Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Does descending health resources reform impact patient low-level hospital selection behavior? Evidence from Zhejiang, China

Does descending health resources reform impact patient low-level hospital selection behavior?... Background: Since 2013, China launched descending resources reform, which is a new attempt to correct unbalanced allocation of health resources through human capital spillovers and brand implantation from high-level hospitals. The purpose of this paper is to explore the patients’ hospital selection response to this reform with the focus of low-level hospitals to better understand the effect of this reform on correcting regional inequality of health resources allocation. Methods: The European Consumer Satisfaction Index model (ECSI) was used to design a questionnaire, and cross- sectional data from 17 hospitals were collected through 1287 questionnaires from Zhejiang Province. Patient hospital selection (loyalty) is measured using ordinary variables by considering patient willingness to choose a low-level hospital when suffering an illness or severe illness. Analysis of variance (ANOVA) and the structure equation model are applied to examine the effect of reforms on patient behavior. Results: The descending resources reform promotes improvements in the capabilities and medical environment of low-level hospitals, and descending doctors also have high accessibility. Perceived quality, patient expectations, and hospital image have significant positive effects on patient satisfaction, and the explanatory power of brand implantation from cooperative high-level hospitals and descending doctors is stronger than the image of the low-level hospital itself. And descending resources reform and patient satisfaction have significant positive impacts on patient’s choice for low-level hospitals with the existence of mediating effect of satisfaction. Conclusions: This paper provides supporting empirical evidence of the descending resources reform’simpact on patients’ low-level hospital selection. This reform has been effective in improving the capabilities of low-level hospitals, and brand implantation of high-level hospitals shows strong explanatory power. China’s reform offers a distinct and valuable approach to correcting the uneven allocation of health resources. Besides, the findings also suggest that policymakers could pay more attention to the importance of information channels in impacting patient awareness, responses, and hospital selection. Keywords: Descending resources reform, Low-level hospitals, Patient satisfaction, Consumer behavior, Hospital selection behavior * Correspondence: wangshh@126.com Department of Stomatology, Tongde Hospital of Zhejiang Province, Hangzhou, China Department of Stomatology, Songjiang Hospital of Shanghai, Shanghai, China Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. SUN et al. Archives of Public Health (2021) 79:179 Page 2 of 14 Background regional equality. One new solution for this problem is Regional inequality of health resources allocation is a the descending resources reform, introduced in Zhejiang global concern, and a key focus is access issues between and other provinces in China in 2013. The core idea of urban and rural areas [1]. Rural areas typically face the this reform is to encourage high-level hospitals to estab- challenge of doctor shortages due to the difficulties with lish cooperative ties with low-level hospitals, thus driving transport and communications that exist in most devel- the high-level hospitals’ human capital to transfer down oping and developed countries [2]. China faces a differ- (i.e., descend) to the low-level hospitals. As part of this ent constraint in terms of its structural congestion policy, the government would provide a subsidy to (at between overcrowded (city) high-level hospitals and idle least partially) compensate hospitals for the reform costs. (county and town) low-level hospitals due to patients’ Overall, the policy aims are to (1) narrow the human biased behavior, which motivates them to choose high- capital gaps among hospitals via spillover effects, and (2) level hospitals [3]. This dynamic stems from China’s imbed (brand) the image of a high-level hospital on a long-lasting price regulation and health resource con- low-level one, which could then help reshape patients’ centration in (city) high-level hospitals, and it generates behavior in terms of hospital selection, with an emphasis medical cost and efficiency losses as well as doctor–pa- on low-level hospitals [10]. This reform makes full use tient conflicts and detrimental social consequences [4]. of the dominant role of China’s public hospital system, Since 2003, following the SARS crisis, the Chinese gov- but there is still little empirical evidence on this reform’s ernment has paid increasing attention to investment in effect on patients’ care-provider choices. This paper ex- the infrastructure of (mainly town) low-level hospitals. plores this issue by using a structural equation model Policies such as higher government health expenditure, (SEM) based on questionnaire data collected in Zhejiang, expansion of medical insurance coverage, and abolition China. of marked-up drug prices have been implemented since Patient hospital choices can be viewed as reflecting pa- 2009. However, increasing health care affordability due tient loyalty to different care providers, and in the con- to expanded medical insurance coverage only worsened text of China’s public hospital system, these choices can the structural congestion [5, 6]; the efficiency of low- be viewed as patient loyalty to different levels of hospital. level hospitals has yet not improved [7]. Different from the traditional literature on patient satis- In China, the dominant public hospital system divides faction [11, 12], this paper uses the term “patient satis- hospitals into three different levels, from first to third. faction with the reform” to measure patients’ response All hospitals run by MOH (Ministry of Health of China) for the reform and its impact on low-level hospitals. This and nearly all hospitals run by provinces and major cities satisfaction would thus reflect patients’ loyalty to and were approved to be ranked at the third level, and are thus their choice behavior in terms of choosing low-level thus viewed as high-level hospitals. Community and hospitals. township hospitals were identified as first level, and dis- In the marketing science literature, a satisfaction index trict- or county-level hospitals were generally approved model can be used to discuss a reform’s effect on patient to be second level; these hospitals are usually viewed as satisfaction and loyalty. The Swedish Customer Satisfac- low-level hospitals with community and township ones tion Barometer (SCSB model) emphasizes the determi- being the lowest. nants of two antecedent factors: customer expectations Previously, the guiding concept and perception was and perceived performance; customer satisfaction then that low-level hospitals could only diagnose and treat a affects customer complaints and ultimately impacts cus- limited range of ailments and offer a lower standard of tomer loyalty [13]. The American Consumer Satisfaction care than high-level hospitals. Presently, underutilization model (ASCI) proposed by Fornell et al. [14] adds the la- and the low capabilities of low-level hospitals are still se- tent variable of perceived quality, but still uses perceived vere challenges for China [8]. These issues reflect the value to measure perceived performance. Brady and Cro- fact that previous investment focused on fixed assets ra- nin [15] emphasized service quality evaluations based on ther than human capital [3]. This kept low-level hospi- the dimensions of results, interaction quality, and phys- tals at a disadvantage in attracting patients due to the ical environment quality, similar to the ACSI model. differences in human capital between different levels of hospitals coupled with regulated medical service prices. The uneven allocation of health resources accelerated The authors thank one anonymous reviewer’s very helpful comment after 2009 due to reforms that regulated prices and ex- for this term difference. The traditional literature usually discusses individual-level response on health service through indicators of the panded insurance coverage, which then enabled more physical environment, patient-friendly environment, response capacity, patients to afford high-level hospitals [9]. communication, privacy, and security. Some other factors, like know- This history indicates that past demand-side and ledge of history, monitoring of health problems, and information, are supply-side reforms were unsuccessful in realizing also discussed. SUN et al. Archives of Public Health (2021) 79:179 Page 3 of 14 The European Consumer Satisfaction Model (ECSI) hospital (LLH) to measure patients’ hospital choice, initiated by the European Commission in 1999 removed which is affected by patient satisfaction [19, 30] and two the latent variable of customer complaint from the ACSI exogenous variables. Satisfaction is affected by three la- and SCSB models because complaint processing has no tent variables: perceived quality, consumer expectations, significant impact on customer satisfaction or loyalty in and hospital image. According to the ECSI model, the empirical research [16]. The ECSI model includes cor- difference between consumer expectations and perceived porate image, a move intended to incorporate customers’ performance is expressed as the expected value, but memory associated with organizations [17]; and satisfac- technical reliability and treatment effects of medical ser- tion mediates between service quality and loyalty [18, vices are difficult for patients to evaluate [31], so the 19]. Meanwhile, some literature has also highlighted the perceived value cannot be directly observed. Donabedian impact of demographic and exogenous policy variables [32] suggested using other non-technical variables such on satisfaction [20, 21]. However, there are few empirical as convenience and information; however, these sug- studies on how health policies affect patient satisfaction gested variables are already included in perceived quality and loyalty. and demographics. Relevant studies in China mainly discussed the impacts The descending resources reform can affect patient of the circa-2009 health reforms on the efficiency of satisfaction in two ways. First, it substantially changes low-level hospitals, but these studies did not use micro- low-level hospitals’ capabilities, which have been in- data collected at the individual level, nor were they in- cluded in the latent variables, as well as impacting per- volved the descending resources reform [6]. Some stud- ceived quality and hospital image. Second, the reform ies used micro-data to explore patient satisfaction in information can be transmitted to patients and impact developing countries including China using demographic those patients’ choices. Latent variables of the reform in- variables including age, gender, and education level [22, clude (1) whether the reform information is correctly 23]; other studies used medical market concentration, recognized by patients, (2) the information channel, and income, and health insurance status [24, 25], as well as (3) related policy on medical service price, differential other factors. Researches on developing countries also medical insurance reimbursement and tiered medical utilized scales to evaluate the relationships among ser- services, which are derived from individual hospitals. vice quality, satisfaction, and loyalty [26, 27]. However, these studies treated the institutional environment as Questionnaire design and data given, an assumption that is inconsistent with the cir- The question items and their definitions of latent vari- cumstances in developing countries experiencing a rap- able are reported in Table 1, where a five-point unbal- idly evolving healthcare system and reforms. anced scale is used for ordered variables. Patient hospital Different from reforms in developing countries, which choice (loyalty) is measured as the intention to choose a focus on designing different health resource formulae local low-level hospital and the intention to choose a and financing mechanisms [28], China’s descending re- local low-level hospital first when suffering a serious ill- sources reform is paving a new way to correct the un- ness, respectively. Latent variables of patient satisfaction even allocation of health resources. Some recent studies include (1) reform satisfaction at the industry level, and discussed the impact of this reform on both doctors or (2) satisfaction with local low-level hospitals. For the patients [25, 29], but the OLS/OLM (ordinary least three exogenous latent variables of the ESCI model, hos- square or ordered logit model) methodology used by pital image refers to patients’ brand recognition percep- Sun et al. [25] could not solve measurement error of tion of the hospital, and patient trust provides the basis survey data. The contribution of this paper is to explore of future cooperation in terms of patients’ future hos- this reform’s effect on patient hospital-selection behavior pital choices [33]. In addition, patients’ awareness of the using a SEM model. Health policy is incorporated into descending (high-level) hospital and descending doctors the patient response model in order to better understand provide a measurement of the degree to which the image the effect of the descending resources reform. Mean- of high-level hospitals is implanted onto low-level hospi- while, different levels of hospitals and cognitive channels tals. The aforementioned three variables are used as are introduced to discuss the heterogeneous effects of measurement variables of hospital image. the reform on patient behavior. Clavolino and Dalsgaard [34] pointed out that patient expectations are related to prior expectations of the ob- Materials and methods served existing services. Because the descending re- The European Consumer Satisfaction Index model is sources reform involves both high-level and low-level used as the basic model for this study, and two exogen- hospitals, we take the diagnosis/treatment capability of ous variables are included: the descending resource re- low-level hospitals and the accessibility of descending form and demographics. We use loyalty to a low-level doctors as measurement variables. Patient expectations SUN et al. Archives of Public Health (2021) 79:179 Page 4 of 14 Table 1 Questionnaire scale, variables, and definitions Variable Latent variables Measurement variables types Symbol Name Symbol Question items Definition Exogenous ξ1 Hospital image X11 Trust for LLH 1 for negative change, 2 for no change, 3–5 for ordinary variables positive change X12 Awareness for the descending 1–5 ordinary variables from very low to very high hospitals X13 Awareness for the descending 1–5 ordinary variables from very low to very high doctors ξ2 Patient X21 Accessibility to the descending 1–5 ordinary variables from very low to very high expectation doctors X22 LLH capability change 1 for negative change, 2 for no change, 3–5 for ordinary positive change X23 Medical cost change 1 for positive change, 2 for no change, 3–5 for ordinary negative change ξ3 Perceived