Hospital admissions are important surrogate measures for disease worsening and increased demand on healthcare resources; few studies have examined hospitalizations in multiple sclerosis (MS). We examined hospital admission rates and patterns in a large Canadian MS cohort.

Data from the British Columbia MS database were linked with hospital separation and registry administrative data, 1986 to 2008. Main outcomes included all-cause hospital admission rates and length of stay. The influence of time and patient characteristics was examined using multivariable regression models.

Overall rate of all-cause admissions was 32.4 per 100 MS patients. Rates decreased by 1.4% (adjusted incidence rate ratio [IRR] 0.986; 95% confidence interval [CI] 0.982-0.990) per year from 1986 onward. Higher admission rates were associated with older age (adjusted IRR 1.011; 95% CI 1.007-1.014), primary progressive MS (adjusted IRR 1.294; 95% CI 1.162-1.441), and a longer disease duration (adjusted IRR 1.030; 95% CI 1.027-1.034). Mean length of inpatient stay was 10.2 (standard deviation [SD] 24.8) days, and increased over time. Hospital stays were longer for older patients and those with a longer disease duration, but were not influenced by sex or disease course.

Admission rates for MS patients have decreased since 1986, but length of stay has increased. Patients with a longer disease duration and those with primary progressive MS had higher rates of admission and longer stays. Understanding the impact of time and patient characteristics on hospitalizations is important for resource allocation planning and designing future research studies examining interventions and treatments for MS.

Patients with multiple sclerosis (MS) are relatively high users of the healthcare system, but little is known about one of the most financially burdensome elements of healthcare use—hospitalizations.

		Hospitalization rates have decreased over time, but length of stay has increased.

		Both patient (age) and disease (duration and course) characteristics were associated with the number or length of hospital stays.

		These study results will help aid in appropriate resource allocation planning, and are useful in planning future studies examining the impact of various MS interventions and treatments.

Multiple sclerosis (MS) is a chronic disease and the most common cause of non-traumatic neurological disability in young adults

 placing a considerable burden on patients, families, and the healthcare system. In fact, MS patients are more than twice as likely to visit a medical professional or be hospitalized as individuals without MS.

Hospital admissions are measured outcomes in both clinical trials and observational studies,

 and are often considered important surrogate measures for disease worsening and overall demand on healthcare resources.

 Thus, understanding hospitalization patterns is imperative to facilitate appropriate resource allocation, can aid in the evaluation of disease-management strategies, and can provide an indication of drug effectiveness. Despite this, few studies have specifically examined hospital admission patterns in MS.

 One American study evaluating admissions captured by a large inpatient care database found MS-related hospital stays average 5 days, remaining stable from 1998 to 2006.

However, much remains unknown; to date, few others have investigated hospital admission rates over time in MS, or done so in a setting where healthcare is universally covered for all residents. The purpose of this study was to examine patterns, temporal trends, and patient characteristics influencing hospital admissions in a large MS cohort in British Columbia, Canada, from 1986 to 2008.

 criteria) registered at one of the 4 British Columbia MS clinics by December 31, 2004, were identified through the British Columbia MS (BCMS) database. Established in 1980, the database contains prospectively collected clinical information on approximately 80% of the British Columbia MS population during the study period. We linked this clinical data with the British Columbia Ministry of Health’s Discharge Abstract (hospital separations) database and Registry and Premium Billing Files (to confirm residency in British Columbia

) at the individual level using unique health identification numbers. The latter linkage provided registration dates for all British Columbia residents enrolled in the universal healthcare plan, providing confirmation of a patient’s residence in British Columbia throughout the study period. These databases have served as the basis for many observational studies, and are considered to be of high quality.

Data were extracted for all hospital admissions from 1986 (the first full year both hospital and registry data were available) to 2008 and included admission and discharge dates, length of stay, primary diagnostic classification (ie, reason for admission), and up to 24 additional diagnostic classifications (

 code of G35 indicated an admission due to MS. Only admissions occurring after onset of MS symptoms (determined by the MS neurologist and recorded in the BCMS database) were included in the analyses.

The main outcome in this descriptive study was annual all-cause hospital admission rates from 1986 to 2008. Hospital admissions were categorized as “inpatient” (acute admission requiring at least 1 overnight stay), “day care surgery” (admitted and discharged on the same day), and “all-cause” (all admissions [inpatient and day care surgery], regardless of reason, level of care, or length of stay). Day care surgery admissions typically involve a medical procedure, but do not include drug infusions (eg, intravenous corticosteroids or natalizumab). Secondary outcomes included admission rates attributed to MS (ie, with a primary or secondary diagnostic code for MS [

Annual admission rates (all cause and MS specific) were estimated by dividing the total number of yearly hospital admissions by the number of MS cohort patients who were resident in British Columbia during that year, and reported per 100 patients with MS. Quasi Poisson regression, fitted using the method of generalized estimating equations (GEEs) with an exchangeable working correlation structure, was used to examine changes in all-cause admission rates over time. The following covariates were determined to be clinically relevant or statistically important based on univariate analyses (

 <.1) and were therefore included in the models: sex, age at observation year, disease course (relapsing-onset or primary progressive), disease duration, and calendar year. To account for differences in the British Columbia residency time between patients, the logarithm of time registered with the British Columbia Ministry of Health for each year was included as an offset in the model. No meaningful first-order interactions were detected (

 <.1). Findings were reported as adjusted incidence rate ratios (IRRs), with 95% confidence intervals (CIs).

Admissions were further examined only in patients with at least 1 hospital admission during the study period to identify potential factors associated with admission with a primary diagnosis of MS. A multivariable logistic regression model fitted by the GEE method, with an exchangeable working correlation structure, was used, with the same covariates as above. An interaction between sex and age at observation year (

 <.001) necessitated separate models for males and females. Findings were reported as adjusted odds ratios with corresponding 95% CIs.

