Changes in Chinese Patients' Trust in Doctor:a Systematic Review and Meta-Analysis

Background: To understand how patient’s trust has changed over time in China and explore the in�uencing factors of the changes. Methods: Databases were systematically searched from inception to Jan 1, 2020 for studies of patient’s trust in China investigating the change of the patient’s trust over time and the factors which in�uence the trust. Results: The cross-temporal meta-analysis results showed that the total trust, benevolence trust and technical competence trust of Chinese patients decreased rapidly across the year of data collection. The related meta-analysis results showed that trust and patient demographic variables have a weak but signi�cant correlation (gender: r = 0·032; 95% CI 0·012, 0·052; p = 0·002; age: r = 0·079; 95% CI 0·047, 0·111; p < 0·001; education: r = 0·092; 95% CI 0·054, 0·131; p < 0·001). The medical policy stage and patient type had a signi�cant in�uence on changes of trust. Conclusion: Patients’ trust in doctors have declined with increasing years, and farmer-friendly policy, con�icts and individuals’ characteristics in�uence patients’ trust. Practice implication: Chinese government


Introduction
Trust, the core of the harmonious doctor-patient relationship, is related to the process and the nal results of medical diagnosis and treatment. 13][4] In contrast, due to mistrust, the patient suspects the doctor's motivation and even treats the therapeutic staff aggressively. 5During the process, necessary self-protection measures by the doctor lead to conservative medical treatment and di culty of communication. 6Therefore, a good patient's trust is not only the basis of effective treatment but also a vital link to build a harmonious doctor-patient relationship.

Patient's trust
Patient's trust is the willingness of the patient to be vulnerable to the actions of the doctor based on the expectation that the doctor will perform a particular action important to the patient, irrespective of the ability to monitor or control that the doctor. 7,8][11][12] With regard to the types of the patient's trust, Hall classi ed it into two categories: technical competence trust, which is quali ed practice skills to make correct decisions and avoid mistakes, and benevolence trust, which is interpersonal and communication skills to care the patient's interests and tell the truth. 8sides, some researchers have proposed that patient's trust consists of affective trust and cognitive trust. 13,146][17] We adopted the Hall's classi cation in this study which was used by the most researchers. 2,8ere are a variety of the patient's trust scales, especially in the developed countries.In 1990, Anderson explored the rst English-language scale for the systematic assessment of patient trust which has good reliability and established a standard for such measurement tools in the future. 18Since then, many researchers have begun systematic studies on patient's trust scales, such as the Safran Primary Care Scale, 19 the Kao Patient Trust Scale, 20 and the Wake Forest Physician Trust Scale (WFPTS). 823] Patient's trust in China At present, the status of the patient's trust in the world is still unknown.The global comparisons of trust attitudes around the world suggested that the share of people reporting to trust others in China is much higher than that in most developed countries. 24However, in the past decade, doctorpatient con icts occurred frequently and the negative reports about the relationship between doctors and patients have been overwhelming in China, which led to the original harmonious doctor-patient relationship has been becoming tense. 12fore 2009, the patient's trust in China was high. 257][28][29][30] Meanwhile, after comparing the level of interpersonal patient's trust between China in 2011 and 2016, it was found that the trust in 2016 was signi cantly lower than that in 2011. 23These ndings have prompted calls for increased attention to the patient's trust in China.
In 2002, Chinese researchers started exploring doctor-patient trust from the perspective of ethics.There is quite more theoretical research on the concept, model construction, evolution process and evaluation system, while few empirical studies focus on improving patient's trust. 31As far as the research content is concerned, the current empirical research of patient's trust has been lagging behind the investigation of the characteristics and overall situation of both doctors and patients, which lacks in-depth medical research with the dynamic and systematic construction.With regard to the measurement tools, we lack a universal scale that is fully applicable to the Chinese patient's trust research.These studies in China, which should have supported the government in improving patient's trust, are still in their infancy.Therefore, it is urgent for Chinese organizations and researchers to take better measures to improve the patient's trust.

Factors in uencing patient's trust in China
In order to improve the patient's trust, we need to explore the main in uencing factors of the trust in China.Some researchers summarize the antecedents of the patient's trust, including social environment, treatment situations and individual characteristics. 12At present, few research studies in China focus on treatment situations and doctors' characteristics.Therefore, we summarized the in uencing factors of the patient's trust from two aspects: patients' characteristics and social environment.

