Influence of patient and hospital characteristics on inpatient satisfaction in China's tertiary hospitals: A cross‐sectional study

Abstract Background Patient satisfaction has been seen as a key criterion when evaluating hospitals and is one of the main focuses of the current health‐care reform in China. This paper aimed to explore patient‐ and hospital‐level factors associated with inpatient satisfaction, which can provide policy implications for the evaluation and development of a patient‐oriented health‐care system. Methods The paper analyses data from the 2017 China National Patient Survey which includes 20 300 inpatients from 131 tertiary hospitals across 31 provinces. Descriptive analysis and multivariable logistic regressions are conducted to identify key factors related to satisfaction. Results Patient sociodemographic characteristics, including gender, age, income and insurance type, are found to be strongly associated with their satisfaction of inpatient experience. In terms of institutional characteristics, hospital type, size, staffing and financial performance are also significantly correlated with inpatient satisfaction. Patients are more satisfied with specialist hospitals and large hospitals measured by the number of beds and surgeries. Hospitals with higher nurse‐to‐bed ratio also receive more satisfaction. The financial performance of hospitals, however, is negatively associated with satisfaction. Conclusion Patient satisfaction contains unique information on service quality and thus should be incorporated into the matrix of hospital evaluation. Meanwhile, differences in patient composition must be adjusted to make fair comparisons across hospitals. Moreover, future reform needs to put greater efforts in the design of comprehensive public insurance scheme, efficient hospital structure and an overall well‐functioning health‐care delivery system in order to better serve patients in China.


| INTRODUC TI ON
Patient satisfaction has long been recognized as a key measurement of hospital quality. Apart from observable health outcomes, patients' assessments of their experience with health-care providers can capture unique information on the provision of care, for example, complications that are hard to quantify, patient involvement in treatment decisions and physicians' explanation of procedures. Over the past 20 years, many developed countries have been trying to include patient satisfaction surveys as a quality improvement tool in healthcare markets. [1][2][3][4] China is also putting a great emphasis on improving the quality of care and patient satisfaction in her current health care system reform. 5,6 The 'Healthy China 2030' initiative raised a set of requirements, which include promoting service appointment systems, optimizing ward structure and enhancing health information system to improve patients' hospital experience. 7 In order to evaluate the improvement of health-care quality, Peking Union Medical College School of Public Health was commissioned by the Chinese government as a third-party academic institution to perform annual surveys of patients, known as the China National Patient Survey. 7 This survey collects data from 136 tertiary hospitals, which are defined as institutions with more than 500 beds in China. Tertiary hospitals are a major provider of inpatient service in Chinese health-care system. In 2017, there were 2340 tertiary hospitals nationwide, which accounted for 7.5% of hospitals and served 44.4% of inpatient visits. 8 Among these tertiary institutions, over 90% are public hospitals which are self-financing entities responsible for their own balance sheets while receiving some government funding. [8][9][10] In this paper, we use data collected in 2017 to explore key factors related to inpatient satisfaction, which will help to have a better understanding of how to incorporate patient satisfaction into hospital evaluation and guide future reforms in improving hospital service quality.
There has been a large strand of literature trying to identify the relationship between hospital structure and patient satisfaction in developed countries. Most of these studies have shown that higher staff-to-patient ratio and better hospital environment were associated with higher patient satisfaction rates. 11,12 Large hospitals and teaching hospitals, however, received lower scores. [12][13][14][15][16][17][18][19] In China, although there has been increasing attention on healthcare quality from both the government and the public, there is a lack of rigorous studies examining how hospital characteristics correlate with inpatient satisfaction. Most of the studies were restricted by the scope of their surveys which only involved a small number of hospitals in a single geographic region. [20][21][22][23][24][25][26][27][28][29] The focuses of these studies are patient demographic characteristics, 20 20,27 ; and the correlation between different dimensions of satisfaction. 27 Among the few studies that used data from nationwide surveys, it has been shown that higherlevel hospitals, lower competition in providers' market and higher market share of private hospitals were negatively correlated with inpatient satisfaction. 30 Higher patient-to-nurse ratio, however, had ambiguous relationship with patient rating. 31,32 There are also studies looking at how patient-doctor relationship correlates with patient satisfaction. 7,33 However, such measurements are more of reflections than the causes of patient satisfaction. For example, patients tend to have more trust towards their doctors when they feel satisfied with their treatment experience. Therefore, enhancing patient-doctor relationship is more like setting a goal instead of providing practical suggestions for hospitals to improve service quality.
It is also a concern that most of papers discussing relationship between hospital structure and patient satisfaction used correlation analysis without controlling for patient characteristics. Since it has been shown in both foreign and Chinese settings that patients' age, insurance type and health status have strong relations with their evaluation, 14,16,[20][21][22]30,34 such correlation studies might be biased by omitted patient-level factors. For example, if old patients are more likely to visit large hospitals and also tend to rate high in satisfaction surveys, there would be a positive correlation between large hospital and satisfaction score, while we can hardly conclude from it that larger hospitals have better services.
By including both patient-and hospital-level variables into regression analysis, we are able to explore key factors associated with satisfaction holding other confounding factors from the other side constant. The results of both sides can also have policy implications of two directions. On the one hand, our results on the relationship between patient characteristics and satisfaction can help to construct future hospital evaluation matrix which adjust for differences in patient composition across hospitals. Major institutional factors identified in this study, on the other hand, will provide guidance for the improvement of hospital quality and promote the development of a patient-oriented health-care system.

