Ethical behaviour in clinical practice: a multidimensional Rasch analysis from a survey of primary health care professionals of Barcelona (Catalonia, Spain)

Abstract

Purpose

Normative ethics includes ethical behaviour health care professionals should uphold in daily practice. This study assessed the degree to which primary health care (PHC) professionals endorse a set of ethical standards from these norms.

Methods

Health care professionals from an urban area participated in a cross-sectional study. Data were collected using an anonymous, self-administered questionnaire. We examined the level of ethical endorsement of the items and the ethical performance of health care professionals using a Rasch multidimensional model. We analysed differences in ethical performance between groups according to sex, profession and knowledge of ethical norms.

Results

A total of 452 Professionals from 56 PHC centres participated. The level of ethical performance was lower in items related to patient autonomy and respecting patient choices. The item estimate across all dimensions showed that professionals found it most difficult to endorse avoiding interruptions when seeing patients. We found significant differences in two groups: nurses had greater ethical performance than family physicians (p < 0.05), and professionals who reported having effective knowledge of ethical norms had a higher level of ethical performance (p < 0.01).

Conclusions

Paternalistic behaviour persists in PHC. Lesser endorsement of items suggests that patient-centred care and patient autonomy are not fully considered by professionals. Ethical sensitivity could improve if patients are cared for by multidisciplinary teams.

Appendix 1: Rasch models formulae

The Rating Scale Model (RSM)

$$\pi_{nix} = \frac{{\exp \sum\limits_{j = 0}^{x} {\left[ {\theta_{n} - \delta_{i} - \tau_{j} } \right]} }}{{\sum\limits_{k = 0}^{m} {\exp \sum\limits_{j = 0}^{k} {\left[ {\theta_{n} - \delta_{i} - \tau_{j} } \right]} } }}$$

On ethical behaviour latent trait continuum, π _nix is the probability of professional n to score category x of item i; θ _n is professionals ethicality, and δ _i represents the extent of professional’s ethical endorsement for item i. Categories are ordered from 0 to m, and the τ _j are the rating scale structure parameters (e.g. thresholds) and represent the points on the continuum of behaviour in clinical practice, where adjacent categories are equally probable [23].

The Partial Credit Model (PCM)

$$\pi_{nix} = \frac{{\exp \sum\limits_{j = 0}^{x} {\left[ {\theta_{n} - \delta_{ij} } \right]} }}{{\sum\limits_{k = 0}^{m} {\exp \sum\limits_{j = 0}^{k} {\left[ {\theta_{n} - \delta_{ij} } \right]} } }}$$

π _nix is the probability of professional n to score category x of item i ¡. θ _n is ethicality of professional n, and δ _ij represents the extent of professional’s ethical endorsement for item i with a j particular thresholds from item categories. Thus, the PCM allows each item to vary its number of categories an estimate the probability of the threshold for each item instead that all entirely [28].

Multidimensional Random Coefficients Multinomial Logit Model (MRCML)

$$P(X_{n} ;\delta \left| {\theta_{n} } \right.) = \frac{{\exp \left[ {x_{n}^{'} \left( {B\theta_{n} - A\delta } \right)} \right]}}{{\sum\limits_{z \in \varOmega } {\exp \left[ {z_{n}^{'} \left( {B\theta_{n} - A\delta } \right)} \right]} }}$$

The MRCML assumes that a set of dimensions determines ethical endorsement. In the formula, the position of the professionals n in each dimension is described by the D × 1 column vector θ _n  = (θ _n1, θ _n2, θ _nD ), δ is the vector of ethical endorsement corresponding to each dimension, and Ω is the set of all possible response vectors. Z denotes a vector coming from the full set of response vectors, while x _n denotes the vector of interest. Matrices A and B are known as the design and scoring matrices, respectively. Scoring matrix B allows the description of the score that is assigned to each response category k on each of the D component ethical behaviour latent traits. Design matrix A is used to specify the linear combinations of the D component parameters δ to describe the ethical performance to each item [27].

The discrepancy index (DI) formulation

$${\text{DI}}_{n} = \sum\limits_{d = 1}^{{D_{i} }} {\left( {\overline{\theta } - \theta_{d} } \right)^{2} }$$

where D is the number of dimensions, n the number of professionals, $\overline{\theta }$ the endorsement of each item in a given dimension and θ the mean estimate of endorsement across all dimensions. The percentage of PHC professionals showing discrepant measures between dimensions would show how each dimension was providing differing information on ethical performance [27].