Introduction

Pneumonia is a major cause of morbidity and mortality worldwide. In 2013, lower respiratory tract infections (primary pneumonia) were the fifth largest causes of death in Taiwan [1–6]. In 2008, pneumonia became the leading infectious cause of death in the United States [7, 8]. In 2006 and 2007, pneumonia was the third largest cause of death in Germany [9].

Patients with pneumonia present with several signs and symptoms, including cough, dyspnea, sputum production, and pleuritic chest pain. Severity assessment is a crucial component of managing patients with pneumonia. The fatality rate associated with untreated bacteremic pneumonia is nearly 80 %. Serum therapy can decrease the fatality rate by 50 % and antimicrobial therapy can reduce the mortality rate by up to 80 % [10].

The National Health Insurance (NHI) program is the backbone of the health care system in Taiwan and covers 99 % of the total population. National health care expenditure in Taiwan increased from 5.3 % to 6.0 % of the gross national product during 1995–2001. The NHI operated on a fee-for-service (FFS) basis between 1995 and 2001, during which period the health care expenditure increased by approximately 50 %. For preventing the unrestricted and rapid growth of health care expenditure, the Bureau of National Health Insurance (BNHI) implemented a global budget system for modifying the FFS mechanism in 2002. Global budgeting in Taiwan is an overall spending target, designed to limit the volume and expenditure of health care service [11–15]. The global budget system adopted in Taiwan is an expenditure-cap system with a floating-point-value mechanism. The national budget for a given year is determined before the end of the previous year by consultation between the BNHI and hospital representatives. The aggregate expenditure in the previous year and the age structure of the population determine the national budget. The monetary value of each point is obtained by dividing the fixed budget by the number of points claimed by all hospitals. Thus, it fluctuates to precisely match the predetermined budget. Reimbursement to providers is based on an existing FFS schedule, which lists a relative value or the number of points for each item of the service. The degree of reimbursement cuts is measured on the basis of the quarterly monetary value of each point.

This study compared the health service utilization [length of hospital stay (LOS)], health care expenditures (e.g., total costs, drug costs, diagnostic costs, and therapy costs), and quality of care [e.g., the risk of readmission within 14 days and risk of revisiting the Emergency Department (ED) within 3 days] before and after implementing the global budget system among patients with pneumonia.

Materials and methods

This pre–post comparison study was based partly on data from the National Health Insurance Research Database (NHIRD), provided by the BNHI and the Department of Health. The NHIRD is managed by the National Health Research Institutes (NHRI) and comprises registration files and original claims data for reimbursement; this database is released to researchers in Taiwan. Each year, the BNHI categorizes the NHI data into data files, which are deidentified by scrambling the identification codes assigned to both patients and medical facilities. Subsequently, the files are forwarded to the NHRI as the original claims data for inclusion in the NHIRD [16].

Specific data subsets can be reconstructed for research by using the registration files and original claims data in the NHIRD. Two categories of expenditure after systematic sampling were used in this study: inpatient expenditures by admission (DD file) and ambulatory care expenditures by visit (CD file). The interpretation and conclusions drawn herein do not represent those of the BNHI, Department of Health, or NHRI [16].

Systematic sampling of CD and DD files

0.2 % of the ambulatory care expenditures, by visit, CD file extracted by the systematic sampling method on a monthly basis, together with the related records in details of ambulatory care orders form the systematic sampling CD file. 5 % of the inpatient expenditures, by admission, DD file extracted by the systematic sampling method on a monthly basis, together with the related records in details of inpatient orders form the systematic sampling DD file [16]. In total, 1 million cases of the systematic sampling dataset were used in this study.

Study population

Patients included in the CD and DD files and diagnosed with pneumonia (International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes 480–488) in 2000, 2001, 2004, and 2005 were recruited for comparison and analysis.

The global budget system in Taiwan was fully implemented in 2002. Therefore, data collected in 2000 and 2001 were used as the baseline data, and data collected in 2004 and 2005 were used as the postintervention data. In 2003, an outbreak of severe acute respiratory syndrome (SARS) occurred in Taiwan. SARS reportedly affected health care utilization [17]. Therefore, data collected in 2003 were excluded.