X31 Convenience changes 1–5 ordinary variables from very low to very high quality X32 LLH Environment change 1–5 ordinary variables from very low to very high ξ4 Reform policy X41 Reform Recognition 1–5 ordinary variables from very low to very high X42 Reform information Channels 1 for public channels (newspaper, TV and hospital), private channels being 0 X43 Medical service price 1–5 ordinary variables from very low to very high X44 Insurance reimbursement 1–5 ordinary variables from very low to very high X45 Tiered medical service 1–5 ordinary variables from very low to very high ξ5 Socio- X51 Gender 1 for male and 0 for female demographics X52 Age 1 for ≤30, 2–5 for 31–40, 41–50, 51–60, and ≥ 61, respectively X53 Education level 1 for primary or below, 2–5 for junior, high school, college or university, graduate degree, respectively Mediating ξ6 Satisfaction ME1 Reform satisfaction 1–5 ordinary variables from very low to very high variable ME2 LLH satisfaction 1–5 ordinary variables from very low to very high Endogenous η Hospital Y1 Intention to choose LLH 1–5 ordinary variable from very low to very high variable selection Y2 Intention to choose LLH when 1–5 ordinary variables from very low to very high (Loyalty) suffering serious illness Source: The authors are also related to medical costs, which are included as a The data used in this study cover 17 public hospitals in measurement variable of patient expectations. The latent Zhejiang Province, China, including six tertiary hospitals, variables of perceived quality are related to associated eight secondary hospitals, and three primary hospitals. In services, and we further considered the medical environ- each hospital, face-to-face interviews were performed in the ment and convenience of low-level hospitals as measure- outpatient department, where questionnaires were randomly ment variables. distributed to patients by trained independent investigators. Socio-demographic variables include gender, age, and edu- All interviewees that finished the questionnaire received a cation level. Awareness of reforms, cognition channels, and small gift worth $1.00 (7 RMB Yuan) for their time. From reform-related policy evaluations regarding medical service November 2018 to October 2019, we collected 1354 ques- prices, differentiated insurance reimbursements, and tiered tionnaires, among which 1287 were valid, an effective rate of medical services are incorporated to measure the latent vari- 95.05%. able of reform. Of these, differentiated insurance reimburse- ment policies will incentivize patients to choose low-level Empirical methods hospitals by reducing/increasing the reimbursement ratio The samples used in this paper cover three different when choosing high- or low-level hospitals, respectively. The levels of hospitals and different information channels, so tiered medical services policy requires patients to choose a we first use one-way analysis of variance (ANOVA) to low-level hospital first, and then be referred to a high-level examine whether significant differences exist among pa- hospital if suffering a serious illness, which impacts patients’ tients at different levels of hospitals and information satisfaction and hospital choice behavior. channels. If the ANOVA results (F-test) reach the SUN et al. Archives of Public Health (2021) 79:179 Page 5 of 14 threshold value (α = 0.05), a significant difference exists. Following the ESCI model, we establish a SEM model Then, a multiple posteriori comparison will be per- incorporating descending resources reform and hospital formed to compare the differences by using the LSD selection (Fig. 1). Detailed variables are given in Table 1. (least significant difference) test. The empirical analysis is performed using two steps: (1) Because the data used in this paper come from question- estimating the impact of different exogenous latent vari- naires, patients’ cognitions of and responses to the reform ables on patient satisfaction; and (2) exploring the im- are subjective and difficult to directly measure, meaning it is pact of patient satisfaction and the reform on patients’ hard to avoid subjective measurement errors. The structural hospital selection (loyalty). In addition, subsamples of equation model (SEM) has the advantage of handling mul- different levels of hospitals will be discussed to test the tiple variables and measurement errors of variables. In robustness of the results. addition, this method can estimate both the factor structure Finally, in order to measure the existence of a mediating ef- and factor relationships, which makes it suitable to process fect, referring to Baron and Kenny [36], we estimate β , β , c a ’ ’ and analysis questionnaire data. According to Qiu and Lin β ,and β showninFig. 2 in turn. If β is insignificant and b c c [35], this model consists of a measurement equation and the others are significant, then complete mediating effects structure equation as follows: exist; however, if the estimated value of β is significant and its absolute value is less than that of β , then there is a partial Y ¼ Λ η þ ε ð1Þ mediating effect. X ¼ Λ ξ þ δ ð2Þ Results We used SPSS 23.0 software and reliability tests to assess η ¼ Bη þ Γξ þ ζ ð3Þ the reliability and consistency of the scale and data. The Equations (1) and (2) are measurement equations to results show that the Cronbach’s α coefficient is 0.915(> describe the relationship between latent variables and 0.800), indicating that the scales and data have a good measurement variables. Y and X are the observable vari- internal consistency. Then, factor analysis was utilized to ables of endogenous and exogenous latent variables re- test the validity of the questionnaire. The KMO (Kaiser- spectively, η and ξ are endogenous and exogenous latent Meyer-Olkin) value of 0.901 and Bartlett spheroid test variables respectively, and Λ and Λ are the factor load- value of 9223.199 (Sig. = 0.0001) show that the question- y x ing matrix. Equation (3) provides the structure model, naire has good structural validity, so the scale and data which is used to describe the relationship between latent are suitable in performing empirical estimations. Using variables, where in the structure coefficient matrix, B AMOS 21.0 software, the absolute fit indices of patient and Γ represent the relationship between endogenous la- loyalty model shows that RMSEA (Root Mean Square tent variables and the impact of exogenous latent vari- Error of Approximation) =0.067(< 0.10), GFI (Goodness ables on endogenous latent variables, respectively; and ζ of Fit Indices) =0.934(> 0.90), AGFI (Adjusted Goodness is the residual term matrix. of Fit) =0.846(> 0.80), and PGFI (Parsimony Goodness of Fig. 1 The theoretical model of patient hospital choice (loyalty). Source: The ESCI model is shown inside the dotted line, this model referred to Ref. [25] SUN et al. Archives of Public Health (2021) 79:179 Page 6 of 14 Fig. 2 The estimation method of mediating effects. Sources: Baron and Kenny (1986) Fit Indices) =0.545(> 0.50), suggesting that the data can reached (3.35 ± 0.87), where 87% of interviewees be used for SEM model estimation. responded “positive,”“high,” and “very high,” 10% re- ported unchanged satisfaction levels, and only 1% re- ANOVA and multiple comparison results ported that their satisfaction had declined. As for the Table 2 presents the summary statistics for the sample hospital selection (loyalty) variable, the average score for variables. It can be seen that women account for 58% of “intention to choose local low-level hospital” is 3.30, and the sample. The average age is between 2= “31-40 years the loyalty score when suffering a serious illness reached old” and 3= “41-50 years old,” with a mean of (2.35 ± 3.20 (3 = “positive” and 4 = “high”). Both of these results 1.25). The average education level is close to 3 = “high indicate that following the descending resources reform, school” (2.88 ± 1.11). The mean reform satisfaction is patients will prefer to choose a low-level hospital. (2.70 ± 1.39), which is between 2 = “low” and 3 = “fair”. For measurement variables, trust in local low-level However, satisfaction with local low-level hospitals hospitals is (3.28 ± 0.98), showing that the descending Table 2 Descriptive statistics of key variables, ANOVA, and multiple comparison results Latent variables Measured Overall Hospital level Information channel variables Tertiary Secondary Primary Public Private *,1 *,1 *,1 *,1 Hospital Image X11 3.28 ± 0.98 3.29 ± 0.95 3.34 ± 1.01 3.11 ± 0.87 3.49 ± 0.94 2.93 ± 0.95 *,2 *,1 *,1 *,1 *,1 X12 2.55 ± 1.26 3.05 ± 1.40 2.49 ± 1.29 2.48 ± 1.06 2.98 ± 1.18 1.78 ± 1.02 *,2 *,1 *,1 *,1 *,1 X13 2.50 ± 1.25 3.02 ± 1.46 2.41 ± 1.27 2.55 ± 1.02 2.93 ± 1.18 1.75 ± 0.99 *,1 *,1 *,1 *,1 Patient expectation X21 3.10 ± 1.10 2.96 ± 1.14 3.16 ± 1.14 2.97 ± 0.96 3.35 ± 1.01 2.65 ± 1.11 *,1 *,1 *,1 *,1 X22 3.35 ± 0.85 3.28 ± 0.90 3.41 ± 0.86 3.22 ± 0.78 3.48 ± 0.83 3.12 ± 0.84 *,1 *,1 X23 2.80 ± 0.99 2.88 ± 1.07 2.78 ± 0.99 2.80 ± 0.94 2.95 ± 1.02 2.52 ± 0.86 *,1 *,1 *,1 *,1 Perceived quality X31 3.27 ± 1.00 3.29 ± 0.98 3.33 ± 1.03 3.10 ± 0.92 3.48 ± 0.96 2.91 ± 0.97 *,1 *,1 X32 3.38 ± 0.91 3.35 ± 0.94 3.37 ± 0.92 3.41 ± 0.85 3.53 ± 0.89 3.10 ± 0.87 *,2 *,1 *,1 * * Reform X41 2.36 ± 1.29 2.90 ± 1.44 2.30 ± 1.32 2.30 ± 1.09 2.78 ± 1.27 1.63 ± 0.96 policy X42 0.64 ± 0.48 0.72 ± 0.45 0.62 ± 0.49 0.65 ± 0.48 –– *,2 *,1 *,1 *,1 *,1 X43 3.30 ± 0.87 2.99 ± 0.94 3.35 ± 0.86 3.30 ± 0.87 3.46 ± 0.87 3.02 ± 0.81 *,2 *,1 *,1 *,1 *,1 X44 3.20 ± 0.86 2.97 ± 0.93 3.25 ± 0.87 3.16 ± 0.80 3.34 ± 0.86 2.95 ± 0.80 *,1 *,2 *,1 *,1 *,1 X45 3.29 ± 0.90 3.15 ± 0.99 3.35 ± 0.88 3.20 ± 0.89 3.41 ± 0.91 3.08 ± 0.84 *,1 *,1 Socio-demographics X51 0.42 ± 0.49 0.32 ± 0.47 0.45 ± 0.50 0.39 ± 0.49 0.43 ± 0.50 0.40 ± 0.49 X52 2.35 ± 1.25 2.41 ± 1.11 2.34 ± 1.28 2.34 ± 1.21 2.38 ± 1.24 2.29 ± 1.26 *,2 *,1 *,1 * X53 2.88 ± 1.11 3.34 ± 1.05 2.81 ± 1.09 2.87 ± 1.13 2.89 ± 1.11 2.87 ± 1.10 *,2 *,1 *,1 *,1 *,1 Satisfaction ME1 2.70 ± 1.39 3.26 ± 1.49 2.65 ± 1.42 2.59 ± 1.19 3.22 ± 1.24 1.79 ± 1.15 *,1 *,1 *,1 *,1 ME2 3.35 ± 0.87 3.29 ± 0.88 3.40 ± 0.88 3.23 ± 0.81 3.50 ± 0.84 3.08 ± 0.84 *,1 *,1 *,1 *,1 Loyalty Y1 3.30 ± 0.90 3.29 ± 0.88 3.36 ± 0.91 3.13 ± 0.84 3.43 ± 0.89 3.08 ± 0.86 *,1 *,1 *,2 *,1 *,1 Y2 3.20 ± 0.96 3.42 ± 1.01 3.26 ± 0.99 2.94 ± 0.82 3.33 ± 0.97 2.97 ± 0.91 Sample size 1287 1287 130 851 306 819 Note: [1] Asterisks and n = 1 or 2 (*, n) denote a statistically significant difference among the different hospital groups with ANOVA (α =0.05), and the number of differences by using multiple posteriori comparison SUN et al. Archives of Public Health (2021) 79:179 Page 7 of 14 resources reform has improved the image of low-level Regarding the other exogenous latent variables, scores hospitals. However, respondents reported average scores for trust, accessibility, capability, and convenience are between 2 = “low” and 3 = “fair” for awareness of de- significant higher for secondary hospitals compared with scending high-level hospitals/doctors. In contrast, acces- primary hospitals. In terms of socio-demographics vari- sibility of the descending doctors reached (3.10 ± 1.10, ables, the education level of patients visiting tertiary hos- where 3 = “positive” and 4 = “high”), indicating that re- pitals (3.34 ± 1.05) is significantly higher than those of spondents found the descending doctors easy to access. patients attending primary and secondary hospitals; no The variables of environment, capability, and conveni- significant difference was found in education levels of ence for local low-level hospitals all have means between patients visiting the latter two. Accordingly, the reform 3.2 and 3.4 (3 = “positive” and 4 = “high”), suggesting that awareness of patients in tertiary hospitals is significantly the reform had a positive impact from the patients’ per- higher than that among patients attending other levels spective. In addition, the response for medical cost is of hospitals; however, their evaluation of reform-related (2.80 ± 0.99, where 2 = “no change” and 3 = “slight de- policies like medical service prices and insurance reim- crease”), indicating that the descending resources reform bursement amounts is significantly lower. has lowered medical costs in general. For the reform policy latent variable, 64% of patients SEM estimation results of patient satisfaction obtain information from public channels such as news- The above discussion has offered some preliminary in- papers, television, and the hospital, with a mean of vestigations into the effect of the descending resources (2.36 ± 1.29, where 2 = “low” and 3 = “fair”). However, reform on patient responses to the reform, but more evi- reform-related policies like medical service prices, (dif- dence is needed to understand the marginal effect of dif- ferential) insurance reimbursement levels, and tiered ferent latent variables. Therefore, AMOS 21.0 software medical services are evaluated high, with an average of is used to test the theoretical model established in Fig. 1. 3.2–3.4. This section reports the results of the patient satisfaction The ANOVA results in Table 2 indicate that except model. Following an iteration procedure using the boot- for the variables of medical cost, environment, informa- strap method, the resulting path diagram of the SEM tion channel, and age, the variables all show significant model is shown in Fig. 3. Perceived quality, patient ex- differences among different levels of hospitals (α = 0.05). pectations, and hospital image all have a significant posi- For information channel groups, except for socio- tive impact on patient satisfaction, and their normalized demographics, the public channel scores of other vari- path coefficients of 0.577, 0.711, and 1.014, respectively, ables are significantly higher than those for the private indicate that the ECSI model can better explain the fac- channel group; in particular, scores for getting informa- tors influencing patient satisfaction. Overall, this model tion through public channels, hospital selection (loyalty), is significant. At the same time, the coefficient of the re- and satisfaction are significantly higher. form policy is 0.140, which confirms the reform’s posi- The results of multiple comparisons show that post re- tive effect on patient satisfaction. form, patients’ satisfaction in tertiary hospitals is signifi- The relationships between the variables and estimated cantly higher than that in primary and secondary coefficients are reported in Table 3. It can be seen that hospitals, but no significant difference exists between the effects of socio-demographics on patient satisfaction primary and secondary hospitals. In addition, no signifi- do not pass the significance test, but other latent vari- cant difference exists in patient satisfaction with low- ables are significant at α = 1%, which demonstrates the level hospitals between tertiary and other low-level hos- existence of a causal relationship between the variables. pitals, whereas patient satisfaction in secondary hospitals This in turn indicates that the structural equation model is significantly higher than that in primary hospitals. of patient satisfaction is appropriate. In this model, hos- The results for the two hospital selection (loyalty) variables pital image exerts the biggest influence on patient satis- are different: (1) although patients visiting a tertiary hospital faction, which