What combinations of less prevalent combinations of chronic comorbidities are most critical to address in terms of care utilization and cost? What are the future research considerations for MCC research?
While a number of studies in recent years have examined patients with MCC, patients with less common combinations have largely been overlooked. Given the growth and aging of the United States population, and the continual rise in healthcare costs, the “long tail” will likely only become larger and more complex and costly over time. Studying the population with less prevalent MCC combinations represents a shift from studying more prevalent chronic conditions to focusing on chronic disease complexity at a finer level of detail. Many combinations of MCC make up the long tail and, in the aggregate, a substantial number of patents with less prevalent MCC combinations and above-average costs and healthcare needs, are excluded from clinical research studies. Understanding how the long tail impacts healthcare costs and quality and improved treatment of MCC patients is a clear need.
Development of a research agenda for the population with low prevalence MCC should be guided by consideration of issues related to the data available for conducting this research and the types of research questions to be explored:
- Data for conducting research on this patient population: How well do existing data sources support research of this patient population? What types of additional types of data are needed? What types of additional studies are needed to understand the types of research on the long tail that existing data sources that can support?
- Research questions: What types of analyses are needed to improve our understanding of the long tail? What types of research are most likely to contribute to improvements in the effectiveness and efficiency of care for patients with MCC in the long tail? How can this research influence the way MCC patients are managed and treated, both in terms of patients with common combinations of chronic conditions and those with less prevalent combinations? What types of research studies can be informative for MCC patients?
As in all MCC research, different stakeholders will have different requirements in the level of clinical detail their disease models must support. Improved understanding of the MCC population has implications for quality of care, disease management, reimbursement, and the design of research studies. Below, we discuss potential topics for future research and initiatives on patients with low prevalence of MCC, organized by stakeholder.
Researcher & Interventionist Stakeholders
- Reproducing the long tail. To date, the long tail distribution of low-prevalence chronic disease combinations has been observed using Medicare claims data only. Other large, detailed sources of diagnostic information should be analyzed to determine if the long tail can be reproduced and if differences in distributions are evident among varying populations. Due to their large sample size and comprehensiveness, databases such as HCUP’s NIS, Medicaid’s MAX, and the National Ambulatory Care Survey (NAMCS) would be viable candidates for this type of research. Only recently has HCUP’s NIS and NAMCS been leveraged to study MCC prevalence and healthcare utilization (Ashman et al., 2013 & Steiner et al., 2013).
- Improving our understanding of how MCC prevalence and outcomes vary by patient characteristics. Because they have not been the focus of many prior studies, our understanding of patients with complex combinations of MCC who comprise the long tail is limited. Basic descriptive studies that examine the number of patients with various MCC combinations, the number of possible combinations, and the costs incurred by this population would be useful. Given the clinical detail (e.g., ICD-9 codes from claims data) needed to identify patients in the long tail, claims data (from Medicare, Medicaid, or private payers) have been the main data source for studies of the MCC population. Other potential data sources, such as surveys that collect information on self-reported conditions, typically lack the clinical detail needed to support research on the less prevalent MCC. While claims data have the large sample size and clinical detail that such studies require, they have several limitations that affect the types of MCC research that they can support:
- Lack of demographic and socioeconomic variables. Medicare claims data can be linked to administrative data with information on enrollee characteristics, but the administrative data contains limited demographic information (e.g., gender, age, and race) and no real socioeconomic information.
- Inclusion of patients only if treatment for a condition occurs in the specified time period. Only iif the patient is treated for the condition (i.e., has a claim with the ICD-9 code listed) during the period that the claims data cover will the diagnosis be included, and thus the data may underestimate prevalence. Some patients lack access to appropriate healthcare and claims data will not include all of their medical conditions. For other patients who have access, differences in screening, diagnosis, and coding practices can lead to differences in the types of diagnoses that are recorded in claims data.
- Claims data are not representative of the United States population. For example, Medicare claims data are only available for Medicare fee-for-service beneficiaries; HCUP’s full NIS database represents about 90% of hospitals and 95% of discharges, but has unique limitations; and all-payer databases are only available for certain states.
