What methodologies and analytic techniques have been used to study MCC? What are the potential limitations of these approaches in considering less prevalent combinations of MCC?
In the section below we discuss the methodological and analytic concepts to consider when conducting MCC research, with a special emphasis on less prevalent combinations of chronic conditions. We discuss the methodologies and analytic techniques that have been used to conduct MCC research to-date, the potential strengths and limitations of these approaches and how they relate to studying less prevalent combinations of MCC.
-
Defining Diagnosis of Chronic Condition
-
There are two main sources of information about patients’ chronic conditions: 1) surveys that collect self-reported disease status, and 2) claims and clinical systems that contain diagnosis codes (e.g., International Classification of Disease, 9th edition [ICD-9], ICD-10, Systematized Nomenclature of Medicine Clinical Terms [SNOMED CT]). Other sources of information, such as pharmaceutical prescription or laboratory data, can also be used to identify patients’ chronic conditions. However, these additional modalities are not thoroughly discussed in this paper.
MCC research has been conducted using both primary sources of diagnostic information noted above. For example, Schoenberg and colleagues analyzed Health and Retirement Study (HRS) data to understand the relationship between chronic disease constellations and out-of-pocket medical expenditures. In the study, chronic conditions were identified using eight self-reported chronic conditions from the HRS (Schoenberg et al. 2008). Similarly, Bae and Rosenthal used 177 ICD-9 codes derived from self-reported chronic conditions from the Medical Expenditure Panel Survey to study MCC and quality of care (Bae & Rosenthal, 2008). Conversely, Sorace et al., used approximately 3,000 ICD-9 codes derived from the HCC model to study the complexity of disease combinations in the Medicare population (Sorace et al., 2011).
There are strengths and weaknesses of self-reported versus claims-based information for identifying chronic conditions (See Exhibit 9). Claims-based diagnosis codes allow researchers to study a large number of chronic conditions at a very fine level of granularity and to understand the full range of patients’ diagnoses, including which specific diagnoses are present ( e.g., primary malignant neoplasm of the lung or carcinoma in situ of the lung vs. simply lung cancer). Sensitivity is critically important in enabling the study of less prevalent or rare chronic disease combinations. Claims are usually provider-generated and based on a differential diagnosis and supporting clinical documentation, eliminating potential error associated with patient self-reported information and other survey-related biases, such as recall and selection concerns. However, there are systematic limitations associated with ICD-9 codes, such as misspecifications, unbundling, and upcoding by providers and coders (O’Mailley et al., 2005). There is also a tendency for providers and billers to under-report diagnoses that lack payment incentive, such as mental health conditions. These issues can lead to inaccurate estimates of chronic disease prevalence and imprudent results. Diagnosis coding using ICD-9 and ICD-10 codes has also been shown to misestimate the prevalence of certain conditions.
Exhibit 9: Strengths and Weaknesses of Self-Reported versus Claims-Based Chronic Conditions
Coding Type Strengths Limitations Self-Report Easy to collect, used to identify prevalent conditions, patient-derived. Subject to recall, sampling and selection bias. Few diagnoses studied and at a coarse level of granularity. Limited number of patients surveyed/studied. ICD-9 A large number of diagnoses are considered at a fine level of granularity. Commonly used in the United States. Used in large administrative databases; large sample size. There are a number of well documented limitations, such as over and underestimation of certain diseases, as well as inaccuracies due to malicious coding behavior ICD-10 Associated with improved coding accuracy. Greater number of diagnoses considered and at a more granular level. Used in large administrative databases; large sample size. Not in widespread use in the United States and won’t be for a number of years. Limited research available on coding inaccuracies and other shortcomings. SNOMED CT Greatest number of diagnosis codes considered at the finest level of granularity. Limited research available on coding inaccuracies and other shortcomings. Potentially too granular for use in certain healthcare settings. Underestimation is a concern when a significant proportion of the population may not have a claim during the study period; overestimation may occur for conditions that lead to higher payment rates if they are reported as being present. Woo et al. found that obesity identified by discharge ICD-9 codes underestimated the true prevalence of obesity in an inpatient pediatrics population (Woo et al., 2009), while Kern et al. found that ICD-9-CM codes failed to identify the majority of veteran patients with comorbid chronic kidney disease (Kern et al., 2005). ICD-10 codes have also been shown to overestimate the prevalence of certain diagnoses, such as post-traumatic stress disorder (Rosner & Powell, 2009). However, recent evidence suggests that the introduction and use of ICD-10 coding may be associated with improved accuracy of co-morbidity coding for the majority of clinical conditions (Januel et al., 2011). It is unclear whether the improvement is due to the ICD-10 coding system itself or changes in coder and physician behavior.