in turn mainly originates from the trust in have already chosen a high-level hospitals, their willingness local low-level hospitals and the brand implantation of to select a low-level hospitals is not significantly different cooperative high-level hospitals. from patients who attend other level hospitals; (2) those visit- Next, the confirmatory factor analysis method is used ing secondary hospitals reported a higher score for this vari- to conduct the single-factor structural validity analysis able than those visiting primary hospitals; and (3) when a (Table 4). This analysis also serves as a reliability evalu- patients suffers a serious illness, their loyalty to their local ation for the measurement model. The findings indicate low-level hospitals is significantly lower if their most recent that the coefficients for all latent variables except socio- hospital visit was to a primary hospital rather than a second- demographics are significant at α = 1%. This demon- ary or tertiary hospital, but no significant difference was strates that the potential factor structure of the ques- found between the latter two. tionnaire items is reasonable. Several other findings can SUN et al. Archives of Public Health (2021) 79:179 Page 8 of 14 Fig. 3 Patient satisfaction structural equation model path diagram. Source: The authors be drawn from the analysis. First, for the measurement with reform awareness, thus have a strong explanatory variables of perceived quality, the factor loading coeffi- power for the reform latent variable. cients of environment and convenience are quite close, indicating their important explanatory power on pa- SEM estimation results of patient hospital selection tients’ perceptions of quality. Second, for patient expec- (loyalty) tations, the factor loading coefficients of capability of Figure 4 shows the path diagram of the structural equa- low-level hospitals is the highest (=1.285), followed by tion model for patient hospital selection (loyalty). The medical cost and accessibility of descending doctors. normalized path coefficients of the descending resources Third, for hospital image, the factor loading coefficients reform and patient satisfaction to loyalty are 0.450 and of patients’ awareness of cooperative high-level hospitals 0.731, respectively (Table 5). These two latent variables and descending doctors reached 2.295 and 2.444, re- are significant at α = 1%, indicating that a causal rela- spectively, which are far higher than the score for the tionship between variables can be established, and there- variable of trust (=1.000). Finally, for the latent variable fore, this structure equation model of patient hospital of the reform policy, the factor loading coefficients of selection is appropriate. It can be seen that the impact of (differential) insurance imbursement levels and medical patient satisfaction on loyalty is greater than that of the service prices are high (> 1), and the coefficient of tiered reform policy, which indicates patients’ LLH choice is medical services reaches 0.933; these variables, together Table 3 SEM estimation results of patient satisfaction model Number Relationship Normalized path coefficient Standard deviation C.R. value P values 1 Satisfaction←Perceived quality 0.577 0.051 11.246 *** 2 Satisfaction←Patient expectation 0.711 0.071 10.051 *** 3 Satisfaction←Hospital image 1.014 0.084 12.099 *** 4 Satisfaction←Reform policy 0.140 0.040 3.521 *** 5 Satisfaction←Socio-demographics 0.252 0.383 0.658 0.511 Note: *** indicates significance level of α =1% SUN et al. Archives of Public Health (2021) 79:179 Page 9 of 14 Table 4 Confirmatory factor analysis results of patient satisfaction model Number Relationship Factor loading coefficient standard deviation C.R. value P values 1 Reform satisfaction←Satisfaction 1.000 –– – 2 LLH satisfaction←Satisfaction 0.607 0.027 22.134 * * * 3 Convenience←Perceived quality 1.000 –– – 4 Environment←Perceived quality 1.004 0.066 15.197 * * * 5 Accessibility←Patient expectation 1.000 –– – 6 Capability←Patient expectation 1.285 0.090 14.200 * * * 7 Medical cost←Patient expectation 1.102 0.077 14.375 * * * 8 Trust←Hospital image 1.000 –– – 9 Awareness for high-level hospital←Hospital image 2.295 0.116 19.748 * * * 10 Awareness for descending doctors←Hospital image 2.444 0.125 19.536 * * * 11 Reform awareness← Reform policy 1.000 –– – 12 Information channel←Reform policy 0.228 0.023 9.728 * * * 13 Medical service price←Reform policy 1.039 0.060 17.257 * * * 14 Insurance imbursement←Reform policy 1.114 0.065 17.254 * * * 15 Tiered medical service←Reform policy 0.933 0.057 16.326 * * * 16 Gender←Socio-demographis 1.000 –– – 17 Age←Socio-demographis 16.821 6.333 2.656 0.008 18 Education level←Socio-demographis 13.768 4.853 2.837 0.005 Note: *** indicates significance level of α =1% more affected by their own evaluations of the reform variables of loyalty have close factor loading coefficients and local low-level hospitals. (1.000 and 0.963), showing their strong explanatory The confirmatory factor analysis results for the hos- power. Satisfaction with the reform and LLHs also have pital selection (loyalty) model are reported in Table 6. similar factor loading coefficients (1.000 and 1.054). The coefficients of different socio-demographics are still In terms of the measurement variables of the reform significant at α = 1%, but this latent variable itself does policy, the factor loading coefficients of (differential) in- not have a significant impact on loyalty according to the surance imbursement levels and medical service prices results reported in Table 5. The two measurement are still greater than 1 (=1.119 and 1.028), which shows Fig. 4 The path diagram of patient hospital selection (loyalty) . Source: The authors SUN et al. Archives of Public Health (2021) 79:179 Page 10 of 14 Table 5 SEM estimation results of patient hospital selection (loyalty) model Number Relationship Normalized path coefficient Standard deviation C.R. value P values 1 Loyalty←Satisfaction 0.731 0.052 14.041 * * * 2 Loyalty←Reform policy 0.450 0.043 10.570 * * * 3 Loyalty←Socio-demographics 0.034 0.380 0.091 0.928 Note: *** indicates significance level of α =1% their important impact on perceptions of the descending insignificant. Third, socio-demographics still have an in- resources reform. The factor loading coefficients of re- significant effect on patient satisfaction. Finally, sub- form awareness and tiered medical services also reached sample results for patient hospital selection (loyalty) 1.000 and 0.933, showing their explanatory power of ac- show strong robustness in terms of satisfaction and the ceptance of the reform. reform’s positive effect, as well as the insignificant effect of socio-demographic variables. Thus, it can be con- cluded that the subsample empirical analysis supports Robustness test the estimation results of the full-sample SEM model. In order to test the robustness of the above full-sample results, we use different-level hospital subsamples to per- form empirical analyses. The sample sizes of different Mediation effects estimation result level hospitals meet the requirements of carrying out In order to estimate the impact of the reform latent vari- SEM model estimation. able on hospital selection (loyalty) through satisfaction, Table 7 reports the estimation results of different sub- we use the method shown in Fig. 2 to perform the test samples. It can be seen that, firstly, the impacts of per- the mediating effect (Table 8). It can be seen that β is ceived quality, patient expectations, and hospital image significant but its absolute value is less than β in the on patient satisfaction are confirmed to be positive again full-sample estimation, which confirms the existence of for the tertiary and secondary hospital subsamples at the a partial mediating effect. This finding indicates that the 1% significance level. However, for the primary hospital reform latent variable has an indirect impact on hospital subsample, the impact of patient expectations is slightly choice behavior through satisfaction, but it also directly insignificant (P = 0.072). Second, for tertiary and second- affects hospital choice. In order to verify the robustness ary hospitals subsamples, the positive impact of the re- of this result, we also perform hospital subsample esti- form latent variable on patient satisfaction is significant mations, the results of which show that heterogenous re- at α = 5%, although it is insignificant for the primary sults exist for different hospital subsamples. The results hospital subsample. This makes sense because doctors for the tertiary and secondary hospitals subsamples show descending from high-level hospitals in Zhejiang Prov- a complete mediating effect, which is consistent with the ince only descend to secondary hospitals, so this reform previous literature on the impact of satisfaction upon [18.19] latent variable’s impact on primary hospitals is loyalty implied in ECSI model ; however, no Table 6 Confirmatory factor analysis results of patient hospital selection (loyalty) model Number Relationship Normalized path coefficient Standard deviation C.R. value P values 1Y1← loyalty 1.000 –– – 2Y2← loyalty 0.963 0.027 35.719 * * * 3 Reform satisfaction←Satisfaction 1.000 –– – 4 LLH satisfaction←Satisfaction 1.054 0.075 14.000 * * * 5 Reform awareness←Reform policy 1.000 –– – 6 Information channel←Reform policy 0.219 0.023 9.7375 * * * 7 Medical service price ←Reform policy 1.028 0.060 17.165 * * * 8 Insurance imbursement←Reform policy 1.119 0.064 17.399 * * * 9 Tiered medical service←Reform policy 0.963 0.058 16.552 * * * 10 Gender←Socio-demographics 1.000 –– – 11 Age←Socio-demographics 16.773 6.378 2.630 0.009 12 Education level←Socio-demographics 13.699 4.823 2.840 0.005 Note: *** indicates significance level of α =1% SUN et al. Archives of Public Health (2021) 79:179 Page 11 of 14 Table 7 Subsample estimation results of patient satisfaction and hospital selection (loyalty) models Number Relationship Subsample Normalized path coefficient Standard deviation C.R. value P values 1 Satisfaction← Tertiary 0.335 0.087 3.836 * * * Perceived quality Secondary 0.313 0.039 8.085 * * * Primary 0.488 0.119 4.089 * * * 2 Satisfaction← Tertiary 0.271 0.100 2.708 0.007 Patient expectation Secondary 0.502 0.058 8.603 * * * Primary 0.258 0.144 1.799 0.072 3 Satisfaction← Tertiary 0.456 0.080 5.718 * * * Hospital image Secondary 0.571 0.047 12.051 * * * Primary 1.120 0.305 3.667 * * * 4 Satisfaction← Tertiary 0.267 0.070 3.807 * * * Reform policy Secondary 0.051 0.026 2.004 0.045 Primary 0.101 0.116 0.870 0.385 5 Satisfaction← Tertiary 0.073 0.406 0.179 0.858 Socio-demographics Secondary 0.099 0.276 0.358 0.720 Primary 0.677 1.309 0.517 0.605 6 Loyalty← Tertiary 0.540 0.128 4.214 * * * Satisfaction Secondary 0.615 0.058 10.632 * * * Primary 0.885 0.199 4.448 * * * 7 Loyalty← Tertiary 0.599 0.120 5.000 * * * Reform policy Secondary 0.379 0.040 9.372 * * * Primary 0.784 0.335 2.339 0.019 8 Loyalty← Tertiary 1.344 1.147 1.172 0.241 Socio-demographics Secondary 0.157 0.415 0.378 0.706 Primary 1.874 8.400 0.223 0.823 Note: *** indicates significance level of α =1% mediating effect is found for the primary hospital sub- sample, which may be related to the fact that the de- scending resources reform in Zhejiang mainly involves secondary and tertiary hospitals but not primary hospitals. Table 8 Mediation effects estimation result of reform latent Discussion variable China’s descending resources reform offers a unique ap- Exogenous variable: Reform; Mediating variable: Satisfaction; proach to overcoming the uneven allocation of health Exogenous variable: Hospital selection (Loyalty) resources. It works by utilizing the dominant role of the Sample β β β β Conclusion a b c c public hospital system in the Chinese health market Full-sample 0.855*** 1.579*** 0.752*** −0.683* Partial compared with other developing countries. Using ques- mediating effect tionnaire data, the expanded ECSI model, and the struc- ture equation model, we found that the descending Hospital Tertiary 0.758*** 1.148** 0.470*** −0.144 Complete subsample mediating resources reform had a significantly positive impact on effect patients’ satisfaction and their selection of local low-level Secondary 0.846*** 1.439*** 0.760*** −0.535 Complete hospitals. It was further found that the measurement mediating variables of perceived quality, patient expectations, and effect hospital image can also explain patient behavior. Primary 1.010*** 1.439 0.778*** 0.945 No ANOVA and multiple comparison techniques demon- mediating effect strated that significant differences exist among patients Note: ***and** indicate significance level of α =1 and 5%, respectively at different levels of hospitals. SUN et al. Archives of Public Health (2021) 79:179 Page 12 of 14 Thus, it is found that the descending resources reform secondary hospital evaluating them significantly higher contributed to the improvement of convenience, capabil- than those visiting a tertiary hospital, again indicating ities, and the environment of local low-level hospitals, the focus and main beneficiaries of this reform being offering evidence of the effect of this reform from the secondary hospitals. Meanwhile, such polices contribute patients’ perspective. This paper’s SEM estimations also to explaining the reform’s latent variables well and have confirm the significantly positive impacts of perceived significantly positive effects on patient satisfaction and quality, patient expectations, and hospital image on pa- loyalty. These results suggest that the above supporting tient satisfaction. Among these, the treatment/diagnosis policies can help reduce the burden of medical costs in- capabilities of low-level hospitals were found to have the curred by patients in low-level hospitals, and their satis- strongest explanatory power for patient expectations, in- faction with and loyalty to low-level hospitals can be dicating that LLH capabilities are a top factor shaping improved via financial incentives [39]. patients’ expectations, and this issue is also the core The survey results also show the coexistence of low re- focus of the descending resources reform. form satisfaction and high satisfaction with local low- Among the different latent variables, hospital image level hospitals, where patients visiting tertiary hospitals exerts the greatest influence. Of the three measurement have higher reform satisfaction, whereas those visiting variables, i.e., trust in low-level hospitals and awareness secondary hospitals have higher satisfaction with local of cooperative high-level hospitals and descending doc- low-level hospitals than those visiting primary hospitals. tors, the confirmatory factor analysis suggested that the Correspondingly, the reform awareness of patients visit- latter two have higher explanatory power than trust in ing tertiary hospitals is also significantly higher than low-level hospitals. This result shows that embedding those visiting other levels of hospitals. This could reflect the brand image of high-level hospitals and doctors into the fact that one of the reform’s main goals was to re- low-level hospitals is an effective way to improve the duce congestion in high-level hospitals by reallocating image of low-level hospitals, and the role of brand im- health resources to different levels of hospitals. Thus, pa- plantation