Patients' characteristics
A majority of the existing empirical researches in this area have begun to focus on the in uence of patients' characteristics. 12,22,32From the perspective of patient demographic variables which were studied most, some research pointed out that female patients and the patient with the same gender as the doctor trust doctors more; 33,34 some studies found that patient age is negatively related to patient's trust, 33 but some evidence show the opposite result. 1,22,34Additionally, a positive correlation between patients with different educational background and the patient's trust was also demonstrated. 350][41] Inpatients, rural patients and the patient with the serious illness also have a high level of trust. 25,42,43

Social environment
The formation of patient's trust depends not only on patients' characteristics, but also on social factors such as government, organizations and media.[46] In 1949-1977, the doctor-patient relationship was good.People enjoyed free medical services and high public welfare.However, a market-oriented operation adopted in the healthcare industry led to the fragile doctor-patient relationship in 1978-2002.Since 2003, in order to better realize the transition from marketoriented operations to public welfare-oriented operations for the medical industry, the State Council of China launched a new round of medical reform programs in 2009, which aimed to establish and develop the basic coverage system for urban and rural residents.Especially, with the incessant deepening of the healthcare industry reform between 2013 and 2017, a package of measures has been established, including the hierarchical medical system, lowering drug prices, medical insurance fee system, promoting the reform of public hospitals, and equalization of basic public health service. 44The farmer-friendly policy which involved in the interests of a majority of the people may promote the trust level of the public. 24,45wever, in recent years, China's doctor-patient con icts occurred frequently.Statistics from the National Health and Family Planning Commission of China (2001-2016) showed that the number of vicious incidents caused by doctor-patient con icts was 4,914 in 2001, increasing to over 8,000 in 2004, and nearly 10,000 in 2006.[48][49] Violence against the medical personnel has become a terrible wound in Chinese society. 50These con icts exacerbated the mistrust between doctors and patients, 51 which is also the main reason for the deterioration of the doctor-patient relationship. 52though we summarized these important factors on patient's trust, evidence about the role of these in uencing factors in the patient's trust is still unknown.Therefore, it is essential for China to understand the changing trend of patient's trust in these years, explore the effect of the in uencing factors of the changes and determine the relationship between some patient characteristics and patient's trust.To date, to our knowledge, there is no studies comprehensively examining the status of patient's trust, including the status, in uencing factors, existing problems and future research directions of Chinese patient's trust.
4][55] We also investigated whether the policy stage, the type of patients, and the scales differentially in uence the trust.Especially, about the patient type, most studies used t test and F test to analyze the relationship between some patient characteristics (gender, age and education) and patient's trust, rather than separately exploring the trust value of different gender, age and education.In order to clarify the in uence of these patient characteristics on patient's trust, we conducted a related meta-analysis to examine the association between patient demographic variables (gender, age, and education) and patient's trust.

Search strategy and selection criteria
In accordance with the PRISMA guidelines 56 , we began our systematic literature search for studies by some widely used academic databases from database inception to Jan 1, 2020, including Wanfang, CNKI, Google Scholar, PubMed, Web of Science, the Social Sciences Citation Index, Chinese Doctoral Dissertations Full-Text Database, and Chinese Master's Dissertations Full-Text Database.Studies were located by employing combinations of the following search terms with no language restrictions: China, Chinese, trust, physician, doctor, and patient.
We included studies using quantitative data to assess the trust of Chinese patients or the association between patients' trust and patients' demographic variables.We included studies in which su cient data were reported (Mean, Standard Deviation and Sample size; r value or F, t or χ 2 value that can be converted to r wherever possible). 57We excluded studies of case vignettes, interviews, reviews and theory.We required studies to measure interpersonal and intergroup trust to doctors with a valid, reliable and established trust questionnaire.Papers from which data were extracted are marked with an asterisk in the reference section.
Two reviewers (Y.L. and P.W.) independently screened each abstract and full-text article, and excluded those that were not ineligible.Discrepancies about inclusion of full-text reports were resolved by discussion.The study protocol is available online.

