Current level and determinants of inappropriate admissions to township hospitals under the new rural cooperative medical system in China: a cross-sectional study

We used a Delphi expert consultation to build AEP criteria for township hospitals in rural China. All current judgments on clinical indicators were screened based on current AEP criteria in the US [25], the seven aforementioned EU countries [27], the Netherlands, and other countries [32], and China's current clinical pathway and literature studies. We then consolidated and classified the systems to develop a preliminary evaluation of AEP entries. A total of 32 experts were chosen to conduct three rounds of consultation and adjustment. Of the 32, 11 were from top-class national hospitals, 15 from county hospitals, four from township hospitals, and two from NRCMS management agencies. The experts’ deep understanding of China’s existing services in rural areas enabled them to assess the suitability and other indicators of systemic AEP standards, and so build the AEP criteria for inappropriate admissions to township hospitals.

The experts reviewed 400 randomly selected medical records from a township hospital to formulate two standards, a final AEP standard and implicit diagnostic criteria for clinical experts. Both were tested with a kappa test.

We applied stratified cluster sampling. Counties and county-level cities in China’s central (e.g., Jingshan, Songzi, and Dangyang in Hubei Province) and western (e.g., Danjiangkou and Meixian in Shaanxi Province) regions were designated as sample areas.

The sample size was calculated using the inappropriate admission rate in township hospitals. From existing research, the estimated inappropriate admission rate P is 16%, and the expected error d = 0.1 P = 1.6%, and α = 0.05. So,

An additional 10% was added to the final sample, giving a total of 2220 medical records.

Township hospitals were ranked by size and those in the 4^th and 10^th places were chosen for the study. In each hospital, 250 non-delivery medical records from 2012 were randomly selected. A total of 2,100 records were collected (fewer than 250 records were available from Songzi and Dangyang). After screening, 2,044 records were found acceptable. All the sampling patients were covered by NRCMS, and had received reimbursement. Sociological characteristics such as gender, age, complications, admission information including hospital department, admitted severity, length of stay, disease category, and other data like the treatment capacity of the township hospital and the payment were collected to build a database using Epidata 3.2. Diseases were categorized using the International Classification of Diseases 10th revision (ICD-10). Additionally, NRCMS management personnel and doctors from the subject township hospitals were interviewed.

From previous research, we identified possible determinants including gender, age, department, admitted severity, length of stay, disease category, complications and the payment method [15],[17]-[19]. The treatment capacity of a medical institution is an important factor in inappropriate admission. The data we obtained indicate a hierarchical structure. The 2,044 records were aggregated by township hospital. The determinants for inappropriate admission to hospitals were examined using MLwiN 2.30, which was developed by the University of Bristol, UK, for multilevel binomial logistic regression analysis [33],[34].

After three rounds of Delphi evaluation and adjustment, the AEP criteria were identified (Table 1). The criteria comprise nine requirements for services, such as surgery needs, and 20 indicators of disease severity, such as persistent fever over 38°C for 5 days. Patients were considered to be inappropriately admitted if their records did not fit any AEP criteria.

The Cronbach coefficients for the three rounds of expert consultation were 0.676, 0.942, and 0.729. The applicability of the new AEP criteria was investigated by 32 experts, 28 of whom (87.5%) claimed the criteria could be used for more than 80% of the township hospital inpatients. Both the AEP criteria and clinical experts’ implicit diagnostic criteria were applied to judging 400 medical records (see Table 2). The judgments were consistent, and the kappa coefficient of the result was 0.686 (u = 10.34, P < 0.01).

Table 3 shows how the patients from the five counties benefited from the reimbursement policies of NRCMS. These counties shared the same financing policies in 2012, in which an individual who pays 60 RMB receives a government subsidy of 240 RMB. Reimbursement takes two forms, namely, outpatient and inpatient reimbursement and the different payment schemes lead to a sizeable difference in hospitalization benefits.

