Volume 1243, Issue 1 p. 69-87
Open Access

Prioritization of care in adults with diabetes and comorbidity

Neda Laiteerapong

Neda Laiteerapong

Section of General Internal Medicine, Department of Medicine, University of Chicago, Chicago, Illinois

Search for more papers by this author
Elbert S. Huang

Elbert S. Huang

Section of General Internal Medicine, Department of Medicine, University of Chicago, Chicago, Illinois

Search for more papers by this author
Marshall H. Chin

Marshall H. Chin

Section of General Internal Medicine, Department of Medicine, University of Chicago, Chicago, Illinois

Search for more papers by this author
First published: 23 December 2011
Citations: 35
Neda Laiteerapong, M.D., Department of Medicine, University of Chicago, 5841 S. Maryland Avenue, MC 2007, Chicago, IL 60637. [email protected]

Abstract

Approximately half of adults with diabetes have at least one comorbid condition. However, diabetes care guidelines focus on diabetes-specific care, and their recommendations may not be appropriate for many patients with diabetes and comorbidity. We describe Piette and Kerr's typology of comorbid conditions, which categorizes conditions based on if they are clinically dominant (eclipse diabetes management), symptomatic versus asymptomatic, and concordant (similar pathophysiologic processes as diabetes) versus discordant. We integrate this typology with clinical evidence and shared decision-making methods to create an algorithmic approach to prioritizing care in patients with diabetes and comorbidity. Initial steps are determining the patient's goals of care and preferences for treatment, whether there is a clinically dominant condition or inadequately treated symptomatic condition, and the risk of cardiovascular disease. With these data in hand, the clinician and patient prioritize diabetes treatments during a shared decision-making process. These steps should be repeated, especially when the patient's clinical status changes. This patient-centered process emphasizes overall quality of life and functioning rather than a narrow focus on diabetes.

Introduction

Approximately half of adults with diabetes and nearly 60% of elderly adults with diabetes have at least one comorbid chronic disease.1–3 As many as 40% of elderly adults with diabetes have four or more comorbid diseases.4 However, diabetes care guidelines typically focus on decreasing diabetes-related microvascular and cardiovascular complications5–8 and may not apply to many patients with diabetes who are also burdened by other comorbid conditions.1, 9–11

Comorbidity is defined by the presence of two or more medically diagnosed conditions in the same person.12 Comorbidity includes conditions that are highly symptomatic with large effects on quality of life (e.g., depression)13 and those that have no immediate effects on quality of life (e.g., hyperlipidemia).14 When comorbidity is present, the status and treatment of one disease can improve,15 worsen,15–17 or have little effect on the other disease.

A comorbid chronic condition can influence the health outcomes of a patient with diabetes through multiple mechanisms. By their mere presence, some comorbid chronic conditions can directly worsen health-related quality of life (HRQL).13,18 Comorbid chronic conditions, especially depression and chronic pain, can also impair the ability of patients to perform diabetes self-management and to adhere to treatments; these changes in self-care behavior can increase the probabilities of adverse health outcomes.19–24 Having multiple chronic conditions or functional disabilities may also decrease the potential health benefits of intensive glucose control by increasing the risk of background mortality.25,26

The interactions of diabetes and comorbid conditions are becoming increasingly important due to the increasing prevalence of individuals with multiple chronic conditions. This trend can be partially attributed to progress in public health and medicine, which has increased life expectancy and led to an older population at greater risk for chronic diseases.27 We have also become increasingly successful at decreasing the risk of mortality for specific chronic diseases such as diabetes, heart disease, and cerebrovascular disease, thus extending the life expectancy of those living with chronic disease.28–30 At the same time, the recent obesity epidemic has increased the prevalence of diabetes.31,32 The combined effect of these trends is that people are living longer with more chronic diseases, which increases the prevalence of patients with diabetes and comorbid conditions.

In light of the rising prevalence of diabetes and comorbidity, significant need exists for guidelines to address the interaction of diabetes with other conditions. The Veterans Affairs/Department of Defense (VA/DoD)7 and the California Healthcare Foundation/American Geriatrics Society (CHF/AGS)8 diabetes care guidelines include detailed discussions about specific comorbid conditions. The VA/DoD guideline recommends considering the patient's total health status when evaluating their diabetes management. As part of that consideration, the guideline identifies comorbid diseases that occur more frequently among patients with diabetes (e.g., coronary artery disease and hyperlipidemia), conditions that affect diabetes management (e.g., substance use disorder and depression), and diseases that may cause secondary diabetes (e.g., Cushing's disease). The CAF/AGS guidelines recommend screening patients for comorbid conditions, known as geriatric syndromes (depression, chronic pain, incontinence, and falls), which are highly prevalent in the elderly and are associated with lower HRQL.13 Although the previously mentioned guidelines are unusual because they discuss specific comorbid illnesses, many more guidelines do recommend considering life expectancy when setting glucose control targets.5–8

While these guidelines introduce some important considerations in patients with diabetes and comorbidity, numerous questions remain: How does one prioritize treatment decisions in adults with diabetes and comorbidity? Are there certain comorbid diseases that should be treated prior to others? And what key issues should be considered when caring for adults with diabetes and comorbidity?

In this review paper, we address these questions by suggesting an approach for prioritizing care decisions in patients with diabetes and comorbidity that integrates a typology for understanding comorbid conditions with the tenets of shared decision making. We describe the typology using three clinical conditions that highlight how comorbid conditions may guide diabetes treatment decisions (Table 1). We then outline a general approach to prioritizing care based on shared decision making—a marriage of the patient's preferences, population-based evidence, and the clinician's judgment (Fig. 1).33

