The Effect of Cuts in Reimbursement on Stroke Outcome
A Nationwide Population-Based Study During the Period 1998 to 2007
Background and Purpose— As healthcare costs keep rising, cuts in reimbursement such as the Balanced Budget Act in the United States or global budgeting have become the key to healthcare reform efforts. Limited information is available, however, concerning whether reimbursement cuts are associated with changes in stroke outcomes. The objective of this study is to determine whether 30-day mortality rates for patients with ischemic stroke changed under increased financial strain from global budgeting in Taiwan.
Methods— We analyzed all 258 167 patients with ischemic stroke admitted to general acute care hospitals in Taiwan over the period 1998 to 2007 through Taiwan’s National Health Insurance Research Database. Multilevel logistic regression analysis was used to examine whether 30-day stroke mortality rates varied after the implementation of hospital global budgeting since July 2002 adjusted for patient, physician, and hospital characteristics.
Results— The magnitude of payment reduction on overall hospital net revenues was between 4.3% and 10.0%. The 30-day mortality rates for patients with ischemic stroke in Taiwan increased after the implementation of hospital global budgeting after adjustment for patient gender and age, comorbidities, surgery, physician age and volume, specialty, hospital volume, ownership, accreditation level, bed size, geographic location, competition, and trend.
Conclusions— The mortality rate of patients with stroke rose under increased financial strain from cuts in reimbursement. Therefore, stroke outcomes are more likely to be affected by hospital financial pressures. It is imperative to monitor stroke outcomes and develop strategies to maintain levels of stroke care as cuts in reimbursement are adopted.
Stroke is the third most common cause of death in developed countries.1 Stroke mortality has also been regarded as one important performance or outcome indicator to compare countries and hospitals.2,3⇓ Finding the determinants that influence stroke mortality is essential for developing effective initiatives to improve stroke outcomes. Previous studies regarding health system factors have identified associations of weekend admissions, physician specialty and volume, simultaneous contribution of physician and hospital volume, and competition with stroke mortality using nationwide cross-sectional population-based data.4–6⇓⇓ To our knowledge, however, there has been little research on the impact of cuts in reimbursement and the previously mentioned health system factors on changes in stroke mortality using nationwide longitudinal population-based data.
As healthcare costs continue to rise, cuts in reimbursement have become the main focus of healthcare reform efforts such as the Balanced Budget Act in the United States or global budgeting in some other Organization for Economic Cooperation and Development countries and Taiwan.7,8⇓ An important but largely unanswered question is whether such reductions are associated with changes in patient outcomes.7 Moreover, limited information is available regarding stroke.
As hospitals come under increasing financial stress owing to the adoption of the Balanced Budget Act or global budgeting, they may reduce operating expenses to preserve profits. Strategies to reduce operating expenses may include reducing the quantity or quality of nursing staff, reducing efforts to train and improve staff performance, curtailing investment in infrastructure, and reducing levels and availability of services.7,9,10⇓⇓ The strategies responding to cuts in reimbursement may lead to deficiencies in quality of care, which may result in poor patient outcomes. The effect that the strategies may have on patient outcomes will depend on the illness severity of the patient population, because patients with more severe diseases, like patients with stroke,3 may be less able to withstand changes in their care.7,9⇓
The primary goal of this study was to use nationwide population-based data from Taiwan from 1998 to 2007 and apply a multilevel model to examine whether cuts in reimbursement affected 30-day stroke mortality. Our secondary objective was to identify which other health system determinants influenced 30-day stroke mortality over a 10-year period.
We identified all patients with ischemic stroke, aged ≥18 years, admitted to acute care hospitals between January 1, 1998, and December 31, 2007, through the National Health Insurance Research Database provided by the Bureau of National Health Insurance and managed by the National Health Research Institutes. The National Health Insurance Research Database is a national database that contains patient-level demographic, diagnostic, and administrative information across Taiwan. The major inclusion criterion was admission with a principal diagnosis of ischemic stroke as identified through the patient’s principal diagnosis recorded using the International Classification of Diseases, 9th Revision, Clinical Modification codes 433 and 434. When a patient had more than one ischemic stroke admission during the study period, only the first event was included in this analysis. The study patients needed to have a minimum 30-day follow-up period after the initial diagnosis of stroke. Patients with missing data were excluded.