The mean length of hospital stay was examined for inpatient admissions only; a separate mean was also estimated for inpatient admissions where the primary admission reason was MS. To identify factors associated with a longer hospital stay for all inpatient admissions, a negative binomial regression model

 fitted by the GEE method, with an exchangeable working correlation structure, with the same covariates as above, was used. No significant interactions were detected (

 <.1). Findings were reported as adjusted IRRs with corresponding 95% CIs. Statistical analyses were carried out using SPSS version 18.0 (SPSS Inc, Chicago, Illinois) and R: A Language and Environment for Statistical Computing V.2.13.2 (R Foundation for Statistical Computing, Vienna, Austria; 2011). This study was approved by the University of British Columbia’s Clinical Research Ethics board (study number: H08-01544).

A total of 6859 patients with a neurologist-confirmed diagnosis of definite MS were identified from the BCMS database, and 6601 (96.2%) were successfully linked to the British Columbia Ministry of Health’s databases. Overall, 72.0% of patients were female and 90.7% had relapsing-onset MS. The mean age at symptom onset was 32.0 (standard deviation [SD] 10.2) years. Between January 1, 1986, and December 31, 2008, 5138 patients had 31,375 hospital admissions (

The overall rate of all-cause admissions was 32.4 admissions per 100 patients with MS, which decreased from an annual rate of 39.2 to 29.0 per 100 patients with MS between 1986 and 2008 (

). Admission rates, estimated using multivariable regression, decreased by 1.4% (adjusted IRR 0.986, 95% CI 0.982-0.990) per year from 1986 onward (

). When examined separately, the rate of inpatient admissions decreased over time while the rate of day care surgery admissions increased (Figure 1). Higher IRRs were associated with primary progressive MS (adjusted IRR 1.294, 95% CI 1.162-1.441), an older age at hospital admission (adjusted IRR per year 1.011, 95% CI 1.007-1.014), and a longer disease duration (adjusted IRR per year 1.030, 95% CI 1.027-1.034). Women had a lower rate of admission compared with men (adjusted IRR 0.875, 95% CI 0.811-0.945) (Table 2).

Rates of admissions for MS decreased over time, although they differed considerably depending on whether MS was listed as the primary or as a secondary reason for admission. Where the primary reason for admission was MS, the overall rate per 100 MS patients was 5.9, decreasing from an annual rate of 14.1 in 1986 to 1.6 in 2008. When MS was listed as a secondary reason for admission, the overall rate was 10.7 per 100 patients with MS, decreasing from an annual rate of 12.1 in 1986 to 8.2 in 2008 (

). However, the odds of admission to hospital where MS was the primary reason have increased by approximately 9% per year when sex, age at observation, disease course, and disease duration are taken into account (

). Older men were more likely to have a primary admission due to MS compared with younger men (adjusted odds ratio [OR] per year 1.026, 95% CI 1.016-1.036); this association with age was not significant for women (adjusted OR 0.995, 95% CI 0.988-1.001) (Table 3). Disease course and disease duration had no association when MS was the primary reason for admission (Table 3).

The mean length of stay for all inpatient hospital admissions between 1986 and 2008 was 10.2 (SD 24.8) days, and 13.2 (SD 25.9) days when restricted to only inpatient admissions where the primary reason was MS (

). The length of hospital stay for MS patients increased with age (adjusted IRR per year 1.013, 95% CI 1.009-1.018). Patients with a longer disease duration had relatively longer hospital stays (adjusted IRR 1.010 per year, 95% CI 1.004-1.016). Sex or disease course were not associated with an extended hospital stay (

In this retrospective British Columbian cohort study, 78% of the MS patients had at least 1 hospital admission between 1986 and 2008, with an overall admission rate of 32 per 100 patients. While admission rates have decreased over the last 2 decades, the mean length of stay has increased. MS-related clinical characteristics of the cohort appeared to influence our results. There are few other MS-specific studies with which to compare our findings, and although similar general patterns in hospitalizations are evident in the overall Canadian population,

 we did observe some unanticipated differences.

Sex was associated with admissions in unexpected ways. Men with MS were more likely to have a hospital admission compared with women with MS, although length of stay was similar for men and women after adjusting for factors such as age and disease course. Findings were in contrast to those of a recent study evaluating the geography of hospital admissions for MS in England,

 as well as those typically found in the general population, where women tend to be higher users of the healthcare system,

 Possible explanations might include the more rapid disease progression (from MS onset) observed in men,

or the higher odds of physical comorbidity reported in men with MS.

Articles

MANAGERIAL

An examination of hospitalization patterns in patients with multiple sclerosis with a focus on the association with time and patient characteristics.

Hospital Admissions and MS: Temporal Trends and Patient Characteristics

Although we recognize that implementation is challenging, we recommend a value strategy including market based and regulatory activity rather than limiting coverage for individuals through high deductibles.

Benefit design and coverage should reward the highest performing providers.

Purchasing strategies fostering delivery system reform (eg, patient-centered medical home, accountable care organizations) should align into multipayer efforts to get the greatest impact; this does not mean that all the measures and payment methods need to be identical.

Purchasers should reward the health plans that deliver on better quality and lower costs.

The rise in healthcare spending is affecting business, government, and individuals. Too often the reaction is to shift large chunks of financial risk to beneficiaries, which lowers cost but can work against the agenda to address the underuse of needed services and better health. Also, the shift does nothing to send price and quality signals to the delivery system. With more use of reference pricing with quality built in, for example, providers who deliver care more efficiently would receive greater market share. This would make efficiency a “business issue” for providers. Although we do not underestimate the challenges of healthcare system reform, change to delivery systems, incentives, and expectations offers a better path to getting more value for healthcare spending.

So often, health plans, providers, and beneficiaries have no good reason to obtain better value. Health plans can have higher revenues if they do less to manage total spending; indeed, new rules limiting profits, administration, and utilization management may reward the plans that do the least to manage spending. As has been amply discussed elsewhere, fee-for-service payment systems to doctors, hospitals, and other providers reward volume over value. Current payment norms also reward specialty care over primary care and procedures over cognitive services, which heavily affects the mix of what is being delivered. Providers that are successful in commanding high fees through market power or cache also are rewarded with higher revenues. Due to the way insurance coverage— both the premiums and cost sharing&mdash;is generally structured, the consumer who seeks out more cost-effective care receives no financial reward for doing so. Moreover, there is usually little financial consequence for people who make unhealthy choices; eg, have a poor diet, lack exercise, do not take charge of chronic conditions.

