Patterns of Multimorbidity in Primary Care Patients at High Risk of Future Hospitalization

Introduction

M ultimorbidity, defined as the co-occurrence of 2 or more (chronic) conditions within 1 person, ¹ is found to be the rule rather than the exception in primary care. ² As a result of advances in medical care, the proportion of multimorbid patients is expected to increase during the next decades. ³ Multimorbidity negatively affects quality of life, ⁴ life expectancy, ⁵ and physical functioning. ⁶ Multimorbid patients are more likely to be hospitalized for a “potentially avoidable cause,” thus accounting for substantial health care spending. ⁷ However, most of the current guidelines, and subsequently the majority of evidence-based disease management programs, focus on single diseases. ⁸

Based on emerging research in the field of multimorbidity, complex interventions, such as care management programs, have been invented to address the care needs of multimorbid patients. ⁹ These programs are designed to assist patients and their support systems in managing medical and nonmedical conditions via individualized care planning and monitoring. ¹⁰ Targeting these programs to patients at highest risk for cost-intensive care offers the greatest opportunity to improve quality of care and reduce health care costs. ¹¹

Since the 1980s, several statistical models, commonly known as “predictive models,” have been developed to predict future health care utilization and costs. ¹² These models are based on data about morbidity, prior health care utilization, and costs, which can be obtained easily from health insurance claims data. Predictive models can be used to precisely identify “high-risk” individuals as future participants of intensive care programs. ¹³ These models indicate that predicted high-utilizers are more likely to be complex patients with a higher level of multimorbidity.

However, tailoring care interventions to the needs of this highly select population requires that underlying patterns of (multi)morbidity be explored; type and severity of co-occurring conditions are known to affect self-care capabilities ¹⁴ and medical treatment. ¹⁵ Until now, few studies have explored the co-occurrence of chronic conditions in different populations, focusing either on large disease clusters, ^16,17 pairs, ¹⁸ or triads. ¹⁹ Expanding on evidence from these epidemiological studies, we aim to inform care providers about the type, numbers, and mutually exclusive patterns of co-occurring chronic conditions in patients with a high predicted risk of hospitalization who are expected to benefit from care management.

Methods

This analysis is part of a retrospective cohort study of 6026 patients from 10 general practices in southwest Germany. ²⁰ All patients are beneficiaries of the German General Regional Health Fund (Allgemeine Ortskrankenkasse, AOK). We analyzed insurance claims data from January 1, 2007 to December 31, 2008. The study protocol was approved by the institutional review board of Heidelberg University Hospital.

Results

Discussion

Tailoring care interventions for a high-risk population requires the exploration of underlying (multi)morbidity. In this study we determined single prevalence of chronic conditions and explored mutually exclusive patterns of multimorbidity using insurance claims data, including diagnoses from hospital and ambulatory care. Case finding via predictive modeling identified multimorbid older patients with complex care needs that result from heterogeneous patterns of chronic diseases.

Risk selection led to high single prevalence of severe chronic conditions such as malignant disorders or chronic heart failure. Compared to epidemiological studies on multimorbidity in elderly patients, we found notably higher prevalence of the 10 most frequent conditions under study. ^{16 –19} As the predictive model is designed to concentrate patients at high risk for future hospitalizations, it is not surprising that 6 of the 10 most frequent chronic conditions intersect with the most frequent causes for hospital admissions in Germany in 2009 ²⁴ (hypertension, coronary heart disease, type 2 diabetes mellitus, chronic heart failure, osteoarthritis of the hip/knee, and bronchial cancer). Four of these conditions (hypertension, coronary heart disease, type 2 diabetes mellitus, and chronic heart failure) are known to be ambulatory care-sensitive conditions, meaning that optimal (primary) care could help to reduce the number of hospitalizations. ²⁵