Self-reported diagnoses from surveys or those that are mapped to ICD-9 or ICD-10 codes from surveys provide a much smaller number of chronic conditions for analysis, at a very coarse level of detail. Typically surveys do not include the breadth of chronic conditions a patient has or the specific types of chronic conditions (e.g., a specific type of cancer). For example, the HRS only allows researchers to investigate eight chronic conditions (hypertensions, diabetes, cancer, chronic lung disease, heart conditions, arthritis, stroke and psychiatric/emotional problems) and it does not allow them to drill down to what specific types of conditions a patient has (e.g., what type of cancer?). Thus, the use of surveys limits the ability to understand the true complexity of chronic disease combinations a patient is experiencing as well as the occurrence of less prevalent chronic conditions. In addition, self-reported diagnoses can be limited due to survey-related biases, such as recall, ascertainment and selection bias. For example, those individuals who avoid or who do not have access to healthcare may not be evaluated for potential chronic conditions of interest. Although evidence suggests that self-reported chronic conditions may be reasonably valid (Martin et al., 2000), self-reported diagnoses are not provider generated, may be subject to recall error by patients, and may not be captured in a sufficiently structured and systematic manner for analysis. Biases in self-reported diagnoses may be reduced through survey question structure; many surveys typically ask patients, “Has the doctor told you….?”. Overall, self-reported conditions can lead to non-uniform and inaccurate diagnosis categories and errors when mapping self-reported information to ICD-9 or ICD-10 codes.
In addition to the considerations described above, it is also important to note that validity of the presence of chronic conditions and reliability of reporting/detecting chronic conditions are two key issues that challenge MCC research. Researchers have attempted to improve validity by examining diagnoses across care settings and determining if patients have two or more claims reporting a specific diagnosis code over a given period of time to confirm disease occurrence. However, validity and reliability will remain a challenge given the vastness and complexity of many of the large databases and systems used to collect and analyze diagnostic information.
It is important to recognize that the trajectory of diagnosis coding in the United States is moving away from ICD-9 codes and towards larger, more detailed coding schemes, such as ICD-10 and SNOMED. In fact, on January 16th, 2009 the Department of Health and Human Services published a final rule specifying an anticipated ICD-10 implementation date of October 1, 2013 (although this may be delayed). The World Health Organization (WHO) has already begun work on developing ICD-11. It is inevitable that diagnosis coding will continue to become more refined over time, providing researchers with the ability to study disease complexity at a level of detail not currently possible. Although “new” coding schemes will improve our ability to identify specific diagnoses of individuals with MCC, they will have some limitations.
The transition from ICD-9 to ICD-10, as well as to other future coding schema, will present challenges to researchers. During coding transition periods back-coding ICD-10 codes to ICD-9 and forward-coding ICD-9 codes to ICD-10 will be necessary for longitudinal analyses and comparative investigations. ICD-9 based indexes and measures, such as the Charlson Comorbidity Index and AHRQ’s Patient Safety Indicators, will also need to be translated to ICD-10 systems to support their continued use. There may be a “lag time” associated with re-specifying these tools, which researchers will need to be aware of. Additionally, there will most likely be a “testing” period after new coding systems are implemented, as researchers will need to explore the nuances and limitations of new systems prior to conducting analyses (Iezzoni, 2010). Researchers may also need to observe a data “black out” period as clinicians learn, perfect and then settle into new coding behaviors associated with the transition to ICD-10 (Januel et al., 2011). This “black out” period may also be needed by individual health systems and providers. The transition from ICD-9 to ICD-10 in the United States will not be smooth and universal. Health systems and providers will “go live” with ICD-10 at various points in time with different levels of success.