Table 1. Clinical evidence for relationships between diabetes and specific comorbid conditions
Author Study design Study population Intervention Measurements Outcomes
Clinically dominant condition: end-stage diseases—limited life expectancy
UKPDS 33 study35 Randomized controlled trial 3,867 U.K. adults with new-onset type 2 diabetes Intensive (FPG <6 mmol/L) with sulfonylurea or insulin versus conventional glucose therapy (diet alone) Any diabetes-related endpoint, diabetes-related death, all-cause mortality Differences in diabetes-related death and all-cause mortality were not significant between groups. Compared to the conventional group, the intensive group had a 12% lower risk (CI: 1–21, P= 0.03) for any diabetes-related endpoint, especially due to a 25% risk reduction (CI: 7–40; P= 0.01) in microvascular endpoints. Kaplan–Meier plots for microvascular endpoints separated significantly only after nine years of follow-up. After 12 years of follow-up, results from most surrogate endpoints (ankle reflexes, cardiomegaly, and silent myocardial infarction) and end-stage complications (blindness and peripheral vascular disease) did not differ between groups.
Huang et al.26 Diabetes microsimulation model Hypothetical population of adults 60–80 years old with type 2 diabetes and varied life expectancies estimated from a validated population-level mortality index Intensive (A1C 7.0%) versus moderate glucose control (A1C 7.9%) Lifetime difference in complications incidence and average quality-adjusted days Expected benefits of intensive control declined as the mortality index increased (indicated by increasing levels of comorbid illness and functional impairment) for all age groups. For example, for adults 60–64 years old with new-onset diabetes, the benefits declined from 106 quality-adjusted days (CI: 95–117 days) at baseline good health to 44 days (CI: 38–50 days) with three additional mortality index points and eight days (CI: 5–10 days) with seven additional index points.
Greenfield et al.25 Five-year longitudinal observational study 3,074 patients with type 2 diabetes categorized into high- and low-to-moderate comorbidity based on responses to the TIBI questionnaire Total mortality and five-year incident cardiovascular events. HR adjusted for age and sex. A1C ≤6.5% at baseline was associated with lower cardiovascular events in the low-to-moderate comorbidity group (TIBI ≤12) (adjusted HR 0.60; CI: 0.42–0.85; P= 0.005). A1C <7.0% was associated with fewer cardiovascular events in the low-to-moderate comorbidity group (adjusted HR 0.61; CI: 0.44–0.83; P= 0.001). Neither A1C ≤6.5% nor A1C ≤7.0% was associated with lower cardiovascular events in the high comorbidity group (TIBI >12 points).
Symptomatic condition: chronic pain
Laiteerapong et al.13 Cross-sectional survey 6,317 patients with diabetes receiving care through Kaiser Permanente Northern California HRQL measured with a short survey based on the SF-8 Patients with chronic pain had lower physical HRQL scores (β-coefficient, –5.6; CI: –6.1 to –5.0; P < 0.001) to the same degree as participants who had myocardial infarction (β, –5.3; CI: –7.0 to –3.5; P < 0.001) stroke (β, –4.7; CI: –7.6 to –1.8; P= 0.002), or hypoglycemia (β, –4.7; CI: –8.2 to –1.3; P < 0.001).
Bair et al.45 Cross-sectional survey 11,689 participants with diabetes participating in a multisite prospective cohort study HRQL measured by the SF-12 Patients with diabetes and chronic pain were more likely to report being depressed than patients without chronic pain (41.3% vs. 16.2%; P < 0.001). Patients who reported moderate pain were 1.3 times (OR, CI: 1.1–2.0) more likely to report depression, and those with extreme pain were 2.0 times (OR, CI: 1.6–2.5) more likely to report depression.
Krein et al.21 Cross-sectional survey 993 patients with diabetes receiving care through the Department of Veterans Affairs (VA) Self-report of chronic pain, (pain present most of the time for ≥six months during past year) and diabetes self-management Compared to patients who did not report chronic pain, patients with chronic pain (n= 557) had poorer overall diabetes self-management (β-coefficient, –5.0; CI: –7.8 to –2.2; P= 0.002) and greater difficulty following exercise (adjusted OR, 3.0; CI: 2.1–4.1) and eating plans (adjusted OR, 1.6; CI: 1.2–2.1). Patients who reported severe or very severe pain had significantly greater difficulty taking their diabetes medications (adjusted OR, 2.0; CI: 1.2–3.4) and exercising (adjusted OR, 2.5; CI: 1.3–5.0).
Krein et al.46 Prospective cohort study 1,169 patients with diabetes and a BP ≥140/90 prior to a PCP visit For each visit, PCPs reported the top three issues discussed and whether hypertension medications were intensified or not intensified A discussion of pain during visits significantly lowered the likelihood that blood pressure medications were intensified (35% vs. 46%; P= 0.02).
Concordant disease: cardiovascular disease
ACCORD trial54,55 Randomized controlled trial 10,251 participants with type 2 diabetes aged ≥40 years with CVD or ≥55 years with a high risk of CVD Intensive (A1C <6.0%) versus standard glucose therapy (A1C 7.0–7.9% Nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes After a mean follow-up of 3.5 years, the intensive therapy group had a higher risk of death (HR, 1.22; CI: 1.01– 1.46) compared to the standard therapy group. At five-year follow-up, the intensive therapy group still had a higher risk of death (HR, 1.19; CI: 1.03–1.38) compared to the standard therapy group.
ADVANCE trial56 Randomized controlled trial 11,140 participants with type 2 diabetes ≥55 years with a history of major macrovascular or microvascular disease or at least one risk factor for vascular disease Intensive (A1C <6.5%) versus standard glucose therapy based on local guidelines Composite of macrovascular events and composite of microvascular events After a mean follow-up of 5.6 years, the intensive therapy group had a 10% lower risk of the composite outcome of major macrovascular and microvascular events (HR, 0.90; CI: 0.82–0.98), due mostly to a reduction in nephropathy (HR, 0.79; CI: 0.66–0.93).
VADT57 Randomized controlled trial 1,791 military veterans with a suboptimal response to therapy for Type 2 diabetes Intensive (A1C reduction of 1.5%) versus standard glucose therapy Time to any CVE: death, CVE, and microvascular events Intensive and standard glucose therapy groups did not differ in their time to CVE (HR, 0.88; CI: 0.74–1.05), mortality rates (HR, 1.07; CI: 0.81–1.42), CVE, or microvascular events
VADT substudy57 Substudy of VADT Subsample of 301 participants with type 2 diabetes in the VADT Intensive (A1C reduction of 1.5%) versus standard therapy lowering Baseline coronary artherosclerosis measured by CAC, CVE Among participants randomized to intensive treatment, 11 of 62 individuals with a high CAC score (>100) had events, as compared to 1 of 52 individuals with low CAC scores (≤100). Participants with high CAC scores had a greater risk of CVE compared to those with low CAC scores (multivariable HR, 0.74; CI: 0.46–1.20; P= 0.21 vs. HR, 0.08; CI: 0.008–0.77; P= 0.03).
  • A1C, hemoglobin A1c; ACCORD, Action to Control Cardiovascular Risk in Diabetes Trial; ADVANCE, Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation; CAC, coronary artery calcium; CI, 95% confidence interval; CVE, cardiovascular events; FBG, fasting blood glucose; HR, hazard ratio; HRQL, health-related quality of life; OR, odds ratio; PCP, primary care physician; RR, relative risk; TIBI, Total Illness Burden Inventory; UKPDS, United Kingdom Prospective Diabetes Study; VADT, Veterans Affairs Diabetes Trial.
Details are in the caption following the image