Value-based purchasing offers a set of tools to achieve 2 important goals:

It pays providers for what consumers want the healthcare system to do, ie, reward value and quality.

It engage consumers to be sophisticated and knowledgeable users of healthcare and therefore encourages them to do more to improve their health.

We see purchasers—employers and government programs&mdash;as being the primary actors for pushing these changes; health plans can be partners. To the degree possible, purchasers should send consistent signals to plans, providers, and consumers about what they want the healthcare system to achieve. Providers facing different incentives and measures from different payers often have no way to respond effectively. Ironically, if all pile on in a similar direction but in different ways, the signal to noise ratio goes down. Alignment among payers is very important. This does not necessarily mean that every initiative needs to use the same measures and payment methods&mdash;innovation is important in this area and we still need to learn more about what works best. But we have seen examples of multipayer initiatives, eg, the patient-centered medical home, where collaboration among payers has sent a strong signal to primary care physicians, resulting in impressive outcomes.

Specific recommendations include the following:

Aligning cost sharing with value. Cost sharing would be lower for high-value services (eg, maintenance care for chronic conditions) and higher for low-value services.

Another example of value-based purchasing is the use of “centers of excellence.” Health plans can use cost sharing to guide patients to high-value hospitals and providers. They can identify hospitals that are centers of excellence for treating high-cost or high-risk conditions, and make them preferred providers with the lowest cost sharing for patients. This approach could also reward hospitals or other providers that have strong patient safety records. The high-value plan of the future should rely heavily on value metrics to select its network. Ensuring that members have access to physicians with good credentials is an important part of consumer protection and of the National Committee for Quality Assurance (NCQA) health plan accreditation. But excellent health plans must also measure the performance of providers. They must use that information to build networks and report the information to consumers.

The last 3 years have seen an encouraging trend toward delivery system reform, and away from traditional fee-forservice payment methods that reward volume over value. The growth in the number of practices and providers involved in patient-centered medical home programs is a great example of this trend. This growth has been stimulated by the many purchasers—both public and private&mdash;that have embraced this effective model of care. Public purchasers include many state Medicaid programs and federal government agencies like the Health Resources and Services Administration and the Department of Defense. Medicare is following the lead of these types of innovators and making payments for Medicare beneficiaries who see clinicians participating in existing multipayer pilots.

This strategy steers physicians and patients to the most cost-effective treatments by tying reimbursement for an item or service to the price of the most effective treatment. Less cost-effective treatments are still covered by the plan, but members pay more for them. Reference pricing needs to be done in a way that builds in quality, so there would need to be a quality threshold as well, except for items and services where quality is indistinguishable.

Now that there is a firm start toward delivery system reform made for primary care, the next challenge is building from this model and developing others that include specialists and hospitals. One way is through the accountable care organization (ACO), with Medicare leading development. Other purchasers, including private insurance companies and Medicaid programs, are beginning to experiment with the concept; however, employers are not driving forces in this area. To make it more worthwhile for providers to invest in the systems and other organizational changes needed to run successful ACOs, payers should work together to set the same requirements for ACOs. Based on our experience with the patient-centered medical home, we also recommend that payers go beyond the set of performance measures in these initiatives to include standards that articulate, in clear and unambiguous terms, what program elements are needed to lead to success and can be evaluated consistently. Too often these programs require lengthy descriptions that can be difficult to evaluate and thus not terribly useful.

Another approach for building from the patient-centered medical home is in the “patient-centered home neighborhood” idea, aspects of which the NCQA is building into its specialty practice recognition program. This program is designed to recognize the commitment specialists make to not only coordinate the care of their patients inside their practice but also to coordinate with primary care providers. Although health reformers have developed more complex payment models with stronger incentives than the simpler versions of the pay-for-performance schemes of the 1990s, rigorous and meaningful measurement is still critical to help both plans and providers focus on what they need to do to be successful and to reassure consumers that any economies achieved will not be at the price of lower quality.

Employers should further the value agenda by awarding higher payments to plans that demonstrate high value. Many state Medicaid agencies have such programs; the Medicare Advantage program has one as well and we are beginning to see early signs of performance increases. Reference pricing can be used for choosing health plans also. This would make the patient pay the difference between the “benchmark plan” based on cost and quality and higher-priced options.

While we support many of the policies regarding health plans in exchanges, we wish there were greater emphasis on using exchanges as a strategy to pursue a value agenda. Apart from a few states, the emphasis seems to be more on operational issues. While this emphasis is understandable, given the enormity of the task, we would like for the pace to pick up on articulating quality measures and encouraging health plans participating in exchanges to participate in statewide delivery reform initiatives. More investment should be made in developing useful measures of value—overuse, misuse, and waste. In an effort to keep subsidies relatively affordable, the benefit packages offered in the exchanges wil likely have high deductibles. We encourage both federal and state policy makers to look for flexibility in the law to use elements of value-based insurance design to keep premiums down in place of high cost-sharing requirements. For example, consumers using exchanges could be allowed to reduce the deductible if they agree to get care in a patient-centered medical home, or participate in wellness or care management programs. Plans and purchasers are developing benefit designs that waive cost sharing for effective services like those measured by the Healthcare Effectiveness Data and Information Set.

As the Congress contemplates permanent changes to the physician payment system in Medicare, now is the time to use the powerful payment incentives in payment systems to motivate better outcomes in quality and cost across all types of physicians. Medicare has taken some steps to rebalance payment in favor of primary care, but more should be done to reward the challenging work of primary care. Medicare should also adopt payment strategies that create incentives for physicians to provide care outside of face-to-face visits, including the work of coordinating care among different providers. Given the upcoming shortage of primary care providers and some specialists, capacity can only grow if the payment system creates innovative ways to deliver care through use of professional teams and through other methods of communication.

