OUP user menu

Cardiologist service volume, percutaneous coronary intervention and hospital level in relation to medical costs and mortality in patients with acute myocardial infarction: a nationwide study

C.-Y. Liu, Y.-N. Lin, C.-L. Lin, Y.-J. Chang, Y.-H. Hsu, W.-C. Tsai, C.-H. Kao
DOI: http://dx.doi.org/10.1093/qjmed/hcu044 557-564 First published online: 24 February 2014

Abstract

Background: We explore whether cardiologist service volume, hospital level and percutaneous coronary intervention (PCI) are associated with medical costs and acute myocardial infarction (AMI) mortality.

Methods: From the 1997–2010 Taiwan National Health Insurance Research Database of the National Health Research Institute, we identified AMI patients and performed multiple regression analyses to explore the relationships among the different hospital levels and treatment factors.

Results: We identified 2942 patients with AMI in medical centers and 4325 patients with AMI in regional hospitals. Cardiologist service volume, performing PCI and medical costs per patient were higher in medical centers than in regional hospitals (P < 0.0001). However, the two hospital levels did not differ significantly in in-hospital mortality (P = 0.1557). Post hoc analysis showed significant differences in in-hospital mortality rate and in medical costs among the eight groups subdivided on the basis of hospital level, cardiologist service volume, and whether PCI was performed (P < 0.001 and P = 0.001, respectively).

Conclusions: These results highlight the importance of encouraging hospitals to develop PCI capability and increase their cardiologist service volume after taking medical costs into account. Transferring AMI patients to hospitals with higher cardiologist service volume and PCI performed can also be very important.

Introduction

Acute myocardial infarction (AMI) is a very important and common disease for which different treatment options are available. In recent years, the development of new medical technology in clinical heart care, in particular, percutaneous coronary intervention (PCI), has improved the quality of care for AMI patients.

Higher-volume hospitals are associated with a reduction in mortality for AMI, heart failure and pneumonia, but there is a volume threshold above which an increased condition-specific hospital volume is no longer significantly associated with reduced mortality.1 An excessive average physician workload may cause increased patient transfers, morbidity, or even deaths.2 A meta-analysis supports the view that patients receiving PCI in high-volume hospitals exhibit lower in-hospital mortality than those receiving PCI in low-volume hospitals.3 Early intervention does not significantly differ from delayed intervention in the primary outcome, but it does reduce the rate of secondary outcomes of death, myocardial infarction, or refractory ischemia and is superior to delayed intervention in high-risk patients.4 A study on hospitals in Ontario, Canada, finds that higher spending intensity is associated with lower mortality, readmissions and event rates.5

A national health system is implemented in the UK, Canada, Australia, New Zealand, Ireland, Denmark, Iceland, Sweden, Norway, Finland, Italy, Spain, Portugal and other countries, where medical expenses are borne mostly by the national treasury; these countries also retain full health insurance databases. In the East, Taiwan’s National Health Insurance (NHI) is a compulsory program that has extended its coverage to 99.6% of the population in Taiwan. The NHI has a complete, national and highly accurate health insurance database. However, because many countries are facing the serious problem of aging populations, the national per capita cost of medical treatment is bound to increase year by year. Taiwan’s current payment system contains a more hidden crisis of inappropriate use of medical resources. The limited government support for health insurance financing and the need to limit health care costs while delivering the highest-quality patient care are becoming critically important issues.

Donabedian’s 1966 framework of structure, process and outcomes has guided many studies on the elements needed to evaluate and compare medical care quality. In this study, we used ‘cardiologist service volume’ and ‘hospital level’ as the structural factors, ‘performing PCI’ as the process factor, and ‘mortality’ and ‘medical costs’ as the outcome factors to conduct a more complete assessment of the quality of care for patients with AMI in Taiwan.