is greater than that of the image (trust) of tients visiting high-level hospitals should have a real un- low-level hospitals from the patient perspective. Further- derstanding of the reform’s effects, benefit from the more, improving the image of low-level hospitals can reform, and have higher satisfaction with reform. Mean- also significantly contribute to higher patient satisfaction while, in this reform, descending doctors mainly flow and loyalty to low-level hospitals, which is fully consist- into secondary hospitals, which receive the greatest ent with the goal of the descending resources reform. amount of resources inflow and capability improve- This study also confirmed the significantly positive im- ments, thus contributing to their patients’ satisfaction pact of the reform and satisfaction on patients’ hospital compared with those visiting primary hospitals. selection (loyalty) of low-level hospitals and the exist- It is worthy pointing out the importance of informa- ence of mediating effects of satisfaction. Previous litera- tion channels on the reform’s effect. The ANOVA re- ture has found that medical service quality affects sults indicate that reform awareness and patient patient satisfaction, which in turn affects their behavioral behavior are related to the information channel. Infor- choices [37]; high service satisfaction has a significantly mation access via public channels can avoid information positive impact on customer loyalty [38]. The empirical transmission distortion and thus help promote a positive results on the mediating effect confirm that in secondary response from patients. Although according to the SEM and tertiary hospitals, the reform latent variables affect estimation, information channels do not have a strong patients’ hospital choice behavior through the mediating explanatory power for the reform latent variable, the factor of satisfaction. However, such effects disappear for ANOVA results showed that scores of perceived quality, primary hospitals, a difference that highlights the reality patient expectation, hospital image, satisfaction and hos- that descending resources reform in Zhejiang focused pital selection (loyalty), are all significantly higher for the more on secondary and tertiary hospitals. This paper public channel subsample than the private channel sub- provides new evidence based on patient behavior in the sample, indicating that a more unbiased information context of China’s latest healthcare reform, and evidence supply and effective transmission are essential in enhan- for the appropriation of expanded ECSI model used in cing patients’ positive responses and reshaping their hos- this paper. pital choices. Zhejiang’s reform is characterized by descending doc- tors from high-level hospitals to low-level hospitals, to- Conclusion gether with important supporting policies like medical China’s reforms carried out since 2003 have offered evi- service prices, differential insurance imbursement levels, dence that traditional approaches focusing on higher and tiered medical services. These policies all received public health expenditure and expansion of medical in- higher scores from patients, with patients visiting a surance coverage are inadequate in correcting the SUN et al. Archives of Public Health (2021) 79:179 Page 13 of 14 uneven allocation of health resources. The descending interpreted data; ZD and SJ performed the empirical work with software. All authors read and approved the final manuscript. resources reform launched in 2013 is an attempt to im- prove the capability of low-level hospitals and re-attract Funding patients by using human capital spillovers of doctors de- The authors thank for the financial support from National Social Science Fund of China (20NDJC244YB), Soft Science Fund of Zhejiang Province scending from high-level hospitals together with brand (2018C35028) and Development Research Center of Shanghai(2020-YJ-C05). implantation of these high-level hospitals. Using ques- tionnaire survey data collected from patients in Zhejiang Availability of data and materials The datasets during the current study available from the corresponding Province, this paper provides supporting empirical evi- author on reasonable request. dence of the reform’s impact on patient behavior. The results indicate that the reform has been effective in im- Declarations proving the capabilities of low-level hospitals, and brand Ethics approval and consent to participate implantation of high-level hospitals shows strong ex- Not applicable. planatory power. The findings also suggest that policy- makers could pay more attention to the importance of Consent for publication Not applicable. information channels in impacting patient awareness, re- sponses, and hospital selection. Competing interests For developing countries where public hospitals play a The authors declare that they have no competing interests. dominant healthcare role, China’s reform offers a dis- Author details tinct and valuable approach to correcting the uneven al- School of Finance and Business, Shanghai Normal University, Shanghai, location of health resources. The approach indicates that China. Department of Stomatology, Tongde Hospital of Zhejiang Province, Hangzhou, China. Department of Stomatology, Songjiang Hospital of greater investment and demand-side reforms might not Shanghai, Shanghai, China. School of Finance, Zhejiang Gongshang necessarily incentivize patients to respond as the govern- University, Hangzhou, China. ment intends. In addition to affordability, building the Received: 31 March 2021 Accepted: 28 September 2021 capabilities of hospitals in rural/remote areas remains a key factor, which can be done by cooperation between high- and low-level hospitals, human capital spillovers, References and brand implantation. Meanwhile, some demand-side 1. Gwatkin D. Trends in health inequalities in developing countries. Lancet Glob Health 2017; 5(4): e371–e372. https://doi.org/https://doi.org/10.1016/ policies like differential medical insurance and price reg- S2214-109X(17)30080-3. ulations could also create incentives for patients to re- 2. Strasser R. Rural health around the world: challenges and solutions. Fam shape their hospital choices so as to resolve the existing Pract 2003; 20(4): 457–463. https://doi.org/https://doi.org/10.1093/fampra/ cmg422. structural congestion or uneven allocation problems. In 3. Sun ZS, Wang SH, Barnes SR. Understanding congestion in china’s medical addition, access to full and unbiased information is of market: an incentive structure perspective. Health Policy Plann 2016; 31(3): great importance in realizing the full effect of these re- 391–400. https://doi.org/https://doi.org/10.1093/heapol/czv062. 4. Zhang L, Cheng G, Song S, Yuan B., Zhu W., He L., Ma X., Meng Q. Efficiency forms. Therefore, the findings of this paper highlight performance of china’s health care delivery system. Int J Health Plann Manag that the government should pay attention to public in- 2017; 32: 254–263. https://doi.org/https://doi.org/10.1002/hpm.2425,3. formation channels in improving the effect of healthcare 5. Cai Y, Gong G, Zhang Q. Getting rid of the medicine-selling dependent hospital by value returning from the demonstration of a theoretical model reforms. Besides, after the data collection of this paper, (in Chinese). Nankai Econ Res. 2013;1:40–52. the outbreak of COVID-19 pandemic encourages pa- 6. Tao C, Guo T, Xu K. Measure of equalization level of medical and health tients to choose local low-level hospitals more so as to resources allocation in China (in Chinese). Statist Decis-making. 2019;24:42–6. 7. Jiang S, Min R, Fang P. The impact of healthcare reform on the efficiency of avoid travelling and infection risk to high-level hospitals, public county hospitals in China. BMC Health Serv Res 2017; 17(1): 838. so this phenomenon could enhance the existing effect of https://doi.org/https://doi.org/10.1186/s12913-017-2780-4. the above reform on patients’ choice to low-level hospi- 8. Meng Q, Mills A, Wang L, et al. What can we learn from China’s health system reform? Br Med J 2019; 365: 3–7. https://doi.org/https://doi.org/1 tals, and it does not harm the applicability and useful- 0.1136/bmj.l2349. ness of our findings. 9. Huang G, Qiu Y. A study on the structural reform of the supply side of the public hospitals at county level (in Chinese). Soc Secur Res. 2018;1:50–7. Abbreviations 10. Zhejiang Province Government. Implementation Opinions on Promoting ECSI: The European consumer satisfaction index model; ANOVA: Analysis of the Construction of Long-term Mechanism for Reform of Double variance; SEM: Structure equation model; SARS: Severe acute respiratory Descendings and Two Promotions. 2015. [Accessed 2019 May 12] http:// syndrome; KMO: Kaiser-Meyer-Olkin test; OLS: Ordinary least square; www.zj.gov.cn/art/2015/9/22/art_32431_245673.html. OLM: Ordered logit model 11. Valls Martinez M del C, Ramirez-Orellana a. patient satisfaction in the Spanish national health service: partial least squares structural equation Acknowledgements modelling. Int J environ res pub Health 2019; 16: 4886. https://doi.org/ Not applicable. https://doi.org/10.3390/ijerph16244886, 24. 12. Fatima T, Malik S A, Shabbir A. Hospital healthcare service quality, patient Authors’ contributions satisfaction and loyalty: an investigation in context of private healthcare SZ provides the conception and design of the work. WS contributes to the systems. Int J Qual Reliab Manage 2018; 35(6): 1195–1214. https://doi.org/ data collection, analysis, and drafted the work; ZH and ZX analyzed and https://doi.org/10.1108/IJQRM-02-2017-0031. SUN et al. Archives of Public Health (2021) 79:179 Page 14 of 14 13. Fornell C. A national customer satisfaction barometer: the Swedish 36. Baron RM, Kenny DA. The moderator-mediator variable distinction in social experience. J Marketing 1992; 56: 6–21. https://doi.org/https://doi.org/10.11 psychological research: conceptual, strategic and statistical considerations. J 77/002224299205600103,1. Pers Soc Psychol 1986; 51: 1173–1182. https://doi.org/https://doi.org/10.103 14. Fornell C, Johnson MD, Anderson EW, Cha J, Bryant BE The American 7//0022-3514.51.6.1173,6. consumer satisfaction index: nature, purpose, and findings. J Marketing 1996; 37. Naidu A. Factors affecting patient satisfaction and healthcare quality. Int J 60: 7–18. https://doi.org/https://doi.org/10.1177/002224299606000403,4. Health Qual Assur 2009; 22(4): 366–381. https://doi.org/https://doi.org/10.11 08/09526860910964834. 15. Brady MK, Cronin JJ. Some new thoughts on conceptualizing perceived 38. Rust RT, Zahorik AJ. Customer satisfaction, customer retention, and market service quality: a hierarchical approach. J Marketing 2001;, 65: 34–49. https:// share. J Retail 1993; 69: 193–215. https://doi.org/https://doi.org/10.1016/ doi.org/https://doi.org/10.1509/jmkg.65.3.34.18334,3. 0022-4359(93)90003-2,2. 16. Johnson MD, Gustfasson A, Andreassen W, et al. The evolution and future of 39. Liu D. China’s new rural cooperative medical scheme: evolution, design and national customer satisfaction index models. J Econ Psychol 2001; 22(2): impacts. International Verlag der Wissenschaften: Frankfurt am Main, 217–245. https://doi.org/https://doi.org/10.1016/S0167-4870(01)00030-7. Germany; 2013. 17. Bayol M, Foye A, Tellier C, et al. Use of PLS path modeling to estimate the European consumer satisfaction index (ECSI) model. Stat Applicata. 2000; 12(3):361–75. Publisher’sNote 18. Caruana A. Service loyalty: the effects of service quality and the medicating Springer Nature remains neutral with regard to jurisdictional claims in role of customer satisfaction. Eur J Mark 2002; 36(7/8): 811–827. https://doi. published maps and institutional affiliations. org/https://doi.org/10.1108/03090560210430818. 19. Fullerton G, Taylor S. Mediating, interactive, and nonlinear effects in service quality and satisfaction with services research. Can J Admin Sci 2002; 19(2): 124–136. https://doi.org/https://doi.org/10.1111/j.1936-4490.2002.tb00675.x. 20. Strasser S, Aharony L, Greenberger D. The patient satisfaction process: moving toward a comprehensive model. Med Care Res Rev 1993; 50: 219– 248. https://doi.org/https://doi.org/10.1177/107755879305000205,2. 21. Banka G, Edgington S, Kyulo N, Padilla T., Mosley V., Afsarmanesh N., Fonarow G.C., Ong M.K. Improving patient satisfaction through physician education, feedback, and incentives. J Hosp Med 2015; 10(8): 497–502. https://doi.org/https://doi.org/10.1002/jhm.2373. 22. Jiao W, Jiang H, Yu M, et al. Investigation and analysis of inpatient satisfaction in a third-level hospital in Beijing (in Chinese). J Nurs Adm. 2010; 10(9):630–2. 23. Ma P. Inpatient satisfaction factors and countermeasures (in Chinese). Med Inf Med Surg. 2009;22(3):259–60. 24. Pan J, Liu D, Ali S. Patient dissatisfaction in China: what matters. Soc Sci Med 2015; 143: 145–153. https://doi.org/https://doi.org/10.1016/j. socscimed.2015.08.051. 25. Sun Z, Wang S, Zhao H, Yu H Does descending resources reform improve patient satisfaction and reshape choice of care providers. Inquiry 2020; 57: 1–13. https://doi.org/https://doi.org/10.1177/0046958020956899. 26. Senarath U, Fernando D, Rodrigo I. Factors determining client satisfaction with hospital-based perinatal care in Sri Lanka. Tropical Med Int Health 2006; 11(9): 1442–1451, https://doi.org/https://doi.org/10.1111/j.1365-3156.2 006.01698.x. 27. Ahmed S, Tarique K, Arif I. Service quality, patient satisfaction and loyalty in the Bangladesh healthcare sector. Int J Health Care Qual Assur 2017; 30(5): 477–488. https://doi.org/https://doi.org/10.1108/IJHCQA-01-2017-0004. 28. Love-Kol J, Griffin S, Kataika E, et al. Methods to promote equity in health allocation in low- and middle-income countries: an overview. Glob Health 2020; 16: 6. https://doi.org/https://doi.org/10.1186/s12992-019-0537-z,1. 29. Zhang LD, Liu L, Nan DH, et al. The descending resources reform and doctors’ response: an SEM estimation based on Shanghai survey data. J Fujian Medical Univ (Soc Sci). 2021;22(3):1–17. 30. Vinagre M, Neves J. The influence of service quality and patients’ emotions on satisfaction. Int J Health Care Qual Assur 2008; 21(1): 87–103. https://doi. org/https://doi.org/10.1108/09526860810841183. 31. Alrubaiee L, Alkaa'ida F. The mediating effect of patient satisfaction in the patients’ perceptions of healthcare quality-patient trust relationship. Int J Mark Stud 2011; 3(1): 103–127. https://doi.org/https://doi.org/10.5539/ijms.v3 n1p103. 32. Donabedian A. The quality of care: how can it be assessed? J Am Med Ass 1988; 260: 1743–1748. https://doi.org/https://doi.org/10.1001/jama.260.12.1 743, 12. 33. Dwyer RF, Schurr PH, Oh S. Developing buyer-seller relationship. J Marketing 1987; 51: 11–27. https://doi.org/https://doi.org/10.1177/0022242987051002 02,2. 34. Clavolino E, Dahlgaard JJ. ECSI- customer satisfaction modeling and analysis: a case study. Total Qual Manage. 2007;18(5):545–54. https://doi.org/10.1 080/14783360701240337. 35. Qiu H, Lin B. The theory and its application of structure equation model. Beijing, China: China Light Industry Press; 2012. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Archives of Public Health Springer Journals

Does descending health resources reform impact patient low-level hospital selection behavior? Evidence from Zhejiang, China

Loading next page...
