In the United States, much of the research on multiple chronic conditions focuses on highly prevalent conditions ( e.g., obesity, hypertension, and diabetes) because they affect a large number of individuals, may be successfully managed or controlled, and they are included in major national surveys and other data collection efforts. The purpose of the literature was two-fold: 1) to characterize the burden of MCC across various populations, and 2) to identify MCC populations associated with increased healthcare utilization and poorer quality of care, so that patients can be targeted for intervention by providers, health plans and public health officials.
To conduct the work, researchers have employed methods such as basic prevalence and incidence calculations, and regression modeling and odds ratios to predict healthcare utilization and cost based on either the occurrence of MCC or the number of MCC a patient has. MCC prevalence has commonly been measured as the percent of patients in a population with two or more chronic conditions, while chronic disease complexity has been assessed through examining the distribution of MCC patients across an increasing number of chronic conditions ( e.g., 2, 3, 4, 5 + conditions). Predictive statistics are then used to estimate healthcare utilization, cost and quality of care based on the occurrence of MCC or the number of MCC for a specific patient.
A number of studies and initiatives have investigated both the prevalence and complexity of MCC across a variety of different populations (See Exhibit 5). One of the most well-known is the CMS Chronic Conditions Data Warehouse (CCW), which not only provides a database of patients with chronic conditions for research purposes, but also an interactive dashboard to investigate chronic condition prevalence, condition counts, and utilization information using a variety of different demographic filters ( e.g., gender, geographic area, and dual eligibility status). Information contained in the CCW database originates from Medicare and Medicaid beneficiary claims and assessment data from different healthcare settings across the continuum of care (CMS, 2013). By using the dashboard (which can be applied to Medicare fee-for-service beneficiaries only), users can compare prevalence estimates between states or between a state and national benchmarks. As an example, 37% of Medicare beneficiaries in New Hampshire have five or more chronic conditions, compared to 41% in Alabama and 43% nationwide.
The Faces of Medicaid publications have articulated the prevalence of MCC among Medicaid beneficiaries. The 2007 publication The Faces of Medicaid II: Recognizing Needs of People with Multiple Chronic Conditions estimated that 10% of non-disabled, adult Medicaid beneficiaries had three or more chronic condition categories, compared to 35% of adults with a disability and 39% of elderly Medicaid beneficiaries (Kronick et al., 2007). Similarly, in 2012 the National Center for Health Statistics reported that 21% of non-elderly, United States adult civilians have two or more chronic conditions, and that the rate of MCC in populations is increasing over time (Fried et al., 2012). As shown in Exhibit 5 below, prevalence estimates differ by study, depending on the population being studied, the number of chronic conditions per person included in the study, and the number of combinations of MCC.
Exhibit 5: Comparison of MCC Prevalence Estimates by Study; for Over and Under Age 65
|Author||Country||Population||Primary Data Source||Grouping System||# of CCs||MCC Prevalence||# of Comb.|
|Legend: Primary Data Source, what data was analyzed in each study (NHIS=National Health Interview Survey, MEPS=Medical Expenditure Panel Survey, HRS=Health & Retirement Study, EMR=Electronic Medical Record, BRFSS=Behavioral Risk Factor Surveillance System); Grouping Systems, system used to aggregate diagnosis codes together (ACG=Adjusted Clinical Groups Case-mix System, CDPS=Chronic Illness Disability Payment System, CCS=Clinical Classification System, CCW ALGM=Chronic Conditions Data Warehouse Algorithm, ADG=Aggregated Diagnosis Groups; # of CCs, number of chronic conditions categories studied; # of Comb. (Combinations), how researchers examined complexity by stratifying patients into categories representing the occurrence of different numbers of chronic conditions ( e.g., e.g., 2, 3, 4, 5 + conditions, etc.).|
|Average Population Age < 65 Years|
|Fortin et al. (a) 2010||Canada||Adult Civilians||Community Survey||n/a||7||14%||≥ 2|
|Fried et al. (a) 2012||U.S.||Adult Civilians||NHIS||n/a||9||21%||≥ 2|
|Machlin & Soni 2013||U.S.||Adult Civilians||MEPS||n/a||20||25%||1 to > 4|
|Ward et al. 2013||U.S.||Adult Civilians||NHIS||n/a||10||26%||1 to > 4|
|Prados-Torres et al. (a) 2012||Spain||Primary Care Patients||EMR Data (ICD-9)||ACG||264||26%||1 to 14|
|RWJF 2010||U.S.||Adult Civilians||MEPS||n/a||9||28%||1 to 5|
|Chen et al. 2011||U.S.||Adult Civilians||BRFSS||n/a||8||29%||1 to ≥ 3|
|Kronick et al. 2007||U.S.||Medicaid Patients||Medicaid Claims||CDPS||20||39%||1 to > 7|
|Lee et al. 2007||U.S.||VA Patients||VA Databases||CCS||11||41%||1 to > 4|
|Yoon et al. 2011||U.S.||VA Patients||VA Databases||n/a||16||48%||1 to > 4|
|Yu et al. 2003||U.S.||VA Patients||VA Databases||n/a||29||52%||1 to > 3|
|Naessens et al. 2011||U.S.||Adult Employees & Dependents||Insurance Claims||CCS||259||54%||≥ 2|
|Fortin et al. (b) 2010||Canada||Family-Practice Patients||Family Practice- based Sample||n/a||7||58%||≥ 2|
|Schneider et al. 2012||Sweden||Adult Inpatients||EMR Data (ICD-10)||n/a||22||93%||≥ 2|
|Average Population Age ≥ 65 Years|
|Ford et al. 2013||U.S.||Adult Civilians||NHIS||n/a||9||15%||1 to 5|
|Schneider et al. 2009||U.S.||Medicare Patients||CMS CCW||CCW ALGM||9||20%||1 to > 3|