Table 4 shows the characteristics and inpatient admissions data for the subject patients. A total of 26.5% of the records indicated inappropriate admissions. The incidence rate varied in different regions (P < 0.001), which ranged from 10% to 40%. More than half of the cases, 57.5%, were women, but appropriate admissions were only 51.7% (P = 0.021). The average age of the 2,044 hospitalized patients was 54.7 years, whereas the mean age of inappropriate admissions was 57.8 years (P < 0.001). Elderly patients (52.1%) were more likely to be inappropriately admitted (P < 0 .001). Patients in internal medicine departments were more likely to have been inappropriately admitted than other departments (P < 0.001). Inappropriate admission also occurred mostly in non-urgent patients (P < 0.001). Respiratory, circulatory, and skeletal muscle diseases were more likely in those inappropriately admitted (P < 0.001). No statistical difference between the two groups was found in terms of length of stay (LOS) and whether the patients had complications.

The two-level logistic regression used was by MLwiN 2.30, the medical records were level 1, and the township hospitals level 2. The level 2 variance of the zero model was statistically significant (χ² = 4.917, P = 0.027), with aggregation of information at the township level.

The specific results of the introduction of the explanatory variables to fit two variance components models are shown in Table 5. The major determinants of inappropriate admissions to township hospitals were age, hospital department, severity of illness on admission, and disease category. If other factors remain constant, inappropriate admission was higher among groups aged 20 and above. The accidental estimation of the group aged 20 and above was about 2.6, which is considerably higher than younger groups. No difference existed between pediatrics and internal medicine, and inappropriate admissions for surgical and other departments were far below those for internal medicine. More severe conditions had lower rates of inappropriate admission. Meanwhile, digestive, symptomatic, and respiratory diseases had the same rate of inappropriate admission. The rates for circulatory and skeletal muscle diseases were higher than respiratory diseases (P < 0.001), whereas the rates for urinary tract diseases, trauma, and poisoning were much lower (P < 0.001).

Based on the existing management capacity and supervision scope of NRCMS, AEP is the most effective regulatory tool. The AEP developed in this study is based on the service capacity and need for hospitalization in rural China. The standard applies to common diseases, and mainly covers surgery, gynecology, and pediatrics. A few specialist diseases are exceptions. A kappa coefficient of 0.686 for AEP and clinical experts’ implicit diagnostic criteria indicates high consistency. The former could reflect the results of the latter at 0.686, which is higher than other evaluated tools in previous studies [35]. From the studies published in the past 20 years, no approach could guarantee total reliability in evaluating inappropriate admission [1]. The present AEP criteria, which were developed based on the physical condition and medical needs of the general public, could be implemented in large rural areas in China.

Otherwise, it is worth noting that the standard of whether hospitalization was appropriate is based on disease development and treatment demands, but social factors were not taken into account. These might include older people living alone, the need for multiple outpatient visits, or distance from medical facilities and difficulties traveling. Although such hospitalizations are clinically unreasonable, their rationality is worth discussing from the point of view of humanization and social care.

The two-level logistic regression shows an evident hierarchy among the inpatient data (township hospitals–patients). Patients hospitalized at the township level are clustered. The township hospitals of different counties differ in NRCMS reimbursement, payment policies, service concept, and medical capability, which affect the incidence of inappropriate admission. Because of limited information, the results which these factors affect inpatient service reflected from a qualitative point of view.

First, the design of outpatient and inpatient NRCMS reimbursement influences the rate of inappropriate admission. International experience indicates that decreasing the out-of-pocket (OOP) share of inpatients in hospital fees induces excess hospitalization demand and increases costs [29]. If the difference between inpatient and outpatient reimbursement is not large, patients prefer inpatient treatment. For example, those paying through a family account or via individual reimbursement could afford only a few visits to the clinic, because they would have to fund any more themselves. The reimbursement for outpatients is lower and there is a limit per day, whereas the corresponding inpatient reimbursement is higher, as much as 70%. As a result, patients exaggerate their outpatient needs to enjoy a more economical inpatient service. Dangyang County uses a fund-pooling system for outpatient service, with reimbursement of up to 40% and a limit of 100 RMB per capita. The incidence of inappropriate admission reaches 39.9% because the 100 RMB only pays for a few outpatient visits, compared with inpatient reimbursement, through which all expenses exceeding 100 RMB will be reimbursed. Thus, for patients with chronic diseases, who need more frequent clinic visits, the tendency for inappropriate admission increases, up to 39.9%. Jingshan reimburses 6 RMB per outpatient, with the balance to be paid by the patient. Those suffering from chronic diseases or requiring multiple visits tend to choose hospitalization after paying 100 RMB, because 90% of inpatient costs can be reimbursed. Meixian County has no reimbursement limitations for outpatients and has a lower reimbursement for inpatient services than other counties, so residents have to pay more to be admitted. It therefore has a lower hospitalization rate than other counties, and the inappropriate admission rate is only 10.3%.