Materials and methods

Data source

Data were obtained from the Taiwan National Health Insurance Research Database (NHIRD). The database covers over 99% of Taiwan’s 23 million residents6,7 and is the largest and most comprehensive population-based medical benefit claims program in Taiwan since the beginning of 1995. A data subset of NHIRD, Longitudinal Health Insurance Database 2000 (LHID2000), contains the original claim data of 1 million beneficiaries enrolled in the year 2000 randomly sampled from the year 2000 Registry for Beneficiaries Health Insurance Database of the NHIRD. All LHID2000 data can be interlinked through encryption of resident identifications and be provided to the public in Taiwan for research purposes. Information including gender, birth date, occupation, hospital level, geographical area and discharge date have been described in previous studies.8,9 The accuracy and validity of diagnoses identified by the specialists in the NHIRD have been strictly verified and certificated.10,11 The NHRID encrypts the patients’ personal information for privacy protection and provides researchers with anonymous identification numbers associated with the relevant claim information, which includes the patient’s sex, date of birth, registry of medical services and medication prescriptions. Patient consent is not required for accessing the NHIRD. This study was exempted by the Institutional Review Board of China Medical University in central Taiwan (CMU-REC-101-012).

Identification of AMI cases

The coding system of the International Classification of Disease, Ninth Revision (ICD-9), was used for all diagnoses. We selected all first-time health care diagnoses (admission, emergency, or outpatient) of AMI (ICD-9 code 410) from 1997 through 2010 that met the following criteria. The new AMI diagnosis was confirmed by a cardiology specialist, with the index date being the date of AMI registration. Comprehensive supporting information was provided and a rigorous cardiology assessment was made. We excluded patients with incomplete age or sex information. All included AMI patients were stratified into the two major hospital levels in Taiwan: medical centers and regional hospitals.

Variables of interest

Potential cofactors were identified on the basis of established comorbidities, and analyses were performed to establish whether these variables were substantially associated with AMI. Baseline comorbidities that have been known to affect AMI episodes include hypertension (ICD-9 codes 401–405), diabetes mellitus (DM, ICD-9 code 250), hyperlipidemia (ICD-9 code 272), renal disease (ICD-9 codes 403.01, 403.11, 403.91, 404.02, 404.03, 404.12, 404.13, 404.92, 404.93, 582, 583.0, 583.1, 583.2, 583.4, 583.6, 583.7, 585, 586, 588.0, V42.0, V45.1 and V56), congestive heart failure (ICD-9 codes 398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 425.4, 425.5, 425.7, 425.8, 425.9 and 428), peripheral vascular disease (ICD-9 codes 093.0, 4373, 440, 441, 4431, 4432, 4438, 4439, 4471, 5571, 5579 and V43.4), chronic pulmonary disease (ICD-9 codes 416.8, 416.9, 490–496, 500–505, 506.4, 508.1 and 508.8) and rheumatologic disease (ICD-9 codes 710, 714 and 725). Therefore, patients were assessed for these factors at the start of AMI. We calculated the number of comorbidities to represent disease severity. In addition, we stratified our AMI patients into three occupational groups: white-collar workers performing professional, managerial, or administrative work; blue-collar workers referred to as farmers, fishermen, or industrial laborers; and others consisting of mainly retired, unemployed and low income individuals. We also calculated the average medical costs of the AMI patients.

Statistical analysis

The distributions of categorical sociodemographic characteristics and comorbidities were compared between AMI patients in medical centers and AMI patients in regional hospitals, and the differences were examined with the chi-square test. Differences in the distributions of continuous variables were analyzed for the two hospital levels with Student’s t-test. The Taiwan Ministry of Health stratified hospitals into three levels. The academic medical centers are affiliated to colleges of medicine, with high volume of medical personnel and much greater resources with at least 500 beds. The academic medical centers are teaching hospitals. In this study, the academic medical centers were defined as the medical centers. The second level hospitals are metropolitan hospitals with at least 250 beds, but not accredited as teaching hospital. The third level hospitals are local community hospitals with less health care personnel and resources. In this study, the second and third level hospitals were defined as the regional hospitals. In-hospital mortality rate was calculated as (deaths/n)*100. Cardiologist service volume, defined as the number of patients cared for by each cardiologist in a hospital per day, was divided into two levels: high and low; the demarcation level was set at the median value of 1.5, meaning one cardiologist caring for 1.5 patients per day. To define cardiology service volume, we used the below mentioned two definitions, ‘the ward service capacity multiplies occupancy rate and then divides into the number of doctors’ and ‘the number of patients discharged from CV ward per month divided 30 days and then further divided into the number of doctors’. Performing PCI was defined as whether PCI was performed per AMI patient in each group. The two hospital levels, the two levels of cardiologist service volume, and whether PCI was performed combine to produce a total of eight groups. Differences in medical costs among the eight groups were analyzed by one-way analysis of variance (ANOVA) with Bonferroni analysis as the post hoc test. Trend analysis was performed by using a linear regression model. All statistical analyses were performed with the SAS package (Version 9.2 for Windows; SAS institute, Inc., Cary, NC, USA), and a two-tailed P-value lower than 0.05 was considered statistically significant.