- Claims have limited information on patient outcomes. Alternatives or supplements to claims data will need to be explored to understand the relationship between MCC and patient characteristics and outcomes associated with MCC and the different treatment patterns for them. Linking claims to other data sets is one way that our understanding can be improved, at least for specific patient populations. For Medicare beneficiaries who are nursing home residents, linking Medicare claims data to the Minimum Data Set assessment tool would allow more detailed exploration of how MCC patterns differ based on patient characteristics and also support analysis of the relationship between MCC and patient outcomes. Linking claims data to the Outcome and Information Assessment System (OASIS) would allow similar exploration for patients receiving home health services. While only feasible for small numbers of patients, chart review for a sample of complex patients may be useful for better understanding and defining complexity. Given the unique disease combinations that one tends to find on the long tail, the generalizability of such results to other patients may be limited.
- Developing a reporting and theoretical framework. MCC researchers have utilized a variety of different systems of diagnostic classification and analytical methods. In the context of each individual research paper, these choices may have been reasonable and appropriate, but these choices can strongly influence the findings of research calculations. For example, an analysis with more diagnostic categories automatically finds more chronic disease combinations. Because findings depend on methods, it may be difficult or near impossible to combine research from diverse sources in order to synthesize consistent results. Different stakeholder groups are concerned with different aspects of MCC research; consequently, differences in diagnostic classification are likely to persist in future research. Standardization cannot be demanded merely for the sake of making literature review and synthesis easier. However, it can be suggested that authors of papers relevant to the MCC field begin considering how to cast their results in ways that facilitate comparison with the rest of the MCC literature, and that the scientific community address the development of a theoretical framework that would support more systematic reporting of MCC findings. In particular, methods developed for producing results that are invariant across methodologies, and for distinguishing clinically important combinations from those that are the inevitable result of arithmetic may be beneficial.
- Understanding how study conclusions are impacted by the classification system that is used. Little is known about how robust study findings are to the disease classification system that they use. As part of this study, we examined the ICD-9 codes used in three widely used classification systems: the CCS, HCCs, and the CCW. These systems vary with respect to the number of disease categories that they include—the CCS includes 285 categories, there are 70 HCC categories, and the CCW includes 27 chronic condition categories. These differences may contribute to differences in study findings that are purely driven by the classification system—a study that uses the CCS would presumably have more MCC combinations than one that uses the CCW just due to the difference in the number of categories in the two systems. The number of combinations actually observed in the data is an artifact of the classification plus the sample size.
The classification systems also vary with respect to the number of ICD-9 codes that they use. The HCUP system includes virtually all of the 14,573 ICD-9 codes, the HCC system uses around 3,000 ICD-9 codes, while the CCW uses approximately 600 ICD codes. As a result, a higher proportion of patients would be classified into a disease category for studies that use the CCS than for studies that use HCCs or the CCW. Additional research is needed to understand the robustness of study findings to the classification system that is used.
- Cost patterns for those with MCC: Additional research on the healthcare costs incurred by patients in the long tail is important for understanding the potential savings from programs targeted at this population. There has been little research on the cost and utilization patterns for patients with specific combinations of MCC; the large number of possible combinations is a limiting factor. But identification of specific combinations associated with high costs is important for shaping development of cost effective programs for MCC treatment.
- Analysis of disease combinations (or clusters). For the most part, disease classification systems focus on individual conditions rather than specific combinations or clusters of conditions. As a result, few studies have examined the clustering of MCC, particularly for less prevalent MCC combinations for which there are a very large number of possible combinations. The lack of research on disease clusters is related to the large amounts of data that such studies require. Rare clusters cannot be identified without large amounts of data with detailed information on patient diagnoses (i.e., claims data). Analysis of such large data files will identify more disease combinations than it is possible to analyze. Additional research is needed to identify the disease clusters that should be the focus of future research efforts—for example, combinations associated with high-risk patient populations.
- Comparing MCC Studies Across Countries. MCC studies have primarily been conducted in the United States and Europe. Assessment of the data sources and methods used in these studies should be conducted to determine whether the results of these studies are comparable. That is, do data quality or infrastructure concerns suggest that research from one country may be more reliable than another? What do such comparisons suggest about how information and analytic techniques can be leveraged across international borders? Are there any potential implications for the treatment of MCC patients? Are there any studies of patients on the long tail that can be compared across countries or are there too few of these studies to draw any meaningful comparisons?
- Understanding the impact of transition to ICD-10. While this question will affect many types of research and transcends research on the long tail, the impact on MCC research resulting from efforts to map ICD-9 to ICD-10 is not known; at a minimum, this transition is likely to limit researchers’ ability to measure changes in disease prevalence and patient complexity over time. The ICD-10 transition will also affect classification systems such as HCUP and HCCs which are an important component of MCC research.