| Data
This study uses data from the 2017 China National Patient Survey which was collected from 136 tertiary hospitals across 31 provinces during December 2017-January 2018. In order to take into account different types of hospitals, one provincial general hospital, one provincial traditional Chinese medicine (TCM) hospital, and one maternal and child health hospital were selected from each province. In addition, 43 hospitals affiliated with National Health and Family Planning

K E Y W O R D S
Chinese health care, demographic characteristics, hospital performance, hospital structure, patient satisfaction Commission, including 28 general hospitals and 15 specialist hospitals, were also included. All hospitals in the sample are public hospitals.
Both inpatients and outpatients were interviewed in the survey.
However, in this paper, we only focus on inpatient satisfaction and factors related to it. At least 150 inpatient respondents who were to be discharged on the survey days were selected from each hospital (for hospitals with not sufficient discharges, discharged patients were all selected for the survey; for hospitals with sufficient discharges, patients were stratified by specialties/wards), generating a total of 21 125 respondents among which 21 092 were effective respondents. The interviews were conducted on-site in the wards by a group of pre-trained medical students. The inpatient questionnaire has five domains, including Process management, Hospital environment, Nursing care, Physician care and Overall satisfaction (detailed contents are listed in Table 1). A total of 20 questions were asked using a Likert scale from 1 through 5, corresponding to 'very unsatisfied', 'unsatisfied', 'neutral', 'satisfied' and 'very satisfied', respectively. The questionnaire was validated by small-scale multidisciplinary expert consultations, patients' cognitive interviews and pilot field tests.
Information on patients' characteristics, such as age, gender, education, income, insurance type and length of stay, was also collected.
Hospital-level information includes number of doctors, nurses, beds, total cost and revenue in 2017. These data were collected from the statistical department of the hospital.

| Analysis
Due to missing data for cost/revenue, five hospitals are excluded from the sample, together with 792 patients from these hospitals, leaving a sample of 131 hospitals and 20 300 patients. For each of the five satisfaction domains, we calculate the average score of the included items as domain-level satisfaction score and test for internal consistency using Cronbach's α coefficient. As shown in Table 1, the average satisfaction scores range from 4.56 to 4.77, with Cronbach's α all at acceptable level.
In order to show the relationship between patient characteristics, hospital characteristics and the satisfaction score in each domain, we first present descriptive evidence using Kruskal-Wallis and Mann-Whitney tests for difference in satisfaction scores between different sociodemographic groups and hospital categories. In order to focus more on the factors associated with patients being satisfied, we then proceed to multivariable logistic regression where we translate satisfaction scores for each domain into binary variables which equal to 1 if the score is higher than 4 (indicating 'satisfied' or 'very satisfied' in the original questionnaire) and 0 otherwise. Independent variables of interest including age, gender, education, income category, insurance type, length of stay of each inpatient, and specialty type, number of beds, number of surgeries, doctor-to-nurse ratio, nurse-to-bed ratio, revenue-to-cost ratio of each hospital were all put in the regressions using enter method. All data analysis was performed using SPSS version 22.0.

