INTRODUCTION

Patient-physician relationships have been widely studied in the context of chronic health problems. For people living with HIV (PLWH), being seen by different doctors each time or exposure to negative attitudes, like stigma, discrimination or prejudices, shown by medical staff can lead to poor outcomes such as inadequate adherence to antiretroviral treatment (ART) or loss of follow-up (^{Kuznetsova et al., 2016}; ^{Westergaard, Hess, Astemborski, Mehta, & Kirk, 2013}). On the other hand, positive experiences with health professionals, health professionals who respond to the emotional needs of patients, satisfaction with one’s doctor and belief in the physician’s competence, have been reported as facilitators for an adequate medical follow-up (^{Brincks, Feaster, Burns, & Mitrani, 2010}; ^{Hurley et al., 2018}; ^{Kuznetsova et al., 2016}; ^{Yehia et al., 2015}).

As a general construct, patient-physician relationships can include different components, such as patient’s satisfaction with care, physician’s communication style or agreement between the patient and physician (^{Brincks et al., 2010}; ^{Honavar, 2018}). A determining factor in the patient-physician relationship is patients’ trust in their doctors (^{Hall, Dugan, Zheng, & Mishra, 2001}; ^{Hall et al., 2002}). In this sense, in individuals with chronic illness, trust in their doctors has been associated with positive health outcomes, treatment compliance and adherence to doctors’ recommendations (^{Kerse et al., 2004}; ^{Piette, Heisler, Krein, & Kerr, 2005}). Similar associations between trust in physicians and adherence to medications and treatment have been found in people living with HIV (^{Graham, Shahani, Grimes, Hartman, & Giordano, 2015}; ^{Roberts, 2002}; ^{Schneider, Kaplan, Greenfield, Li, & Wilson, 2004}; ^{Whetten et al., 2006}), in which adherence to treatment have an impact not only on individual health (^{Gebrezgabher et al., 2017}) but also on public health, preventing transmission of the virus (^{Skarbinski et al., 2015}) or development of drug resistance (^{Almeida et al., 2014}).

The Trust in Physician Scale (TPS; ^{Anderson & Dedrick, 1990}) is a psychometric test that measures the patient’s trust in their doctor and consists of 11 items measured in a five-point Likert format with response options ranging from “strongly agree” to “strongly disagree.” The instrument has shown adequate internal consistency (Cronbach’s alphas between .85 - .90) and construct and predictive validity (^{Thom, Ribisl, Stewart, & Luke, 1999}). When used in PLWH, it has shown adequate psychometric characteristics (^{Piette et al., 2005}). Trust in doctor measured with this instrument was significantly related to patients’ desires for control in their patient-physician interactions and to subsequent satisfaction with care (^{Anderson & Dedrick, 1990}).

The instrument has been used in different studies (^{Aloba, Mapayi, Akinsulore, Ukpong, & Fatoye, 2014}; ^{Freburger, Callahan, Currey, & Anderson, 2003}; ^{Kalsingh, Veliah, & Gopichandran, 2017}) but to date, it has not been validated in Spanish for use in a Mexican population. Considering the important relationship between this construct, adherence to ART and quality of life in PLWH, and the fact that in Mexico, inadequate adherence rates between 40% and 50% have been reported (^{Peñarrieta et al., 2009}; ^{Plascencia de la Torre et al., 2019}), it is considered relevant to improve knowledge about trust in physician for which it is essential to have adequate and reliable instruments to evaluate it. We thus adapted and validated the TPS to Spanish and evaluated its psychometric properties in a sample of people living with HIV in Mexico.

METHOD

RESULTS

In the sample used to conduct EFA, responses were received from 251 participants who answered the TPS. A total of 36 questionnaires were eliminated: 26 due to lack of variability in the answers, six because they had a response pattern, and four because they belonged to participants who were not living with HIV. Final analyses were carried out with responses from 215 participants. The vast majority (n = 207, 96.3%) were male participants, in line with the sex distribution of PLWH receiving care at the clinic where the study was carried out. Mean age was 32.45 years (SD = 7.73). More than half were single (n = 156, 72.6%) and had higher education (n = 129, 60.1%). The majority was taking ART (n = 210, 97.7%) and the median of months after diagnosis was 41 (IQR = 20-89). Table 1 describes the sociodemographic variables of the participants.    

Construct validity

Adequacy of the first sample (n = 215) was tested using EFA trough the KMO value of .902, and Bartlett’s test of sphericity of χ²(15) = 989.47, p < .001, thus confirming adequacy and variability of the sampling. EFA results showed one factor with eigenvalue > 1, suggesting a unidimensional structure. Factor loading of all six items was over .40, so no more items were excluded. A one-factor solution explained 68.9% of the total variance, which is considered very acceptable. In Table 3, factor loading scores are shown, all of them ranging from .756 to .894.    