Definition of Variables
The degree of cuts in reimbursement was measured according to the quarterly monetary value of each point. In Taiwan, since July 2002, a hospital global budgeting system (fixed budget) has been adopted to control medical expenditures, which increased rapidly after the implementation of National Health Insurance for the entire population in March 1995. Reimbursement to providers is based on an existing fee-for-service schedule, which lists a relative value or number of points for each item of service. The monetary value of each point was fixed (NT $1 [US $0.03]) until the adoption of global budgeting. Since then, the monetary value of each point has been equal to the fixed budget divided by the number of points, so it fluctuates. The value of NT $0.9460 per point for the fourth quarter of 200711 meant that hospitals had an average decease in National Health Insurance reimbursement of NT $0.0540 per point, which corresponded to a 5.4% reduction in net revenues (Table 1). Thus, the monetary value per 100 points represents the percentage of reimbursement (with 100% indicating the preglobal budget period).
Patient covariates were gender, age, comorbid conditions, administration of surgical operation (yes/no), and weekend admission (yes/no). The Charlson-Deyo index was used to quantify patients’ comorbidities.12 The index is a widely used comorbidity index in stroke outcome studies4–6,13–16⇓⇓⇓⇓⇓⇓ and provides as good a prediction of stroke outcome as stroke-specific comorbidities or other indices developed specifically for stroke.15,16⇓ Thus, the Charlson-Deyo index was selected to facilitate comparisons across studies. This index is the sum of weighted scores based on the presence or absence of 17 different medical conditions. Cerebrovascular disease and hemiplegia were excluded, however, because they are reflected in the condition being evaluated.14 The higher the scores, the greater the comorbidity burden.
Physician covariates were age, specialty (neurology versus others), and physician volume. In Taiwan, there are 23 official physician specialties, including neurology, neurosurgery, internal medicine, surgery, gynecology, pediatrics, and so on. For physician volume, each patient admission was linked with the number of patients with ischemic stroke treated by that physician in the calendar year before the year of the patient’s admission.
Hospital covariates were ownership (public, not-for-profit, private), accreditation level (academic medical center, regional, district), bed size, hospital volume, geographic location (Taipei, northern, central, southern, Kao-Ping, eastern), and competition. For hospital volume, each patient admission was linked with the number of patients with ischemic stroke admitted to that hospital in the calendar year before the year of the patient’s admission. In Taiwan, every patient is free to go to any hospital because all hospitals have contracts with the Bureau of National Health Insurance. Additionally, because stroke care is not regionalized or centralized, patients with acute stroke are generally sent to the nearest hospital. Thus, Taiwan’s healthcare system provides an excellent opportunity to examine the volume–outcome relationship for stroke. Hospital competition was measured by the Herfindahl–Hirschman Index (HHI). We calculated HHI for the 17 medical area networks (markets) demarcated by Taiwan’s Department of Health. HHI is the sum of squared market shares of each hospital in the market, and the market share is the ratio of bed counts in each hospital divided by the total beds in the market. The trend variable (linear time trend) was included to capture all omitted trending variables such as advances in medical technology and to separate them from the effect of hospital global budgeting.17,18⇓
We used 30-day mortality, which the Centers for Medicare and Medicaid Services and the Agency for Healthcare Research and Quality have suggested.3,19⇓ The advantage of using 30-day mortality is that variation in lengths of stay does not have an undue influence on mortality rates. Without a standardized period, institutions would have an incentive to adopt strategies that would shift deaths out of the hospital without improving quality of care. Quality of care can affect patient outcomes in this timeframe. The period of 30 days after admission is used commonly in stroke studies to assess short-term mortality.2,6,13,20⇓⇓⇓ Thirty-day mortality was calculated by linking inpatient admission records with withdrawal certificate records.6,21⇓ Withdrawal dates are the same as death dates based on death certificate records. Because Taiwan’s National Health Insurance is a compulsory single-payer program and, for Taiwanese residents, the only reason for being withdrawn from the National Health Insurance coverage within 30 days of hospital admission would be that they had died (the other 2 conditions for withdrawal being jailed for >2 months or disappearing for >6 months would not be possible reasons for withdrawal within 30 days of hospital admission).
