Data source
A cross-sectional inpatient survey was conducted in 47 tertiary public hospitals (32 general hospitals and 15 specialty hospitals) in Shanghai in July and August 2018. Only three tertiary public hospitals in Shanghai (one mental health center, one hospital specializing in infectious disease, and one with no inpatient care) were excluded from the study. Because 90% of all patients receive medical care at public hospitals in China [28], patient care in public hospitals can generally represent patient care in China.
A random sample of the inpatients who had completed their main medical care (such as surgeries or therapeutic procedures) was selected from each of the sampled tertiary public hospitals during one workweek. The number of sampled inpatients in each hospital was 55 on average (52 to 79). All the voluntary investigators, who were mainly senior medical students from the major medical colleges in Shanghai, were trained on the inpatient survey. The survey was conducted via an e-questionnaire administered on iPads. Oral informed consent was obtained before patients’ participation in the survey.
In the questionnaire survey, the data related to inpatient satisfaction, inpatients’ perceived SDM, public hospital type (general vs. specialty), inpatient characteristics (such as gender, age, residence (Shanghai vs. non-Shanghai), education, family monthly income (<5k, 5k-, 10k-, 20k-, 50k RMB)], cancer experience (yes vs. no), having surgery (yes vs. no) and admitting clinical department (internal medicine, surgery, gynecology, pediatrics, traditional Chinese medicine, ENT, others)) were collected.
Measures
SDM scale Thirteen items were used to assess four aspects of SDM in inpatient care (Table 1). The four aspects included “Patients’ information preference”, “Patients’ active involvement in SDM”, “Patients’ perceived encouragement from their physicians to achieve SDM” and “Informed consent”. Of the four aspects, the former two aspects reflected patients’ desire for autonomy, while the latter two aspects reflected patients’ perceived autonomy support [14]. The items in the aspect “Patients’ active involvement in SDM” were based on PICS [22], and items in the aspect “Patients’ perceived encouragement from their physicians to achieve SDM” were based on the SDM-Q-9, SDM-Q-Doc, CollaboRATE and PICS [22-25]. The items in the aspects “Patients’ information preference” and “Informed consent” were developed by the authors according to important relevant issues with respect to patient autonomy. Experts in medical care quality were consulted for all the items in the SDM scale.
Each item of the SDM assessment was rated using a 5-point Likert scale: 1 for “strongly disagree”, 2 for “disagree”, 3 for “neither agree nor disagree”, 4 for “agree” and 5 for “strongly agree”. In this study, the percentage of inpatients who rated an item on the SDM scale as “strongly agree” or “agree” was referred to as the positive response rate (PRR) for this item.
Inpatient satisfaction scale Based on our previous inpatient satisfaction scale and consultation with experts in medical care quality, four dimensions with 35 items were used to assess inpatient satisfaction (Supplement 1). The four dimensions of the inpatient satisfaction scale were “Facilities and equipment”, “Physician services”, “Nonphysician services” and “Medical care process and effectiveness”. To assess the association of inpatients’ PRRs for SDM with their satisfaction, overall inpatient satisfaction with medical care, the dimension of “Physician services” (termed “physician services” hereafter) and two items, “Medical expenses are reasonable” and “I was satisfied with medical care outcomes” (referred to as “medical expenses” and “treatment outcomes” hereafter, respectively) were used.
Each item of the inpatient satisfaction scale was scored using a 5-point Likert scale: 1 for “very dissatisfied”, 2 for “dissatisfied”, 3 for “neither satisfied nor dissatisfied”, 4 for “satisfied” and 5 for “very satisfied”. If an item was irrelevant to a surveyed inpatient, this item was treated as a missing value for this patient. In the analyses, the missing value of an item was replaced by the average score of the item. The percentage of inpatients who rated medical care equal to or greater than 4 is referred to as the inpatient satisfaction rate (SR), while the percentage of inpatients who rated medical care equal to 5 is referred to as the inpatient high satisfaction rate (HSR).
The psychometric analysis indicated that the inpatient satisfaction measure used in this study had relatively good construct validity based on standard tests of goodness of fit using a confirmatory factor analysis model (GFI=0.8530; AGFI=0.9030; SRMR=0.0366; RMSEA=0.0501) and had high internal reliability (overall Cronbach’s α=0.9477).
Statistical analyses
We computed the average PRR for items of a given aspect as the PRR for each aspect of SDM. We also calculated the average PRR for the 13 items of the SDM scale in the survey as a summary statistic, which we refer to as “PRR overall”. PRR overall was computed as the average of all responses received and then computed separately for general hospitals and specialty hospitals, for inpatients with cancer and without cancer, and for inpatients who had surgery and those who did not have surgery.
We computed both SRs and HSRs for medical care overall, physician services, medical expenses and treatment outcomes. The SRs and HSRs for physician services and overall inpatient care were the average SRs and HSRs of the items in the “Physician services” dimensions and of all items of the inpatient satisfaction scale.
To examine whether hospital type, admission department, cancer experience, and having surgery during admission affected inpatient PRRs for the four aspects of SDM and SDM overall, we applied t tests and two-level mixed linear regression models.
In the two-level mixed linear regression models, random effects (intercepts) were specified at the hospital level to take into account the fact that inpatients were nested within hospitals. In the above models, hospital type, admitting department, cancer experience, surgery during admission and inpatient characteristics (gender, age, residence, education and family monthly income) were used as fixed effects. The two-level mixed linear regression models could control potential confounders to some extent. To test the appropriateness of using two-level mixed linear regression models to account for the nesting of individuals within hospitals, we first ran the empty model. The results revealed significant differences in inpatient perceptions of SDM among hospitals (P<0.05). We also tried three-level mixed linear regression models that accounted for the nesting of the SDM questions within inpatient individuals within hospitals. However, the empty models did not support the use of three-level mixed linear regression models.
To illustrate the differences in adjusted PRRs between groups of inpatients, we used the coefficients in the two-level mixed linear regression models to predict the adjusted PRRs while holding all other variables constant at their means and graphically presented relevant predictions.
To test the differences in inpatient SRs and HSRs overall and for physician services, medical expenses and treatment outcomes between inpatients with or without positive responses for SDM overall and the four aspects of SDM, we used three-level mixed linear regression models that accounted for the nesting of the satisfaction questions within inpatients within hospitals. In the models, positive SDM responses referred to the average PRRs for each dimension of SDM or SDM overall that were equal to or greater than 80%. These analyses were conducted separately for the model responses for SRs and HSRs. To determine the appropriateness of three-level mixed linear regression models, we examined the empty models of inpatient SR and HSR overall and inpatient SRs and HSRs for physician services, medical expenses and treatment outcomes. The results showed significant differences in SRs and HSRs among hospitals (P<0.01) and among individuals nested in hospitals (P<0.05), except for the inpatient SR for medical expenses among hospitals (P>0.05). In these models, hospital type, admitting department, cancer experience, surgery during admission and inpatient characteristics (gender, age, residence, education and family monthly income) were used as the fixed effects. The three-level mixed linear regression models could also control potential confounders to some extent.
This study was approved by the Institutional Review Board of the School of Public Health, Fudan University (IRB#2018-05-0683).