Data collection
We formed a questionnaire based on that of the EURONIC study10 (with the kind permission of the authors), with the addition of some further questions which we estimated that they would reflect our national values. The study questionnaire is provided as supplementary material. To confirm agreement for equivalence and validation purposes, the questionnaire was translated from English to Greek and then again to English, for the final comparison of the original and translated version.
Participants
An anonymous structured questionnaire was sent to all Neonatologists employed in the 28 NICUs across the country with a prepaid return envelope. A cover letter explaining the purpose of the study was also included; a reminder after 4 weeks was sent to those who had not responded promptly. Data was collected between September 2018 to January 2019.
Questionnaire
The questionnaire included 16 questions and consisted of four sections:
Section 1 (Q 1-13) included information on professional and demographic characteristics. Participant's professional group, job rank, qualifications/education, employment contract, working experience in the field, working hours per week, daily duties, participation to follow up and involvement in research programs, gender, age, parenthood, religious background and importance of religion were collected.
Section 2(Q 14-15) included self-reported practices of EoLDs in certain neonatal groups (those with severe neurological prognosis, at end stage, with poor prognosis and extreme prematurity). EoLDs included withdrawing treatment, avoiding emergency treatment, withholding treatment, continuation of ongoing treatment without adding further therapeutic interventions, withdrawal of mechanical support and administration of drugs even at the risk of respiratory depression and death as reported by Cuttini et al.10
Section 3 (Q 16) included personal views regarding Greek law reform.
In section 4(Q 16) data referred to attitudes on limitation or continuation of intensive care The 12-item questionnaire indexed scale, was used to assess attitudes by respondent’s agreement, on a 5-point Likert scale (from 1 “strongly agree” to 5 “strongly disagree”) developed by Rebagliato et al.1
The research protocol was approved by the Medical Research Ethics Committee, Aretaieion Hospital, National and Kapodistrian University of Athens (Ref. number 112/13-02-19).
Statistical analysis
All questionnaires were collected and coded in Microsoft Excel for subsequent analysis. Data analysis was performed via the SAS for Windows 9.4 software platform (SAS Institute Inc., NC, U.S.A.). Statistical analysis was also performed with SPSS 17.0v for Windows (SPSS Inc., Chicago, Ill, USA). P value < 0.05 was considered statistically significant.
Numerical data was expressed as mean± standard deviation, and for completeness as median, interquartile range, minimum and maximum values; for categorical data, the relevant percentages within individual groups were used. For numerical data with non-normal distribution, non-parametric tests were used, specifically the Kruskal-Wallis test; for the comparison of categorical variables for differences between groups frequencies, χ2 test was used.
Factor analysis was used to identify the underlying dimension of the 12 attitude items. In the 12-item attitude questionnaire, participants were asked to grade on a scale of 1 to 5 their attitude for the value of life (see appendix A). The reliability of the complete test was tested by Cronbach’s alpha (standardized)23 and was calculated at 68.7%. In order to reveal correlation between items of the questionnaire Spearman correlation coefficient was used. Subsequent factor analysis revealed that a single factor could explain 69.9% of the total variance. The items contributing to this factor according to the highest loadings were questions (attitude items) 1, 2, 3, 6, 7 & 8 and had an acceptable Cronbach’s reliability α=78.0%. A score was created based on the sums of these items after weighting by their loadings (i.e. each item was multiplied by the loading and subsequently all products were added); this score is subsequently called attitude score. A low attitude score was indicative of medical decisions towards sustaining life at any cost, despite potential severe morbidities (sanctity of life approach) while a high attitude score towards withdrawing intensive care in cases of poor prediction (quality of life approach). Moreover, the score was normalized in order to produce comparative results to other studies1 between 0 (indicative for sanctity for life approach) and 10 indicative for quality of life approach. Reported EoL practices were evaluated in relation to the attitude score and a binary logistic regression model predicting a positive response on the basis of attitude score was subsequently constructed.
We used factor analysis in the preset study, because it creates a theoretical model of latent factors that cause the observed variables (i.e. the questionnaire items). The main factor of this analysis was used as the attitude score and explains very high percentage of the responses, thus it produced a single number that is related to the participants’ behavior. A second reason for preferring factor analysis instead of principal component analysis (PCA) is the fact that this technique had already been used in another study, thus this report produces compatible and comparable results.
A univariate analysis was used to identify the variables associated with a physician's attitude score, with the score as the dependent variable. Independent variables included demographic and professional characteristics of the respondents. The variables retained in the final multivariate analysis (by a generalized linear model) were correlated with the attitude score at p<0.10, while variables not significantly related to the attitude score were removed from the model. In all analyses, data were based on valid responses for each group or subgroup, since not all respondents answered all questions.