Results

Baseline characteristics of AMI patients stratified by hospital level

From the claims data of the 1997–2010 periods, we identified 2942 patients with AMI in medical centers and 4325 patients with AMI in regional hospitals. There were more men than women, and almost half of the patients were in northern Taiwan (Table 1). The mean age of AMI patients in medical centers (66.2 ± 13.9 years) was 0.5 year greater than in regional hospitals (65.7 ± 14.3 years). There were significantly more white-collar workers in medical centers and more blue-collar workers in regional hospitals (P < 0.0001). Almost two-thirds of AMI patients had two or more comorbidities, with the number of comorbidities being higher in medical centers (P = 0.028). DM and hyperlipidemia were more prevalent among AMI patients in medical centers than in regional hospitals.

View this table:
Table 1

Baseline characteristics in patients with AMI, by hospital level

VariableNumber of AMI patientsP-value
Medical centers (N = 2942)Regional hospitals (N = 4325)
n%n%
Mean age, year (SD)a66.2(13.9)65.7(14.3)0.2107
Sex0.1642
    Female97433.1150034.7
    Men196866.9282565.3
Occupation<0.0001
    White collar121541.3160137.0
    Blue collar105635.9178341.2
    Other67122.894121.8
Area<0.0001
    Northern153852.3189243.8
    Central58920.064114.8
    Southern and eastern81527.7179241.4
Comorbidity
    DM122641.7169839.30.0396
    Hyperlipidemia122241.5170239.40.0624
    Hypertension216373.5319673.90.7217
Number of comorbidities0.0285
    035512.153912.5
    159220.199022.9
    263621.688620.5
    ≥3135946.2191044.2
Performing PCI133145.2168238.9<0.0001
In-hospital mortality903.11593.70.1557
Medical costs per patient, NTD199 432(186 293)156 678(126 071)<0.0001
Cardiologist service volume per day per cardiologist1.09(0.47)0.89(0.18)<0.0001

Relationships between hospital level, cardiologist service volume, performing PCI, mortality and medical costs

As seen in Table 1, a higher percentage of AMI patients in medical centers received PCI than that in regional hospitals (45.2% and 38.9%, respectively; P < 0.0001). In-hospital mortality rate was higher in AMI patients in regional hospitals than in medical centers (3.7% and 3.1%, respectively). Cardiologist service volume was significantly lower in regional hospitals than in medical centers (1.09 ± 0.47 and 0.89 ± 0.18, respectively), whereas medical costs was significantly higher in medical centers than in regional hospitals (NTD 199 432 and NTD 156 678, respectively).

Medical costs stratified by cardiologist service volume, performing PCI and hospital level

The data in Table 2 show statistically significant differences in medical costs between the high and low cardiologist service volume groups, between the PCI and non-PCI groups, and between medical centers and regional hospitals. Medical costs were significantly different among the eight groups derived from the triple-variable combinations (P < 0.0001). The post hoc test showed that the cost for Group 6, in which AMI patients received non-PCI treatment in medical centers with a high cardiologist service volume, was significantly higher than that for Group 1, 2, 3, 5, or 7 (P < 0.0001). A high cardiologist service volume means that one cardiologist cared for 1.5 or more AMI patients per day. Analysis also found Group 1 (the low cardiologist service volume, non-PCI and regional hospital combination) to have the lowest costs (P-value for trend < 0.0001). For clarity, the eight groups are arranged in descending order of costs per patient in Figure 1, from the highest (Group 6) to the lowest (Group 1) (P-value for trend < 0.0001).