 
/lp/springer-journals/does-descending-health-resources-reform-impact-patient-low-level-YnBY78JkRb

References (74)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
eISSN
2049-3258
DOI
10.1186/s13690-021-00700-6
Publisher site
See Article on Publisher Site

Abstract

Background: Since 2013, China launched descending resources reform, which is a new attempt to correct unbalanced allocation of health resources through human capital spillovers and brand implantation from high-level hospitals. The purpose of this paper is to explore the patients’ hospital selection response to this reform with the focus of low-level hospitals to better understand the effect of this reform on correcting regional inequality of health resources allocation. Methods: The European Consumer Satisfaction Index model (ECSI) was used to design a questionnaire, and cross- sectional data from 17 hospitals were collected through 1287 questionnaires from Zhejiang Province. Patient hospital selection (loyalty) is measured using ordinary variables by considering patient willingness to choose a low-level hospital when suffering an illness or severe illness. Analysis of variance (ANOVA) and the structure equation model are applied to examine the effect of reforms on patient behavior. Results: The descending resources reform promotes improvements in the capabilities and medical environment of low-level hospitals, and descending doctors also have high accessibility. Perceived quality, patient expectations, and hospital image have significant positive effects on patient satisfaction, and the explanatory power of brand implantation from cooperative high-level hospitals and descending doctors is stronger than the image of the low-level hospital itself. And descending resources reform and patient satisfaction have significant positive impacts on patient’s choice for low-level hospitals with the existence of mediating effect of satisfaction. Conclusions: This paper provides supporting empirical evidence of the descending resources reform’simpact on patients’ low-level hospital selection. This reform has been effective in improving the capabilities of low-level hospitals, and brand implantation of high-level hospitals shows strong explanatory power. China’s reform offers a distinct and valuable approach to correcting the uneven allocation of health resources. Besides, the findings also suggest that policymakers could pay more attention to the importance of information channels in impacting patient awareness, responses, and hospital selection. Keywords: Descending resources reform, Low-level hospitals, Patient satisfaction, Consumer behavior, Hospital selection behavior * Correspondence: wangshh@126.com Department of Stomatology, Tongde Hospital of Zhejiang Province, Hangzhou, China Department of Stomatology, Songjiang Hospital of Shanghai, Shanghai, China Full list of author information is available at the end of the article © The Author(s). 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. SUN et al. Archives of Public Health (2021) 79:179 Page 2 of 14 Background regional equality. One new solution for this problem is Regional inequality of health resources allocation is a the descending resources reform, introduced in Zhejiang global concern, and a key focus is access issues between and other provinces in China in 2013. The core idea of urban and rural areas [1]. Rural areas typically face the this reform is to encourage high-level hospitals to estab- challenge of doctor shortages due to the difficulties with lish cooperative ties with low-level hospitals, thus driving transport and communications that exist in most devel- the high-level hospitals’ human capital to transfer down oping and developed countries [2]. China faces a differ- (i.e., descend) to the low-level hospitals. As part of this ent constraint in terms of its structural congestion policy, the government would provide a subsidy to (at between overcrowded (city) high-level hospitals and idle least partially) compensate hospitals for the reform costs. (county and town) low-level hospitals due to patients’ Overall, the policy aims are to (1) narrow the human biased behavior, which motivates them to choose high- capital gaps among hospitals via spillover effects, and (2) level hospitals [3]. This dynamic stems from China’s imbed (brand) the image of a high-level hospital on a long-lasting price regulation and health resource con- low-level one, which could then help reshape patients’ centration in (city) high-level hospitals, and it generates behavior in terms of hospital selection, with an emphasis medical cost and efficiency losses as well as doctor–pa- on low-level hospitals [10]. This reform makes full use tient conflicts and detrimental social consequences [4]. of the dominant role of China’s public hospital system, Since 2003, following the SARS crisis, the Chinese gov- but there is still little empirical evidence on this reform’s ernment has paid increasing attention to investment in effect on patients’ care-provider choices. This paper ex- the infrastructure of (mainly town) low-level hospitals. plores this issue by using a structural equation model Policies such as higher government health expenditure, (SEM) based on questionnaire data collected in Zhejiang, expansion of medical insurance coverage, and abolition China. of marked-up drug prices have been implemented since Patient hospital choices can be viewed as reflecting pa- 2009. However, increasing health care affordability due tient loyalty to different care providers, and in the con- to expanded medical insurance coverage only worsened text of China’s public hospital system, these choices can the structural congestion [5, 6]; the efficiency of low- be viewed as patient loyalty to different levels of hospital. level hospitals has yet not improved [7]. Different from the traditional literature on patient satis- In China, the dominant public hospital system divides faction [11, 12], this paper uses the term “patient satis- hospitals into three different levels, from first to third. faction with the reform” to measure patients’ response All hospitals run by MOH (Ministry of Health of China) for the reform and its impact on low-level hospitals. This and nearly all hospitals run by provinces and major cities satisfaction would thus reflect patients’ loyalty to and were approved to be ranked at the third level, and are thus their choice behavior in terms of choosing low-level thus viewed as high-level hospitals. Community and hospitals. township hospitals were identified as first level, and dis- In the marketing science literature, a satisfaction index trict- or county-level hospitals were generally approved model can be used to discuss a reform’s effect on patient to be second level; these hospitals are usually viewed as satisfaction and loyalty. The Swedish Customer Satisfac- low-level hospitals with community and township ones tion Barometer (SCSB model) emphasizes the determi- being the lowest. nants of two antecedent factors: customer expectations Previously, the guiding concept and perception was and perceived performance; customer satisfaction then that low-level hospitals could only diagnose and treat a affects customer complaints and ultimately impacts cus- limited range of ailments and offer a lower standard of tomer loyalty [13]. The American Consumer Satisfaction care than high-level hospitals. Presently, underutilization model (ASCI) proposed by Fornell et al. [14] adds the la- and the low capabilities of low-level hospitals are still se- tent variable of perceived quality, but still uses perceived vere challenges for China [8]. These issues reflect the value to measure perceived performance. Brady and Cro- fact that previous investment focused on fixed assets ra- nin [15] emphasized service quality evaluations based on ther than human capital [3]. This kept low-level hospi- the dimensions of results, interaction quality, and phys- tals at a disadvantage in attracting patients due to the ical environment quality, similar to the ACSI model. differences in human capital between different levels of hospitals coupled with regulated medical service prices. The uneven allocation of health resources accelerated The authors thank one anonymous reviewer’s very helpful comment after 2009 due to reforms that regulated prices and ex- for this term difference. The traditional literature usually discusses individual-level response on health service through indicators of the panded insurance coverage, which then enabled more physical environment, patient-friendly environment, response capacity, patients to afford high-level hospitals [9]. communication, privacy, and security. Some other factors, like know- This history indicates that past demand-side and ledge of history, monitoring of health problems, and information, are supply-side reforms were unsuccessful in realizing also discussed. SUN et al. Archives of Public Health (2021) 79:179 Page 3 of 14 The European Consumer Satisfaction Model (ECSI) hospital (LLH) to measure patients’ hospital choice, initiated by the European Commission in 1999 removed which is affected by patient satisfaction [19, 30] and two the latent variable of customer complaint from the ACSI exogenous variables. Satisfaction is affected by three la- and SCSB models because complaint processing has no tent variables: perceived quality, consumer expectations, significant impact on customer satisfaction or loyalty in and hospital image. According to the ECSI model, the empirical research [16]. The ECSI model includes cor- difference between consumer expectations and perceived porate image, a move intended to incorporate customers’ performance is expressed as the expected value, but memory associated with organizations [17]; and satisfac- technical reliability and treatment effects of medical ser- tion mediates between service quality and loyalty [18, vices are difficult for patients to evaluate [31], so the 19]. Meanwhile, some literature has also highlighted the perceived value cannot be directly observed. Donabedian impact of demographic and exogenous policy variables [32] suggested using other non-technical variables such on satisfaction [20, 21]. However, there are few empirical as convenience and information; however, these sug- studies on how health policies affect patient satisfaction gested variables are already included in perceived quality and loyalty. and demographics. Relevant studies in China mainly discussed the impacts The descending resources reform can affect patient of the circa-2009 health reforms on the efficiency of satisfaction in two ways. First, it substantially changes low-level hospitals, but these studies did not use micro- low-level hospitals’ capabilities, which have been in- data collected at the individual level, nor were they in- cluded in the latent variables, as well as impacting per- volved the descending resources reform [6]. Some stud- ceived quality and hospital image. Second, the reform ies used micro-data to explore patient satisfaction in information can be transmitted to patients and impact developing countries including China using demographic those patients’ choices. Latent variables of the reform in- variables including age, gender, and education level [22, clude (1) whether the reform information is correctly 23]; other studies used medical market concentration, recognized by patients, (2) the information channel, and income, and health insurance status [24, 25], as well as (3) related policy on medical service price, differential other factors. Researches on developing countries also medical insurance reimbursement and tiered medical utilized scales to evaluate the relationships among ser- services, which are derived from individual hospitals. vice quality, satisfaction, and loyalty [26, 27]. However, these studies treated the institutional environment as Questionnaire design and data given, an assumption that is inconsistent with the cir- The question items and their definitions of latent vari- cumstances in developing countries experiencing a rap- able are reported in Table 1, where a five-point unbal- idly evolving healthcare system and reforms. anced scale is used for ordered variables. Patient hospital Different from reforms in developing countries, which choice (loyalty) is measured as the intention to choose a focus on designing different health resource formulae local low-level hospital and the intention to choose a and financing mechanisms [28], China’s descending re- local low-level hospital first when suffering a serious ill- sources reform is paving a new way to correct the un- ness, respectively. Latent variables of patient satisfaction even allocation of health resources. Some recent studies include (1) reform satisfaction at the industry level, and discussed the impact of this reform on both doctors or (2) satisfaction with local low-level hospitals. For the patients [25, 29], but the OLS/OLM (ordinary least three exogenous latent variables of the ESCI model, hos- square or ordered logit model) methodology used by pital image refers to patients’ brand recognition percep- Sun et al. [25] could not solve measurement error of tion of the hospital, and patient trust provides the basis survey data. The contribution of this paper is to explore of future cooperation in terms of patients’ future hos- this reform’s effect on patient hospital-selection behavior pital choices [33]. In addition, patients’ awareness of the using a SEM model. Health policy is incorporated into descending (high-level) hospital and descending doctors the patient response model in order to better understand provide a measurement of the degree to which the image the effect of the descending resources reform. Mean- of high-level hospitals is implanted onto low-level hospi- while, different levels of hospitals and cognitive channels tals. The aforementioned three variables are used as are introduced to discuss the heterogeneous effects of measurement variables of hospital image. the reform on patient behavior. Clavolino and Dalsgaard [34] pointed out that patient expectations are related to prior expectations of the ob- Materials and methods served existing services. Because the descending re- The European Consumer Satisfaction Index model is sources reform involves both high-level and low-level used as the basic model for this study, and two exogen- hospitals, we take the diagnosis/treatment capability of ous variables are included: the descending resource re- low-level hospitals and the accessibility of descending form and demographics. We use loyalty to a low-level doctors as measurement variables. Patient expectations SUN et al. Archives of Public Health (2021) 79:179 Page 4 of 14 Table 1 Questionnaire scale, variables, and definitions Variable Latent variables Measurement variables types Symbol Name Symbol Question items Definition Exogenous ξ1 Hospital image X11 Trust for LLH 1 for negative change, 2 for no change, 3–5 for ordinary variables positive change X12 Awareness for the descending 1–5 ordinary variables from very low to very high hospitals X13 Awareness for the descending 1–5 ordinary variables from very low to very high doctors ξ2 Patient X21 Accessibility to the descending 1–5 ordinary variables from very low to very high expectation doctors X22 LLH capability change 1 for negative change, 2 for no change, 3–5 for ordinary positive change X23 Medical cost change 1 for positive change, 