Data analysis
We designed two data extraction tables to extract data on the average patient's trust score (including total trust, benevolence trust and technical competence trust), sample size, publication year and data collection year for the cross-temporal meta-analysis (table 1), and the r value (the association between trust and gender, age, education), sample size, publication year and data collection year for the related meta-analysis (table 2) (appendix).9][60] If data from two studies were reported in one publication, we extracted data from each study independently.Data extraction was conducted by two reviewers (Y.L. and P.W.).
Meanwhile, the results of the cross-temporal meta-analysis will be affected by the characteristics of participants. 54,61So, patient type was also recorded in the rst code.In order to further determine the effect of medical policy stages on patient's trust, combined with the classi cation of medical policy stages, 44 we further divided the data collection years into two stages in view of different process of new medical reform, one is the early stage of reform (2008-2012), the other is the later stage which developed comparatively mature (2013-2017).Besides, less literature used the same scale (only half of the studies use the Wake Forest Patient Trust Scale), in order to prevent publication bias due to the missing data, we tried to include researches that using valid self-created and adapted questionnaires with similar research content while ensuring the quality of literature, and conducted subgroup analysis on these scales in the results.Thus, we also recorded medical policy stages and patient's trust measurement dimensions in two sheets.
Risk of bias was assessed by two reviewers (Y.L. and P.W.) using the Newcastle-Ottawa Quality Assessment scale for observational study designs. 62The quality tiers used for categorization of studies as 'Low', 'Intermediate' and 'High' quality were presented in the appendix.
In the cross-temporal meta-analysis, we conducted the weighted regression analysis to explore the change in the patient's trust (including total trust, benevolence trust and technical competence trust) over time.Then, we did moderator analysis on medical policy stages, patient type, and patient's trust measurement dimensions; In the related meta-analysis, we measured heterogeneity using I 2 statistic and assessed publication bias by funnel plots and Egger test in separate models.Then, we carried out the meta-analysis to assess correlations between patient's trust and patient demographic variables (gender, age, and education), and performed subgroup analysis on medical policy stages and patient's trust measurement.All statistical analyses were done using Excel 2018, SPSS 24•0 and Stata 14•0. 63

Role of the funding source
The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.The corresponding author had full access to all the data in the study and Y.L. and P.W. had nal responsibility for the decision to submit for publication.

Results
Search results and quality assessment Our search strategy identi ed 544 articles, of which 40 met eligibility criteria for full-text review ( gure 1).Among the 40 studies that were published between 2009 and 2019 with 48,376 participants, 33 studies with 45,711 participants were used to conduct the cross-temporal meta-analysis, 2,21-23,25-29,32,34,35,37,38,41- 43,64-79 and 17 studies with 9,976 participants were used to conduct the related meta-analysis. 2,21,26,32,34- 38,65,71,80-85 f the 40 qualitative studies, 28 were deemed to be of 'high' quality and 12 were deemed to be of 'Intermediate' quality according to our evaluation by using the Newcastle-Ottawa Quality Assessment scale.No studies were excluded on the basis of quality.

Cross-temporal meta-analysis
The trends in different types of patient's trust According to the shape of the Scatter chart ( gure 2, 3), a generally linear decline in patient's trust from 2008 to 2017 and two large uctuations separately at 2009 in gure 2 and 2013 in gure 3 were shown.In order to clarify the differences in trust between years, we conducted variance analysis in total trust, benevolence trust and technical competence trust

Correlations between trust and years
Since these patients' trust levels were all showing a downward trend after 2009, it was more meaningful to analyze the correlations from 2009 to 2017.When sample size was weighted, the total trust scores has a signi cant correlation with year of data collection, β = -0•445, p = 0•001, the benevolence trust scores also has a signi cant correlation with year of data collection, β = -0•044, p = 0•034, and the correlation between technical competence trust and year of data collection was -0•390, p = 0•066.These suggested that patients' trust scores decreased over time.
To calculate the magnitude of change in all trust scores, we predicted the mean scores of the rst (2009) and last year (2017) of the included studies by conducting the weighted regression analysis with patient's trust scores and year of data collection as the dependent and independent variables and averaged all the standard deviations reported in the studies through calculating the average standard deviation. 86,87The regression equation is: y = Bx + C, y is the average score of various types of trust, B is the partial regression coe cient, x is the year of data collection, and C is a constant.The regression equations for total trust, benevolence trust, and technical competence trust were: y = -0

Moderator analysis
The results indicated that the medical policy stage had a signi cant in uence on trust, and patient's trust questionnaires had no signi cant effect on the relationship between trust and year (table 3).All trust values of the previous stage were signi cantly higher than the later stage.What's more, in the patient type, it was found that inpatients had higher results in total trust than outpatients signi cantly and rural patients had higher trust levels than urban patients by using independent t-test.

