Understanding the High Prevalence of Low-Prevalence Chronic Disease Combinations: Databases and Methods for Research. Chronic Condition Clusters and Co-occurring Conditions

09/20/2013

Research on chronic condition clusters and conditions that co-occur with a primary or “index” disease is increasing and leading to understanding of patterns of chronic disease combinations and how MCC co-occur or spread across populations in clinically and statistically meaningful ways. Knowing which chronic diseases tend to co-occur together and manifest over time offers clinicians the ability to develop multi-disease clinical guidelines and to identify opportunities for longitudinal disease prevention for patients. Conditions that co-occur may be statistically associated with one another with no known causal relationship or have an underlying pathophysiological connection (van den Akker et al., 1998). Although more research may be warranted to further investigate non-causal disease relationships, understanding which chronic conditions tend to cluster together provides clinicians with the opportunity to more accurately target disease prevention efforts and understand multimorbid complexity on a more granular scale. Metabolic syndrome is an example of a cluster that is widely recognized in the United States as well as internationally.

Two methodological approaches have been used to study patterns of chronic disease combinations and MCC co-existence. The more simplistic of these two approaches is to calculate the most common dyads and triads of co-occurring chronic conditions by determining what chronic conditions co-occur with an index disease, or by simply examining the percentage of patients in a population with a given combination of chronic diseases (Marengoni et al., 2009). For example, Lochner and colleagues found that hypertension and hyperlipidemia was the most common dyad among Medicare patients of all age groups (Lochner et al., 2013) while diabetes, hypertension, and hyperlipidemia was the most prevalent triad among younger Medicare patients, and ischemic heart disease, hypertensions, and hyperlipidemia, and arthritis, hypertension, and hyperlipidemia were the most common triads among older Medicare patients. As shown in Exhibit 6, a number of different dyads and triads have been reported in the literature to-date. Most studies report two-way and three-way combinations that include chronic diseases such as hypertension, hyperlipidemia, heart disease, diabetes and arthritis. Low-prevalence chronic disease combinations have not been included within reported dyads and triads.

Exhibit 6: Research on Co-occurring Chronic Condition Dyads and Triads

Author Country Population Mean Age (≥65) # of CCs # of Clusters Description of Chronic Disease Clusters
Legend: # of CCs, number of chronic conditions categories studied; # of clusters, the number of chronic condition clusters observed by researchers; Description of chronic disease clusters, how authors characterized the chronic condition clusters they observed.
CMS Chartbook 2012 U.S. Medicare Patients Yes 15  

Dyads

  • High cholesterol & high blood pressure
  • High cholesterol & ischemic heart disease

Triads

  • High cholesterol, high blood pressure, & ischemic heart disease
  • High cholesterol, high blood pressure, & diabetes

(Most prevalent clusters listed)

Fried et al. (a) 2012 U.S. Adult Civilians No 9 3

Dyads

  • Hypertension & diabetes
  • Hypertension & heart disease
  • Hypertension & cancer
Fried et al. (a) 2012 U.S. Adult Civilians Yes 9 3

Dyads

  • Hypertension & diabetes
  • Hypertension & heart disease
  • Hypertension & cancer
Kronick et al. (2007) U.S. Medicaid Patients No 20 5

Triads

  • Cardiovascular-Pulmonary-Gastrointestinal
  • Cardiovascular-Central Nervous System- Pulmonary
  • Central Nervous System -Pulmonary- Gastrointestinal
  • Cardiovascular-Central Nervous System- Gastrointestinal
  • Cardiovascular-Pulmonary-Psychiatric
Lochner et al. (a) 2013 U.S. Medicare Patients No 15 5

Dyads

  • Hypertension & hyperlipidemia (M)
  • Hypertension & hyperlipidemia (F)

Triads

  • Diabetes, hypertensions & hyperlipidemia (M)
  • Diabetes, hypertensions & hyperlipidemia (F)

(Two most prevalent clusters listed by sex)

Lochner et al. (b) 2013 U.S. Medicare Patients Yes 15 5

Dyads

  • Hypertension & hyperlipidemia (M)
  • Hypertension & hyperlipidemia (F)

Triads

  • Ischemic heart disease, hypertension, & hyperlipidemia (M)
  • Arthritis, hypertension, & hyperlipidemia (F)

(Two most prevalent clusters listed by sex)

Machlin & Soni 2013 U.S. Adult Civilians No 20 12

Dyads

Hypertensions & hyperlipidemia

  • Diabetes & hypertension
  • Diabetes & hyperlipidemia

Triads

  • Hypertension, hyperlipidemia, & diabetes
  • Hypertension, hyperlipidemia, & coronary artery disease

(Most prevalent clusters listed)

Schoenberg et al. 2007 U.S. Adult Civilians Yes 8 7

Dyads

  • High blood pressure & arthritis
  • High blood pressure & heart disease
  • High blood pressure & diabetes

Triads

  • High blood pressure, heart disease, & arthritis
  • High blood pressure, heart disease, & diabetes