2 for no change, 3–5 for ordinary negative change ξ3 Perceived X31 Convenience changes 1–5 ordinary variables from very low to very high quality X32 LLH Environment change 1–5 ordinary variables from very low to very high ξ4 Reform policy X41 Reform Recognition 1–5 ordinary variables from very low to very high X42 Reform information Channels 1 for public channels (newspaper, TV and hospital), private channels being 0 X43 Medical service price 1–5 ordinary variables from very low to very high X44 Insurance reimbursement 1–5 ordinary variables from very low to very high X45 Tiered medical service 1–5 ordinary variables from very low to very high ξ5 Socio- X51 Gender 1 for male and 0 for female demographics X52 Age 1 for ≤30, 2–5 for 31–40, 41–50, 51–60, and ≥ 61, respectively X53 Education level 1 for primary or below, 2–5 for junior, high school, college or university, graduate degree, respectively Mediating ξ6 Satisfaction ME1 Reform satisfaction 1–5 ordinary variables from very low to very high variable ME2 LLH satisfaction 1–5 ordinary variables from very low to very high Endogenous η Hospital Y1 Intention to choose LLH 1–5 ordinary variable from very low to very high variable selection Y2 Intention to choose LLH when 1–5 ordinary variables from very low to very high (Loyalty) suffering serious illness Source: The authors are also related to medical costs, which are included as a The data used in this study cover 17 public hospitals in measurement variable of patient expectations. The latent Zhejiang Province, China, including six tertiary hospitals, variables of perceived quality are related to associated eight secondary hospitals, and three primary hospitals. In services, and we further considered the medical environ- each hospital, face-to-face interviews were performed in the ment and convenience of low-level hospitals as measure- outpatient department, where questionnaires were randomly ment variables. distributed to patients by trained independent investigators. Socio-demographic variables include gender, age, and edu- All interviewees that finished the questionnaire received a cation level. Awareness of reforms, cognition channels, and small gift worth $1.00 (7 RMB Yuan) for their time. From reform-related policy evaluations regarding medical service November 2018 to October 2019, we collected 1354 ques- prices, differentiated insurance reimbursements, and tiered tionnaires, among which 1287 were valid, an effective rate of medical services are incorporated to measure the latent vari- 95.05%. able of reform. Of these, differentiated insurance reimburse- ment policies will incentivize patients to choose low-level Empirical methods hospitals by reducing/increasing the reimbursement ratio The samples used in this paper cover three different when choosing high- or low-level hospitals, respectively. The levels of hospitals and different information channels, so tiered medical services policy requires patients to choose a we first use one-way analysis of variance (ANOVA) to low-level hospital first, and then be referred to a high-level examine whether significant differences exist among pa- hospital if suffering a serious illness, which impacts patients’ tients at different levels of hospitals and information satisfaction and hospital choice behavior. channels. If the ANOVA results (F-test) reach the SUN et al. Archives of Public Health (2021) 79:179 Page 5 of 14 threshold value (α = 0.05), a significant difference exists. Following the ESCI model, we establish a SEM model Then, a multiple posteriori comparison will be per- incorporating descending resources reform and hospital formed to compare the differences by using the LSD selection (Fig. 1). Detailed variables are given in Table 1. (least significant difference) test. The empirical analysis is performed using two steps: (1) Because the data used in this paper come from question- estimating the impact of different exogenous latent vari- naires, patients’ cognitions of and responses to the reform ables on patient satisfaction; and (2) exploring the im- are subjective and difficult to directly measure, meaning it is pact of patient satisfaction and the reform on patients’ hard to avoid subjective measurement errors. The structural hospital selection (loyalty). In addition, subsamples of equation model (SEM) has the advantage of handling mul- different levels of hospitals will be discussed to test the tiple variables and measurement errors of variables. In robustness of the results. addition, this method can estimate both the factor structure Finally, in order to measure the existence of a mediating ef- and factor relationships, which makes it suitable to process fect, referring to Baron and Kenny [36], we estimate β , β , c a ’ ’ and analysis questionnaire data. According to Qiu and Lin β ,and β showninFig. 2 in turn. If β is insignificant and b c c [35], this model consists of a measurement equation and the others are significant, then complete mediating effects structure equation as follows: exist; however, if the estimated value of β is significant and its absolute value is less than that of β , then there is a partial Y ¼ Λ η þ ε ð1Þ mediating effect. X ¼ Λ ξ þ δ ð2Þ Results We used SPSS 23.0 software and reliability tests to assess η ¼ Bη þ Γξ þ ζ ð3Þ the reliability and consistency of the scale and data. The Equations (1) and (2) are measurement equations to results show that the Cronbach’s α coefficient is 0.915(> describe the relationship between latent variables and 0.800), indicating that the scales and data have a good measurement variables. Y and X are the observable vari- internal consistency. Then, factor analysis was utilized to ables of endogenous and exogenous latent variables re- test the validity of the questionnaire. The KMO (Kaiser- spectively, η and ξ are endogenous and exogenous latent Meyer-Olkin) value of 0.901 and Bartlett spheroid test variables respectively, and Λ and Λ are the factor load- value of 9223.199 (Sig. = 0.0001) show that the question- y x ing matrix. Equation (3) provides the structure model, naire has good structural validity, so the scale and data which is used to describe the relationship between latent are suitable in performing empirical estimations. Using variables, where in the structure coefficient matrix, B AMOS 21.0 software, the absolute fit indices of patient and Γ represent the relationship between endogenous la- loyalty model shows that RMSEA (Root Mean Square tent variables and the impact of exogenous latent vari- Error of Approximation) =0.067(< 0.10), GFI (Goodness ables on endogenous latent variables, respectively; and ζ of Fit Indices) =0.934(> 0.90), AGFI (Adjusted Goodness is the residual term matrix. of Fit) =0.846(> 0.80), and PGFI (Parsimony Goodness of Fig. 1 The theoretical model of patient hospital choice (loyalty). Source: The ESCI model is shown inside the dotted line, this model referred to Ref. [25] SUN et al. Archives of Public Health (2021) 79:179 Page 6 of 14 Fig. 2 The estimation method of mediating effects. Sources: Baron and Kenny (1986) Fit Indices) =0.545(> 0.50), suggesting that the data can reached (3.35 ± 0.87), where 87% of interviewees be used for SEM model estimation. responded “positive,”“high,” and “very high,” 10% re- ported unchanged satisfaction levels, and only 1% re- ANOVA and multiple comparison results ported that their satisfaction had declined. As for the Table 2 presents the summary statistics for the sample hospital selection (loyalty) variable, the average score for variables. It can be seen that women account for 58% of “intention to choose local low-level hospital” is 3.30, and the sample. The average age is between 2= “31-40 years the loyalty score when suffering a serious illness reached old” and 3= “41-50 years old,” with a mean of (2.35 ± 3.20 (3 = “positive” and 4 = “high”). Both of these results 1.25). The average education level is close to 3 = “high indicate that following the descending resources reform, school” (2.88 ± 1.11). The mean reform satisfaction is patients will prefer to choose a low-level hospital. (2.70 ± 1.39), which is between 2 = “low” and 3 = “fair”. For measurement variables, trust in local low-level However, satisfaction with local low-level hospitals hospitals is (3.28 ± 0.98), showing that the descending Table 2 Descriptive statistics of key variables, ANOVA, and multiple comparison results Latent variables Measured Overall Hospital level Information channel variables Tertiary Secondary Primary Public Private *,1 *,1 *,1 *,1 Hospital Image X11 3.28 ± 0.98 3.29 ± 0.95 3.34 ± 1.01 3.11 ± 0.87 3.49 ± 0.94 2.93 ± 0.95 *,2 *,1 *,1 *,1 *,1 X12 2.55 ± 1.26 3.05 ± 1.40 2.49 ± 1.29 2.48 ± 1.06 2.98 ± 1.18 1.78 ± 1.02 *,2 *,1 *,1 *,1 *,1 X13 2.50 ± 1.25 3.02 ± 1.46 2.41 ± 1.27 2.55 ± 1.02 2.93 ± 1.18 1.75 ± 0.99 *,1 *,1 *,1 *,1 Patient expectation X21 3.10 ± 1.10 2.96 ± 1.14 3.16 ± 1.14 2.97 ± 0.96 3.35 ± 1.01 2.65 ± 1.11 *,1 *,1 *,1 *,1 X22 3.35 ± 0.85 3.28 ± 0.90 3.41 ± 0.86 3.22 ± 0.78 3.48 ± 0.83 3.12 ± 0.84 *,1 *,1 X23 2.80 ± 0.99 2.88 ± 1.07 2.78 ± 0.99 2.80 ± 0.94 2.95 ± 1.02 2.52 ± 0.86 *,1 *,1 *,1 *,1 Perceived quality X31 3.27 ± 1.00 3.29 ± 0.98 3.33 ± 1.03 3.10 ± 0.92 3.48 ± 0.96 2.91 ± 0.97 *,1 *,1 X32 3.38 ± 0.91 3.35 ± 0.94 3.37 ± 0.92 3.41 ± 0.85 3.53 ± 0.89 3.10 ± 0.87 *,2 *,1 *,1 * * Reform X41 2.36 ± 1.29 2.90 ± 1.44 2.30 ± 1.32 2.30 ± 1.09 2.78 ± 1.27 1.63 ± 0.96 policy X42 0.64 ± 0.48 0.72 ± 0.45 0.62 ± 0.49 0.65 ± 0.48 –– *,2 *,1 *,1 *,1 *,1 X43 3.30 ± 0.87 2.99 ± 0.94 3.35 ± 0.86 3.30 ± 0.87 3.46 ± 0.87 3.02 ± 0.81 *,2 *,1 *,1 *,1 *,1 X44 3.20 ± 0.86 2.97 ± 0.93 3.25 ± 0.87 3.16 ± 0.80 3.34 ± 0.86 2.95 ± 0.80 *,1 *,2 *,1 *,1 *,1 X45 3.29 ± 0.90 3.15 ± 0.99 3.35 ± 0.88 3.20 ± 0.89 3.41 ± 0.91 3.08 ± 0.84 *,1 *,1 Socio-demographics X51 0.42 ± 0.49 0.32 ± 0.47 0.45 ± 0.50 0.39 ± 0.49 0.43 ± 0.50 0.40 ± 0.49 X52 2.35 ± 1.25 2.41 ± 1.11 2.34 ± 1.28 2.34 ± 1.21 2.38 ± 1.24 2.29 ± 1.26 *,2 *,1 *,1 * X53 2.88 ± 1.11 3.34 ± 1.05 2.81 ± 1.09 2.87 ± 1.13 2.89 ± 1.11 2.87 ± 1.10 *,2 *,1 *,1 *,1 *,1 Satisfaction ME1 2.70 ± 1.39 3.26 ± 1.49 2.65 ± 1.42 2.59 ± 1.19 3.22 ± 1.24 1.79 ± 1.15 *,1 *,1 *,1 *,1 ME2 3.35 ± 0.87 3.29 ± 0.88 3.40 ± 0.88 3.23 ± 0.81 3.50 ± 0.84 3.08 ± 0.84 *,1 *,1 *,1 *,1 Loyalty Y1 3.30 ± 0.90 3.29 ± 0.88 3.36 ± 0.91 3.13 ± 0.84 3.43 ± 0.89 3.08 ± 0.86 *,1 *,1 *,2 *,1 *,1 Y2 3.20 ± 0.96 3.42 ± 1.01 3.26 ± 0.99 2.94 ± 0.82 3.33 ± 0.97 2.97 ± 0.91 Sample size 1287 1287 130 851 306 819 Note: [1] Asterisks and n = 1 or 2 (*, n) denote a statistically significant difference among the different hospital groups with ANOVA (α =0.05), and the number of differences by using multiple posteriori comparison SUN et al. Archives of Public Health (2021) 79:179 Page 7 of 14 resources reform has improved the image of low-level Regarding the other exogenous latent variables, scores hospitals. However, respondents reported average scores for trust, accessibility, capability, and convenience are between 2 = “low” and 3 = “fair” for awareness of de- significant higher for secondary hospitals compared with scending high-level hospitals/doctors. In contrast, acces- primary hospitals. In terms of socio-demographics vari- sibility of the descending doctors reached (3.10 ± 1.10, ables, the education level of patients visiting tertiary hos- where 3 = “positive” and 4 = “high”), indicating that re- pitals (3.34 ± 1.05) is significantly higher than those of spondents found the descending doctors easy to access. patients attending primary and secondary hospitals; no The variables of environment, capability, and conveni- significant difference was found in education levels of ence for local low-level hospitals all have means between patients visiting the latter two. Accordingly, the reform 3.2 and 3.4 (3 = “positive” and 4 = “high”), suggesting that awareness of patients in tertiary hospitals is significantly the reform had a positive impact from the patients’ per- higher than that among patients attending other levels spective. In addition, the response for medical cost is of hospitals; however, their evaluation of reform-related (2.80 ± 0.99, where 2 = “no change” and 3 = “slight de- policies like medical service prices and insurance reim- crease”), indicating that the descending resources reform bursement amounts is significantly lower. has lowered medical costs in general. For the reform policy latent variable, 64% of patients SEM estimation results of patient satisfaction obtain information from public channels such as news- The above discussion has offered some preliminary in- papers, television, and the hospital, with a mean of vestigations into the effect of the descending resources (2.36 ± 1.29, where 2 = “low” and 3 = “fair”). However, reform on patient responses to the reform, but more evi- reform-related policies like medical service prices, (dif- dence is needed to understand the marginal effect of dif- ferential) insurance reimbursement levels, and tiered ferent latent variables. Therefore, AMOS 21.0 software medical services are evaluated high, with an average of is used to test the theoretical model established in Fig. 1. 3.2–3.4. This section reports the results of the patient satisfaction The ANOVA results in Table 2 indicate that except model. Following an iteration procedure using the boot- for the variables of medical cost, environment, informa- strap method, the resulting path diagram of the SEM tion channel, and age, the variables all show significant model is shown in Fig. 3. Perceived quality, patient ex- differences among different levels of hospitals (α = 0.05). pectations, and hospital image all have a significant posi- For information channel groups, except for socio- tive impact on patient satisfaction, and their normalized demographics, the public channel scores of other vari- path coefficients of 0.577, 0.711, and 1.014, respectively, ables are significantly higher than those for the private indicate that the ECSI model can better explain the fac- channel group; in particular, scores for getting informa- tors influencing patient satisfaction. Overall, this model tion through public channels, hospital selection (loyalty), is significant. At the same time, the coefficient of the re- and satisfaction are significantly higher. form policy is 0.140, which confirms the reform’s posi- The results of multiple comparisons show that post re- tive effect on patient satisfaction. form, patients’ satisfaction in tertiary hospitals is signifi- The relationships between the variables and estimated cantly higher than that in primary and secondary coefficients are reported in Table 3. It can be seen that hospitals, but no significant difference exists between the effects of socio-demographics on patient satisfaction primary and secondary hospitals. In addition, no signifi- do not pass the significance test, but other latent vari- cant difference exists in patient satisfaction with low- ables are significant at α = 1%, which demonstrates the level hospitals between tertiary and other low-level hos- existence of a causal relationship between the variables. pitals, whereas patient satisfaction in secondary hospitals This in turn indicates that the structural equation model is significantly higher than that in primary hospitals. of patient satisfaction is appropriate. In this model, hos- The results for the two hospital selection (loyalty) variables pital image exerts the biggest influence on patient satis- are different: (1) although patients