Despite the challenges, more refined coding systems will greatly enhance our ability to conduct research on less prevalent combinations of MCC. New coding systems will provide a very detailed level of diagnostic information.
-
-
Data Aggregation and Grouping Systems
-
Grouping systems, such as AHRQ’s clinical classification system and CMS’s Hierarchical Conditions Categories, are used to organize and aggregate diagnosis codes into different disease categories. These systems serve a variety of different purposes (e.g., research, risk-adjustment, etc.) and vary significantly in terms of which clinical conditions are considered and the number of diagnosis codes that are included in each disease group, as well as the number of groups (See ICD-9 Comparison Excel File). Regardless of their original intent or grouping methodology, however, many different types of grouping systems have been used to conduct MCC research, raising concerns about interpreting research results and comparing findings across MCC studies.
The decision to use specific grouping systems for MCC research should be informed by four key considerations: 1) the function, purpose and original intent of the grouper, 2) the behavior change that is desired by using the grouper to produce actionable information, 3) the end-users and their data needs ( e.g., data granularity), and 4) the research question. Researchers should not assume that a grouping system designed by and for one stakeholder group for one purpose is appropriate for another purpose. In fact, none of the currently available groupers are meant to serve multiple purposes (e.g., clinical decision support and risk-adjustment). Grouping systems are carefully designed and statistically calibrated to serve a specific aim. Using a grouping system for a different aim than intended can lead to meaningless results and misguided interpretation. MCC research which aggregates diagnosis codes should use grouping systems that are well documented, produce useful information for end-users (e.g., fine granularity for clinical decision support), and provide information that is meaningful, actionable and promotes provider behavior change (e.g., to reduce cost or improve care for specific groups). Grouping systems should be in alignment with the research questions at hand; research questions should ultimately drive MCC research designs (Wallace & Salive, 2013).
In choosing which grouping system to use for MCC research, stakeholder agendas matter. Each stakeholder group needs different types of information at varying levels of granularity. For example, those interested in clinical decision support needs a finer level of diagnostic information than risk-adjusters. Similarly, healthcare economists may need more detailed data than public health interventionists. Thus, it is important to consider the degree of coding granularity needed by each stakeholder. Understanding which stakeholder aims can be supported at specific levels of diagnostic granularity may be a beneficial area for investment for MCC researchers.
To determine which clinical classification systems exist and have been used for MCC or disease complexity research, a comprehensive grouping systems review was conducted. Grouping systems were identified through the literature review as well as input form the Co-Project Officers, TAG and key informants. Full descriptions of each classification system and the methodological issues to consider when using the grouper can be found in Appendix C. A condensed version of the results is shown in Exhibit 10 below.
