“They are interrelated, one feeds off the other”: A taxonomy of perceived disease interactions derived from patients with multiple chronic conditions

Abstract

Objective:

To understand patients’ experiences with condition interactions and develop a taxonomy to inform care for patients with multiple chronic conditions.

Methods:

We conducted qualitative and quantitative analysis of free-text data from patient surveys in which respondents were asked to indicate their most bothersome chronic condition and describe how their other conditions affect their self-care for that condition. Using standard content analysis, we developed a taxonomy comprising how patients perceive interactions among their conditions, and examined cross-cutting themes that reflect qualities of these interactions.

Results:

Among 383 eligible survey respondents, the mean (SD) number of chronic conditions was 4 (2); common conditions included hypertension (60%), chronic pain (49%), arthritis (41%), depression (32%), diabetes (29%), and post-traumatic stress disorder (26%). Patients’ perceived condition interactions took four broad forms: 1) unidirectional interactions among conditions and/or treatments, 2) cyclical or multidimensional interactions, 3) uncertain or indistinct interactions, and 4) no perceived interaction. Cross-cutting themes included beliefs about causal relationships between conditions, identification of interactions as negative vs. positive, and interactions between physical and mental health.

Conclusion:

This study presents a novel taxonomy of condition interactions from the patient perspective.

Practice Implications:

Understanding perceived condition interactions may support patient self-management and shared decision-making efforts.

1. Introduction

Patients with multiple chronic conditions (MCCs) frequently face complex self-management demands and treatment regimens. Each additional condition increases a patient’s medications, recommended lifestyle changes, home monitoring requirements, and physician visits, [1-3] the volume of which can become overwhelming [4]. Greater multimorbidity is also associated with higher rates of hospitalization, increased mortality risk, and greater out-of-pocket expenses [1,5], driving intense demand for effective clinical interventions and self-management support for these patients.

There are a number of ways in which MCCs create unique challenges for patients. Studies have demonstrated that for patients with diabetes, hypertension, and heart failure, the presence of comorbidities is associated with self-management challenges and deficiencies. [6-8] When a patient has multiple diagnoses, clinical practice guidelines for each condition may lead to unwieldy self-care regimens where patients are tasked with remembering medications and self-care activities throughout the day. [9] Certain comorbidities such as depression and chronic pain may pose particular challenges for patients with physically or emotionally taxing self-care needs [10,11].

While early multimorbidity investigations focused on the number of co-occuring conditions, several studies have now examined challenges specific to comorbidity interrelatedness (i.e., the ways in which condition interactions influence self-management and treatment). [12] Qualitative studies have described patients’ experiences with medication interactions [13] and the stress generated by needing to prioritize, monitor, and distinguish between symptoms from different diseases [14-17]. Some patients also experience anxiety due to the perception that adhering to self-care behaviors or treatments for one illness may have negative consequences for a comorbid condition [8,14,18]. Despite these early qualitative findings, the majority of studies focusing on condition interactions have examined a population with a specific index condition such as heart failure or hypertension [7,13,17-19] and there have been few systematic undertakings to understand the multidimensionality of patient experiences with condition interactions.

We analyzed qualitative and quantitative survey data from a previous study [4] to codify the ways in which patients experience interactions among their multiple chronic conditions and patient characteristics associated with perceived interactions. The survey included free-text data from patients describing the interactions between their conditions. While the objective of the larger survey was to examine self-management needs of patients with multiple chronic conditions, the objective of this study is to develop a taxonomy reflecting types of condition interactions, to identify themes about the nature of these interactions and their effects on self-management, and to perform exploratory analyses to understand characteristics associated with perceived interactions.

2. Methods

This paper presents secondary qualitative and quantitative analyses of responses to an open-ended question on a survey of patients with multiple chronic conditions, described previously [4]. The original study recruited individuals who receive care at an academic medical center or at a Veterans Affairs (VA) healthcare facility in Northern California identified in person or through clinic registries. We approached patients in clinic, by mail, and by e-mail with the goal of identifying individuals with a wide range of chronic conditions who could speak to the diverse experiences of patients with MCCs. Once patients responded to the survey, we limited analyses to respondents who indicated that they have 2 or more chronic conditions. The study was approved by the Stanford University Institutional Review Board.

Survey respondents were asked to indicate on the survey their current health conditions from a list of common diseases and to hand-write any additional conditions of theirs that were not on the list. Analyses reported here focused on a pair of survey questions: the first question was used for quantitative analyses and the second question was used for qualitative analyses. After completing the survey indicating their health conditions, respondents were asked to circle the health problem that bothers them the most due to symptoms, discomfort, or challenging self-management requirements. Then they were asked to describe potential interactions among their conditions by responding to the open-ended question: “How do your other health problems affect your ability to take care of the health problem that you circled above?” Survey respondents were also asked about sociodemographics, including age, sex, race/ethnicity, education, annual household income, and current employment status.