Shared decision-making framework for prioritization of care in adults with diabetes and comorbidity. SDM, shared decision making.

Typology of comorbid chronic conditions

For patients with diabetes, Piette and Kerr published a typology of comorbid chronic conditions that serves as our starting point (Table 2).34 They described three major types of comorbid diseases: clinically dominant comorbid conditions, symptomatic versus asymptomatic conditions, and concordant versus discordant conditions. Clinically dominant conditions are “conditions that are so complex or serious that they eclipse the management of other health problems.” Examples include conditions that significantly shorten life expectancy, are severely symptomatic, or are recently diagnosed. Symptomatic conditions are conditions for which patients experience notable symptoms, as opposed to asymptomatic conditions, which are treated with the sole purpose of preventing future morbidity. Chronic pain is a classic symptomatic condition, whereas a typical asymptomatic condition is hyperlipidemia. Lastly, concordant conditions are conditions that are “parts of the same overall pathophysiologic risk profile,” whereas discordant conditions have a less direct relationship to the condition of interest. In the case of diabetes, concordant conditions include hypertension and cardiovascular disease (CVD), whereas discordant conditions include cancer, chronic obstructive pulmonary disease, and anemia. This typology provides a simplified framework of the complex relationships between diabetes and comorbid conditions.

Table 2. Typology of comorbid chronic conditionsa
Clinically dominant conditions
 • Comorbid chronic conditions that are so complex or so serious that they eclipse the management of other health problems.
 • Examples:
  • End-stage diseases: metastatic cancer, New York Heart Association Class 4 heart failure, advanced dementia, chronic obstructive pulmonary disease with severe hypoxia, severe decompensated liver disease, renal failure
  • Severe acute symptoms: chest pain, profound vomiting, unintentional weight loss (>5% body weight)
  • New serious diagnoses: cancer, pulmonary embolism, cerebrovascular accident
Symptomatic versus asymptomatic chronic conditions
 • Treatment for symptomatic chronic conditions focuses on improving patients’ symptom profiles, functioning and quality of life, and may also delay or prevent poor long-term outcomes.
  • Examples: chronic pain, depression, incontinence, falls/fear of falling, functional disability
 • Treatment of asymptomatic chronic conditions focuses almost exclusively on preventing downstream adverse events and early mortality.
  • Examples: hyperlipidemia, mild hypertension
Concordant versus discordant chronic conditions
 • Concordant conditions represent parts of the same overall pathophysiologic risk profile and are more likely to be the focus of the same disease and self-management plan.
  • Examples: cardiovascular disease, cerebrovascular disease
 • Discordant treatments are not directly related in either their pathogenesis or management.
  • Examples: chronic obstructive pulmonary disease, cancer, anemia
  • a Adapted from Piette and Kerr.34

We use this typology to demonstrate how specific types of comorbid chronic conditions can guide the prioritization of care for diabetes treatments. For the clinically dominant condition, we describe end-stage diseases, since these diseases shorten life expectancy and thus may eclipse the potential benefit of intensive glucose control. We provide the example of chronic pain to show how a symptomatic condition can alter the priorities of diabetes care. Lastly, we describe the example of CVD, a key concordant condition that may increase mortality in patients who are treated with very intensive glucose targets. The clinical evidence for the relationships between diabetes and specific conditions are presented in Table 1.

Clinically dominant condition: end-stage diseases—limited life expectancy

For a patient with diabetes, clinically dominant conditions are those that determine prognosis and dictate everyday experiences. These include end-stage diseases, severely symptomatic conditions, and recently diagnosed diseases. Examples of end-stage diseases include metastatic cancer, New York Heart Association Class 4 heart failure, advanced dementia, chronic obstructive pulmonary disease with severe hypoxia, and severe decompensated liver disease (moderate ascites, bilirubin >3 mg/dL, albumin <2.8 g/dL, or elevated international normalized ratio >2.3). In these patients, end-stage diseases severely limit life expectancy to such an extent that the long-term benefits of optimally managed glycemic control would not be observed. In the United Kingdom Prospective Diabetes Study (UKPDS) intensive A1C treatment did not lower rates of diabetes-related mortality or CVD even after 10 years of follow-up. Intensive A1C control did significantly lower rates of microvascular disease (risk reduction, 25%; P= 0.001); however, the Kaplan–Meier plots only separated significantly after nine years of follow-up.35 Even after 12 years of follow-up, results from most of the surrogate endpoints (ankle reflexes, cardiomegaly, and silent myocardial infarction) and end-stage complications (blindness and peripheral vascular disease) did not differ. Thus, if nearly a decade of intensive glucose control is needed prior to seeing improvements, then limited life expectancy may prevent patients from ever receiving the potential benefits of intensive glucose control.