Figure 1.

Medical costs for patients with AMI stratified by cardiologist service volume, performing PCI and hospital level.

View this table:
Table 2

Costs and in-hospital mortality in patients with AMI, stratified by cardiologist service volume, PCI and hospital level

GroupaVariableMedical costsc,d,ePost hoc testfMortalityg,h
Cardiologist service volumebPerforming PCIHospital levelMeanSDN%
1LowNoRegional hospitals117 826132 7571 < 4; 1 < 6; 2 < 6; 3 < 6; 5 < 6; 7 < 6128.22
2LowNoMedical center122 203112 1001011.4
3LowYesRegional hospitals166 63577 675112.59
4LowYesMedical center193 762100 41141.75
5HighNoRegional hospitals161 512137 5531263.70
6HighNoMedical center211 373212 198693.01
7HighYesRegional hospitals143 666113 121102.87
8HighYesMedical center179 948121 89772.08
  • aDifferences among groups shown by ANOVA to be statistically significant. bHigh or low cardiologist service volume means one cardiologist caring for at least 1.5 or <1.5 patients a day, respectively. cDifferences between high and low health care load groups significant by t-test. dDifferences between PCI and non-PCI groups significant by t-test. eDifferences between medical centers and regional hospitals significant by t-test. fBonferroni post hoc test. gMortality means (deaths/n)*100. hDifferences among groups significant by chi-square test.

In-hospital mortality stratified by cardiologist service volume, performing PCI and hospital level

The data in Table 2 also show statistically significant differences in in-hospital mortality rates among the eight triple-variable combination groups. For either hospital level, the highest in-hospital mortality rate was observed in the low cardiologist service volume and non-PCI combination, namely, Groups 1 and 2 (11.4% for medical centers and 8.22% for regional hospitals, respectively; P-value for trend = 0.0001). Lower mortality rates were observed in patients receiving PCI: in ascending order of mortality magnitude, Groups 4, 8, 3 and 7 (Figure 2). In particular, Group 4, which exhibited the lowest in-hospital mortality rate (1.75%), represents PCI performed in medical centers with a low cardiologist service volume (Figure 2).

Figure 2.

In-hospital mortality in patients with AMI stratified by cardiologist service volume, performing PCI and hospital level.

Discussion

Our study reveals that the mortality rate of AMI varied significantly in different hospital settings in Taiwan from 1997 through 2010. Low cardiologist service volume medical centers with more PCIs performed were associated with the lowest mortality, whereas the highest mortality was observed in the same hospital settings with fewer PCIs performed. Apparently, patient mortality was reduced if the hospitals boosted their total cardiologist service volume and PCIs performed. However, hospital level seems not to be a factor that affected the mortality. In terms of cost, our analysis showed that Taiwan National Institute paid the most per patient in high cardiologist service volume medical centers with fewer PCIs performed, but paid the least in low cardiologist service volume regional hospitals with fewer PCIs performed. No significant differences in medical costs were found for different hospital settings.

In clinical practice, mortality from AMI is influenced by many factors, including age, Killip class, time delay to treatment, mode of treatment, history of prior myocardial infarction, DM, renal failure, number of diseased coronary arteries and ejection fraction.12 In parallel with the greater use of reperfusion therapy, primary PCI, modern antithrombotic therapy and secondary prevention, several recent studies have highlighted a fall in acute and long-term mortality following AMI.13,14 Aside from patient and disease associated factors, time delay to treatment and mode of treatment are closely related to characteristics of health care and hospital. Previous literature disclosed that admission to higher-volume hospitals was associated with a reduction in mortality for AMI.1 This phenomenon may be attributable to better physician experience, routine treatment algorithms and thorough discharge programs. In additional to volume, experience in performing PCI may shorten the door-to-balloon time and reduce AMI mortality.15,16 Not surprisingly, this study revealed that medical centers or regional hospitals with either a high cardiologist service volume or a greater number of PCIs performed had reduced AMI mortalities. These results are compatible with the findings of previous studies.