Charlson comorbidity index

A Charlson comorbidity index (CCI) score of 1 was assigned for the following comorbidities: myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, peptic ulcer, mild liver disease, and diabetes without end-organ damage. A CCI score of 2 was assigned for the following comorbidities: hemiplegia, moderate or severe renal disease, diabetes with end-organ damage, nonmetastatic tumor, leukemia, and lymphoma. A CCI score of 3 was assigned for moderate and severe liver disease. A CCI score of 6 was assigned for the following comorbidities: metastatic solid tumors and acquired immunodeficiency syndrome (AIDS). The total CCI score was derived by summing the weighted scores of all comorbidities. For patients older than 40 years, a CCI score of 1 was added for each decade to the total score [18–20].

Income state index

In Taiwan, health insurance premiums are calculated as a percentage of the monthly salary of an individual. Seven income levels were defined in this study on the basis of monthly salary income: Level 0 (no monthly income), Level 1 (<NT$17 ,280), Level 2 (NT$17,281–22,800), Level 3 (NT$22,801–28,800), Level 4 (NT$28,801–36,300), Level 5 (NT$36,301–45,800), Level 6 (>NT$45,801) [21].

Statistical analysis

Data are presented as the mean ± standard deviation (SD). The t-test was used for comparing the differences in the mean values of the LOS, diagnostic costs, drug costs, therapy costs, total costs, rate of revisiting the ED within 3 days, and rate of readmission within 14 days before and after implementing the global budget system.

Multilevel and mixed-effects models were used for determining the effect of several independent variables of LOS, diagnostic costs, drug costs, therapy costs, total costs, the risk of revisiting the ED within 3 days, and the risk of being readmitted within 14 days after discharge. In mixed-effects modeling, a regression analysis was conducted by considering two types of effect sizes: fixed effects (intercepts and slopes that described the population as a whole) and random effects (intercepts and slopes that could vary across subgroups of the sample) [22].

Two nested levels were used in this study: hospital accreditation levels and geographic areas of Taiwan. The three hospital accreditation levels in Taiwan are medical centers, regional hospitals, and local hospitals. Taiwan is divided into six geographical areas: Taipei city, northern Taiwan, central Taiwan, southern Taiwan, Kaohsiung, and eastern Taiwan. The independent variables evaluated in this study were pre–post global budgeting, age, sex, income state index, Charlson comorbidity index, the three hospital accreditation levels, and the six geographic areas of Taiwan.

All statistical analyses were performed using the statistical package STATA for Windows (Version 11.0). A p-value < 0.05 was considered statistically significant.

Results

Data on 32,535 patients with pneumonia (11,352 at baseline and 21,183 after implementing the global budget system) were used in this study. The mean age of patients in the prebudget group was 23.30 ± 0.26 years and that of the patients in the postbudget group was 41.56 ± 0.24 years. In the prebudget and postbudget groups, 56 % and 59 % were males, respectively. The mean CCI score was 0.58 ± 0.01 and 1.48 ± 0.01 in the prebudget and postbudget groups, respectively. Both the groups demonstrated significant differences in age, sex, and CCI scores (all p < 0.001). Both groups showed no significant difference in the income state index (p = 0.27).

The mean LOS values in the prebudget and postbudget groups were 6.36 ± 0.07 days and 10.78 ± 0.09 days, respectively. The mean values of diagnostic costs, drug costs, therapy costs, and total costs before implementing the global budget system were NT$570.92 ± 24.48, 4947.74 ± 145.15, 3554.55 ± 124.76, and 22,697.82 ± 542.40, respectively, and those after implementing the system were NT$1401.78 ± 36.01, 13,065.87 ± 233.59, 13,043.83 ± 203.91, and 62,016.7 ± 793.19, respectively. The mean rate of revisiting the ED within 3 days before implementing the global budget system was 5.5 ± 0.2 % and that after implementation was 4.6 ± 0.1 %. The mean rates of readmission within 14 days before and after implementing the system were 6.1 ± 0.2 % and 8.2 ± 0.2 %, respectively. The LOS, diagnostic tests, drug costs, therapy costs, total costs, mean rate of revisiting the ED within 3 days, and mean rate of readmission within 14 days before and after implementing the global budget system differed significantly for both groups (all p < 0.001; Table 1).