Heterogeneity
The random effect model and the xed effect model are two main models in meta-analysis. 89Higgins believed that I 2 is 25%, 50%, and 75%, which means low heterogeneity, medium heterogeneity, and high heterogeneity, respectively. 90Hedges proposed that the random effect model is more appropriate when the Heterogeneity test result is signi cant, otherwise, the xed effect model is selected. 89Combined with the above two criteria, the heterogeneity tests were conducted on the meta-analysis data of the relationship between patients' gender, age, education and patient's trust.The results showed that the effect size of the gender data was homogeneous but the effect size of the age and education data were heterogeneous.Therefore, it is appropriate for the gender data to choose the xed-effect model on the grounds that 31•5% of the observed variation in the gender data was caused by the real difference between the two, and yet the random effect model was suitable for the age and gender data due to 50•3% and 58•8% observed variation (table 2). 91

Publication bias
Publication bias tests were performed by using a funnel plot and Egger test. 92From the funnel plot, the studies involved in this study were distributed evenly on both sides of the total effect.According to the Egger test, 93 there was no publication bias for gender (t = 1•73, p = 0•104) and education (t = 2•07, p = 0•061).Although there was publication bias for age (t = 2•50, p = 0•025), the number of missing studies was estimated to be 5 after 5 iterations with the Linear method, and 5 similar studies were added around the center of the funnel plot at the opposite mirror position by the trim-and-ll analysis. 94Then, the metaanalysis was performed again using the random effects model.The new combined effect index was 0•055 ([95% CI 0•024-0•085]; p < 0•001).Because the combined effect index before trimming was 0•079 ([95% CI 0•047-0•111]; p < 0•001), and there was not much difference between the two combined results.Therefore, it could be considered that although there was a slight publication bias in the age data, the results were still valid (appendix). 95

Association between trust and patient demographic variables
Across all outcomes, we found a small but signi cant correlation between trust and patient demographic variables (gender: r = 0

Subgroup analysis
Strati ed analysis showed some variation in associations between gender, age and trust when studies were analyzed at different medical policy stages.In the medical policy stage 1, the results of association between gender, age and trust were insigni cant.However, in the stage 2, the results of these association were signi cant, which indicated that women and the seniors report more trust to their doctors.No variation in associations between patient demographic variables and trust when different trust questionnaires were used (table 3).