(Most prevalent clusters listed)

Steinman et al. (2012) U.S. VA Patients Yes 23 30

Triads

  • Hypertension, hyperlipidemia, & CHD (M)
  • Hypertension, hyperlipidemia, & arthritis (F)

(Two most prevalent clusters listed by sex)

The second methodological approach that has been used to study patterns of chronic disease combinations and MCC is cluster analysis. Cluster analysis is a type of statistical approach that groups relatively homogenous or similar patients into clinically relevant groupings based on calculated correlations between diagnoses. Cluster analysis is a relatively “novel” statistical method and as a result, specific methods employed vary significantly across studies. For example, researchers have used techniques such as agglomerative hierarchical clustering, factor analysis, and multiple correspondence analysis, among other approaches, to examine correlations between diagnoses. The variability in these approaches makes it difficult to interpret chronic condition clustering research, as differences in analytic approach may influence results.

The number of chronic disease clusters vary by study, reporting anywhere from three to thirty clinically significant chronic disease clusters or patterns that warrant attention or further investigation (Schafer et al. 2010 & Steinman et al., 2012). In a study by Prados-Terros and colleagues, five patterns of chronic disease clustering were observed in a primary care population: cardio-metabolic, psychiatric-substance abuse, mechanical-obesity-thyroidal, psychogeriatric, and depressive disorders (Prados-Torres et al., 2012). Similarly, John and colleagues found four clusters among a rural community-dwelling population which included cardiopulmonary, sensory-motor, depressive and arthritic disorders (John et al., 2003). As shown in Exhibit 7, the majority of chronic condition clusters include diagnoses related to cardiovascular, metabolic, neurological and mental health conditions, which are common conditions. Low-prevalence chronic disease combinations that would be found in the “long tail” have not been reported as outputs of cluster analysis studies to-date.

To-date studies on chronic condition clusters have primarily been conducted outside of the United States, in countries such as Sweden, Spain and Germany. The international tendency speaks to the quality and granularity of data available in the United States compared to other countries. European countries in-particular have more standardized and robust healthcare data infrastructures compared to the United States (OECD, 2013).

Exhibit 7: MCC Research Studies Using Cluster Analysis by Author

Author Country Population Mean Age (≥65) # of CCs # of Clusters Description of Chronic Disease Clusters
Legend: # of CCs, number of chronic conditions categories studied; # of clusters, the number of chronic condition clusters observed by researchers; Description of chronic disease clusters, how researchers characterized the chronic condition clusters they observed.
Garcia- Olmos et al. (2012) Spain GP Patients No n/a 4
  • Cardiac arrhythmias, hyperlipidemia, hypertension, & diabetes.
  • Ischemic heart disease, CVD, chronic renal failure, & CHF.
  • Asthma, thyroid disease, anxiety or depression, & schizophrenia.
  • Obesity, osteoporosis, deafness, malignancy, & degenerative joint disease
John et al. (2003) U.S. Community- resident American Indians Yes 11 4
  • Cardiopulmonary
  • Sensory-motor
  • Depression
  • Arthritis
Marengoni et al. (2009) Sweden Stockholm Community Members Yes 15 5
  • Hypertension, heart failure, chronic atrial fibrillation, & CVD.
  • Thyroid dysfunction, COPD, & CHD.
  • Diabetes, visual impairments, & deafness.
  • Dementia, depression & hip facture
  • Malignancy & anemia
Newcomer et al. (2011) U.S. KPCO Insurance Members No 17 10
  • Chronic pain & mental health conditions
  • Diabetes, obesity & mental health conditions
  • Kidney disease, diabetes & obesity
  • Mental health conditions & obesity
  • Mental health conditions, diabetes, obesity, & stroke.
  • Cardiac disease, obesity, & diabetes
  • COPD, obesity & mental health conditions
  • Gastrointestinal bleeding, obesity, & mental health conditions.
  • Abdominal surgery, orthopedic surgery, & obesity
  • Cancer, obesity, & mental health conditions.
Prados- Torres et al. (2012) Spain Primary Care Patients No 264 5
  • Cardio-metabolic
  • Psychiatric-substance abuse
  • Mechanical-obesity-thyroidal
  • Psychogeriatric
  • Depressive
Schafer et al. (2010) Germany Ambulatory Care Patients Yes 46 3
  • Cardiovascular/metabolic disorders
  • Anxiety/Depression/Somatoform disorders, & pain
  • Neuropsychiatric disorders

Although chronic condition clustering and co-occurring conditions research is relatively new, it is a promising means by which to study patterns of chronic disease combinations and the full complexity of disease in various populations. However, the variability in analytic methods used to study co-existing MCC (e.g., dyads, triads, cluster analysis) make the results of these studies difficult to interpret and generalize to other populations. Also, clustering research has primarily been conducted on chronic conditions that are prevalent and/or aggregated into large groups (e.g., all cancers and mental illness); studies have not reported “long tail” distributions of potential disease clusters.

View full report

Preview
Download

"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®