea (33), and internationally, particularly among Norwegian doctors (8),(2)(13) and Australia (6,34,35). There are many more recent studies on mental disorders, addiction, burn-out and suicide in the medical professions (36)(37)(38)(39), but less on the impact of any illnesses and basic preventive activities among physicians which may in uence patients as well (40,41). We included some health practices items from Spanish National Health Survey 2017 in order to compare with the Spanish population of the same level of education and group of age.
The main design limitation was the di culty in conveying the qualitative information from the conceptualisation of main results from focal groups and semi-structured interviews with ill doctors into closed questions without losing relevant information (content validity). Content advice was provided by a very diverse group of experts including professionals from the Fundació Galatea and PAIME, Spanish provincial medical associations and the research group.
The fact that there was not a speci c keyword that can better orientate the literature search was also a limitation. For this reason, it should be considered to create a MeSH term such as 'sick doctor' or 'ill doctor' to facilitate identi cation of references on this important topic on healthcare organizations.
Moreover, the limitations of the instrument cannot be overlooked: a self-report survey accessible on a websites. Selfreported information is limited by the fact that it can rarely be independently veri ed. However, this factor has a lower weighting in individuals with high levels of education (21). In addition, the fact that it makes use of the Internet may have facilitated participation from younger professionals (22) and the need to be registered physician ensured the suitability of respondents. Availability of data and materials:

Conclusion
All data and materials are available upon request to researchers. Nevertheless, questionnaire database is included as a supplementary material with the manuscript submission.
Authors' contributions: All the authors are members of the Bioethics Research Group of Aragon that is leading the project Getting Sick is Human: When the Doctor is the Patient.
This project involves the writing of four doctoral theses related to the stages of the medical practitioner career: medical student, junior doctor, practising doctor and retired doctor.
Dr Bárbara Marco-Gómez is conducting the research focusing on practising doctors.
Dr Rogelio Altisent is the lead researcher on the project and coordinator of all projects undertaken by the group.
Dr Candela Pérez-Álvarez is conducting the research focusing on junior doctors.
Dr Teresa Martínez-Boyero is conducting the research focusing on retired doctors.
Dr Alba Gallego-Royo is conducting the research focusing on medical students.
Dr María Teresa Delgado-Marroquín and Dr María-Pilar Astier-Peña are joint advisors of the doctoral theses derived from this study in conjuntion with Dr Altisent.
All the authors actively participated in every stage of the research: data collection, statistical analysis, discussion of results and writing of the present manuscript.

Competing interests:
On behalf of all the authors, Dr María-Pilar Astier-Peña states that there are no con icts of interest in relation to the research presented involving any of the authors.
Dr María Pilar Astier-Peña, corresponding author on behalf of the other signees guarantees the accuracy, transparency and honesty of all the data and information contained in the study; that no relevant information has been omitted; and that any discrepancies between authors have been properly resolved and described. Funding: This study has been nanced by competitive grant from Spanish Government, Carlos III Institute. Project Number: PI18 /00968. IIS-Aragón: "Building Europe"