Current evidence on the effects of life expectancy on glucose targets are based on microsimulation models and observational data. Microsimulation is a computer modeling technique that is used to predict outcomes for individuals.36 Predictions for patient outcomes are calculated from transition probabilities that are generated from results of randomized controlled trials, epidemiologic studies (e.g., case-control and cohort), meta-analyses, and expert opinions. Microsimulation models often combine data from multiple sources to create a set of rules that can answer policy-relevant questions. Microsimulation models can be used to predict outcomes for populations that are not typically enrolled in clinical trials (e.g., the elderly and patients with multiple comorbid diseases). Additionally, some microsimulation models are flexible and allow researchers to alter individual patient characteristics (e.g., age, sex, and comorbidity). Several microsimulation models have been developed for diabetes, including the UKPDS Outcomes Model,37 Centers for Disease Control and Prevention/Research Triangle Institute Diabetes Cost-Effectiveness Model,38 Archimedes,39 and the Michigan Model for Diabetes.40

Huang et al. used a diabetes microsimulation model to evaluate how life expectancy would affect the potential benefits of intensive glucose control (A1C level of 7.0%) compared to moderate control (A1C level of 7.9%).26 This decision analytic study was performed on a hypothetical population of adults aged 60–80 years with type 2 diabetes and no history of diabetes-related complications. The population was based on the diabetes population aged >60 years old in the National Health and Nutrition Examination Surveys (NHANES) from 1999 to 2002. Huang et al. combined transition probabilities from the UKPDS37 with a geriatric mortality prediction tool.41 Transition probabilities from the UKPDS were based on a sample of 3,642 patients with new-onset type 2 diabetes who were randomized patients to intensive glucose control versus conventional therapy. The geriatric mortality prediction tool was developed from data based on community-dwelling U.S. adults older than 50 years who participated in the 1998 wave of the Health and Retirement Study (HRS). This tool accurately predicted four-year mortality in older adults. The mortality prediction tool used easy-to-collect clinical data (age, sex, comorbid conditions, and functional measures) to estimate a mortality index score, ranging from 1 to 26 points with higher values corresponding to greater risk of mortality.

Combining the diabetes model and the geriatric mortality tool, Huang et al. found that the expected benefits of intensive control were inversely related to the level of comorbid illness and functional impairment for all age groups.26 For example, for adults aged 60–64 years with new-onset diabetes, the benefits declined from 106 quality-adjusted days (95% confidence interval (CI): 95–117 days) at baseline good health to 44 days (CI: 38–50 days) with three additional mortality index points and eight days (CI: 5–10 days) with seven additional index points. For patients with greater duration of diabetes, the expected benefits of intensive glucose control were also negatively associated with life expectancy. For adults aged 60–64 years with diabetes for 10–15 years, the expected benefits of intensive glucose control decreased from 116 quality-adjusted days (CI: 103–129 days) to 36 days (CI: 29–43 days) with four additional index points and to eight days (CI: 6–11 days) with eight additional index points.

To verify whether the benefits of intensive A1C control were affected by comorbidity, Greenfield et al. conducted an observational study of 3,074 patients with type 2 diabetes who were characterized into high and low-to-moderate comorbidity groups at baseline and observed for five years.25 The baseline comorbidity level was determined from patient responses on the Total Illness Burden Index (TIBI) questionnaire. The TIBI questionnaire assesses the presence and severity of eight comorbid conditions (atherosclerotic heart disease, lung disease, heart failure, arthritis, genitourinary disease, vision loss, gastrointestinal conditions, and foot disease). High comorbidity was defined by scores ≥12, whereas low-to-moderate comorbidity was defined by scores <12. The TIBI has been validated as a predictor of 3.5-year mortality42 and HRQL,43 and a score of 12 points can discriminate greater and lesser risk for death.42 Greenfield et al. found that patients in the low-to-moderate comorbidity group with baseline A1C levels ≤6.5% had a lower five-year incidence of cardiovascular events (adjusted hazard ratio (HR), 0.60; CI: 0.42–0.85; P= 0.005). However, patients in the high comorbidity group experienced no significant benefit from A1C levels ≤6.5%. Similarly, only the low-to-moderate comorbidity group had fewer cardiovascular events after attaining an A1C level of 7.0% (adjusted HR, 0.61; CI: 0.44–0.83; P= 0.001). Together, these studies strongly suggest that moderate glucose targets may be reasonable in patients with diabetes and limited life expectancy.

Symptomatic condition: chronic pain

Symptomatic conditions cause patients to experience symptoms that negatively impact HRQL. Several symptomatic conditions are more prevalent among patients with diabetes compared to patients without diabetes, including chronic pain, depression, incontinence, falls/fear of falling, and functional disability.8 To describe this typology, we highlight chronic pain as a classic example of a symptomatic condition.

Chronic pain is highly prevalent in patients with diabetes.44 Chronic pain has been reported in over 40% of elderly adults with diabetes13 and up to 60% of the VA population with diabetes.22 In patients with diabetes, chronic pain may be caused by diabetes (e.g., peripheral neuropathy and peripheral vascular disease) or by common diseases (e.g., osteoarthritis and headache).

Chronic pain seriously impacts HRQL in patients with diabetes. It has been associated with similar declines in HRQL as cardiovascular and microvascular disease. In a study of 6,317 adults with diabetes, Laiteerapong et al. found that patients with chronic pain had lower physical HRQL scores (β-coefficient, –5.6; CI: –6.1 to –5.0; P < 0.001) to the same degree as participants who had myocardial infarction (β, –5.3; CI: –7.0 to –3.5; P < 0.001), stroke (β, –4.7; CI: –7.6 to –1.8; P= 0.002), or hypoglycemia (β, –4.7; CI: –8.2 to –1.3; P < 0.001).13 Furthermore, a study by Bair et al. of 11,689 participants with diabetes found that patients with diabetes and chronic pain were more likely to report being depressed (41.3% vs. 16.2%; P < 0.001).45 They also reported a dose-response relationship between chronic pain and depression with patients who report moderate pain being 1.3 times (odds ratio (OR), CI: 1.1–2.0) more likely to report depression and those with extreme pain being 2.0 times (OR, CI: 1.6–2.5) more likely to report depression.