Table 1 Comparison of the differences in the mean values of length of stay (LOS), diagnostic costs, drug costs, therapy costs, total costs, rate of revisiting the Emergency Department (ED) within 3 days, and 14-day readmission rate before and after implementation of the global budget system (GB)
Full size table

Length of stay

The mixed-effects Poisson regression model was used for clustered count data analysis [23]. The Poisson regression model was fitted using the option incidence rate ratio (IRR) for obtaining exponentiated estimates. A 26 % increase in the LOS was observed after implementing the global budget system (IRR = 1.26, p < 0.001). The LOS was significantly longer in males than in females (IRR = 1.07, p < 0.001). A significantly positive correlation was observed between age and the LOS (IRR = 1.02, p < 0.001). In addition, a significantly negative correlation was observed between the income state index and CCI and LOS (IRR = 0.92, p < 0.001; β = 0.93, p < 0.001). No significant difference in the LOS was observed between patients in regional hospitals and patients in local hospitals (IRR = 1.00, p = 0.79; IRR = 1.03, p = 0.08, respectively). Furthermore, the LOS was 5 %, 6 %, and 4 % longer in central Taiwan, southern Taiwan, and Kaohsiung city, respectively, than that in Taipei city (IRR = 1.05, p = 0.02; β = 1.06, p = 0.01; β =1.04, p = 0.03, respectively). However, no significant difference in the LOS among patients in the hospitals located in northern Taiwan, eastern Taiwan, and Taipei city was observed (all p > 0.05; Table 2).

Table 2 The impact of several independent variables on LOS in patients with pneumonia (mixed-effects Poisson model)
Full size table

Diagnostic, drug, and therapy costs

A mixed-effects linear model was used to analyze diagnostic, drug, and therapy costs. Because diagnostic, drug, and therapy costs were skewed to the right, data were converted to log base 10 for analysis. The global budget system significantly positively affected the diagnostic, drug, and therapy costs (β = 0.30, β = 0.10, β = 0.40; all p < 0.001). Significantly higher diagnostic, drug, and therapy costs were associated with males than with females (β = 0.07, β = 0.05, β = 0.11; all p < 0.001). Significantly positive correlations among age, diagnostic costs, and drug costs were observed (β = 0.02, β = 0.01; both p < 0.001). However, a significantly negative correlation between age and therapy costs was observed (β = −0.02, p < 0.001). Furthermore, a significantly negative correlation between the CCI, diagnostic costs, and drug costs was observed (β = −0.18, β = −0.09; both p < 0.001). However, a significantly positive correlation between the CCI and therapy costs was noted (β = 0.57, p < 0.001). Compared with medical centers, no significant differences in diagnostic costs, drug costs, or therapy costs between regional hospitals and local hospitals were observed (all p > 0.05). Compared with Taipei city, no significant differences in diagnostic, drug, or therapy costs incurred in hospitals in northern Taiwan, central Taiwan, southern Taiwan, Kaohsiung city, or eastern Taiwan were observed (all p > 0.05).

Total costs

Global budgeting significantly positively affected the total costs (β = 0.32, p < 0.001). Males were significantly associated with higher total costs than females (β = 0.01, p = 0.01). A significantly positive correlation between age and total costs was observed (β = 0.001, p < 0.001). In addition, a significantly negative correlation between the income state index, CCI, and total costs was observed (β = −0.01, p < 0.001; β = −0.01, p = 0.02). Compared with medical centers, no significant difference in total costs at regional hospitals or local hospitals was observed (both p > 0.05). Compared with Taipei city, no significant difference in total costs in northern Taiwan, central Taiwan, southern Taiwan, Kaohsiung city, or eastern Taiwan was observed (all p > 0.05; Table 3).

Table 3 The impact of GB, age, sex, income index, Charlson comorbidity index (CCI), accreditation hospital level, and regional level on total costs (log10) using a mixed-effects linear model
Full size table

Risk of revisiting the ED within 3 days

A generalized linear binary regression model was used for this analysis. The risk of revisiting the ED within 3 days was significantly lower after implementing the global budget system [odds ratio (OR) = 0.80, p < 0.001]. The risk was significantly greater in males than in females (OR = 1.11, p = 0.05). The risk did not correlate significantly with age, income state index, and CCI (all p > 0.05). Compared with the risk in medical centers, the risks of revisiting the ED within 3 days at regional hospitals and local hospitals did not differ significantly (both p > 0.05). Compared with the risk in hospitals located in Taipei city, the risks in hospitals located in northern Taiwan, central Taiwan, southern Taiwan, Kaohsiung city, and eastern Taiwan did not differ significantly (all p > 0.05; Table 4).