A total of 1800 individuals were invited and 550 individuals responded to the survey (31% response rate). For the analyses presented here, respondents were excluded if they did not report 2 or more conditions (n = 60), if they were missing information (n = 94), or had inappropriate information (n = 13) in the open-ended question of interest. A total of 383 (69%) survey responses met all inclusion criteria. There were no significant differences in demographics or chronic conditions among respondents whose responses were and were not included in analyses.

2.1. Data analysis

We used STATA 12.0 (StataCorp LP, College Station, TX, USA) to analyze quantitative survey data, and Atlas.ti 7.0 (Atlas.ti Scientific Software Development GmbH, Berlin, Germany) to analyze free-text data.

Chronic condition prevalence was determined based on patient self-report in the survey. Conditions written in under “Other” were grouped together to create meaningful categories, which were verified by a physician team member for accuracy. When respondents wrote in a health issue that seemed to overlap with a listed condition, we categorized the condition based on the patient’s primary definition/description (e.g., if a person selected “chronic pain” from the condition list and specified in free text that the pain was secondary to arthritis, we categorized the condition as “chronic pain” rather than “arthritis”).

Data analysis was guided by established qualitative research principles [20,21] and overseen by a qualitative research advisor (EL). Four team members, including two physicians (DZ, JS) and two research staff (CS, KB) reviewed survey responses and used content analysis methods to generate a taxonomy of condition interactions as experienced by patients [22,23]. The taxonomy reflects mutually exclusive categories that capture all of the types of interactions described by participants. After coding all descriptions of condition interactions, coders identified cross-cutting themes that reflected qualities of interactions across multiple categories. The interaction categories and themes were iteratively refined by the coding team, and then each coder applied an interaction category and one or more themes to each survey response. Discrepancies in interaction categories and themes were resolved by consensus with the entire coding team. Once consensus was achieved for all survey responses, representative quotes were identified for each interaction category and theme, and approved by the entire research team. In post-hoc analyses, we used quantitative survey data to examine bivariate associations between patient characteristics and a perceived condition interaction (Appendix 3).

To validate study findings, we distributed preliminary results that described the categories in the proposed taxonomy to 44 participants who indicated in the original study [4] that they were open to being re-contacted; 23 completed an online survey (response rate 52%) and provided feedback on the taxonomy and themes (Appendix 2).

3. Results

Characteristics of survey respondents are described in Table 1. The mean (standard deviation) age of the study population was 62.5 (12), over 83% were male, and 16% were non-white. More than half (55%) had an income under $50,000/year. The majority of patients (75%) had between 2 and 5 chronic conditions, 26% had between 6 and 9 conditions, and 3% had 10 or more. Table 2 lists the conditions the study population endorsed. Conditions that were reported by more than 10% of survey respondents include: hypertension (60%), chronic pain (49%), arthritis (41%), depression (32%), diabetes (29%), post-traumatic stress disorder (PTSD) (26%), lung/breathing problem (21%), prostate problem (18%), headaches/migraines (17%), and cancer (12%).

Table 1.

Demographics of Survey Respondents (N = 383).

	n	%
Age, mean (SD) (n = 377)	62.5(12)
Gender (n = 377)
Male	312	83
Race (n = 379)
White	317	85
Asian	24	6
African American	20	5
Native Hawaiian or other Pacific Islander	4	1
American Indian or Alaska Native	14	4
Ethnicity (n = 378)
Hispanic/Latino	31	8
Education (n = 378)	378
High school degree or less	39	12
Some college	149	39
College degree or more	190	51
Annual Household Income (n = 363)
<$50,000	199	55
$50,001-$75,000	71	20
>$75,001	93	26
Employment
Full time	69	18
Part time	39	10
Retired	198	52
Unemployed	44	12
Disabled	101	27
Student	8	2
Number of Conditions
2 to 3	167	44
4–5	119	31
6–7	65	17
8–9	19	5
10 or more	13	3

Table 2.