The presence of chronic pain negatively impacts diabetes self-management. In a study by Krein et al., 993 patients with diabetes were asked to report if they had chronic pain and if they had difficulty with diabetes self-management.22 Diabetes self-management scores could range from 0 to 100, with higher scores indicating less difficulty with self-management. Even after adjustment for several factors including depressive symptoms and general health status, patients with chronic pain (n= 557) had poorer overall diabetes self-management (β-coefficient, –5.0; CI: –7.8 to –2.2; P= 0.002) compared to patients who did not report chronic pain. Patients with chronic pain also reported greater difficulty following exercise plans (adjusted OR, 3.0; CI: 2.1–4.1) and eating plans (adjusted OR, 1.6; CI: 1.2–2.1) compared to patients without chronic pain. Furthermore, compared to those with moderate pain, patients with diabetes who reported severe or very severe pain had significantly greater difficulty taking their diabetes medications (adjusted OR, 2.0; CI: 1.2–3.4) and exercising (adjusted OR, 2.5; CI: 1.3–5.0).

Because of the serious consequences of chronic pain to HRQL and diabetes self-management, deprioritizing chronic pain may impede both diabetes-specific and overall care. However, symptomatic conditions, like chronic pain, can be very challenging to treat. For example, Krein et al.46 found that addressing pain significantly lowered the likelihood that blood pressure medications were intensified (35% vs. 46%, P= 0.02) during clinic visits. Their results may not be surprising since addressing any single clinical condition may take attention away from the care of any other clinical condition. This study is a reminder that it is important to consider managing diabetes while addressing symptomatic conditions. The severity of the patient's symptoms, degree of diabetes control, and patients’ preferences should guide decisions to treat diabetes and symptomatic conditions simultaneously or sequentially.

Other examples of symptomatic conditions include depression, incontinence, falling/fear of falling, and functional disability. Compared to patients without diabetes, patients with diabetes have a two times greater odds of depression,47 and symptomatic depression has been associated with a two times greater odds of poor adherence to diabetes medications among a cohort of 4,117 patients with diabetes.19 Incontinence, falls, and functional disability are also especially prevalent in the elderly with diabetes and are associated with lower HRQL.13,48–51 However, further evidence is needed to clarify whether these symptomatic conditions also negatively impact diabetes management.

Concordant condition: CVD

Concordant conditions have similar pathophysiological profiles as the disease of interest. As a result, the treatment of a concordant condition generally improves the status of both the concordant condition and the primary disease. Physicians may appropriately prioritize the treatment of concordant conditions over discordant conditions because a single treatment can improve the status of more than one condition.

Several studies have found that discordant conditions may be undertreated compared to concordant conditions. In a study of over one million elderly adults by Redelmeier et al.,16 patients with diabetes were less likely to receive prescriptions for estrogen-replacement therapy (2.4% vs. 5.9%; P < 0.001; at a time when hormone replacement therapy was commonly prescribed in postmenopausal women). Another study by Turner et al. similarly reported that among 15,459 patients with uncontrolled hypertension, patients with more discordant conditions were less likely to have their blood pressure treatment intensified than patients with no discordant condition (one condition vs. none: adjusted OR, 0.85; CI: 0.80–0.90; ≥seven conditions vs. none: adjusted OR, 0.59; CI: 0.51–0.69).17

In patients with diabetes and CVD, both diseases have very similar treatment goals.5,52,53 Increased physical activity and low cholesterol diets benefit both diabetes and CVD. Blood pressure (<130/80 mmHg) and cholesterol targets (low-density lipoprotein <100 mg/dL) are identical for patients with diabetes and patients with CVD. Aspirin therapy is very frequently recommended for patients with diabetes and almost universally recommended for patients with CVD. Thus, for the majority of patients, CVD management enhances the management of diabetes.

Although the benefits of cardiovascular management in diabetes patients with comorbid CVD are well accepted, the effects of glycemic management on CVD are far more uncertain. After the UKPDS trial showed improved outcomes in patients with new-onset diabetes from intensive glucose control,35 several trials attempted to reproduce its results in patients with a history of diabetes, many of whom had CVD. Instead of confirming results, trial results were mixed and found either an increased mortality risk, a decreased risk of microvascular disease, or no benefit with intensive glucose therapy.

In the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, 10,251 study participants with type 2 diabetes (mean duration, 10 years) were randomly assigned to intensive glucose therapy (A1C <6.0%) or standard therapy (A1C 7.0–7.9%).54 In addition to having diabetes, participants were either ≥40 years old with CVD or ≥55 years old with a high risk of CVD. On average, the intensive therapy group achieved an A1C of 6.4% and the standard therapy group achieved an A1C of 7.5%. Unexpectedly, the trial ended early after a mean follow-up of 3.5 years because the intensive therapy group had a higher mortality rate than the standard therapy group (HR 1.22; 95% CI: 1.01–1.46). Recent five-year follow-up results from the ACCORD trial reconfirmed a greater mortality rate in the intensive glucose therapy group (HR, 1.19; CI: 1.03–1.38).55

Results from the Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation (ADVANCE) trial differed from those of the ACCORD trial.56 The ADVANCE trial included 11,140 participants with type 2 diabetes aged ≥55 years and randomized them to intensive glucose therapy (A1C <6.5%) or standard glucose therapy. Participants were required to have a history of major macrovascular or microvascular disease or at least one risk factor for vascular disease. Their median duration of diabetes was about eight years. The intensive therapy and standard therapy groups in ADVANCE achieved similar A1C levels as in the ACCORD trial (6.5% and 7.3%, respectively) at five years of follow-up. However, unlike the ACCORD trial, the intensive glucose therapy group had a 10% relative reduction in the combined outcome of major macrovascular and microvascular events (HR, 0.90; CI: 0.82–0.98), mostly due to a 21% relative reduction in nephropathy (HR, 0.79; CI: 0.66–0.93), and there were no significant effects on major macrovascular events or death.