|Erdem et al. (a) 2013||U.S.||Medicare Part A Patients||CMS CCW||CCW ALGM||27||37%||1 to 10|
|Erdem et al. (b) 2013||U.S.||Medicare Part B Patients||CMS CCW||CCW ALGM||27||41%||1 to 10|
|Fried et al. (b) 2012||U.S.||Adult Civilians||NHIS||n/a||9||45%||≥ 3|
|Schoenberg et al. 2007||U.S.||Adult Civilians||HRS||n/a||8||58%||1 to > 5|
|Salisbury et al. 2011||U.K.||Adult GP Patients||GP Database||ACG||260||58%||1 to ≥ 5|
|Wolff et al. 2002||U.S.||Medicare Patients||Medicare Claims||ADG||24||65%||1 to ≥ 4|
|Glynn et al. 2011||U.K.||Family-Practice Patients||Medical Record Data||n/a||147||66%||1 to > 4|
|Salive 2013||U.S.||Medicare Patients||Medicare Claims||CCW ALGM||15||67%||≥ 2|
|Prados-Torres et al. (b) 2012||Spain||Primary Care Patients||EMR Data (ICD-9)||ACG||264||67%||1 to 14|
|Lochner et al. 2013||U.S.||Medicare Patients||Medicare Claims||CCW ALGM||15||68%||1 to ≥ 4|
|CMS Chartbook 2012||U.S.||Medicare Patients||CMS CCW||CCW ALGM||15||69%||1 to >6|
|CMS CCW 2013||U.S.||Medicare Patients||CMS CCW||CCW ALGM||27||73%||1 to > 6|
|John et al. 2003||U.S.||American Indians||Community Survey||n/a||11||74%||1 to > 6|
|Steinman et al. 2012||U.S.||VA Patients||VA Databases||CCS||23||90%||1 to > 8|
When studying MCC prevalence among Veteran Affairs (VA) patients, Yu and colleagues found that 52% of VA patients had two or more chronic conditions. Of that number, 17% of patients had two chronic conditions, while 35% had three or more (Yu et al. 2003). Similarly, Steinman and colleagues found that approximately 90% of elderly VA patients had three or more chronic conditions; 44% has three to five chronic conditions, while 32% and 14% had six to eight and greater than eight conditions, respectively (Steinman et al. 2012). These prevalence estimates are considerably higher than the 21% that has been reported for all Americans (Vogeli et al., 2007), but demonstrate that MCC are more prevalent within certain populations and increasing age groups. The studies listed in Exhibit 5 have been stratified by average patient population age (less than or greater than or equal to 65 years) to demonstrate the effect of age on MCC prevalence calculations.
As shown in Exhibit 5 prevalence estimates ranged from 14% in Canadian civilians to 93% in Swedish adult in-hospital patients. Although patient population and setting play important roles in determining prevalence, utilizing different methods and analytic techniques can also lead to inconsistent estimations. Researchers used anywhere from nine to 260 chronic conditions categories to study prevalence and the occurrence of one to fourteen chronic conditions to examine different depths of chronic disease complexity. Various data sources and diagnosis code grouping systems were also used.
Just as MCC prevalence is associated with patient age, it has been well documented that healthcare expenditures are positively associated with an increasing number of MCC (Lehnert et al., 2011). In the study by Yu and colleagues 73% of total costs to the VA healthcare system were found to be attributable to patients with three or more chronic conditions, while only 13% and 9% of costs could be attributed to patients with two or a single chronic condition, respectively (Yu et al. 2003). Likewise, in a study of working-age self-funded health plan enrollees, mean annual cost of MCC increased from $4,442 for patients with one chronic condition to over $23,000 for patients with five or more MCC (Naessens et al., 2011). A similar relationship between MCC and cost has also been observed with regard to out-of-pocket medical expenditures (Schoenberg et al., 2007).
MCC are also associated with increased healthcare utilization and mortality as well as poorer quality of life for patients. In a cross-sectional study of Medicare fee-for-service beneficiaries, Wolff and colleagues found a positive relationship between inpatient admissions and hospitalizations and the number of chronic conditions a patient had (Wolff et al., 2002). Similarly, Glynn and colleagues found a strong association between an increasing number of chronic conditions and the frequency of primary care consultations, hospital admissions and hospital out-patient visits among primary care patients (Glynn et al., 2011). With regard to mortality, research suggests that patients with MCC have higher mortality rates compared to patients without chronic conditions. Lee and colleagues found that the five year mortality rate for patients without chronic conditions (4%) was considerably lower than for patients with one (6%), two (8%), three (11%) or four or more (17%) diseases (Lee et al 2007). Lastly, patients with MCC are known to more frequently report limitations in daily living/instrumental activities and “fair” or “poor” overall health status compared to patients without MCC (Chen et al., 2011 & Gulley et al., 2011). However, arguments have been made that patients with a large number of MCC may actually receive higher quality care than patients with fewer conditions due to the increased number of physician visits these patients make (Bae & Rosenthal, 2008).
Overall, the majority of MCC research conducted to-date describes prevalence and complexity of multimorbidity, as well as the relationships that exist between MCC and healthcare utilization, cost and other related metrics. Findings demonstrate that MCC are common across all populations, but are concentrated in specific patient populations and age groups (e.g., the elderly, disabled and VA patients). Furthermore, MCC are associated with increased healthcare utilization, costs, and mortality, as well as lower quality of life. Finally, MCC research in the United States has primarily been conducted on chronic conditions that are highly prevalent and well-known; low-prevalence conditions have not been well studied.