visiting a tertiary hospital faction, which in turn mainly originates from the trust in have already chosen a high-level hospitals, their willingness local low-level hospitals and the brand implantation of to select a low-level hospitals is not significantly different cooperative high-level hospitals. from patients who attend other level hospitals; (2) those visit- Next, the confirmatory factor analysis method is used ing secondary hospitals reported a higher score for this vari- to conduct the single-factor structural validity analysis able than those visiting primary hospitals; and (3) when a (Table 4). This analysis also serves as a reliability evalu- patients suffers a serious illness, their loyalty to their local ation for the measurement model. The findings indicate low-level hospitals is significantly lower if their most recent that the coefficients for all latent variables except socio- hospital visit was to a primary hospital rather than a second- demographics are significant at α = 1%. This demon- ary or tertiary hospital, but no significant difference was strates that the potential factor structure of the ques- found between the latter two. tionnaire items is reasonable. Several other findings can SUN et al. Archives of Public Health (2021) 79:179 Page 8 of 14 Fig. 3 Patient satisfaction structural equation model path diagram. Source: The authors be drawn from the analysis. First, for the measurement with reform awareness, thus have a strong explanatory variables of perceived quality, the factor loading coeffi- power for the reform latent variable. cients of environment and convenience are quite close, indicating their important explanatory power on pa- SEM estimation results of patient hospital selection tients’ perceptions of quality. Second, for patient expec- (loyalty) tations, the factor loading coefficients of capability of Figure 4 shows the path diagram of the structural equa- low-level hospitals is the highest (=1.285), followed by tion model for patient hospital selection (loyalty). The medical cost and accessibility of descending doctors. normalized path coefficients of the descending resources Third, for hospital image, the factor loading coefficients reform and patient satisfaction to loyalty are 0.450 and of patients’ awareness of cooperative high-level hospitals 0.731, respectively (Table 5). These two latent variables and descending doctors reached 2.295 and 2.444, re- are significant at α = 1%, indicating that a causal rela- spectively, which are far higher than the score for the tionship between variables can be established, and there- variable of trust (=1.000). Finally, for the latent variable fore, this structure equation model of patient hospital of the reform policy, the factor loading coefficients of selection is appropriate. It can be seen that the impact of (differential) insurance imbursement levels and medical patient satisfaction on loyalty is greater than that of the service prices are high (> 1), and the coefficient of tiered reform policy, which indicates patients’ LLH choice is medical services reaches 0.933; these variables, together Table 3 SEM estimation results of patient satisfaction model Number Relationship Normalized path coefficient Standard deviation C.R. value P values 1 Satisfaction←Perceived quality 0.577 0.051 11.246 *** 2 Satisfaction←Patient expectation 0.711 0.071 10.051 *** 3 Satisfaction←Hospital image 1.014 0.084 12.099 *** 4 Satisfaction←Reform policy 0.140 0.040 3.521 *** 5 Satisfaction←Socio-demographics 0.252 0.383 0.658 0.511 Note: *** indicates significance level of α =1% SUN et al. Archives of Public Health (2021) 79:179 Page 9 of 14 Table 4 Confirmatory factor analysis results of patient satisfaction model Number Relationship Factor loading coefficient standard deviation C.R. value P values 1 Reform satisfaction←Satisfaction 1.000 –– – 2 LLH satisfaction←Satisfaction 0.607 0.027 22.134 * * * 3 Convenience←Perceived quality 1.000 –– – 4 Environment←Perceived quality 1.004 0.066 15.197 * * * 5 Accessibility←Patient expectation 1.000 –– – 6 Capability←Patient expectation 1.285 0.090 14.200 * * * 7 Medical cost←Patient expectation 1.102 0.077 14.375 * * * 8 Trust←Hospital image 1.000 –– – 9 Awareness for high-level hospital←Hospital image 2.295 0.116 19.748 * * * 10 Awareness for descending doctors←Hospital image 2.444 0.125 19.536 * * * 11 Reform awareness← Reform policy 1.000 –– – 12 Information channel←Reform policy 0.228 0.023 9.728 * * * 13 Medical service price←Reform policy 1.039 0.060 17.257 * * * 14 Insurance imbursement←Reform policy 1.114 0.065 17.254 * * * 15 Tiered medical service←Reform policy 0.933 0.057 16.326 * * * 16 Gender←Socio-demographis 1.000 –– – 17 Age←Socio-demographis 16.821 6.333 2.656 0.008 18 Education level←Socio-demographis 13.768 4.853 2.837 0.005 Note: *** indicates significance level of α =1% more affected by their own evaluations of the reform variables of loyalty have close factor loading coefficients and local low-level hospitals. (1.000 and 0.963), showing their strong explanatory The confirmatory factor analysis results for the hos- power. Satisfaction with the reform and LLHs also have pital selection (loyalty) model are reported in Table 6. similar factor loading coefficients (1.000 and 1.054). The coefficients of different socio-demographics are still In terms of the measurement variables of the reform significant at α = 1%, but this latent variable itself does policy, the factor loading coefficients of (differential) in- not have a significant impact on loyalty according to the surance imbursement levels and medical service prices results reported in Table 5. The two measurement are still greater than 1 (=1.119 and 1.028), which shows Fig. 4 The path diagram of patient hospital selection (loyalty) . Source: The authors SUN et al. Archives of Public Health (2021) 79:179 Page 10 of 14 Table 5 SEM estimation results of patient hospital selection (loyalty) model Number Relationship Normalized path coefficient Standard deviation C.R. value P values 1 Loyalty←Satisfaction 0.731 0.052 14.041 * * * 2 Loyalty←Reform policy 0.450 0.043 10.570 * * * 3 Loyalty←Socio-demographics 0.034 0.380 0.091 0.928 Note: *** indicates significance level of α =1% their important impact on perceptions of the descending insignificant. Third, socio-demographics still have an in- resources reform. The factor loading coefficients of re- significant effect on patient satisfaction. Finally, sub- form awareness and tiered medical services also reached sample results for patient hospital selection (loyalty) 1.000 and 0.933, showing their explanatory power of ac- show strong robustness in terms of satisfaction and the ceptance of the reform. reform’s positive effect, as well as the insignificant effect of socio-demographic variables. Thus, it can be con- cluded that the subsample empirical analysis supports Robustness test the estimation results of the full-sample SEM model. In order to test the robustness of the above full-sample results, we use different-level hospital subsamples to per- form empirical analyses. The sample sizes of different Mediation effects estimation result level hospitals meet the requirements of carrying out In order to estimate the impact of the reform latent vari- SEM model estimation. able on hospital selection (loyalty) through satisfaction, Table 7 reports the estimation results of different sub- we use the method shown in Fig. 2 to perform the test samples. It can be seen that, firstly, the impacts of per- the mediating effect (Table 8). It can be seen that β is ceived quality, patient expectations, and hospital image significant but its absolute value is less than β in the on patient satisfaction are confirmed to be positive again full-sample estimation, which confirms the existence of for the tertiary and secondary hospital subsamples at the a partial mediating effect. This finding indicates that the 1% significance level. However, for the primary hospital reform latent variable has an indirect impact on hospital subsample, the impact of patient expectations is slightly choice behavior through satisfaction, but it also directly insignificant (P = 0.072). Second, for tertiary and second- affects hospital choice. In order to verify the robustness ary hospitals subsamples, the positive impact of the re- of this result, we also perform hospital subsample esti- form latent variable on patient satisfaction is significant mations, the results of which show that heterogenous re- at α = 5%, although it is insignificant for the primary sults exist for different hospital subsamples. The results hospital subsample. This makes sense because doctors for the tertiary and secondary hospitals subsamples show descending from high-level hospitals in Zhejiang Prov- a complete mediating effect, which is consistent with the ince only descend to secondary hospitals, so this reform previous literature on the impact of satisfaction upon [18.19] latent variable’s impact on primary hospitals is loyalty implied in ECSI model ; however, no Table 6 Confirmatory factor analysis results of patient hospital selection (loyalty) model Number Relationship Normalized path coefficient Standard deviation C.R. value P values 1Y1← loyalty 1.000 –– – 2Y2← loyalty 0.963 0.027 35.719 * * * 3 Reform satisfaction←Satisfaction 1.000 –– – 4 LLH satisfaction←Satisfaction 1.054 0.075 14.000 * * * 5 Reform awareness←Reform policy 1.000 –– – 6 Information channel←Reform policy 0.219 0.023 9.7375 * * * 7 Medical service price ←Reform policy 1.028 0.060 17.165 * * * 8 Insurance imbursement←Reform policy 1.119 0.064 17.399 * * * 9 Tiered medical service←Reform policy 0.963 0.058 16.552 * * * 10 Gender←Socio-demographics 1.000 –– – 11 Age←Socio-demographics 16.773 6.378 2.630 0.009 12 Education level←Socio-demographics 13.699 4.823 2.840 0.005 Note: *** indicates significance level of α =1% SUN et al. Archives of Public Health (2021) 79:179 Page 11 of 14 Table 7 Subsample estimation results of patient satisfaction and hospital selection (loyalty) models Number Relationship Subsample Normalized path coefficient Standard deviation C.R. value P values 1 Satisfaction← Tertiary 0.335 0.087 3.836 * * * Perceived quality Secondary 0.313 0.039 8.085 * * * Primary 0.488 0.119 4.089 * * * 2 Satisfaction← Tertiary 0.271 0.100 2.708 0.007 Patient expectation Secondary 0.502 0.058 8.603 * * * Primary 0.258 0.144 1.799 0.072 3 Satisfaction← Tertiary 0.456 0.080 5.718 * * * Hospital image Secondary 0.571 0.047 12.051 * * * Primary 1.120 0.305 3.667 * * * 4 Satisfaction← Tertiary 0.267 0.070 3.807 * * * Reform policy Secondary 0.051 0.026 2.004 0.045 Primary 0.101 0.116 0.870 0.385 5 Satisfaction← Tertiary 0.073 0.406 0.179 0.858 Socio-demographics Secondary 0.099 0.276 0.358 0.720 Primary 0.677 1.309 0.517 0.605 6 Loyalty← Tertiary 0.540 0.128 4.214 * * * Satisfaction Secondary 0.615 0.058 10.632 * * * Primary 0.885 0.199 4.448 * * * 7 Loyalty← Tertiary 0.599 0.120 5.000 * * * Reform policy Secondary 0.379 0.040 9.372 * * * Primary 0.784 0.335 2.339 0.019 8 Loyalty← Tertiary 1.344 1.147 1.172 0.241 Socio-demographics Secondary 0.157 0.415 0.378 0.706 Primary 1.874 8.400 0.223 0.823 Note: *** indicates significance level of α =1% mediating effect is found for the primary hospital sub- sample, which may be related to the fact that the de- scending resources reform in Zhejiang mainly involves secondary and tertiary hospitals but not primary hospitals. Table 8 Mediation effects estimation result of reform latent Discussion variable China’s descending resources reform offers a unique ap- Exogenous variable: Reform; Mediating variable: Satisfaction; proach to overcoming the uneven allocation of health Exogenous variable: Hospital selection (Loyalty) resources. It works by utilizing the dominant role of the Sample β β β β Conclusion a b c c public hospital system in the Chinese health market Full-sample 0.855*** 1.579*** 0.752*** −0.683* Partial compared with other developing countries. Using ques- mediating effect tionnaire data, the expanded ECSI model, and the struc- ture equation model, we found that the descending Hospital Tertiary 0.758*** 1.148** 0.470*** −0.144 Complete subsample mediating resources reform had a significantly positive impact on effect patients’ satisfaction and their selection of local low-level Secondary 0.846*** 1.439*** 0.760*** −0.535 Complete hospitals. It was further found that the measurement mediating variables of perceived quality, patient expectations, and effect hospital image can also explain patient behavior. Primary 1.010*** 1.439 0.778*** 0.945 No ANOVA and multiple comparison techniques demon- mediating effect strated that significant differences exist among patients Note: ***and** indicate significance level of α =1 and 5%, respectively at different levels of hospitals. SUN et al. Archives of Public Health (2021) 79:179 Page 12 of 14 Thus, it is found that the descending resources reform secondary hospital evaluating them significantly higher contributed to the improvement of convenience, capabil- than those visiting a tertiary hospital, again indicating ities, and the environment of local low-level hospitals, the focus and main beneficiaries of this reform being offering evidence of the effect of this reform from the secondary hospitals. Meanwhile, such polices contribute patients’ perspective. This paper’s SEM estimations also to explaining the reform’s latent variables well and have confirm the significantly positive impacts of perceived significantly positive effects on patient satisfaction and quality, patient expectations, and hospital image on pa- loyalty. These results suggest that the above supporting tient satisfaction. Among these, the treatment/diagnosis policies can help reduce the burden of medical costs in- capabilities of low-level hospitals were found to have the curred by patients in low-level hospitals, and their satis- strongest explanatory power for patient expectations, in- faction with and loyalty to low-level hospitals can be dicating that LLH capabilities are a top factor shaping improved via financial incentives [39]. patients’ expectations, and this issue is also the core The survey results also show the coexistence of low re- focus of the descending resources reform. form satisfaction and high satisfaction with local low- Among the different latent variables, hospital image level hospitals, where patients visiting tertiary hospitals exerts the greatest influence. Of the three measurement have higher reform satisfaction, whereas those visiting variables, i.e., trust in low-level hospitals and awareness secondary hospitals have higher satisfaction with local of cooperative high-level hospitals and descending doc- low-level hospitals than those visiting primary hospitals. tors, the confirmatory factor analysis suggested that the Correspondingly, the reform awareness of patients visit- latter two have higher explanatory power than trust in ing tertiary hospitals is also significantly higher than low-level hospitals. This result shows that embedding those visiting other levels of hospitals. This could reflect the brand image of high-level hospitals and doctors into the fact that one of the reform’s main goals was to re- low-level hospitals is an effective way to improve the duce congestion in high-level hospitals by reallocating image of low-level hospitals, and the role of brand im- health resources to different levels of