Exhibit 10: Summary of Diagnostic Grouping Systems
Grouping System Sponsor Level of Diagnosis Aggregation Number of ICD-9 Codes Included Legend: Sponsor: agency, organization or company that maintains the grouping system; Level of Diagnosis Aggregation: the number of chronic condition categories included in the grouping systems; Number of ICD-( Codes: Grouping systems that are proprietary do not make ICD-9 codes available for public review Adjusted Clinical Groups Case-mix System (ACG) Johns Hopkins University 102 discrete categories Proprietary Aggregated Diagnosis Groups (ADG) Johns Hopkins University 32 discrete categories Proprietary All Patient Refined Diagnosis Related Groups (APR-DRG) 3M Health Information Systems 314 base categories and 1256 subclasses Proprietary Chronic Conditions Data Warehouse Algorithm Centers for Medicare & Medicaid Services 27 chronic condition categories 581 Chronic Illness Disability Payment System (CDPS) University of California, San Diego/Medicaid Programs 96 categories of diagnoses that correspond body systems and specific diagnoses 11603 Clinical Classification System (CCS) Agency for Healthcare Research & Quality 285 mutually exclusive categories 14567 Clinical Risk Groups (CRG) 3M Health Information Systems 272 clinically-based categories and 1,080 subclasses Proprietary Diagnosis Related group (DRG) Centers for Medicare & Medicaid Services 538 categories Not Specified Dyani Diagnosis Grouper Axiomedics Research, Inc. 200-300 categories depending on the criteria being examined Proprietary Hierarchical Condition Categories (HCC) Centers for Medicare & Medicaid Services 70 CMS-HCC categories 2916 International Shortlist for Hospital Morbidity Tabulation (ISHMT) World Health Organization 130 categories Not Specified Major Diagnostic Categories Health Level Seven International 25 categories Not Specified Medicare Severity Diagnosis Related Grouper (MS-DRG) 3M Health Information Systems 745 categories Proprietary Thomson Medstat Medical Episode Grouper Thomson Medstat Inc. 550 disease conditions Proprietary We reviewed fourteen grouping systems which were found to serve a variety of different purposes ranging from risk adjustment to comparing morbidity across hospitals internationally. The grouping methodologies of the systems are remarkably different and vary in level of complexity. For example, diagnosis aggregation ranged from 25 categories for the Major Diagnostic Categories to 272 clinically-based groups with 1,080 subclasses for 3M’s Clinical Risk Groups. The difference has a dramatic consequence for the number of disease combinations that can be explored by researchers because the number of combinations (without replacement) scales as per the following formula: C(n,k)=n!/k!(n-k)! (Ammann 2011). In this formula “C” is the number of disease combinations, “n” is the number of disease groups in the grouping system, “k” is the number of disease groups included in the calculation, and “!” stands for factorial. Applying the formula to the Chronic Illness Disability Payment System (CDPS) for two-way disease combinations would result in the following calculation: C(n,K)=20!/(2!)*(18!); or 190 disease combinations could be studied. Using the same formula, but with three-way and four-way combinations, the CDPS model would provide 1,140 and 4,845 disease combinations respectively.
As shown in Exhibit 11 (logarithmic scale), the number of disease combinations for analysis increases rapidly as the number of chronic condition categories and number of diseases that are included in the combinations are increased. Thus, grouping systems with more chronic condition categories (greater “n”) will generate more chronic disease combinations (“C”) for analysis, especially when the number of diseases allowed in the disease combination calculation (“k”) is not truncated at an arbitrary level (i.e. calculate dyads or triads and then truncate at four or more diseases).
The number of diagnosis codes included in each grouping system could not be evaluated across all systems because the information is proprietary for privately owned grouping systems. The lack of transparency represents a methodological limitation and bias for researchers, as they cannot know which diagnoses were included in analyses and therefore assess the level of complexity captured by the grouping system. Despite their differences, the majority of groupers have been used in some form of multimorbidity research to-date. For example, Sorace and colleagues used the HCC model to study complexity in Medicare patients, while Salisbury and colleagues used John’s Hopkins ACG system to study general practice patients and Steinman and colleagues used the CCS to study VA patients (See Exhibit 5 in Section 6).When interpreting published MCC literature as well as designing future MCC research, the methodological differences between grouping systems should be reviewed and considered. For example, grouping systems that provide the finest level of diagnostic information and the greatest number of chronic condition categories, such as AHRQ’s CCS, would be most appropriate for research on less prevalent chronic disease combinations.
Exhibit 11: Possible Number of Chronic Disease Combinations by Diagnosis Grouping System
It is also important to note that many MCC researchers have designed and employ their own groupers or modify an existing grouper which affects the methodological quality of results. Decisions to include, exclude or aggregate diagnoses often are not reported in author’s methodology sections. Authors may state that the decisions were guided by physician consensus or technical expert panels, but do not list specific diagnosis codes that were included or excluded. The impact of grouping algorithms on other analysis steps and how they may affect the interpretation of results are also missing from studies. For example, authors do not discuss how costs are allocated to disease categories after eliminating certain diagnosis codes from analyses, nor the percentage and types of patients that are excluded from a study.