We strongly support strategies to make healthcare costs, including prices, more visible to policy makers, purchasers, and consumers, and to create rewards to the consumers who act on this information. We recognize this is not a simple task given the wide variety of payment methods and accounting systems. But as health economists often find that higher prices underlie higher costs, it serves a very important public policy objective to make prices more visible as long as the information is presented in a fair yet meaningful way. Once these systems are built, we support using strategies like tiering and reference pricing to give consumers incentives to choose higher-value (lower-cost and higher-quality) alternatives.

Health plans and the new delivery system entities taking responsibility for patients’ care over time and across types of providers should incorporate strategies to improve health and consumer engagement. Examples of these strategies include:

— Promoting wellness and health promotion through health appraisals and other strategies. Using financial incentives to encourage participation in programs designed to improve health; eg, smoking cessation and weight loss programs.

— Publicly reporting provider performance and involvement in delivery system reforms.

— Providing members with incentives like lower copays to use shared decision making to choose therapies.

— Covering palliative and end-of-life care and implementing strategies to ensure that providers know and follow patient preferences and decisions.

— Surveying enrollees about their experiences and how they rate providers. Then, use this information to provide feedback to physicians and construct networks.

None of these strategies alone can solve the problems of poor quality, poor health, and high healthcare costs. But synergies between them will reinforce key policy goals to improve efficiency and results.

All that we have enumerated above are pure “market-based strategies,” but we believe that a number of more classic regulatory strategies would help lower costs without diminishing the value of care. Ideas worth exploring are to simplify administrative systems, crack down on excessively high prices, work to reduce medical malpractice costs by creating safe harbors for high quality providers, and getting rid of barriers that prevent health professionals from practicing at the full scope of their training.

 From Public Policy and Communications (ST), National Committee for Quality Assurance (MOK), Washington, DC.

The authors (ST, MOK) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Concept and design (MOK); drafting of the manuscript (ST); critical revision of the manuscript for important intellectual content (ST); and administrative, technical, or logistic support (ST).

Sarah Thomas, MS, Vice President, Public Policy and Communications, National Committee for Quality Assurance, 1100 13th St NW 1000, Washington, DC 20006. E-mail: thomas@ncqa.org.

It has taken decades for American healthcare to become the expensive, complex, and often dysfunctional enterprise it is today. While there have been attempts to redesign policy and payment to address some of the major drivers of cost and complexity, they have often been relatively timid and narrow attempts to tweak the system, and have achieved very limited results. Too often, the different sectors that pay for healthcare have worked at cross purposes, sending confusing signals to providers, or allowing costs to be shifted onto other payers or patients. We are at a unique historical moment when the costs of care are a cause of widespread concern, when new and promising models of care delivery are beginning to emerge, and when technologies to engage the patient are affordable and widely available. A comprehensive strategy of value purchasing can align the interests of health plans, providers, and patients to achieve greater value for the healthcare dollar.

COMMENTARY

Purchasers-employers and government programs-are primary actors for pushing for payment, benefit design, and transparency initiatives to get better value; health plans can partner.

Value-Based Purchasing

Human epidermal growth factor receptor 2 (HER2) expression is amplified in about 20% of breast cancer tumors, and evaluation of HER2 status should influence therapy selection. A critical gap in our knowledge is the real-world implementation of HER2 testing and its impact on treatment decisions for women diagnosed with breast cancer.

To assess use of HER2 testing, to describe characteristics of patients who do or do not receive HER2 testing, to describe which HER2 tests were used (fluorescence in situ hybridization or immunohistochemistry), and to evaluate trastuzumab use as a function of HER2 results.

The population included 6460 women diagnosed with invasive breast cancer between 1999 and 2007 at 8 geographically distributed Cancer Research Network healthcare delivery systems in the United States.

Electronic records were used to identify patient and tumor characteristics and treatment with trastuzumab. Chart abstraction was performed for 400 women (50 per site) to identify receipt of HER2 testing and results.

More than 90% of study participants received HER2 testing. Everyone who received trastuzumab had a HER2 test, and nearly all (>95%) who received trastuzumab had a positive HER2 test result recorded in their medical chart. Most (77%) eligible patients with a positive HER2 test result diagnosed after 2005 received trastuzumab. This study expands upon previous work in individual health plans.

HER2 status has been successfully incorporated into medical practice to guide treatment decisions for breast cancer patients in diverse integrated healthcare delivery settings.

We evaluated real-world implementation of human epidermal growth factor receptor 2 (HER2) testing and its impact on treatment decisions for women diagnosed with breast cancer between 1999 and 2007 at 8 US integrated delivery systems. There was consistent success in implementation of HER2-guided treatment decisions for breast cancer patients in these settings.

More than 90% of study participants received HER2 testing.

Everyone who received trastuzumab had a HER2 test, and more than 95% of these women had a positive HER2 test result.

Most (77%) eligible patients with a positive HER2 test result diagnosed after 2005 received trastuzumab.

In 2011, about 230,000 women in the United States were newly diagnosed with invasive breast cancer.

About 40,000 of these women are expected to die of this disease. Recent advances in medicine utilizing genetic-based technologies better predict cancer recurrence or response to various treatments and present an important opportunity for patients and their physicians to individualize therapy. However, there are limited reports in the literature about how such technologies are being incorporated into clinical practice and whether they are being used appropriately.

 As genetic-based technologies proliferate, understanding how and when these applications should be used in clinical practice becomes increasingly important.

One such genetic technology targets human epidermal growth factor receptor 2 (HER2). HER2 is a gene that influences cell growth, division, and repair. A normal cell has 2 copies of HER2. About 18% to 20% of breast cancers have HER2 gene amplification (more than 2 copies of this gene).

 HER2 amplification is associated with rapid tumor proliferation, shorter disease-free survival, and poorer overall survival.

The drug trastuzumab (Herceptin, Genentech Inc, San Francisco, California) is indicated for individuals with breast cancer that either overexpresses the HER2 protein or amplifies the HER2 gene to prevent the growth of HER2-positive cancer cells.