Self-Reported Chronic Condition Prevalence and Frequency.

	n	%
High Blood Pressure	230	60
Chronic pain	189	49
Arthritis/rheumatism	158	41
Depression	123	32
Diabetes	112	29
Post-Traumatic Stress Disorder	101	26
Lung/breathing problem	79	21
Prostate Problems (non-cancerous, e.g., BPH)	68	18
Headaches/Migraines	66	17
Cancer	45	12
Heart Failure	36	9
Thyroid Problems/Graves	35	9
Lower leg edema/venous insufficiency	33	9
Musculoskeletal and Neuropathy*	34	9
Kidney Problem	32	8
Chronic Fatigue Syndrome	21	5
Osteoporosis	20	5
Cardiovascular Disease and Arrhythmia*	18	5
Stroke History	20	5
Spinal Cord Injury	20	5
Other Mental Health	24	6
Gastrointestional*	15	4
Insomnia and Sleep Disorders*	15	4
Vision and hearing*	15	4

Footnotes: Respondents could check off more than one condition. Not shown are conditions selected by fewer than 15 people: Dementia/Alzheimer’s, Fibromyalgia, Hepatitis C, History of alcohol or drug abuse, HIV/AIDS, Manic Depression, Multiple sclerosis, Parkinson’s disease, Schizophrenia, Seizures/epilepsy, and PVD/Vascular disease.

^*denotes a category created from “Other”/write-in results; a total of n = 49 respondents selected “Other” conditions that were not otherwise categorized.

3.1. Categorization of perceived interactions

In codifying the ways in which patients experience condition interactions, we identified four broad categories: 1) unidirectional interactions among conditions and/or condition treatments (including lifestyle changes and other resources that patients draw on to manage their health and well-being), 2) cyclical or multidimensional interactions among multiple conditions, 3) uncertain or indistinct interactions, and 4) no interaction. Below we describe each of these scenarios with representative quotes from patients, with indications for gender and number of conditions in parantheses. The full taxonomy is presented in Fig. 1.

Fig. 1.

Taxonomy of Condition Interactions Derived from Patients with Multiple Chronic Conditions.

3.1.1. Unidirectional condition and treatment interactions

Many patients described interactions among their conditions and/or treatments; in fact, all possible combinations were present in patient responses, including condition-condition, treatment-treatment, and condition and treatment interactions in either direction (i.e., condition affecting treatment, or treatment affecting condition). Examples of condition-condition interactions included statements like, “emphysema leads to high blood pressure,” (P250, male, 53y, 4 conditions) “the constant worry about prostate cancer recurrence aggravates my anxiety and depression,” (P200, male, 67y, 5 conditions) and “chronic pain causes loss of sleep and restless leg syndrome.” (P78, male, 68y, 5 conditions) Treatment-treatment interactions were less common, but included statements like “cancer treatments have caused fatigue and interrupted [my] spine care,”P17, (male, 64y, 10 conditions) and “diabetes makes it easier to take care of my manic depression by making me take all my medications for all my conditions once in the morning and once at bedtime.” (P309, male, 46y, 4 conditions)

Descriptions of condition-treatment interactions were also common. A number of individuals described how treatment for one condition affected other conditions, for example, “The epidurals I had [for chronic pain] caused extremely high blood sugar levels.” (P61, male, 66y, 3 conditions) This type of interaction often extended to non-medication therapies, for example when a patient described that exercise was being used to treat depression but caused “pain in joints, back, nerves, muscles.” (P374, male, 75y, 4 conditions) In the other direction, respondents described how conditions influenced their treatment for other issues. This relationship was sometimes direct, for example when osteoarthritis made “getting to pills hard.” (P22, male, 67y, 4 conditions) More commonly, certain conditions impeded activities that were important for managing another disease, for example, “My back pain keeps me from getting enough exercise which makes it difficult to manage my weight which ultimately is bad for my already damaged heart.” (P27, male, 74y, 4 conditions)

3.1.2. Cyclic/multidimensional interactions

Several patients reported cyclical interactions among multiple conditions, reporting, for example, that their conditions interact “in every way.” (P17, female, 60y, 4 conditions) One person stated, “I believe they are interrelated, one feeds off the other;” (P22, male, 70y, 4 conditions) another reported a cyclical interaction: “Lack of exercise leads to testosterone problems, and [decreases] motivation to exercise.” (P250, male, 53y, 4 conditions)

3.1.3. Uncertain or indistinct interaction

Some patients reported that they were uncertain whether their conditions interact. In some cases this was because of ambiguity regarding which condition was causing their symptoms, as one person wrote, “[I] sometimes cannot tell which one is the problem causing the pain.” (P281, male, 54y, 2 conditions) Another person said, “[it’s] hard to distinguish between being bipolar and the PTSD symptoms.” (P268, male, 69y, 3 conditions) Another person stated, “[It’s] not yet clear how and if back and leg pain are related to Multiple Sclerosis.” (P191, male, 71y, 3 conditions) There were also individuals who reported that their conditions feel indistinct from one another, for example, because they had two conditions that shared similar symptoms.