We applied multilevel logistic regression (also known as the hierarchical generalized linear model) to inpatient admissions data over a period of 10 years (1998 to 2007 inclusive) to analyze trends in 30-day ischemic stroke mortality.17 The patient was the unit of analysis. When data comprising patients treated by physicians who practice in hospitals are structured hierarchically, multilevel modeling is appropriate because it provides more accurate results in terms of precise estimates than conventional regression modeling, which ignores the possible correlation of outcomes within a given physician or hospital.22,23⇓ Multilevel modeling also avoids the problem of atomistic fallacy, in which inferences about the groups are incorrectly drawn from individual-level information.24,25⇓ Patients (Level 1) were considered to be nested within physicians at Level 2 and then nested within hospitals at Level 3.
Multilevel logistic regression was used to examine the effects of cuts in reimbursement and other health system factors on 30-day mortality adjusted for patient demographics and complexity, physician age, geographic location, and trend. A cross-level interaction term of physician volume and hospital volume was also included to examine whether the relationship between physician volume and 30-day mortality differed by hospital volume.24,25⇓ The SAS statistical software (Version 9.1) and HLM (Version 6.02) were used for analysis. A 2-sided P value of less than 0.05 was considered statistically significant.
A total of 289 972 patients hospitalized for first-ever ischemic stroke in Taiwan between 1998 and 2007 was used for analysis. After removal of patients with missing data, 258 167 patients were available for analysis. The number of patients with first-ever ischemic stroke was 22 744 from 1299 physicians in 1998, gradually increasing to 27 376 from 1739 physicians in 2007.
The percentage of male patients increased slightly from 57.4% to 58.6%. The mean age of patients increased steadily from 67.8 to 68.7 years. The mean Charlson-Deyo index increased from 0.53 to 0.65. The percentage of patients receiving surgery decreased slightly from 1.6% to 1.5%. The proportion of weekend admissions was almost constant (approximately 23.0%). The mean age of physicians increased from 39.9 to 41.8 years. The percentage of neurologists grew from 65.8% to 77.7%. Mean physician volume increased from 81 cases in 1997 to 94 cases in 2002, but then decreased to 82 cases in 2006. The proportion of not-for-profit hospitals increased from 49.3% to 54.9%. The proportion of academic medical centers increased from 29.2% to 36.7%. The mean number of hospital beds rose from 852 to 1120. Mean hospital volume increased from 420 to 569 cases. The mean HHI declined from 0.055 to 0.042. The 30-day mortality rate decreased from 5.8% to 3.7% (Table 2).
Table 3 presents the results of the multilevel logistic regression analysis examining the influences of reimbursement cuts and other health system factors on 30-day ischemic stroke mortality. After adjusting for patient, physician, and hospital characteristics, there were significant relationships of cuts in reimbursement, weekend admissions, physician volume, competition, and trend with 30-day mortality. There was also a significant interaction between physician and hospital volume on adjusted 30-day mortality.
Owing to cuts in reimbursement, there was deterioration in 30-day ischemic stroke mortality subsequent to the introduction of hospital global budgeting. The analysis found that an increase in reimbursement of NT $1 per 100 points would reduce the odds of 30-day ischemic stroke mortality by 1.2% (OR=0.988; 95% CI=0.979 to 0.997). In other words, for every NT $1 decrease per 100 points, which corresponded to every 1% reduction in average hospital net revenues, the odds of 30-day ischemic stroke mortality rose by 1.2%.
Regarding other health system determinants, patients admitted on weekends had 12.8% higher odds of 30-day death compared with those admitted during weekdays (OR=1.128; 95% CI=1.087 to 1.172). Patients had 24.0% lower odds of 30-day mortality per additional 100 patients treated by a physician (OR=0.760; 95% CI=0.700 to 0.827). Physician volume had a stronger relationship with 30-day mortality in higher-volume hospitals (OR=0.977 with each additional 100 patients in hospital volume; 95% CI=0.963 to 0.990). Lower competition (higher HHI) led to higher mortality (OR=2.104; 95% CI=1.570 to 2.820). The trend was monotonic (P<0.05). There was a trend toward decreased ischemic stroke mortality per year (OR=0.941; 95% CI=0.928 to 0.955).
Our results using nationwide longitudinal population-based data show that larger cuts in reimbursement from hospital global budgeting of July 2002 led to higher 30-day ischemic stroke mortality. In terms of other health system determinants, weekend admissions, lower physician volume, simultaneous contribution of lower physician volume and lower hospital volume, and lower competition also increased 30-day ischemic stroke mortality.