 It was approved by the US Food and Drug Administration (FDA) for use in the treatment of metastatic breast cancer in 1998 based on the findings from 2 clinical trials, which showed improvement in the median time to progression of 3 to 5 months, and an increase in the overall response rate of 10% to 30%.

 Despite the fact that trastuzumab is also well established as a highly efficacious (adjuvant) antineoplastic agent for HER2-positive breast cancer, it is expensive ($44,000-$65,000 per year

) and may have serious cardiotoxic effects.

 Because of these factors, selecting patients for whom trastuzumab treatment is appropriate remains vital. Trastuzumab treatment should be cautiously used for patients with preexisting cardiac dysfunction or deteriorating cardiac function even if they have a positive HER2 test result, especially older women with prior cardiac history or prior cardiotoxic therapies.

Two types of tests are FDA approvedto select patients for treatment with trastuzumab. The first test is a based on immunohistochemistry (IHC) (eg, Dako HercepTest, Ventana PATHWAY), which measures membrane expression of HER2 receptor protein. The second type of test (eg, Vysis PathVysion, Ventana INFORM HER2 probe) is based on the fluorescence in situ hybridization (FISH) technology to measure the underlying gene amplification in tumor cells. In this approach, fluorescent probes are used to quantify the number of HER2 gene copies in a tumor cell. If there are 2 or more copies of the HER2 gene per chromosome 17, then a gene amplification has occurred, and the breast cancer is considered HER2 positive.

The American Society of Clinical Oncology and College of American Pathologists HER2 testing guidelines currently recommend initial testing by IHC testing and reflex testing of equivocal result (2+) by FISH testing.

 Although some reports have suggested that the FISH technology more accurately predicts response to trastuzumab than IHC technology,

 a recent summary report indicates that both technologies perform equally well if carefully validated testing is performed.

 Cost may also be a factor in the selection of test modality, because the FISH test is generally more expensive to perform than the IHC test. For both testing methods, underlying quantitative measures can be translated into a dichotomous positive/negative finding. Both tests are considered equivocal when the quantitative measures are near the threshold for a positive result. Retesting might resolve these equivocal findings.

 Increased error rates and inconsistent results are also introduced by decentralized testing, particularly for centers with a low volume of testing. Thus, Wolff and colleagues

 recommend a quality assurance program for HER2 testing, and suggest various accreditation, proficiency, and competency measures for laboratories conducting HER2 testing.

Trastuzumab was initially approved by the FDA in 1998 for use in breast cancer patients with metastatic cancer. In November 2006, the FDA approved the use of trastuzumab in the adjuvant setting for women with lymph node—positive and HER2-positive breast cancer, following several major trials.

 Professional organizations, including a joint guideline from the American Society of Clinical Oncologists and College of American Pathologists, currently recommend that HER2 testing should be performed for all invasive breast cancers regardless of lymph node disease status.

The National Comprehensive Cancer Network has also issued physician practice guidelines endorsing HER2 testing.

 Despite these evidence-based practice guidelines, there are few reports on how HER2 testing has been utilized in realworld settings. In particular, Phillips and colleagues

 highlighted the fact that little is known about how many patients are tested for HER2, which testing methods are used, whether patients are retested to confirm indeterminate results, and how many patients with negative or equivocal results receive trastuzumab. Additionally, it is not clear how trastuzumab is utilized across important subgroups such as patient age, race, and cancer stage.

To answer these questions, we conducted a large, multiinstitutional study in 8 Cancer Research Network (CRN) integrated healthcare delivery systems across the United States. The organizations participating in this study provide a unique resource to address this research question by providing access to longitudinal and comprehensive electronic medical record data for populations totaling more than 6 million covered lives. We describe the real-world implementation of HER2 testing and trastuzumab prescribing in this setting.

The study population included 6460 women who were diagnosed with an invasive breast cancer that was 2+ cm in size or who had positive nodes at diagnosis from January 1, 1999, to December 31, 2007, and who were enrolled at one of the 8 participating CRN study sites for at least 12 months prior to diagnosis. These tumor size and lymph node status criteria were used to restrict the population to women similar to those in the adjuvant trastuzumab clinical trials. Subjects were followed for up to 12 months following diagnosis. Follow-up periods of less than 12 months were due to death or disenrollment during that period. The 8 participating study sites are geographically distributed within the United States and represent diverse populations: Group Health Cooperative (Washington), Kaiser Permanente Northwest (Oregon and Washington), Kaiser Permanente Northern California, Kaiser Permanente Colorado, Kaiser Permanente Georgia, Marshfield Clinic (Wisconsin), Henry Ford Health System (Michigan), and Harvard Pilgrim Health Care Institute with clinical data from Harvard Vanguard Medical Associates (Massachusetts). The study population included all eligible subjects at each site except for Kaiser Permanente Northern California, which contributed data on a 10% random sample of their eligible population.

A subset of 400 cases (50 from each site) was sampled from the original cohort for detailed chart review using stratified sampling (

) to ensure adequate numbers of subjects with chemotherapy exposure and the outcome of heart failure. Weights were used for some analyses, and were calculated as 1 over the probability of being sampled (ie, 1/(1/N), where N is the number of subjects in that stratum; N 

1, since at minimum the sampled subject belongs to the stratum). The primary purpose of the chart review was to validate electronic administrative data for chemotherapy exposure to anthracyclines or trastuzumab and the outcome of heart failure, as measured by clinical criteria and measurement of left ventricular ejection fraction, which was previously reported in a separate publication.

 However, the availability of data on HER2 genetic testing from the chart abstraction allowed us to conduct this secondary analysis. This study was approved by the institutional review boards (IRBs) at Group Health Cooperative, Marshfield Clinic Research Foundation, and Henry Ford Health System, and did not require written informed consent. The IRBs for the remaining sites ceded authority to the Group Health Cooperative IRB. A small number of health system members at each site have elected not to participate in anonymous or unconsented research protocols, and these subjects were excluded from this research study.