Patient and Provider Stakeholders
- Disease management programs. Clinical approaches often focus on individual diseases, without considering how the presence of MCC may affect healthcare needs. This is particularly true for patients in the long tail. As a result, clinicians have a very limited body of evidence-based knowledge for approaching the care for these patients. A focus of additional research should be improvements in disease management programs that are effective for patients with multiple conditions and prioritizes the role of care coordination. For example, how many different providers do MCC patients visit during the course of one year? Who do patients consider to be their “primary” physician? How many different physician offices and healthcare facilities do patients visit? How many different combinations of pharmaceutical drugs are MCC patients prescribed? What are the different types of systems indicators that can be used to monitor MCC patients?
- Patient perspectives on living in the long tail. There is a large patient stake in MCC research. For example, PatientsLikeMe expanded their list of potential diagnoses from 300 to 2,000 due to patient demand, a list that continues to increase. The “long tail” is not just a conceptual problem, but a problem that affects many patients. How do we bring the patient voice to MCC research? What would MCC patients like to know? How do we focus MCC efforts on patients and not a research paradigm or list of chronic conditions? One option is to provide opportunities through digital media for individuals with multiple chronic conditions to provide information about “a day in their lives” and their medical and health needs so that we can better understand what information is needed to better care for those with MCC. This would provide insights that cannot be obtained via data analyses, although it is not clear how generalizable findings would be to other MCC patients. Another option would be to develop patient-reported outcomes specific to MCC patients and to leverage patient-reported information that is collected through EHR systems. As EHRs continue to advance and online patient portals become more widely available, electronic information that is patient-derived may be a robust source of data that helps bring the patient voice to the forefront of MCC research.
- Understanding the different types of interactions between low-prevalence chronic disease combinations. When chronic diseases co-occur they can have additive, multiplicative or even protective effects. For example, body mass has been found to have a paradoxical effect on mortality in patients with rheumatoid arthritis (Escalante et al., 2005). Understanding the different types of interactions between chronic diseases can allow providers to better target groups of MCC patients for intervention (e.g., patients with chronic diseases that have a multiplicative effect).
- Payers. Reimbursement systems may fail to recognize the incremental costs associated with MCC, particularly for the less prevalent MCC combinations that comprise the long-tail. As a result, the full costs of caring for these patients may not be reflected in payment rates, potentially impacting quality and access to care for these patients. Additional research on patients with MCC combinations may lead to improvements in the ability of payment systems to recognize the incremental costs associated with specific MCC combinations, thus promoting appropriate reimbursement rates for these patients, promoting access to care. Some examples of potential research questions may include: How can patient diagnoses be more accurately identified and costs more accurately predicted? How can “active” diagnoses be determined compared to those patients are no longer seeking treatment for? What risk stratification levels may be warranted for persons with different combinations of chronic disease?
- Quality Measures and value-based purchasing programs. Quality measures may show skewed calculations due to inaccurately classified individuals if low-prevalence MCC are not accounted for. For example, a person with type 2 diabetes and Alzheimer’s disease may not be a good candidate for tight glycemic control. Exclusion of patients with specific MCC combinations is one option for dealing with this issue, but this would reduce the incentives to provide high quality care to this patient population, and also lead to a lack of relevant information on provider quality for MCC patients. Focusing on applicable quality measures that can be applied broadly across both MCC and non-MCC patients (e.g., related to patient-centeredness or care coordination, or self-management) is a better option. Development of MCC-disease-specific quality measures seems impractical for those on the long tail given the many MCC combinations and small sample sizes that would be available for measure calculation.
Similarly, value-based purchasing programs may not account for disease complexity, as many metrics used in adjusting reimbursement are focused on single diseases and related clinical processes. The quality of care coordination and the ability to manage complexity may be more accurately assess by examining MCC patients, including those with low-prevalence conditions.
As is clear from the discussion above, there are many gaps in our knowledge of patients with less prevalent combinations of MCC. These gaps are partly a reflection of the data and analytic-related challenges that must be resolved to conduct research on this population and partly due to the inclination to focus on patients with individual conditions or on the more prevalent combinations of MCC. There are, however, a number of opportunities for future research that would improve our knowledge of the long-tail and perhaps lead to improvements in the care for this population. These potential research questions differ by stakeholder perspective. However, opportunities to share information, ideas and initiatives should be pursued across these perspectives to cultivate a community of professionals focused on improving care for all types of MCC patients.