Table 3  Factor loadings of items of TPS (n = 215)  

Item	Mean	SD	Factor loading	Communality
TPS2	3.42	1.276	.894	.650
TPS3	3.85	1.192	.864	.732
TPS4	3.66	1.208	.855	.640
TPS6	3.71	1.253	.806	.799
TPS9	3.66	1.305	.800	.747
TPS10	3.40	1.363	.756	.572

SD: Standard deviation

In order to confirm the factor structure of TPS, a model with six items and a single factor was tested by CFA in a second sample of 140 participants. An χ²(7) = 10.19 p = .178 was obtained, indicating good fit. The rest of the indices were then calculated to check the model fit; the RMSEA value was .057 p = .376, the SRMR was .0015 and the CFI was .997, all of them showed an adequate model fit. All factorial weights were high, ranging from .84 to .96. In figure 1 the model obtained with CFA is showed.    

Figure 1  CFA of Trust in Physician Scale.  

DISCUSSION AND CONCLUSION

The TPS translated and adapted in a sample of people living with HIV from Mexico was shown to be a valid and reliable instrument, consistent with the original scale (^{Anderson & Dedrick, 1990}). The translated and adapted version was still an unifactorial scale, although it was composed by six of the 11 items of the original scale. The score that can be obtained with this adaptation ranges from 6 to 30, with higher scores indicating greater confidence in the doctor. Scores of adapted version correlated with the ones obtained with the original instrument, suggesting that they measure the same psychological construct. The final internal consistency index was also adequate (Cronbach’s alpha = .93), even higher than those found in other studies (^{Aloba et al., 2014}; ^{Freburger et al., 2003}; ^{Kalsingh et al., 2017}) and similar to that of the original instrument (^{Anderson & Dedrick, 1990}). The adapted version of TPS remained unifactorial, like the original instrument (^{Anderson & Dedrick, 1990}), and explained a very high percentage of variance (almost 70%). All items were correlated with each other and had high factorial loads. These results were confirmed in a second sample by confirmatory factor analysis, which obtained adequate fit indices, showing the adapted TPS to be a reliable and valid instrument to measure patients’ trust in their doctors. This may be because the population of this study is similar to the population in which the original instrument was developed, which were adults with diabetes, another chronic disease that involves regular treatment and in which the relationship with the physician is key.

In our adapted scale, items that had a very low correlation with the total scale, which curiously coincided with the inverse items of the original scale, were removed. Inverse items (which contain negative statements, or they measure the construct in the opposite direction to the rest of the instrument) may create confusion in certain populations, especially those with low levels of education or scarce resources, leading to inconsistencies with the rest of the scale, and this may have happened for our study population. Although these items do not show lower correlation with the total test in the original instrument, other authors have also had issues with these items, Aloba and collaborates concluded they may belong to a different factor than the rest of the items (^{Aloba et al., 2014}). To tease out these hypotheses, the translated scale could be tested with the inverted items in the same direction as the rest. Despite the elimination of these five items, the correlation between the two versions (adapted and original) is very high, suggesting that they still measure the same construct, and thus making the adapted TPS version an even simpler and shorter instrument, which can be especially useful for use in resource-poor settings, as is the situation in most of Mexico.

The instrument showed a variable spread of answers across the trust spectrum. No differences were found in the level of trust in physician between men and women, although participants were mostly male and the study did not have the power to detect differences between genders. Likewise, no differences were found in the trust score by age group, similar with what was found in other studies (^{Kayaniyil et al., 2009}; ^{Muir et al., 2009}), probably because age does not affect the level of trust in the physician, and it is probably other variables included in the doctor-patient relationship that have a greater effect on this construct. The study did not capture data on the patients’ main physician or other potential key variables that could be associated with the observed variability, and it would be important for future studies to address them.

The level of trust in the physician found in this study is considered generally adequate, although at the moment there are no cut-off points for this construct. There are more patients who fully trust their doctors than those who scored the lowest. This would contribute to explain the very good rates of adherence to ART in this population, where the vast majority showed ART adherence of more than 95%. “Trust in the doctor” and “patient-physician relationship” in people living with HIV are very relevant elements because they are related to better commitment to care (^{Freburger et al., 2003}), better health outcomes, better adherence to treatment and better clinical follow-up (^{Hall et al., 2002}; ^{Hurley et al., 2018}; ^{Mechanic & Meyer, 2000}; ^{Safran et al., 1998}), which implies important benefits for individual and public health.

Limitations of the study include that it had to be carried out online due to the COVID-19 pandemic. This can bias the study by excluding people with no Internet access, who don’t follow the clinic’s social networks, or had not provided their email. Also, in online surveys it cannot be controlled who answers the questionnaire, because of that, and to minimize the risk of responses were not from the target population, an HIV self-reporting question was included. Another limitation could be the relatively small sample size, it could be important to test the translated and adapted version of TPS in other Spanish-speaking populations of people living with HIV in Mexico.

The Spanish for Mexican translated and adapted version of the TPS has proven to be a reliable and valid instrument, with adequate psychometric properties and good model fit indices in people living with HIV. A reliable instrument to measure this variable can be key to detect and resolve trust issues in patient-physician relationships in a timely manner. When trust issues will be detected, they may benefit from interventions by mental health professionals, trainings, and workshops to improve patient-physician relationship and confidence in doctors and health system. While more studies are necessary to verify the construct and convergent validity of this instrument, results suggest that the adapted and translated TPS can be a useful, brief and easy-to-use tool to assess trust in doctors among Mexican people living with HIV.