Consequently, researchers are creating unique diagnostic categories that may be fundamentally different from one another making it difficult to interpret how one researcher’s disease category for “cancer” compares to another. If researchers utilized publicly available, well documented grouping systems (standardization) such as AHRQ’s CCS, the challenges of interpreting results across studies would be minimized. However, it is not practical and may not make clinical sense to use only publicly available grouping systems. For example, some diagnosis codes may warrant exclusion from analyses because they are ambiguous (physician consensus does not yet exist on the diagnostic criteria for a particular condition) and over time grouping systems will become obsolete as new coding systems are adopted ( e.g., ICD-10) and new, more robust groupers are developed. Regardless of the future of grouping systems in MCC research, providing researchers and readers with the ability to understand how disease categories are constructed across studies will help make methodologies more transparent and results more interpretable.
-
-
Study Designs and Analytic Methods
-
As discussed above, most studies examine chronic conditions with the highest prevalence, costs, utilization, hospitalizations, and adverse events. For example, to study chronic disease prevalence in male Medicare patients, Black and colleagues limited their analyses to the “top ten” most prevalent diseases (Black et al., 2007). Other researchers have examined a somewhat larger number of conditions, but have purposely excluded less prevalent diseases (Schafer et al. 2010). It is critical to take the number of chronic conditions being investigated into account because prevalence estimates of multimorbidity are dependent on the number of diseases that are examined. This limitation was recently discussed by Salive, who found a prevalence estimate of 17.1% for 25–44 year old primary care patients when considering a list of seven conditions, and 73.9% when considering all possible conditions (Salive, 2013). Similarly, Fortin and colleagues found prevalence estimates of 47.3% among 45–64 year old primary care patients when considering seven conditions, and 93.1% when considering an open list (Fortin et al., 2010). Schneider and colleagues found that over 20% of Medicare beneficiaries had two or more chronic conditions when using the CMS Chronic Conditions Warehouse and a list of nine potential diseases, (Schneider et al., 2009). A considerably larger figure (52%) was reported for Veteran Affairs (VA) patients when almost triple the number of potential diseases (29 conditions) was considered (Yu et al., 2003). Thus, MCC prevalence can be under-estimated when fewer chronic conditions are investigated.
In addition to the number of chronic conditions that are studied, the specific types of chronic conditions that are examined across studies differ (e.g., cardiovascular conditions are studied vs. all possible chronic conditions). The “filtering” phenomenon can be observed when comparing a list of the chronic conditions that are investigated in two separate studies. For example, comparing the chronic conditions that were studied by Newcomer and colleagues (2011) (17 chronic conditions) to Chen and colleagues (2011) (8 chronic conditions), only three conditions were found to overlap. Although prevalence estimates for single conditions may be comparable across different data collection systems and surveys (Li et al., 2012), multimorbidity prevalence estimates across studies that include different conditions complicated the interpretation, generalizability and comparability of results.
MCC research has been conducted using a variety of different study designs (See Exhibit 12). However, the majority of MCC studies used retrospective cohort and cross-sectional designs, including secondary data analyses of data, due to the need for large sample sizes. It is important to note that these study designs have systematic limitations. For example, although retrospective cohorts are longitudinal and usually contain information on a large number of patients, they are often subject to attrition bias and bias due to changes in data collection procedures over time. This is an important concern for MCC studies, as prevalence estimates may be directly impacted by changes in data collection procedures, for example sampling strategies that change in terms of periodicity and population observed over time. Similarly, cross-sectional designs are not longitudinal and provide a “snap-shot” of information at one point in time. Future MCC research may benefit from employing longitudinal, prospective studies that provide researchers with large sample sizes, but also the ability to appropriately assess potential biases and study limitations as they occur. Preferred study designs for research on less prevalent combinations of MCC produce large sample sizes, are longitudinal, and provider researchers with the ability to assess the accuracy of diagnostic coding over time. Therefore, large prospective cohorts are advantageous for research on less prevalent combinations of MCC, although they are usually very expensive. The research questions that need to be answered may also dictate which study designs are most appropriate for certain MCC studies.