All genetic test results used in this study are from testing that occurred as part of routine medical care at each of the 8 participating sites. The laboratory performing the testing varied by site; however, the interpretation was similar. IHC testing for HER2 was classified as negative (0 or 1+), equivocal (2+), or positive (3+), except at Henry Ford Health System where 2+ was considered positive. FISH testing for HER2 was classified as negative (HER2/CEP 17 ratio <1.8), equivocal (HER2/CEP 17 ratio between 1.8 and 2.2), or positive (HER2/CEP 17 ratio >2.2). At Kaiser Permanente Northern California, the cutoff for FISH positive was a HER2/CEP 17 ratio of >2.0 for all but the last year of this study, following the American Society of Clinical Oncology and College of American Pathologists recommendation to change the cut point to 2.2 in January 2007.

We identified the internal practice guidelines for HER2 testing at each site. Since 2000, guidelines for all sites specify that all new invasive breast cancer patients are automatically ordered a HER2 test using IHC testing. Reflex testing is performed using FISH to clarify or confirm equivocal IHC findings (at all sites) plus additional subsets including all positive IHC findings at 2 sites, cases at the discretion of the treating oncologist at 2 other sites, or cases with an IHC test result of 1+ or more positive at 1 additional site. At 1 site, the protocol was changed in October 2007 to initially test all invasive tumors for HER2 by FISH.

We used stratified sampling to ensure adequate representation of the chemotherapy exposures and outcomes in the chart-reviewed sample (eAppendix). Identification of prevalent heart failure at the time of diagnosis (a possible contraindication for trastuzumab use) was also obtained exclusively from chart abstraction. The method used to define heart failure is described elsewhere.

 We used standardized abstraction forms and abstractor training protocols to ensure uniform data recording across sites and abstractors. The lead project manager rereviewed 10% of charts from each CRN site for inconsistencies, inaccuracies, and missing data, and questions were referred back to the originating site for clarification. During chart review, both qualitative and quantitative descriptions of HER2 results were collected from the medical record. Any apparent discrepancies were resolved by rereview at the originating site.

Each CRN site maintains electronic health record information on its member population including membership, diagnoses, pharmacy data, and a tumor registry. From these records, we extracted electronic data on patient characteristics (age at diagnosis, sex, race, ethnicity, year of diagnosis), tumor characteristics (stage, lymph node involvement, tumor size), and chemotherapy treatment. The Charlson Comorbidity Index is a quantitative variable that was computed using diagnosis codes as previously described anchored at the date of diagnosis.

Patients were classified as receiving trastuzumab using a dichotomous indicator variable if any of the following national drug codes were in the pharmacy records after their date of diagnosis: 50242013460, 50242013468, 50242005656, or 63552047001, or if procedure code J9355 was in the procedures database. This definition was validated against data from chart abstraction and demonstrated to have high predictive value.

Lymph node status was a dichotomous indicator variable defined as positive (1 or more positive nodes) or negative. Tumor size was a quantitative trait defined as the size of the largest qualifying tumor (a small number of patients had more than 1 primary tumor during the study period). We defined invasive breast cancer using Surveillance Epidemiology and End Results (SEER) staging criteria

 as patients diagnosed with localized, regional, or distant metastatic breast cancer. The primary dependent variable was receipt of HER2 testing, which was treated as a dichotomous indicator variable. Data on receipt of HER2 testing and the test result (categorical variable) were obtained exclusively from chart abstraction of the medical record.

All analyses, including descriptive statistics, summaries, and multivariate linear regression, were produced using R (version 2.11.1; R Foundation for Statistical Computing, www.r-project.org) and SAS version 9.2 (SAS Institute Inc, Cary, North Carolina). We used a Bonferroni correction to adjust the simultaneous confidence intervals. The multivariate model used a generalized linear model with receipt of HER2 testing as the dependent variable and the following independent variables: diagnosis year, race, lymph node status, SEER stage, site, age, presence of heart failure, and Charlson Comorbidity Index.

Of the entire cohort of 6460 women, 7.5% (n = 483) received trastuzumab. Trastuzumab use varied by year of diagnosis and stage of disease. Although the 

 of subjects who received trastuzumab was highest among women with distant stage breast cancer (

), women with distantstage breast cancer represented the smallest 

 who received trastuzumab in every year (

), because most women are diagnosed at an earlier stage of disease. The proportion of subjects who received trastuzumab increased for all stages of disease after 2004 when professional guidelines changed (Figure 1). The proportion of women in the chart review population who received trastuzumab was 18% (72/400), which is higher than the proportion of women who received trastuzumab in the entire study population, because we intentionally overselected trastuzumab recipients for review. However, after adjustment using the weights described above, 6.4% of the chart review population received trastuzumab.

All of the sites performed IHC testing in-house, and most sites used the HercepTest kit (Dako, Carpenteria, California). All except 1 site used a reference laboratory to perform FISH testing, and all sites used the Vysis PathVysion test (Abbott Molecular, Abbott Park, Illinois). Among the 400 subjects in the chart review population, 367 had tumors tested by IHC and 94 had tumors tested by FISH. Only 30 women (7.5%) received neither test.

Consistent with practice guidelines at the 8 sites, nearly everyone who was tested received IHC testing, with only 4 women receiving FISH testing alone. FISH testing occurred more frequently in the latter years of the study, possibly reflecting a phasing in of FISH testing during this time frame. Of those who received both tests, 74% had a positive or borderline IHC test result, whereas the majority of patients (69%) who received IHC testing alone had a negative HER2 result. A substantial proportion (40%) of subjects with a borderline IHC test result had no evidence of receiving FISH testing, contrary to most site practice guidelines. Among 17 patients with a borderline IHC test result, 3 received trastuzumab. There was also large variation among sites in the proportion of subjects with a borderline IHC test who did not receive FISH testing (range 0%-63%; 

Of those who did not receive either test, 50% (n = 14) were diagnosed in 1999, around the time that testing was initiated at most of the sites (

). Excluding the 28 patients diagnosed in 1999, 96% (356/372) of patients received some form of HER2 testing. The remaining 16 patients who were not tested were more likely to be older than 70 years (

 = .026), to have distant-stage disease (

 = .0032), to have negative lymph nodes (

 = .013), or to have been diagnosed in 2005. We also detected differences in the proportion who did not receive testing across sites ranging from 0% to 14% (

 = .0003), although about one-third of the women had missing data for race/ethnicity.