The third study, the Veterans Affairs Diabetes Trial (VADT), randomized 1,791 veterans to intensive glucose therapy (an absolute reduction of 1.5% in A1C) versus standard therapy.57 Patients had an average diabetes duration of 11.5 years, and 40% had a history of CVD. The A1C levels in both groups were higher in VADT compared to the ACCORD and ADVANCE trials. The intensive therapy group achieved a mean A1C of 6.9%, and the standard therapy group achieved a mean of 8.4%. However, unlike both the ACCORD and ADVANCE trials, there were no significant differences in major cardiovascular events, death, or microvascular events between the two groups after a median follow-up of 5.6 years. Additionally, a VADT substudy of 301 patients found that among patients in the intensive glucose therapy group, patients with higher baseline levels of coronary atherosclerosis experienced significantly more cardiovascular events than patients with lower levels of coronary atherosclerosis (multivariable HR, 0.08; CI: 0.008–0.77; P= 0.03 vs. HR, 0.74; CI: 0.46–1.20; P= 0.21).58

Because of these conflicting results, the optimal glucose target in patients with diabetes and CVD is still debated. Diabetes experts have recently integrated diabetes trial results and have suggested that the A1C target should be 7.0% or greater in patients with CVD or a high risk of CVD.59

Although the presence of concordant conditions may appear to simplify diabetes treatment decisions, concordant conditions, like CVD, may not have simple relationships with diabetes. Achieving the treatment goals of CVD improves diabetes management, but intensive glucose therapy may increase mortality in patients with CVD. Thus, even with concordant conditions like CVD, treatment decisions for diabetes must be carefully considered.

Integrating comorbidities into diabetes care

The example clinical conditions (end-stage disease, chronic pain, and CVD) and the typology (clinically dominant, concordant vs. discordant, and symptomatic vs. asymptomatic) greatly simplify the complex interactions of comorbid diseases in patients with diabetes. For example, there can be significant overlap in the typology by conditions. End-stage diseases are often quite symptomatic (e.g., New York Heart Association Class IV heart failure). CVD is concordant and usually asymptomatic. Chronic pain is symptomatic but can be either concordant if the pain is due to diabetes complications, or discordant if the pain is due to other conditions.

Another reason this typology is frequently an oversimplification is because many adults with diabetes suffer from more than just one additional disease. Weiss et al. studied the prevalence of five major chronic diseases (arthritis, coronary heart disease, chronic lower respiratory tract disease, cerebrovascular accident, and diabetes) in older men and women from the NHANES 1999–2004 (Table 3).60 Diabetes occurred with one comorbid disease in 39% of men and 37% of women, two comorbid conditions in 18% of men and 27% of women, and three comorbid conditions in 3% of men and 9% of women. Regarding specific combinations, in older men diabetes occurred most frequently alone (27%; 703,120 out of 2,592,800 older men with diabetes); in older women, diabetes occurred most frequently with arthritis (31%; 1,049,700 of 3,426,500 older women with diabetes). The second most common combination for diabetes was with arthritis in men (21%; 545,260) and alone in women (17%; 585,720). Among these five major chronic diseases, there were eight different combinations of diabetes and comorbidity.

Table 3. Estimated prevalence of disease patterns in elderly adults with diabetes (n, %)a
Men, aged 65 or older (Weighted N= 2,592,800) Women, aged 65 or older (Weighted N= 3,426,500)
Diabetes only 703,120 (27) 585,280 (17)
One comorbid condition
 Arthritis 545,260 (21) 1,049,700 (31)
 CHD 285,080 (11) 135,540 (4)
 CLRT 110,890 (4) 97,480 (3)
 CVA 82,780 (3)
Two comorbid conditions
 Arthritis and CHD 313,310 (12) 454,630 (13)
 Arthritis and CLRT 83,450 (3) 369,870 (11)
 Arthritis and CVA 79,260 (3) 98,590 (3)
Three comorbid conditions
 Arthritis, CHD, and CVA 79,420 (3) 123,310 (4)
 Arthritis, CHD, and CLRT 192,380 (6)
  • CHD, coronary heart disease; CLRT, chronic lower respiratory tract disease; CVA, cerebrovascular disease.
  • a Prevalence estimates based on Weiss et al.64 Only the chronic diseases of arthritis, CHD, CLRT, and CVA were included.

The complex patterns of diabetes and comorbid chronic conditions likely preclude the development of a single rigid guideline. In situations where few guidelines exist and there is significant clinical uncertainty, shared decision making between patients and clinicians is a useful, and possibly necessary, tool for making individualized treatment decisions.33 Components of shared decision making for clinicians and patients are noted in Figure 1. Key elements include discussions between clinicians and patients about the goals of care, risks of diabetes and comorbid conditions, options for treatments, and patient preferences for treatment. Additionally, patients should be allowed to participate in shared decision making at their desired level.

This strategy has been endorsed by the United States Preventive Services Task Force (USPSTF),61 an independent panel that develops evidence-based recommendations on clinical preventive services. The USPSTF suggests a graded approach to shared decision making. For screening tests with strong evidence, lengthy and frequent discussion is likely unnecessary, but for screening tests with less certain evidence, engaging patients in shared decision making is necessary to assess the net benefit of the screening test. Thus, shared decision making is not necessary for hypertension screening, a recommendation with substantial net benefit. However, shared decision making should be used for mammography decisions in women <50 years because individual women may have a strong desire to be screened,62 even though there may be only a small net benefit.63

Several critical reasons exist for using shared decision-making principles when caring for adults with diabetes and comorbidity. First, patients can vary greatly in their overall health priorities. A study of 81 elderly adults by Fried et al. found that nearly half ranked maintaining independence as their most important outcome;64 however, 27% ranked staying alive, 21% ranked reducing/eliminating pain, and 10% ranked reducing/eliminating other symptoms as their most important outcome. Second, patients can be inconsistent in their overall health priorities. Fried et al. asked 189 elderly adults with advanced disease, if when faced with a fatal disease exacerbation, they would be willing to either undergo or forego high-burden therapy to possibly avoid death.65 They found that 35% had an inconsistent preference trajectory (e.g., becoming more and then less willing over time and vice versa). Third, providers and patients often disagree on the relative importance of different comorbid conditions. In a study by Zulman et al., 28% of providers did not agree with patients in the ranking of their most important health condition.66 Providers were more likely to rank hypertension as the most important condition, and patients were more likely to prioritize symptomatic conditions like pain, depression, and breathing problems.