Exhibit 12: MCC Study Designs and Considerations
Author Study Designs Design Considerations Ben-Noun 2001 Case-Control Small sample size, prone to recall/retrospective and selection bias, suited for rare conditions. Salisbury et al. 2011 Retrospective Cohort Large sample size, prone to attrition bias, potential unknown coding practices and changes in data collection method, longitudinal. Shelton et al. 2000 Prospective Cohort Large sample size, prone to attrition bias, known methodology changes, potential for missing data, longitudinal, highly expensive. Wolff et al. 2002 Cross-sectional Large sample size, not longitudinal, cannot measure changes over time, cannot draw causal inferences, descriptive in nature. Yu et al. 2003 Secondary Data Analysis All type of sample sizes, potential unknown coding practices and data anomalies. Other important considerations for MCC research are the limitations of the databases and algorithms used to house and analyze chronic conditions data. Over and underestimation of chronic disease prevalence may be due to database-specific characteristics. For example, the CMS Chronic Conditions Warehouse algorithm, which is used to estimate chronic disease prevalence, has been shown to underestimate the prevalence of chronic conditions requiring less frequent healthcare utilization, such as arthritis (Gorina & Kramarow, 2011). The underestimation is due to the fact that the reference period (or look back period) used in the CCW algorithm does not go back far enough to capture diagnoses that were reported on early healthcare claims and not on more recent claims. Setting (e.g., inpatient, nursing home, etc.) and other database characteristics also impact prevalence estimates and the interpretation of multimorbidity. For example, Schram and colleagues (2008) found that multimorbidity prevalence significantly varied across settings, from 22% in the inhospital setting to 82% in nursing homes. As expected, given the inherent differences between these populations, Fortin and colleagues (2010) found that MCC prevalence was much smaller in a general civilian population compared to family practice patients. In addition to the effect of “setting” on chronic disease prevalence estimates, Schram et al. (2008) also concluded that prevalence estimates are dependent on the number of chronic conditions being studied, the data collection method used to capture diagnosis information (i.e., ICD-9 vs. survey) and the time-frame being investigated, similar to the concerns raised by Gorina and Kramarow with the CCW’s look back period.
Database comprehensiveness, sampling frame and the patient population being studied all affect results. In drawing conclusions about analyses conducted on CCW data or AHRQ’s National Inpatient Sample (NIS) data, it is important to know that the CCW covers all Medicare patients, while the publically available version of the NIS covers only 20% of hospital discharges. Understanding these types of database characteristics will help researchers interpret the generalizability of their findings. The fact that the occurrence and clustering of MCC is time-dependent as patients grow older means that longitudinal datasets are best positioned to accumulate a patient’s chronic conditions over time and provide more accurate estimates of disease prevalence than cross-sectional assessments (France et al., 2011 & Wong et al., 2011). Time-dependency is an especially important concept for research on less prevalent combinations of MCC, as less common diseases are more likely to manifest over a long period of time, and diseases have different durations. Cross-sectional studies and analyses of longitudinal datasets covering limited time periods may not contain sufficient diagnostic information to study less prevalent combinations of MCC. Database size is important for research on less prevalent combinations of MCC. Large administrative datasets provide the best option due to the sheer volume of data and number of patients available for study. Less prevalent combinations of MCC are less likely to occur in small datasets with a limited number of patients and diagnoses to consider. Rare disease researchers face similar challenges.
Longitudinal databases have limitations. First, false discoveries and associations between chronic disease on the basis of too few observed diagnoses, inconsistent findings, and multiple test corrections need to be addressed (Wong et al. 2011). Additionally, the further back in time you examine longitudinal claims, the less accurately you can predict resource use and cost for a given condition or combination of conditions because of changing illness intensity over time. Although large administrative databases provide useful, current information on financial burden of disease (Riley, 2009), to more accurately predict resource use and cost, researchers need to know which diagnoses are “active” for patients currently receiving care. A laundry list of diagnoses is of little utility without a way to identify “active” conditions. Many patients will have ICD-9 codes on their past claims that represent errors, unconfirmed suspected diseases, and conditions that have been cured or are in remission. “Non-active” ICD-9 codes captured in longitudinal databases can negatively impact predictions of resource use and cost associated with MCC. Solutions may include an active problem list for patients and/or the use of supplemental data ( e.g., pharmacy and laboratory data) to confirm “active” diagnoses.