Trastuzumab Use as a Function of HER2 Status

As expected based on the sampling criteria, a higher proportion (32% [119/373]) of women in the chart review sample who received HER2 testing were HER2 positive compared with population estimates.

Trastuzumab was primarily used by patients with a HER2- positive test result (

). None of the patients in this sample who were not tested for HER2 status used trastuzumab. Of patients with only a HER2-negative or borderline result, 1.6% (4/255) used trastuzumab, and 35% (9/26) of patients with HER2 discordant results used trastuzumab. In these 9 discordant cases, 7 patients had positive FISH tests and 2 had a positive IHC test. Of the 17 who did not receive trastuzumab, 10 had positive FISH tests and 7 had positive IHC tests. Overall, 64% (61/96) of patients with HER2-positive results received trastuzumab; however, this use was dependent on year of diagnosis and stage of disease. In 1999-2005 (when trastuzumab was primarily recommended for patients with distant metastatic disease) only 52% of HER2-positive patients used trastuzumab, whereas in 2006-2007 (by which time recommendations had been adopted for use of trastuzumab in all cases of invasive cancer), 84% of HER2-positive patients used trastuzumab.

Some subjects with positive HER2 status did not receive trastuzumab. Of the 6 subjects with localized or regional disease who were diagnosed after 2005 and who did not receive trastuzumab, one had heart failure at diagnosis. For the remainder, a rationale was not readily apparent based on our assessment of the comorbidity index, age at diagnosis, or other patient characteristics. In addition, 2 subjects with distant metastatic disease who were HER2 positive did not receive trastuzumab. One of these patients had heart failure at diagnosis, while the other case could not be readily explained based on patient characteristics in our data set. Other factors such as patient preference, physician opinion, unrecorded negative or borderline HER2 test results, or treatment outside of the reporting sites could explain these observations.

CLINICAL

Evaluation of real-world implementation of HER2 testing showed that uptake was high (>90%) and trastuzumab treatment was targeted to patients with positive HER2 status.

Utilization of HER2 Genetic Testing in a Multi-Institutional Observational Study

To test the validity of the adapted Diabetes Complications Severity Index (aDCSI), which does not include laboratory test results, as an indicator of diabetes severity.

Retrospective cohort study using 4 years of claims data from 7 health insurance plans.

 Individuals with diabetes mellitus and continuous enrollment were study subjects (N = 138,615). The 2 independent variables—the aDCSI score (sum of 7 diabetes complications graded by severity as 0, 1, or 2; range 0-13) and the aDCSI diabetes complication count (sum of 7 diabetes complications without severity grading; range 0-7)&mdash;were generated using only claims data. We evaluated the numbers of hospitalizations attributable to the aDCSI with Poisson regression models, both categorically and linearly.

The aDCSI score (risk ratio 1.39 to 6.10 categorically and 1.41 linearly) and diabetes complication count (risk ratio 1.67 to 9.11 categorically and 1.65 linearly) were both significantly positively associated with the number of hospitalizations over a 4-year period. Risk ratios from the aDCSI score were very similar to the risk ratios previously reported for the Diabetes Complications Severity Index (DCSI); the absolute difference between risk ratios ranged from 0.01 to 1.6 categorically and was 0.05 linearly.

The aDCSI is a good measure of diabetes severity, given its ability to explain hospitalizations and its similar performance to the DCSI.

The adapted Diabetes Complications Severity Index (aDCSI), which does not include laboratory test results, is a good measure of diabetes severity.

		The aDCSI was significantly positively associated with the number of hospitalizations over a 4-year period.

		Risk ratios of hospitalizations based on the aDCSI score were similar to risk ratios from the DCSI, which does include laboratory data.

		The aDCSI has yet to be validated against mortality, which is the major limitation of this study.

Diabetes mellitus is one of the major healthcare issues in the United States because of its high prevalence and the growing costs of caring for affected patients.

 Complications associated with diabetes drive the escalating costs of diabetes management. Although several risk measures were developed to quantify the severity of diabetes complications, they mostly targeted a specific condition instead of the broad array of diabetes complications.

The Diabetes Complications Severity Index (DCSI), developed by Young and colleagues, uniquely incorporates a wide range of diabetes complications.

 The DCSI incorporates diagnosed complications along with select laboratory results to assess patients’ risks of adverse outcomes, including hospitalizations and death. It uses information from 7 diabetes complication categories.

 Even though the DCSI is a relatively new measure, it has been quickly adopted by researchers.

 However, DCSI’s utility as a risk measure to characterize a diabetic population may be limited because laboratory test results are not readily available to researchers, particularly those who rely on administrative claims.

Risk measures have increasingly relied on claims data because they are inexpensively available for a great number of individuals; these data include claims for diagnoses and procedures and often information on dispensed medications.

 One widely used risk adjustment measure, the Adjusted Clinical Group system, uses an individual’s diagnoses and pharmacy data from 1 year to assign a morbidity level.

 This approach has been validated both domestically

However, claims data usually do not include laboratory information, which makes it difficult for risk measures that require laboratory results to be applied on a large scale.

Therefore, our purpose was to test the validity of the adapted Diabetes Complications Severity Index (aDCSI), which excludes laboratory test results, as an indicator of diabetes severity. We hypothesized that the aDCSI would be comparable to the DCSI (which includes laboratory data) and be a good measure of diabetes severity.

This was a retrospective cohort study using 4 years of claims data in which we tested the value of the aDCSI for explaining the number of hospitalizations.

We accessed claims data from 7 Blue Cross Blue Shield plans; the detailed information was described in a published paper.19 The original data were collected from 2002 to 2005; the data were subsequently updated with additional data on the original individuals through 2006. The following data were acquired: (1) enrollment files for administrative data; (2) benefits information to determine medical and pharmacy coverage; and (3) inpatient, outpatient, and pharmacy claims records containing the 

International Classification of Diseases, Ninth Revision, Clinical Modification

) diagnosis, Current Procedural Terminology codes, and National Drug Code prescription information.