Fourth, patients with diabetes have been shown to vary greatly in their preferences for treatment (Table 4). In a study of 475 elderly adults with diabetes, Chin et al. found that patients varied greatly in their perceptions of diabetes complications and treatments.67 Using utility values (range, 0 [death] to 1 [perfect health]), they found a large range in the utility values for blindness (0.39; standard deviation [SD]: 0.32), lower leg amputation (0.45; SD: 0.34), conventional glucose treatment (0.76; SD: 0.27), and intensive insulin treatment (0.64; SD: 0.32). In a separate study of elderly adults, Brown et al. evaluated whether patient perceptions of diabetes treatments differed according to patient vulnerability, as assessed by the Vulnerable Elders Scale.68 In general, they found that vulnerable patients reported lower utilities than nonvulnerable patients for most treatments, including intensive glucose control (mean, 0.61 vs. 0.72, P < 0.01). However, within-group variation was large for both groups (SD >0.25). Similarly, Huang et al. studied a multi-ethnic sample of 701 elderly and nonelderly adults with diabetes and found large variation in patients' preferences for diabetes complications and treatments (e.g., intensive glucose control, mean 0.67; SD 0.34).69 Last, there is limited evidence and, as a result, significant clinical uncertainty as to the net clinical benefit of diabetes treatments in most patients with diabetes and comorbid chronic conditions.

Table 4. Utility values for diabetes complications and treatmentsa
Study population Chin et al. 67 473 older adults with diabetes Brown et al. 68 332 vulnerable and nonvulnerable older adults with diabetesb Huang et al. 69 701 multi-ethnic adults with diabetes
Vulnerable Nonvulnerable
Diabetes complications
 Neuropathy 0.66 ± 0.34
 Amputation 0.45 ± 0.34 0.55 ± 0.36
 Retinopathy 0.53 ± 0.36
 Blindness 0.39 ± 0.32 0.38 ± 0.35
 Nephropathy 0.64 ± 0.35
 Kidney failure 0.36 ± 0.31 0.35 ± 0.33
 Mild stroke 0.70 ± 0.31
 Major stroke 0.31 ± 0.31
 Angina 0.64 ± 0.31
Diabetes treatments
 Conventional glucose treatment 0.76 ± 0.27 0.71 ± 0.33 0.79 ± 0.28 0.76 ± 0.31
 Intensive glucose therapy 0.61 ± 0.34 0.72 ± 0.32 0.67 ± 0.34
  Intensive pill therapy 0.77 ± 0.27
  Intensive insulin therapy 0.64 ± 0.32
 Conventional blood pressure control 0.73 ± 0.32 0.80 ± 0.28 0.77 ± 0.30
 Intensive blood pressure control 0.69 ± 0.33 0.76 ± 0.29 0.73 ± 0.32
 Aspirin 0.79 ± 0.28 0.83 ± 0.28 0.80 ± 0.29
 Cholesterol-lowering pill 0.72 ± 0.30 0.83 ± 0.26 0.78 ± 0.29
 Diet 0.89 ± 0.23 0.90 ± 0.21 0.88 ± 0.24
 Exercise 0.85 ± 0.26 0.91 ± 0.19 0.89 ± 0.23
 Polypharmacy 0.58 ± 0.35 0.66 ± 0.32
 Comprehensive diabetes care 0.64 ± 0.34
  • a Utility values can range between 0 (death) and 1 (perfect health).
  • b Study population was categorized by vulnerability based on their score on the Vulnerable Elders Scale-13 (VES-13). Patients with scores ≥3 points were defined as vulnerable.

Because patients vary in their health priorities and diabetes treatment preferences, and because clinicians cannot predict their preferences, shared decision making is especially critical to guiding treatment decisions in patients with diabetes and comorbidity. Using the tenets of shared decision making and the typology of comorbidity presented earlier, we present an algorithm for prioritizing care in adults with diabetes and comorbid chronic conditions.

Algorithm for prioritization of care in patients with diabetes and comorbidity (Fig. 2)

Details are in the caption following the image

Algorithm for the prioritization of care in adults with diabetes and comorbidity. A1C, hemoglobin A1c; BP, blood pressure; SDM, shared decision making.

  • 1

    Elicit patient's goals of care and preferences for treatment to develop a care plan using the tenets of shared decision making.

    Patients should be asked what their goals of care are and to what degree they want to participate in shared decision making. Patients can vary greatly in their desire to actively engage in shared decision making. In a review of eight studies with different patient populations in different settings, interest in shared decision making ranged from 19% to 68%.70 There can be a spectrum of participation in medical decision making, ranging from very little to full ownership. In the paternalistic decision-making model, the clinician implements what he/she believes is best for the patient after receiving their assent (not consent). On the other end of the spectrum, patients who have complete autonomy in the decision-making process use clinicians for their clinical acumen, like technicians, and have complete control over their medical care.71 Shared decision making contrasts with these extreme views because it stresses the partnership between the patients’ values and preferences and the clinicians’ knowledge of clinical evidence and judgment. Barriers such as the power differential between clinician and patient may impede shared decision making. Racial and ethnic minorities with diabetes may face additional barriers such as historical mistrust despite wanting to engage in shared decision making,72, 73 and thus clinicians need to be particularly sensitive in discussing patient preferences.74, 75 If patients are interested in participating in shared decision making, then clinicians should elicit the patients’ overall health goals and specific treatment preferences. This conversation can be framed by a discussion of the overall health priorities introduced by Fried et al.64 (e.g., maintaining independence, staying alive, reducing/eliminating pain, and reducing/eliminating other symptoms) and by discussion of diabetes treatment goals and preferences (e.g., preventing diabetes complications, decreasing polypharmacy, avoiding insulin therapy, avoiding hypoglycemia, and improving quality of life).

  • 2

    Assess whether the patient with diabetes has any clinically dominant conditions.