The payment modes of NRCMS determine the way in which township hospitals operate, affecting the incidence of inappropriate admission. For example, payment for single disease leads to inappropriate admissions. Township hospitals sometimes need to accept patients with minor illnesses as inpatients, to lower the average cost of hospitalization, and to get more reimbursement. Counties that pay different rates for different conditions, such as Dangyang, Jingshan and Danjiangkou, have higher rates of inappropriate admissions. Some common conditions, some regular diseases such as respiratory and circulatory problems, have higher inappropriate admission rates than others in payment for single disease.

Geographical environment correlates with inappropriate hospitalization. Meixian and Songzi County are both plains, and their rate is about 10%. Dangyang and Jingshan are hilly, and Danjiangkou is mountainous; they have higher rates, suggesting that convenience of travel to receive treatment also influences hospitalization.

The management of township hospitals also influences the appropriateness of hospitalization. Meixian County is a pilot project of the World Bank Loan/British Government Grants China Rural Health Development Project. Its township hospitals have undergone more comprehensive and standard reforms and have strong administration, translating to fewer inappropriate admissions. Danjiangkou uses a system of exclusions to control inappropriate admission, which defines 14 kinds of medical problems, 13 kinds of surgical problems, and nine other diseases that do not require hospitalization. These include where Hb > 6 g/L in iron deficiency anemia, and a liver cyst of less than 2 cm. The remaining inpatient services are reimbursed. This policy has, to some extent, caused inappropriate admissions for other diseases to rise, and it now has a higher inappropriate admission rate than any other county.

Table 4 shows that neither a patient’s experience of medical complications nor length of stay have any significant effect on inappropriate admissions, which are affected by age, treating department, admitted severity, and disease category, as shown in Table 5. The largest proportion of inappropriate admissions (52.1%) occurred among older people, which is consistent with other research [36], and children had the lowest incidence. Inappropriate admission among older people may be for two reasons. First, a high demand for health care and an inadequate outpatient account could eventually lead to hospitalization. Second, traveling can be difficult for older people, and since most of them no longer need to go to work, hospitalization is more convenient.

Medical wards tend to have distinctly more inappropriate admissions (30.7%) than surgical wards and pediatrics. Surgical cases require hospitalization, so it is easy to distinguish whether hospitalization is needed, while medical wards have less control over hospitalization standards. The lower rate of inappropriate admissions in pediatric departments (18.75%) conflicts with previous studies, where the global inappropriate admission rate among children reached 20% to 40% [31]. The disparity may be associated with parental choice. Parents may be more inclined to choose county hospitals for children, and the small number of cases, just 32, may not be representative. Inappropriate admission generally occurs among patients with minor diseases. Severely afflicted patients more often require hospitalization, and the rate is therefore more rational.

Respiratory, circulatory, digestive, and endocrine conditions have a higher incidence of inappropriate admission, while urinary and other conditions have a lower incidence [37], perhaps because of the differing treatment capabilities and control levels of outpatient departments. Urinary diseases are poorly controlled in clinics, so admissions are more appropriate. The conditions with more inappropriate admissions are those where it is not easy to differentiate whether outpatient or inpatient care is necessary. Children and older people with low immunity and resistance to disease are more vulnerable to respiratory diseases, such as influenza, bronchial pneumonia in children, and chronic bronchitis in older people. These diseases have higher occurrences and longer durations. Their symptoms usually last 5–7 days, and while medication and rest are critical to recovery, many patients find no improvement in village clinics or outpatient treatments after brief periods and request hospitalization. These diseases do not meet the admission standards, either on severity or intensity, so such admissions are considered inappropriate.