Table 4 The impact of several independent variables on the risk of revisiting the ED within 3 days among patients with pneumonia (mixed-effects linear binary regression model)
Full size table

Risk of readmission within 14 days

A generalized linear binary regression model was used for this analysis. The risk of being readmitted within 14 days increased significantly after implementing the global budget system (OR = 1.39; p < 0.001). The risk did not correlate significantly among age, sex, income state index, and CCI (all p > 0.05). Compared with the risk in medical centers, the risks between patients in regional hospitals and patients in local hospitals did not differ significantly (all p > 0.05). Compared with the risk in Taipei city, the risks of being readmitted 14 days after the initial admission among patients in hospitals in northern Taiwan, central Taiwan, southern Taiwan, Kaohsiung city, and eastern Taiwan did not differ significantly (all p > 0.05; Table 5).

Table 5 The impact of GB, age, sex, income index, CCI, accreditation hospital level, and regional level on the risk of readmission within 14 days using a generalized linear binary regression model
Full size table

Discussion

In this study, patients with pneumonia treated after implementing the global budget system were 15 years older (mean) than those treated before its implementation. As of 2009, the birth rate in Taiwan was the lowest in the world, with an average of 8.99 births per 1000 persons [24]. A two-fold increase in the proportion of patients aged >60 years with pneumonia was observed after implementing the global budget system (Appendix Table 6). Pneumococcal vaccination is reportedly effective in reducing the incidence of pneumonia among adults [25]. The possible reasons for an older population in the postbudget group were aging effects and the limited availability of the inoculation vaccine against pneumonia among elders in Taiwan [26].

The global budget system significantly positively affected the LOS among patients with pneumonia in Taiwan. The budget cap for each hospital is enforced quarterly, and health care expenditures are not settled until discharge. To avoid exceeding the assigned quota, patients were often asked to remain in the hospital at the end of the fiscal quarter until the beginning of a new quarter. Another possible reason is that patients during the study period were older and received a poorer quality of care. Why patients in hospitals in Taipei city, central Taiwan, southern Taiwan, and Kaohsiung city had a significantly longer LOS than those in hospitals in Taipei remains unclear.

The global budget system significantly positively affected diagnostic, drug, therapy, and total costs in patients with pneumonia in Taiwan. The health care expenditures increased from 145.5 % to 267.0 %, significantly exceeding the 2.2 % growth in the consumer price index of Taiwan for the same period. Possible causes of the incremental costs include an aging population, high price of modern medical equipment, and prolonged LOS. No significant differences in costs were observed among hospitals having different accreditation levels or hospitals located in different geographic areas of Taiwan.

After implementing the global budget system, the rate of revisiting the ED within 3 days decreased but the rate of readmission within 14 days increased. The global budget system significantly negatively affected the risk of revisiting the ED within 3 days, but significantly positively affected the risk of readmission within 14 days. The results were independent of the hospital accreditation level and geographic area. The reasons for these inexplicable findings require additional studies in the future. However, the crude death rate of pneumonia has been increasing steadily since 2000 [27]. The quality of care provided to patients with pneumonia appears to have diminished after implementing the global budgeting program.

This study had several limitations. First, the NHIRD provides no detailed information on patient-related factors, such as lifestyle, habits, body mass index, physical activity, socioeconomic status, and family history; all these factors were potential confounding factors in this study. Second, the conclusion derived from a cohort study is generally of a lower methodological quality than that derived from a randomized trial because a cohort study is subject to several biases and requires adjustments for confounding factors. Despite its meticulous design and adequate control of confounding factors, the study may still have biases associated with potentially unmeasured and unknown confounding factors. Third, the registries in the NHI claims are primarily used for administrative billing and are not verified for scientific purposes. Because of the anonymity of the identification numbers, patients could not be contacted for additional information. The accuracy of medical coding in the claims data may affect the data validity. However, the diagnostic data in the NHIRD are highly reliable. The insurance system has mechanisms for monitoring insurance claims.

Conclusion

Although the risk of revisiting the Emergency Department (ED) within 3 days decreased, the length of stay (LOS), health care costs, the mortality rate, and the risk of readmission within 14 days increased in patients with pneumonia after implementing the global budget system. Therefore, the global budget system is failing and the quality of care is diminishing in Taiwan, particularly for patients with pneumonia.