The challenges associated with conducting research on less prevalent MCC are very similar to those faced by researchers of rare diseases. Within the United States, a disease is considered to be rare when it affects less than 1 in 1000 individuals. Thus, like researchers studying less prevalent MCC, rare disease researchers are limited by small patient sample sizes and the inability of data sources to collect information on rare diagnoses, making it difficult to design clinical trials and test new treatments. In a research environment constrained by limited resources, rare disease research is given lower priority than conditions affecting more individuals (Griggs et al., 2009 & Ragni et al., 2012). It is important to consider that while any given rare disease by definition does not represent a prevalent illness, there are many rare diseases that may cumulatively affect a significant segment of the population. Finally, the likelihood of coding a rare chronic condition as a mistake may be similar to the likelihood of a patient truly having a rare disease and having this diagnosis coded accurately on a claim. Although not well studied, both research on rare diseases and research on less prevalent combinations of MCC may suffer from difficulty assessing validity.
Lastly, it is important to recognize that traditional statistical approaches may not be applicable to research on low-prevalence MCC. The issue of multiple comparisons is highly relevant for MCC research due to the number of chronic disease combinations that can be considered in the long tail. In fact, there are almost as many chronic disease combinations as there are patients. For example, if working at the three digit ICD-9 code level with approximately 1,000 diagnosis codes, about one-million pair wise comparisons would be possible. In this case, correcting for multiple comparisons using the Bonferroni method would require p-values of less than 0.00000005 to be significant. To understand the differences between low-prevalence MCC new or modified statistical approaches may need to be considered to address the multiple comparison limitation.
-
-
Reporting of MCC Research Methods
-
The amount and level of methodological detail published in MCC research papers varies greatly. Lack of consistency and detail regarding inclusion and aggregation of diagnosis codes hinders our ability to interpret research results and judge methodological quality. For example, in a manuscript describing chronic disease clustering, Schafer and colleagues provide a list of the specific ICD-10 codes they investigated in their study (Schafer et al., 2010). Conversely, in a paper looking at prevalence of chronic conditions in the VA Health Care System, Yu and colleagues did not report the ICD-9 codes that were examined. Instead the authors stated that “the diagnoses and specific codes used to identify each condition are available upon request from the authors” (Yu et al., 2003). For the purpose of developing this paper, we contracted Yu and his colleagues to obtain the list of the diagnoses and ICD-9 codes used in their study. Unfortunately, we were unable to reach the lead author and could not obtain the information.2 However, an inquiry regarding a different, but related investigation (Yoon et al., 2011) resulted in a list of diagnoses and ICD-9 codes (in SAS) that could be examined and compared to other studies.
A lack of consistency and detail in reporting diagnosis codes is only one example of the variability in methods sections in published MCC studies. Variability is also a concern in understanding why specific conditions are examined vs. others, why certain diagnosis codes are excluded from analyses, how chronic condition categories are constructed, how costs are allocated to chronic condition categories after dropping certain diagnosis codes, etc. A repository of author’s ICD-9 codes is a potential mechanism by which authors could explain why certain diagnosis codes were included or excluded from specific analyses. However, to effectively address the variability across MCC studies a reporting framework or set of criteria, such as the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), may be necessary to begin to standardize efforts and reporting across researchers (Moher et al., 2009).
2 Personal communication with available authors of Prevalence and Costs of Chronic Conditions in the VA Health Care System in Medicare Care Research and Review, 2003.
-
View full report
"rpt_LowPrevMCCData.pdf" (pdf, 1.37Mb)
Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®