Prior studies have demonstrated hospital length of stay to be a crucial contributor to increased cost for AMI.17 However, more recent articles indicated that hospital expenditure and hospital length of stay for the treatment of AMI patients may vary widely depending on the characteristics of the hospital.18,19 Lin et al.19 and Kinjo et al.20 had reported that admission to a high-volume hospital was an important predictor of a shorter length of hospital stay, but there were still discrepancies.21 In our study, the cost of AMI hospitalization was consistent with previous findings and depended largely on specific hospital settings. A larger cardiologist service volume and a greater number of PCIs were associated with increased hospital costs.

Taking into consideration the medical costs and clinical outcomes of AMI, we found that a higher mortality was associated a lower hospital costs (Groups 1 and 2). On the other hand, a lower mortality was associated with a higher hospital costs (Groups 3–8). High cardiologist service volume medical centers with fewer PCIs performed were the least cost-effective, incurring higher medical expenses without a concomitant reduction in mortality (Group 6). However, high cardiologist service volume regional hospitals with more PCIs performed cost less money but achieved a greater mortality reduction (Group 7).

Our findings are also consistent with AMI guidelines22 and have three clear implications. First, with respect to clinical practice, a patient with AMI who was sent to a hospital (whether a regional hospital or medical center) with a smaller cardiologist service volume and fewer PCIs performed should be referred to a hospital that performs more PCIs and has a greater cardiologist service volume in order to minimize mortality. Strengthening emergency medical systems like pre-hospital electrocardiogram (EKG) and transferring to PCI capable hospitals may be of great benefit. Second, with respect to quality control, it is reasonable for all hospitals to increase the number of PCIs they perform. However, it may lead to greater medical costs in most hospital settings except high cardiologist service volume medical centers. Thus, government agencies should encourage all hospitals to develop PCI proficiency after making sound financial considerations, especially high cardiologist service volume medical centers. Third, with respect to hospital development, our study suggests that regional hospitals increase their patient care load before developing their PCI proficiency, but that medical centers develop their PCI proficiency before increasing their patient care volume. The implementation of the above ideas can be expected to promote improved care for AMI while minimizing medical costs.

Conclusion

The strength of this study is that it was a nationwide population-based, longitudinal cohort design study to evaluate the relationships between factors related to hospital structure and process (cardiologist service volume, performing PCI and hospital level) and factors related to health care outcomes (medical costs and mortality) in the treatment of AMI. Despite our meticulous study design to control for confounding factors, a key limitation of this study is the potential for bias because of possible, unmeasured or unknown confounders. In addition, in our study, AMI patients were identified from the 1997–2010 Taiwan National Health Insurance Research Database of the National Health Research Institute. A total of 2942 patients were identified in medical centers and 4325 identified from regional hospitals. We used the 1 million population database which is the subset of NHIRD in this study. The 1 million insured subjects were randomly selected for the population of total 23 million people in Taiwan. For the patients was randomly selected from the subset of NHIRD (1/23 population in Taiwan), therefore our study did not include all medical centers, regional hospitals (There were 20 medical centers and 80 regional hospitals in Taiwan). Thus, from our study, we cannot offer acute annual volume in each hospital. Therefore, further study will help extend our findings to bring positive changes to the country’s health care economy.

Author contributions

Conception/design: C.-Y. Liu, Y.-N. Lin, W.-C. Tsai, C.-H. Kao; provision of study materials: C.-L. Lin, Y.-J. Chang, Y.-H. Hsu; data analysis and interpretation: C.-Y. Liu, Y.-N. Lin, W.-C. Tsai, C.-H. Kao; collection and/or assembly of data, manuscript writing and final approval of manuscript: all authors.

Acknowledgment

This work was supported by the study projects (DMR-102-023 and DMR-103-020) in our hospital; Taiwan Ministry of Health and Welfare Clinical Trial and Research Center for Excellence (DOH102-TD-B-111-004), Taiwan Ministry of Health and Welfare Cancer Research Center for Excellence (MOHW103-TD-B-111-03); and International Research-Intensive Centers of Excellence in Taiwan (I-RiCE) (NSC101-2911-I-002-303).

Conflict of interest: The authors individually or collectively have no significant conflicts of interest or financial disclosures.

Footnotes

  • *These authors contributed equally to this work.

References

View Abstract