Discussion
The patient's trust decline in China Our research conducted the cross-temporal meta-analysis on patients' trust over the past ten years and found patients' total trust, benevolence trust and technical competence trust to doctors have declined within these years.This result was in agreement with the ndings of a comparison of interpersonal trust levels in 2011 and 2016. 23e main reasons for the decline in patient's trust in the past decade are as follows.Doctor-patient disputes could exacerbate doctor-patient mistrust. 517][48] Meanwhile, the doctor's negative image was fabricated from 2011 to 2015 due to the media's unreasonable reporting framework on medical disputes, which has had a greater impact on patient's trust. 96,97100] More importantly, different stages of medical policy had a great impact on trust.China's medical system entered the stage of marketization after 1978: the government categorized hospitals as private ownership, excessive market competition led to the gradual decrease of government health investment, the healthcare service gap between urban and rural expanded, and patient medical costs rapidly increased. 44Among them, marketization had a restraining effect on trust. 101Therefore, in 2009, a new round of medical reform was launched to reverse marketization by increasing public welfare and improving the medical insurance system.From the overall trust value in 2009 (Figure 2), it can be seen that this measure has had some effects on trust.However, the development of market rules was still insu cient to fully protect the interests of patients. 102In particular, from the changes of benevolence trust and technical competence trust, it can be seen that the trust of patients in China has declined signi cantly before 2013.After 2013, a series of measures were vigorously carried out to strengthen the construction of public service system, such as improving the primary care system, increasing grassroots medical institutions, optimizing the allocation of medical resources, strengthening medical supplies regulation, improving the pharmaceutical distribution system, and building a universal health care system. 44According to the data of the National Health and Family Planning Commission of the PRC, the new rural cooperative medical insurance accounted for nearly 95%, and the government and medical insurance beard 299•5 billion yuan (71•1%) of China's total health expenditure in 2016, which to some extent reduced the burden of patients, so patient's trust has been stable.It is worth noting that though the Chinese government has invested heavily in the medical industry, new healthcare reform is still faced with numerous di culties and challenges: the dislocation pro t model, the rising of drugs, and the burdensome medical insurance fund.Thus, the trust value of the later stage (2013-2017) was still low.
White Paper: Medical and Health Services in China pointed out that the contradiction among China's current series of policy, public health needs and the coordinated development of the economy and society was still unsuitable and prominent.During the outbreak of novel coronavirus (2019-nCoV), a lot of emotional construction aimed to improve the doctor-patient relationship was carried out throughout the country.However, serious violent incidence between doctors and patients still occurred, since the medical policy could not meet the needs of patients.
Additionally, our results also showed that the benevolence trust decreased rapidly across the years.The possible reasons are as follows: The rigid behavior and the dehumanizing medical attitude of the doctor may affect the patient's perception of benevolence and trust; 103,104 In addition, excessive medical treatment has been widespread in China in recent years: Doctors put patients on drugs excessively in order to avoid medical treatment risk or to get more pharmaceutical rebates. 44It is di cult for the patient to clarify the motivation of the doctor for excessive medical treatment, which leads to suspicion of doctor's professional ethics; 105,106 Furthermore, high medical costs and fee-for-service systems also affect benevolent trust.Since Chinese patients must pay separately at each step of the medical treatment process (consultation, diagnosis, intravenous drip, medication, etc.), that is, money seems to be emphasized throughout all medical processes, this payment term is likely to increase patients' distrust in doctors' benevolence. 107tient's characteristics in uence trust The results of the related meta-analysis showed that patients' gender, age, and education were signi cantly related to patients' trust, though these correlations were low (r gender = 0•032, r age = 0•079, r education = 0•092). 88Of particular relevance to the trust do-main, women are characterized by a desire to form and maintain relationships compared to men. 108,109So, women are more likely to trust their doctor.With regard to age, the older contact doctors more frequently and are more dependent on doctors.Thus, the senior reported to have more trust to the doctor.1][112] In addition, the medical policy stage, as a moderator variable, had positive effects on the association results of gender and age, which indicated that the construction of the public service system was positively promoting the relationship between gender, age and trust to a certain extent.Studies have demonstrated that the female and the senior have made greater use of basic public health services in recent years. 113It can be seen that the construction of the public service system has brought bene ts to relatively vulnerable groups and promoted their perception of trust.
According to the results of the moderator analysis in the cross-temporal meta-analysis, the study also found that inpatients had signi cantly higher trust levels (total trust, benevolence trust and technical competence trust) than outpatients.This result was in line with the ndings of Tam and Jiang. 25,43Based on Service Encounter Theory, 114 the generation of trust has a cognitive basis, and the increased interaction between the doctor and the patient is conducive to the patient's trust. 102Compared with outpatients, inpatients have longer contact time with doctors, so they are more familiar with their doctors, and doctors also have a more detailed understanding of inpatients. 68,102,115What's more, according to Interdependency Theory, 116 two-way dependence situations and high levels of patient dependence promote patients' trust. 117In general, the conditions of inpatients are more serious than those of outpatients, thus inpatients rely on doctors more and give them more hope for health. 25So, the inpatient has a higher level of trust.For the same reason, rural patients' trust in doctors was more than urban patients' trust.Some studies have investigated the differences in the doctor-patient relationship between urban and rural areas, and found that rural patients usually seek medical treatment in rural health clinics or village clinics. 42Villagers are familiar with their doctors, and convenient to seek medical treatment.
However, municipal hospitals have fewer communication opportunities and higher costs.As a result, rural patients have higher levels of trust in doctors.

Implications and limitations
The comprehensive scope this review on the patient's trust in China is a strength, since we combined the cross-temporal meta-analysis and the related meta-analysis to analyze the status of the patient's trust and the effect of the in uencing factors, with low publication bias.However, this study has limitations.
Our original systematic search did not include literature published in 2010, since there was no data for that year, and thus it might affect the results.Also, the number of researches included in this review was not large.However, the quality and data of the included studies were su cient to explore the changing trends of patient's trust.
In Summary, in the past ten years (2008-2017), the trust of Chinese patients has shown a downward trend, and the benevolence trust has fallen fastest.Medical policy, doctor-patient disputes and patients characteristics in uence the trust.Therefore, our government should attach importance to the public welfare-oriented policies and formulate relevant policies to resolve doctor-patient con icts, medical institutions need to pay more attention to the diverse trust levels of different patient groups and strengthen the training of doctors' benevolence skills, and researches ought to improve measurement tools which is applicable to the Chinese patient's trust research and conduct more in-depth medical researches to improve the patient's trust.Forest plot showing associations between patient's trust and gender . The result showed that patient's trust in 2009 (M = 4•154, SD = 0•127) was signi cantly higher than that in 2008 (M = 3•316, SD = 0•470) (t = -3•504, p = 0•017) in total trust, and patient's trust in 2013 (M = 3•049, SD = 0•035) was signi cantly lower than that in 2011 (M = 4•142, SD = 0•371) (t = 3•941, p = 0•029) in benevolence trust.

Figures Figure 1 Flow of literature through the review Figure 2
Figures

Table 3 .
Associations between patient's trust and patient demographic variables strati ed according to the Medical policy stages and Trust Questionnaire.