Despite low single frequencies, the 15 multimorbidity patterns described in this study could be seen as typologies and considered a starting point for the development of a care management intervention. As focusing care on single diseases may lead to suboptimal health outcomes or harm, ⁸ evidence-based care management for this entire population should account for “typical” multimorbidity patterns. Causal comorbidity ²³ may explain in part the observed higher prevalence of P3, P4, and P9 (defined in Table 3). Diabetes, hypertension, coronary heart disease, and chronic heart failure can be seen as a cascade of pathogenetically-related diseases. Neuropathy and visual impairment resulting from retinopathy are common secondary complications of long-term diabetes. However, these patient groups demand different care strategies: P4 patients may profit from self-management support (eg, increased physical activity, smoking cessation, and dietary advice) and optimal medication regimen. In contrast, pain (osteoarthritis) and dyspnea (chronic heart failure) may significantly limit self-management capabilities of P9 patients, and treatment of clinically dominant malignant disorders may interfere with medical treatment of hypertension, diabetes, and coronary heart disease. ²⁶ In addition to pain, depression is a challenging component in 3 of the identified multimorbidity patterns; 2 (P7, P15; defined in Table 3) may require that care focus on depressive symptoms. Patients suffering from P14 may be at high risk of nonadherence to treatment regimens and less capable of managing their diabetes because of co-occurring depression and pain. ²⁷

In 1976, a German general practitioner first explored patterns of multimorbidity in geriatric patients. ²⁸ Since then, several studies have been undertaken to reveal patterns of multimorbidity in different patient populations using different methodologies. ^{16 –19,29} Two of these studies rely on cluster analysis ^{16 –18} ; 1 relies on exploratory factor analysis ¹⁷ ; 3 revealed large disease clusters such as cardiovascular/metabolic disorders or neuropsychiatric disorders. The analysis by Weiss et al ²⁹ focused on mutually exclusive co-occurrence of osteoarthritis, cerebrovascular accident, chronic lower respiratory tract disease, coronary heart disease, and diabetes in individuals older than 65 years of age. They identified 22 patterns with frequencies between 0.5% and 7%, indicating high heterogeneity within the population. Whereas large disease clusters may be of high epidemiological interest, mutually exclusive patterns may offer insight into real-life challenges in clinical practice, because we encounter individuals rather than groups during our daily routine. However, methodological barriers limit the latter approach. Given the astronomical numbers of theoretically possible combinations, large populations are needed to adequately apply these methodologies. In our sample, less than half of all theoretically possible combinations could be observed in a population that is about 1.5 times larger than the total number of possible combinations. The observed pattern frequencies are very low but comparable to similar studies, ^19,29 indicating a dilemma: multimorbidity exploration either results in large disease clusters that are more or less evident on the basis of pathogenetic concepts or it describes highly heterogeneous individual patterns. Exploration of multimorbidity patterns among highly frequent diseases inevitably produces a number of chance findings. ³⁰ Therefore, from the epidemiological perspective, the findings from this study need further external validation through larger studies. However, our results may inform care providers and researchers about “typical” patterns that may occur in select high-risk patients. This will help to tailor intensive care interventions to their needs with regard to clinical management and self-care.

Because of methodological considerations, we decided to limit exploration of multimorbidity patterns to the 10 most frequent conditions in LOH^high subjects. Hence, we omitted conditions such as chronic obstructive pulmonary disease, which is known to be a highly relevant ambulatory care-sensitive condition with a significant proportion of potentially avoidable hospitalizations. ³¹ Furthermore, significant nonmedical factors, which are known to be associated with potentially avoidable hospitalizations, are not explored in this study (eg, social deprivation, area of residence, ethnicity). ³² In Germany, evidence of avoidable hospitalizations is scarce and is focused mainly on geriatric patients. ³³ Further research should determine which medical and nonmedical factors contribute to avoidable hospitalizations in Germany in order to compare them with what is known from other countries and health care systems.

Information bias also may affect our results in that patients with a chronic disease have more contact with health care professionals and therefore have more diseases identified. An earlier exploratory study comparing German insurance claims data with the charts of German general practitioners revealed a high degree of correspondence for most of the most common conditions (eg, hypertension, diabetes, cardiovascular disease). ³⁴

Predictive modeling identified multimorbid elderly patients with highly individual patterns of co-occurring chronic conditions. Most of the co-occurring conditions appear to be sensitive to ambulatory care. However, development of intensive care programs for primary care patients who are at high risk for hospitalizations must take into account frequently co-occurring conditions, such as like chronic pain and depression, that may interfere with medical treatment and self-care.