    Clinically dominant conditions can include end-stage diseases that limit life expectancy, severe acute symptoms, and new serious diagnoses. Patients with diabetes and end-stage diseases may not benefit from intensive glucose control. In this population, less stringent glucose targets are reasonable. Blood glucose levels at least <200 mg/dL are reasonable to avoid symptoms of diabetes.76 In patients with diabetes, the benefits from tight blood pressure control (144/82 mmHg) compared to standard therapy (154/87 mmHg) were discernible in Kaplan–Meier plots after one year of control.77 Similarly, in a large study by the Heart Protection Collaborative Group, participants with diabetes randomized to cholesterol-lowering medication (simvastatin) had a 22% lower risk of vascular events (P < 0.001) compared to those who received placebo therapy, and this difference became apparent within one to two years of follow-up.78 Thus, patients with diabetes and end-stage diseases may still benefit from blood pressure and lipid control, depending on their prognosis. Decisions about blood pressure and cholesterol treatment should be guided based on each patient's individual context, including their current levels, tolerance of medications, interest in therapy, concern about adverse effects, and overall prognosis. For patients with severe acute symptoms or new serious diagnoses, intensification of glucose, blood pressure, and lipid control may need to be postponed until patients are clinically stable.

  • 3

    Screen the patient with diabetes for untreated or inadequately treated symptomatic conditions and then treat.

    All patients with diabetes should be screened for untreated or inadequately treated symptomatic conditions, that is, chronic pain, depression, incontinence, falls/fear of falling, and functional disability. Patients with diabetes and these symptomatic conditions generally have poor HRQL.8, 13 Additionally, patients with chronic pain and depression often have difficulty managing their diabetes.22 Once diagnosed, symptomatic conditions should be treated. Depending on the severity of symptoms and patients’ preferences, diabetes treatments can also be managed alongside symptomatic conditions.

  • 4

    Determine whether the patient has CVD or a high risk for CVD.

    Though most CVD treatments are concordant with the care of diabetes, much controversy exists regarding the appropriate intensity of glucose control in patients with CVD, or a high risk of CVD, as previously described. For this reason, A1C targets at or higher than 7.0% should be considered in these patients.59

  • 5

    Decide which of the diabetes treatment goals should be given priority.

    The comparative benefit of three diabetes treatments (blood glucose, blood pressure, and lipid control) have been reviewed in studies by Vijan and Hayward.79,80 In one study, using results from the UKPDS, Vijan and Hayward found that the effectiveness of tight blood pressure control (achieved blood pressure, 144/82 [tight] vs. 154/87 [control]) was superior to intensive glucose control (A1C, 7.0%[intensive] vs. 7.9%[usual]) for all major clinical endpoints.79 To prevent one diabetes endpoint, 8.9 persons needed to be treated with tight blood pressure control compared to 31.2 persons with intensive glucose control. In a separate review on the effectiveness of pharmacologic lipid-lowering in patients with type 2 diabetes, Vijan and Hayward found that the number needed to treat to prevent one cardiovascular outcome was 34.5 for primary prevention and 13.5 for secondary prevention.80 Thus, on a population level, overall fewer people would need to be treated with tight blood pressure control than cholesterol-lowering therapy, or intensive glucose control, to benefit one person. For individual patients, however, the optimal strategy is still unknown for prioritizing these three treatment decisions. Will patients fare better if each of the three outcomes is addressed one at a time, or the outcome that is most out of range is addressed first, or if all three outcomes are addressed simultaneously? Because of this clinical uncertainty, the tenets of shared decision making should guide discussions about how to prioritize diabetes treatments along with the care of comorbidities. Also considerations should be made for the patients’ cognitive status, financial situation, and social support. Clinicians and patients should discuss the treatment options to develop a care plan.

  • 6

    Prioritization is an iterative process and should be revisited, especially when the patient's clinical status changes significantly.

    Patients’ priorities may change and their priorities may not be predicated on past preferences. Additionally, clinical knowledge and patients’ comorbidity, cognition, and resources may change. Clinicians should frequently revisit patients’ priorities, potential constraints, and past care decisions to ensure patient-centered care.

Discussion

Despite the fact that about half of adults with diabetes have a comorbid chronic condition, diabetes care guidelines have concentrated their recommendations on patients who only have diabetes. This review paper summarizes the current understanding of the complex relationships between diabetes and comorbid diseases. Several comorbid conditions, such as end-stage diseases, chronic pain, and CVD, can help guide diabetes treatment decisions. For patients with diabetes and other comorbid conditions, we provide a general algorithm for prioritizing care, which integrates a typology of comorbid conditions with the framework of shared decision making.

Although our overall understanding of diabetes and comorbidity is still rudimentary, the relationships between diabetes and selected comorbid conditions have been well-studied. Further research on the relationship between diabetes and symptomatic conditions could enhance the care of patients with diabetes and comorbidity. Potential areas of research include determining whether symptomatic conditions should be managed prior to diabetes, or if both conditions can be addressed simultaneously. Additional research could clarify the relationships between diabetes and discordant conditions. For example, recent studies have suggested that patients with diabetes are at an increased risk for cancer and lung disease.81–84 However, it is still unknown how the presence of these discordant conditions should factor into decisions about diabetes management.

Because of the complexity of caring for patients with diabetes and comorbidity, it may not be possible to create a diabetes care guideline that applies to all patients with diabetes. However, shared decision making allows for treatment decisions to be patient-centered, a key principle of the National Quality Strategy for health care.85 Practical strategies for using shared decision making in patients with diabetes and comorbidity would benefit primary care clinicians who are already hampered by the number of guideline recommendations.86 Some possible strategies include using computers to elicit goals and preferences, as well as the development of multicondition decision aids that can integrate treatment decisions for diabetes and several comorbid diseases.

Although the study of diabetes and comorbidities is still in its nascency, clinicians are already faced with the struggle of caring for these complex patients. Clinicians aim to improve their patients’ overall HRQL and functioning; thus, explicitly addressing the role of comorbidities in diabetes care is essential. Future clinical research in diabetes should concentrate on the practical question of how to best care for patients with diabetes and common comorbid conditions.

Acknowledgments

N.L. is supported by Grant Number F32DK089973 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). E.S.H. is supported by Grant Number R01-DK-081796 from the NIDDK. M.H.C. is supported by a NIDDK Midcareer Investigator Award in Patient-Oriented Research (K24 DK071933). This project was also supported by the NIDDK Diabetes Research and Training Center (P60 DK20595) and NIDDK Chicago Center for Diabetes Translation Research (P30 DK092949). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIDDK or the National Institutes of Health.

    Conflicts of interest

    The authors declare no conflicts of interest.