The finding of the impact of cuts in reimbursement on ischemic stroke mortality is similar to Seshamani et al7 regarding 8 postoperative complications but differs from Volpp et al26 concerning acute myocardial infarction and Seshamani et al9 regarding hip fracture, stroke, acute myocardial infarction, and gastrointestinal hemorrhage. The reason why Volpp et al did not verify the effect of Medicare reimbursement cuts on in-hospital acute myocardial infarction mortality may be that in-hospital mortality is a worse outcome measure than 30-day mortality. The disadvantages of using in-hospital mortality are that variation in length of stay has an undue influence on mortality rates and that institutions would have an incentive to adopt strategies that would shift deaths out of the hospital.3,19⇓ As for Seshamani et al’s research, using 30-day mortality rather than in-hospital mortality as the outcome measure, they did not find the effect of cuts in Medicare reimbursement on 30-day stroke mortality in Pennsylvania. One likely explanation is that in the United States, clinical practice guidelines for stroke are well defined and implemented so quality of stroke care may be more likely to be maintained and monitored despite increased financial strain from the Balanced Budget Act.9 Conversely, in Taiwan, guidelines are still being developed.
Another possible explanation for the adverse impact of cuts in reimbursement on ischemic stroke mortality in Taiwan is that hospital operating expenses declined at rates similar to the degree of reduction in hospital net revenues, suggesting that hospitals might cut their operating expenses to maintain profit margins. Decreasing operating expenses may lead to deficiencies in quality of care, which patients with more severe diseases such as stroke may be less able to resist so they have higher mortality.3 The methods of reducing operating expenses include decreasing service provision, providing lower quality of care, reducing nurse staffing, recruiting nurses with low education levels, or others. Previous studies showed that decreased nurse staffing or lower educational level of nurses was associated with higher mortality.27,28⇓ Therefore, there is every likelihood that financial stress affects quality of stroke care without implementation of clinical practice guidelines.
Regarding other health system determinants, we found using a 10-year longitudinal study that patients with stroke admitted on weekends had a higher mortality rate than those admitted on weekdays. The result is consistent with previous cross-sectional studies.4–6⇓⇓ The weekend–outcome relationship for stroke may have lasted for many years because hospitals may have faced shortages of medical staff and specialized services on weekends. Thus, disparities in medical resources, expertise, or staffing during weekends may explain the differences in weekend mortality.4
Our study also confirms the volume–outcome relationship and the synergistic contribution of physician and hospital volume for ischemic stroke. The findings that patients treated by physicians with higher volume in the previous year had lower stroke mortality and higher-volume physicians in higher-volume hospitals showed much better outcomes in the next year are similar to another cross-sectional study.6 The explanation for the volume–outcome relationship is that experience in managing a clinical condition may directly affect patient outcomes and that the skill and experience of individual physicians is a critical determinant for patient outcomes. Meanwhile, higher-volume physicians in higher-volume hospitals can receive more assistance from more skillful and interdisciplinary care teams to treat and care for patients, so they are much more likely to improve patient outcomes.6 Moreover, the findings that higher physician volume led to higher stroke mortality but physician specialty (neurologist versus nonneurologist) did not are consistent with another study using data from the Registry of the Canadian Stroke Network.13 Therefore, physician experience rather than physician specialty may play an important role in improving stroke outcomes.
We found that higher hospital competition led to lower stroke mortality. The finding is similar to previous studies.6,29⇓ Hospitals in more competitive market areas are more likely to compete on quality of care rather than just on price or amenities after the implementation of managed care or National Health Insurance.29,30⇓ Therefore, it is possible that hospitals that face a higher degree of competition are compelled to improve overall quality of care, which results in enhanced stroke outcomes. Lastly, the finding that a long-term trend improved stroke mortality is consistent with the advances in medical technology, indicating a steady improvement in stroke mortality for the period from 1998 to 2007. For example, antithrombotic drugs have been gradually applied to patients with stroke.
There are 2 limitations of our study that deserve comment. First, in common with prior studies using administrative databases, we do not have information on stroke severity for stroke risk adjustment. Nevertheless, we controlled for patient age, comorbid conditions, and surgical operations, which are also important for adjustment of stroke complexity.4–6⇓⇓ Second, owing to lack of information on processes of care and staff level, we could not identify the mechanisms through which cuts in reimbursement influence stroke outcomes. It is also possible that other unavailable variables such as number of concomitant medications taken by the patient may explain the differences in stroke mortality.