hospitals. Thus, pa- plantation is greater than that of the image (trust) of tients visiting high-level hospitals should have a real un- low-level hospitals from the patient perspective. Further- derstanding of the reform’s effects, benefit from the more, improving the image of low-level hospitals can reform, and have higher satisfaction with reform. Mean- also significantly contribute to higher patient satisfaction while, in this reform, descending doctors mainly flow and loyalty to low-level hospitals, which is fully consist- into secondary hospitals, which receive the greatest ent with the goal of the descending resources reform. amount of resources inflow and capability improve- This study also confirmed the significantly positive im- ments, thus contributing to their patients’ satisfaction pact of the reform and satisfaction on patients’ hospital compared with those visiting primary hospitals. selection (loyalty) of low-level hospitals and the exist- It is worthy pointing out the importance of informa- ence of mediating effects of satisfaction. Previous litera- tion channels on the reform’s effect. The ANOVA re- ture has found that medical service quality affects sults indicate that reform awareness and patient patient satisfaction, which in turn affects their behavioral behavior are related to the information channel. Infor- choices [37]; high service satisfaction has a significantly mation access via public channels can avoid information positive impact on customer loyalty [38]. The empirical transmission distortion and thus help promote a positive results on the mediating effect confirm that in secondary response from patients. Although according to the SEM and tertiary hospitals, the reform latent variables affect estimation, information channels do not have a strong patients’ hospital choice behavior through the mediating explanatory power for the reform latent variable, the factor of satisfaction. However, such effects disappear for ANOVA results showed that scores of perceived quality, primary hospitals, a difference that highlights the reality patient expectation, hospital image, satisfaction and hos- that descending resources reform in Zhejiang focused pital selection (loyalty), are all significantly higher for the more on secondary and tertiary hospitals. This paper public channel subsample than the private channel sub- provides new evidence based on patient behavior in the sample, indicating that a more unbiased information context of China’s latest healthcare reform, and evidence supply and effective transmission are essential in enhan- for the appropriation of expanded ECSI model used in cing patients’ positive responses and reshaping their hos- this paper. pital choices. Zhejiang’s reform is characterized by descending doc- tors from high-level hospitals to low-level hospitals, to- Conclusion gether with important supporting policies like medical China’s reforms carried out since 2003 have offered evi- service prices, differential insurance imbursement levels, dence that traditional approaches focusing on higher and tiered medical services. These policies all received public health expenditure and expansion of medical in- higher scores from patients, with patients visiting a surance coverage are inadequate in correcting the SUN et al. Archives of Public Health (2021) 79:179 Page 13 of 14 uneven allocation of health resources. The descending interpreted data; ZD and SJ performed the empirical work with software. All authors read and approved the final manuscript. resources reform launched in 2013 is an attempt to im- prove the capability of low-level hospitals and re-attract Funding patients by using human capital spillovers of doctors de- The authors thank for the financial support from National Social Science Fund of China (20NDJC244YB), Soft Science Fund of Zhejiang Province scending from high-level hospitals together with brand (2018C35028) and Development Research Center of Shanghai(2020-YJ-C05). implantation of these high-level hospitals. Using ques- tionnaire survey data collected from patients in Zhejiang Availability of data and materials The datasets during the current study available from the corresponding Province, this paper provides supporting empirical evi- author on reasonable request. dence of the reform’s impact on patient behavior. The results indicate that the reform has been effective in im- Declarations proving the capabilities of low-level hospitals, and brand Ethics approval and consent to participate implantation of high-level hospitals shows strong ex- Not applicable. planatory power. The findings also suggest that policy- makers could pay more attention to the importance of Consent for publication Not applicable. information channels in impacting patient awareness, re- sponses, and hospital selection. Competing interests For developing countries where public hospitals play a The authors declare that they have no competing interests. dominant healthcare role, China’s reform offers a dis- Author details tinct and valuable approach to correcting the uneven al- School of Finance and Business, Shanghai Normal University, Shanghai, location of health resources. The approach indicates that China. Department of Stomatology, Tongde Hospital of Zhejiang Province, Hangzhou, China. Department of Stomatology, Songjiang Hospital of greater investment and demand-side reforms might not Shanghai, Shanghai, China. School of Finance, Zhejiang Gongshang necessarily incentivize patients to respond as the govern- University, Hangzhou, China. ment intends. In addition to affordability, building the Received: 31 March 2021 Accepted: 28 September 2021 capabilities of hospitals in rural/remote areas remains a key factor, which can be done by cooperation between high- and low-level hospitals, human capital spillovers, References and brand implantation. Meanwhile, some demand-side 1. Gwatkin D. Trends in health inequalities in developing countries. Lancet Glob Health 2017; 5(4): e371–e372. https://doi.org/https://doi.org/10.1016/ policies like differential medical insurance and price reg- S2214-109X(17)30080-3. ulations could also create incentives for patients to re- 2. Strasser R. Rural health around the world: challenges and solutions. Fam shape their hospital choices so as to resolve the existing Pract 2003; 20(4): 457–463. https://doi.org/https://doi.org/10.1093/fampra/ cmg422. structural congestion or uneven allocation problems. In 3. Sun ZS, Wang SH, Barnes SR. Understanding congestion in china’s medical addition, access to full and unbiased information is of market: an incentive structure perspective. Health Policy Plann 2016; 31(3): great importance in realizing the full effect of these re- 391–400. https://doi.org/https://doi.org/10.1093/heapol/czv062. 4. Zhang L, Cheng G, Song S, Yuan B., Zhu W., He L., Ma X., Meng Q. Efficiency forms. Therefore, the findings of this paper highlight performance of china’s health care delivery system. Int J Health Plann Manag that the government should pay attention to public in- 2017; 32: 254–263. https://doi.org/https://doi.org/10.1002/hpm.2425,3. formation channels in improving the effect of healthcare 5. Cai Y, Gong G, Zhang Q. Getting rid of the medicine-selling dependent hospital by value returning from the demonstration of a theoretical model reforms. Besides, after the data collection of this paper, (in Chinese). Nankai Econ Res. 2013;1:40–52. the outbreak of COVID-19 pandemic encourages pa- 6. Tao C, Guo T, Xu K. Measure of equalization level of medical and health tients to choose local low-level hospitals more so as to resources allocation in China (in Chinese). Statist Decis-making. 2019;24:42–6. 7. Jiang S, Min R, Fang P. The impact of healthcare reform on the efficiency of avoid travelling and infection risk to high-level hospitals, public county hospitals in China. BMC Health Serv Res 2017; 17(1): 838. so this phenomenon could enhance the existing effect of https://doi.org/https://doi.org/10.1186/s12913-017-2780-4. the above reform on patients’ choice to low-level hospi- 8. Meng Q, Mills A, Wang L, et al. What can we learn from China’s health system reform? Br Med J 2019; 365: 3–7. https://doi.org/https://doi.org/1 tals, and it does not harm the applicability and useful- 0.1136/bmj.l2349. ness of our findings. 9. Huang G, Qiu Y. A study on the structural reform of the supply side of the public hospitals at county level (in Chinese). Soc Secur Res. 2018;1:50–7. Abbreviations 10. Zhejiang Province Government. Implementation Opinions on Promoting ECSI: The European consumer satisfaction index model; ANOVA: Analysis of the Construction of Long-term Mechanism for Reform of Double variance; SEM: Structure equation model; SARS: Severe acute respiratory Descendings and Two Promotions. 2015. [Accessed 2019 May 12] http:// syndrome; KMO: Kaiser-Meyer-Olkin test; OLS: Ordinary least square; www.zj.gov.cn/art/2015/9/22/art_32431_245673.html. OLM: Ordered logit model 11. Valls Martinez M del C, Ramirez-Orellana a. patient satisfaction in the Spanish national health service: partial least squares structural equation Acknowledgements modelling. Int J environ res pub Health 2019; 16: 4886. https://doi.org/ Not applicable. https://doi.org/10.3390/ijerph16244886, 24. 12. Fatima T, Malik S A, Shabbir A. Hospital healthcare service quality, patient Authors’ contributions satisfaction and loyalty: an investigation in context of private healthcare SZ provides the conception and design of the work. WS contributes to the systems. Int J Qual Reliab Manage 2018; 35(6): 1195–1214. https://doi.org/ data collection, analysis, and drafted the work; ZH and ZX analyzed and https://doi.org/10.1108/IJQRM-02-2017-0031. SUN et al. Archives of Public Health (2021) 79:179 Page 14 of 14 13. Fornell C. A national customer satisfaction barometer: the Swedish 36. Baron RM, Kenny DA. The moderator-mediator variable distinction in social experience. J Marketing 1992; 56: 6–21. https://doi.org/https://doi.org/10.11 psychological research: conceptual, strategic and statistical considerations. J 77/002224299205600103,1. Pers Soc Psychol 1986; 51: 1173–1182. https://doi.org/https://doi.org/10.103 14. Fornell C, Johnson MD, Anderson EW, Cha J, Bryant BE The American 7//0022-3514.51.6.1173,6. consumer satisfaction index: nature, purpose, and findings. J Marketing 1996; 37. Naidu A. Factors affecting patient satisfaction and healthcare quality. Int J 60: 7–18. https://doi.org/https://doi.org/10.1177/002224299606000403,4. Health Qual Assur 2009; 22(4): 366–381. https://doi.org/https://doi.org/10.11 08/09526860910964834. 15. Brady MK, Cronin JJ. Some new thoughts on conceptualizing perceived 38. Rust RT, Zahorik AJ. Customer satisfaction, customer retention, and market service quality: a hierarchical approach. J Marketing 2001;, 65: 34–49. https:// share. J Retail 1993; 69: 193–215. https://doi.org/https://doi.org/10.1016/ doi.org/https://doi.org/10.1509/jmkg.65.3.34.18334,3. 0022-4359(93)90003-2,2. 16. Johnson MD, Gustfasson A, Andreassen W, et al. The evolution and future of 39. Liu D. China’s new rural cooperative medical scheme: evolution, design and national customer satisfaction index models. J Econ Psychol 2001; 22(2): impacts. International Verlag der Wissenschaften: Frankfurt am Main, 217–245. https://doi.org/https://doi.org/10.1016/S0167-4870(01)00030-7. Germany; 2013. 17. Bayol M, Foye A, Tellier C, et al. Use of PLS path modeling to estimate the European consumer satisfaction index (ECSI) model. Stat Applicata. 2000; 12(3):361–75. Publisher’sNote 18. Caruana A. Service loyalty: the effects of service quality and the medicating Springer Nature remains neutral with regard to jurisdictional claims in role of customer satisfaction. Eur J Mark 2002; 36(7/8): 811–827. https://doi. published maps and institutional affiliations. org/https://doi.org/10.1108/03090560210430818. 19. Fullerton G, Taylor S. Mediating, interactive, and nonlinear effects in service quality and satisfaction with services research. Can J Admin Sci 2002; 19(2): 124–136. https://doi.org/https://doi.org/10.1111/j.1936-4490.2002.tb00675.x. 20. Strasser S, Aharony L, Greenberger D. The patient satisfaction process: moving toward a comprehensive model. Med Care Res Rev 1993; 50: 219– 248. https://doi.org/https://doi.org/10.1177/107755879305000205,2. 21. Banka G, Edgington S, Kyulo N, Padilla T., Mosley V., Afsarmanesh N., Fonarow G.C., Ong M.K. Improving patient satisfaction through physician education, feedback, and incentives. J Hosp Med 2015; 10(8): 497–502. https://doi.org/https://doi.org/10.1002/jhm.2373. 22. Jiao W, Jiang H, Yu M, et al. Investigation and analysis of inpatient satisfaction in a third-level hospital in Beijing (in Chinese). J Nurs Adm. 2010; 10(9):630–2. 23. Ma P. Inpatient satisfaction factors and countermeasures (in Chinese). Med Inf Med Surg. 2009;22(3):259–60. 24. Pan J, Liu D, Ali S. Patient dissatisfaction in China: what matters. Soc Sci Med 2015; 143: 145–153. https://doi.org/https://doi.org/10.1016/j. socscimed.2015.08.051. 25. Sun Z, Wang S, Zhao H, Yu H Does descending resources reform improve patient satisfaction and reshape choice of care providers. Inquiry 2020; 57: 1–13. https://doi.org/https://doi.org/10.1177/0046958020956899. 26. Senarath U, Fernando D, Rodrigo I. Factors determining client satisfaction with hospital-based perinatal care in Sri Lanka. Tropical Med Int Health 2006; 11(9): 1442–1451, https://doi.org/https://doi.org/10.1111/j.1365-3156.2 006.01698.x. 27. Ahmed S, Tarique K, Arif I. Service quality, patient satisfaction and loyalty in the Bangladesh healthcare sector. Int J Health Care Qual Assur 2017; 30(5): 477–488. https://doi.org/https://doi.org/10.1108/IJHCQA-01-2017-0004. 28. Love-Kol J, Griffin S, Kataika E, et al. Methods to promote equity in health allocation in low- and middle-income countries: an overview. Glob Health 2020; 16: 6. https://doi.org/https://doi.org/10.1186/s12992-019-0537-z,1. 29. Zhang LD, Liu L, Nan DH, et al. The descending resources reform and doctors’ response: an SEM estimation based on Shanghai survey data. J Fujian Medical Univ (Soc Sci). 2021;22(3):1–17. 30. Vinagre M, Neves J. The influence of service quality and patients’ emotions on satisfaction. Int J Health Care Qual Assur 2008; 21(1): 87–103. https://doi. org/https://doi.org/10.1108/09526860810841183. 31. Alrubaiee L, Alkaa'ida F. The mediating effect of patient satisfaction in the patients’ perceptions of healthcare quality-patient trust relationship. Int J Mark Stud 2011; 3(1): 103–127. https://doi.org/https://doi.org/10.5539/ijms.v3 n1p103. 32. Donabedian A. The quality of care: how can it be assessed? J Am Med Ass 1988; 260: 1743–1748. https://doi.org/https://doi.org/10.1001/jama.260.12.1 743, 12. 33. Dwyer RF, Schurr PH, Oh S. Developing buyer-seller relationship. J Marketing 1987; 51: 11–27. https://doi.org/https://doi.org/10.1177/0022242987051002 02,2. 34. Clavolino E, Dahlgaard JJ. ECSI- customer satisfaction modeling and analysis: a case study. Total Qual Manage. 2007;18(5):545–54. https://doi.org/10.1 080/14783360701240337. 35. Qiu H, Lin B. The theory and its application of structure equation model. Beijing, China: China Light Industry Press; 2012.

Journal

Archives of Public HealthSpringer Journals

Published: Oct 18, 2021

Keywords: Descending resources reform; Low-level hospitals; Patient satisfaction; Consumer behavior; Hospital selection behavior

There are no references for this article.