We required that the enrollees have type 2 diabetes, and full medical and pharmacy coverage in the 4-year period.

We defined individuals as having type 2 diabetes if they had 1 relevant inpatient code or 2 outpatient 

codes separated by at least 30 days. The relevant codes were 250.xx, 648.0 (diabetes mellitus with pregnancy), and 362.0 (diabetic retinopathy) or 266.41 (diabetic cataract). Individuals only with 250.x3 (type 1 diabetes) were not included. Additionally, any individual filling a prescription for a medication for treatment of hyperglycemia was included (

). Combination medications were also identified. If the prescription was for metformin alone, the individual was also required to have an 

ode for diabetes for inclusion in this group. The calendar year of the earliest diagnosis of diabetes was used as the starting point of the observation period.

The number of hospitalizations was obtained from the inpatient claims over a 4-year period. Costs were obtained from the claims over a 4-year period; per person per year total costs and pharmacy costs were presented. Total costs were examined as well as pharmacy costs.

To replicate the DCSI scores and DCSI complication counts, we identified the claims coded with the 

system for individuals during the 4-year study period and applied the classification method developed by Young and colleagues (

 consists of scores (0, 1, or 2) from 7 complication categories: retinopathy, nephropathy, neuropathy, cerebrovascular, cardiovascular, peripheral vascular disease, and metabolic; it ranges from 0 to 13. The DCSI 

is a count of any complication in the 7 categories and ranges from 0 to 7. We did not include laboratory results in constructing these aDCSI scores and aDCSI complication counts.

For number of hospitalizations, we adopted a Poisson model with adjustment for overdispersion, to parallel what Young and colleagues did. Given that there were many people without any hospitalizations, a zero-inflated negative binomial model was also used for a sensitivity analysis. The patterns of results from both models were similar, so we just report the results from the Poisson model.

We calculated risk ratios for hospitalization. We tested inclusion of the main independent variables, aDCSI score and aDCSI complication count, categorically (0, 1, 2, 3, 4, 5+) and linearly. In categorical analysis, the risk ratio of hospitalizations was derived by comparing samples in a given category with those in category 0; in linear analysis, it was the risk ratio of hospitalizations associated with a 1-unit increase in aDCSI score.

Comparison of DCSI With and Without Laboratory Test Results

We compared the risk ratios of hospitalizations obtained in this study with those in the study by Young et al to determine whether the aDCSI (without laboratory information) and the DCSI (with laboratory information) perform similarly. Given the differences in study population and sample size, we decided that if the risk ratios from both sources were similar and showed similar patterns across risk groups, we would conclude that the aDCSI and DCSI performed similarly.

The data were deidentified in accordance with the Health Insurance Portability and Accountability Act’s definition of a limited data set. The Johns Hopkins University Office of Research Subjects deemed the study to be exempt from federal regulations because the research activities were considered to be of minimal risk to subjects, as they were not identifiable.

). The mean age was about 59 years and roughly 51% were male. About 70% of the study subjects had a score of 0 for DCSI score/complication count, and close to 60% had no hospitalization. The mean aDCSI score, aDCSI complication count, and number of hospitalizations were 0.50, 0.37, and 0.94, respectively. The average annual total cost was $11,500 over a 4-year period, of which $3000 (26%) was pharmacy cost.

METHODS

The aDCSI without the inclusion of laboratory data performs similarly to the DCSI with laboratory results, and is a good measure of diabetes severity.

Validating the Adapted Diabetes Complications Severity Index in Claims Data

To examine the long-term relationships between costs, utilization, and patient-centered medical home (PCMH) clinical practice systems.

Clinical practice systems were evaluated at baseline by the Physician Practice Connections-Research Survey (PPC-RS). Annual costs and utilization of a retrospectively constructed cohort of 58,391 persons receiving primary care at 1 of 22 medical groups over a 5-year period (2005-2009) were compared.

Multivariate regressions adjusting for patient demographics, health status, and autoregressive errors compared PPC-RS scores and study outcomes for the entire cohort and 3 subcohorts defined by medical complexity (medication count 0-2 [n = 29,657], 2-6 [n = 19,505], 

7 [n = 9229]). Outcomes (adjusted to 2005 dollars) were total costs, outpatient costs, inpatient costs, inpatient days, and emergency department (ED) use.

For the entire cohort, a 10% increase in PPC-RS scores was associated with 3.9 (medication count: 0-2), 6 (3-6), and 11.6 (

7) fewer ED visits per 1000 in 2005; and 5.1, 7.6, and 13.6 fewer ED visits in 2009. That 10% increase was not associated with the 0-2 medication subcohort’s total (−$22/person in 2005; $184/person in 2009), outpatient (−$11/person in 2005; $42/person in 2009), or inpatient ($26/person in 2005; $29/person in 2009) costs. However, it was associated with significantly decreased total (−$446/person in 2005; −$184/person in 2009) and outpatient (−$241/person in 2005; −$54/person in 2009) costs for the most medically complex subcohort (

From 2005 to 2009, improved clinical practice systems were associated with cost reductions only for medically complex patients.

November 2012

Hospital Admissions and MS: Temporal Trends and Patient Characteristics

Value-Based Purchasing

Utilization of HER2 Genetic Testing in a Multi-Institutional Observational Study

Validating the Adapted Diabetes Complications Severity Index in Claims Data

Patient-Centered Medical Home Cost Reductions Limited to Complex Patients

Automated Telephone Calls to Enhance Colorectal Cancer Screening: Economic Analysis

Active Pharmacovigilance and Healthcare Utilization

A Systematic Review of Reference Pricing: Implications for US Prescription Drug Spending

Letter to the Editors: Response to Payer Coverage for Patients Enrolled Onto Clinical Trials

Care Coordination Agreements: Barriers, Facilitators, and Lessons Learned

How Comorbidities and Preoperative Expenditures Correlate With Postoperative Adverse Outcomes

Physician Factors That Influence Patient Referrals to End-of-Life Care