Our national longitudinal population-based study showed a robust positive relationship between cuts in reimbursement and stroke mortality. As healthcare payers in many countries have gradually implemented reimbursement cuts to control healthcare expenditures, information on whether reductions in reimbursement influence stroke outcomes becomes very important. Our analysis provides a first look into the long-term impact of reimbursement cuts and finds that larger reimbursement reductions led to higher 30-day ischemic stroke mortality. Therefore, it is imperative to monitor processes and outcomes of care for stroke and develop strategies to maintain levels of stroke care as cuts in reimbursement are adopted. For instance, national protocols and guidelines should be implemented not only to assist all physicians to achieve better outcomes for stroke care, but also to aid continuous monitoring of whether stroke care is deficient.
Sources of Funding
This study was supported by grants from the National Science Council (NSC97-2410-H-130-011) in Taiwan and is based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of Bureau of National Health Insurance, Department of Health, or National Health Research Institutes.
- Received September 27, 2009.
- Revision received November 5, 2009.
- Accepted November 26, 2009.
- ↵World Health Organization. The Atlas of Heart Disease and Stroke. Geneva: WHO; 2004.
- ↵Moon L, Moise P, Jacobzone S. Stroke Care in OECD Countries: A Comparison of Treatment, Costs and Outcomes in 17 Countries. Paris: Organization for Economic Cooperation and Development; 2003.
- ↵Agency for Healthcare Research and Quality. Guide to inpatient quality indicators: quality of care in hospitals—volume, mortality, and utilization. 2007. Available at: www.qualityindicators.ahrq.gov/iqi_download.htm. Accessed May 31, 2008.
- ↵Saposnik G, Baibergenova A, Bayer N, Hachinski V. Weekends: a dangerous time for having a stroke? Stroke. 2007; 38: 1211–1215.
- ↵Saposnik G, Baibergenova A, O'Donnell M, Hill MD, Kapral MK, Hachinski V. Hospital volume and stroke outcome: does it matter? Neurology. 2007; 69: 1142–1151.
- ↵Chang YC, Li YF, Wang FK, Cheng SH. The impact of global budgeting of National Health Insurance: a preliminary study on dental and primary care facilities. Taiwan J Public Health. 2006; 25: 152–162.
- ↵Bureau of National Health Insurance. The monetary value of each point under global budgeting. 2009. Available at: www.nhi.gov.tw/webdata/webdata.asp?menu=3&menu_id=470&webdata_id=2933&WD_ID=539. Accessed September 1, 2009.
- ↵Saposnik G, Hill MD, O'Donnell M, Fang J, Hachinski V, Kapral MK. Variables associated with 7-day, 30-day, and 1-year fatality after ischemic stroke. Stroke. 2008; 39: 2318–2324.
- ↵Goldstein LB, Samsa GP, Matchar DB, Horner RD. Charlson index comorbidity adjustment for ischemic stroke outcome studies. Stroke. 2004; 35: 1941–1945.
- ↵Bushnell CD, Lee J, Duncan PW, Newby LK, Goldstein LB. Impact of comorbidities on ischemic stroke outcomes in women. Stroke. 2008; 39: 2138–2140.
- ↵Centers for Medicare & Medicaid Services. Frequently asked questions (FAQs): implementation and maintenance of CMS mortality measures for AMI & HF. 2007. Available at: www.cms.hhs.gov/HospitalQualityInits/07_MortalityMeasures.asp. Accessed June 1, 2008.
- ↵Rice N, Leyland A. Multilevel models: applications to health data. J Health Serv Policy. 1996; 1: 154–164.
- ↵Roux AVD. A glossary for multilevel analysis. J Epidemiol Community Health. 2002; 56: 588–594.
- ↵Hox JJ. Multilevel Analysis: Techniques and Applications. Mahwah, NJ: Lawrence Erlbaum Associates; 2002.
- ↵Volpp KG, Konetzka T, Zhu J, Parsons L, Peterson E. Effect of cuts in Medicare reimbursement on process and outcome of care for acute myocardial infarction patients. Circulation. 2005; 112: 2268–2275.