| Descriptive statistics
Summary statistics for patient characteristics are presented in mains. Doctor-to-nurse ratio and nurse-to-bed ratio are also grouped as above or below the median. However, while hospitals with lower doctor-to-nurse ratio only have slightly higher overall score and satisfaction towards process management, higher nurse-to-bed ratio is strongly associated with higher satisfaction by all five measurements.
Finally, in terms of hospital financial performance, we compare hospitals with revenue-to-cost ratio above and below 1, or in other words, hospitals with financial gain or loss. Mann-Whitney tests indicate that profitable hospitals receive lower score in either satisfaction domain.

| Patient characteristics
Now we turn to multivariable logistic regression to explore key characteristics related to patient satisfaction. From the patient side, as shown in Table 4 However, given the extremely small share of uninsured patients in the sample, it is hard to make any credible inference from this group of correlation. and nurse-to-bed ratio (OR = 1.251 for overall satisfaction), although the former is not statistically significant at 95% confidence interval. Besides overall satisfaction, higher nurse-to-bed ratio is associated with higher satisfaction with nursing care (OR = 1.280) and physician care (OR = 1.26). Hospital financial performance and patients' satisfaction also show strong correlation. As presented in Table 4, hospitals with financial profits are earning significantly lower satisfaction score across all dimensions (OR = 0.748 for overall satisfaction).

| Hospital characteristics
The model performs well in terms of prediction and goodness of fit, with the per cent of correctness higher than 85% and chi-square P-values smaller than .001 for all dimensions. To check the robustness, we also use average length of stay, number of beds and doctorto-nurse ratio in tertiary hospitals collected from National Health Statistical Yearbook 8 as cut-offs, and the regression results remain the same.

| D ISCUSS I ON
Improving patient satisfaction has been emphasized as one of the main objectives in the current reform of China's health-care system.
While devoting great efforts into developing patient-oriented delivery system, we need to have a better understanding on what is associated with patient satisfaction, how to evaluate and what we can do to improve it. In this study, we use descriptive analysis and multivariable logistic regression to explore factors at both patient level and hospital level that are strongly related to patient satisfaction.
From patient side, we find that gender, age, education, income and insurance type are significantly associated with inpatient sat-  to an under-(over-) estimation of the correlation between hospital size and patient satisfaction. Thus, our results cannot be interpreted as casual effects of these factors, and further studies eliminating such confounding factors are still needed to guide future attempts in improving health-care quality.

| CON CLUS ION
This paper explores patient and hospital factors related to inpatient satisfaction. We found that patient gender, age, income and insurance type are associated with their satisfaction towards inpatient stays.
Female and elder patients tend to have higher satisfaction rating, as well as patients with higher income and more comprehensive insurance. Hospital type, size, personnel structure and financial performance also have significant correlation with the ratings of patients.
Specialist hospitals, large hospitals and those with higher nurse-tobed ratio receive more satisfaction. The financial performance of hospitals, however, is negatively associated with satisfaction. These results suggest that, on the one hand, patient characteristics must be adjusted when incorporating satisfaction into hospital evaluation. On the other hand, future health-care reform should focus more on designing better public insurance benefits, efficient hospital structure and a well-functioning health-care delivery system.

ACK N OWLED G EM ENTS
We We also appreciate 31 provincial health authorities and leading experts who helped in recruitment of investigators and coordination with sample hospitals.

CO N FLI C T O F I NTE R E S T
The authors report no conflict of interest in this work.

AUTH O R CO NTR I B UTI O N S
LH and HD designed the research question and structure of the paper together. LH and SL performed data cleaning and statistical analyses. HD was the major contributor in writing the manuscript.
YL provided guidance and overall support for the paper. All authors read and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings presented in this study are avail-