3.1.4. No interaction

Importantly, some patients indicated that they did not perceive that their conditions interact; as one person wrote, “Health issues do not feel related.” (P349, male, 65y, 5 conditions) A person with a history of anxiety wrote, “My physical problems, which are minor, don’t affect my moods.” (P224, male, 62y, 3 conditions) Another person reported that his kidney problem “doesn’t seem to be a problem in taking care of other health problems other than taking a lot of meds.” (P366, male, 75y, 2 conditions) These statements suggest that for these individuals, health issues are perceived as distinct, without appreciably affecting one another.

3.2. Cross-cutting themes that reflect patient experiences with condition interactions

Beyond the 4 categories describing forms of interaction, several themes emerged that illustrate common experiences that patients describe when considering how their conditions interact. These themes include varying beliefs about causal relationships between conditions, identification of interactions as negative vs. positive, and commonly experienced interactions between physical and mental health.

3.2.1. Condition causality

A recurrent theme related to patients’ perceptions of condition interactions was causality with regards to MCCs. Some felt that one condition caused another, as when one person said, “Some of my conditions cause others to exist, e.g., chronic pain helps cause depression,” (P21, male, 60y, 10 conditions) or another reported that his cancer and its required medications “causes the other health problems.” (P253, male, 78y, 3 conditions) Others noted that multiple conditions might be connected in their underlying pathophysiology, as a patient with arthritis said: “Asthma is also a disease of inflammation, so they both most likely have the same root cause.” (P33, female, 42y, 3 conditions) Finally disease interactions could cause and suppress symptoms, for example a patient reported that drug and alcohol abuse “masks and inflames other conditions.” (P409, female, 57y, 12 conditions)

3.2.2. Positive and negative interactions

While most respondents focused on negative interactions, patients described both negative and positive relationships among their conditions. Representative negative interactions include: “[It’s] difficult to concentrate on [my] heart problem with so much pain,” (P204, male, 65y, 5 conditions) and “[I] have to make sure meds for diabetes and high blood pressure don’t interfere [with each other].” (P196, male, 64y, 3 conditions) But respondents also reported there were a number of ways in which conditions could have synergistic effects that were positive, for example, “What I do for diabetes has [a] beneficial effect on my other conditions,” (P429, male, 66y, 3 conditions) and “If I keep exercising daily the other problems subside.” (P421, male, 78y, 3 conditions)

3.2.3. Physical and mental health interactions

Finally, a common theme was the interplay of physical and mental health conditions. A Veteran with a history of spinal cord injury described, “Being limited in movement increases my depression and PTSD symptoms as I feel more vulnerable and a likely target.” (P24, female, 57y, 7 conditions) Mental health conditions were also commonly described as exacerbating chronic medical conditions: “Being depressed makes it challenging to be proactive to improve my physical conditions.” (P144, male, 29y, 5 conditions)

3.3. Patient characteristics associated with perceived condition interactions

In post-hoc, exploratory analyses, a greater proportion of patients reported perceived condition interactions when they were younger (53.3% of patients <65 vs. 30.2% of patients ≥65 years, p < 0.001), female (62.5% vs. 38.3% among males, p < 0.001), had an employment status of disabled (65.3%, vs. 35.2–48.5% in other employment groups, p < 0.001) and had a greater number of chronic conditions (57.9%, 45.3%, and 31.3% among patients with ≥6 conditions, 4–5 conditions, and 2–3 conditions, respectively, p < 0.001). Specific conditions that were associated with higher likelihood of perceived interactions at a significance level of p < 0.001 were pain, depression, and spinal cord injury (Appendix 3).

4. Discussion and conclusion

4.1. Discussion

This study presents a taxonomy that reflects ways in which patients perceive interactions among their different conditions and condition treatments. Whereas prior work in this area tends to focus on specific conditions (e.g. hypertension, diabetes, depression), [7,13] our work aimed to provide a categorization that might serve as a guide and tool, generalizable across multiple types of comorbidities. Our work reveals that patients hold wide and varied perceptions of whether and how their multiple conditions interact. Patient responses also point to the wide-ranging resources they draw on to manage these interactions, and suggest that understanding patient-perceived interactions is a key step towards person-centered care. The richness of interactions is clarified by this taxonomy. For example, even when perceived as a unidirectional interaction, patients describe four distinct subtypes of relationships among conditions and treatments (i.e. condition A → condition B, condition → treatment, treatment → condition, treatment A → treatment B) (Fig. 1).

By generating a taxonomy that is agnostic to specific health conditions, this work can inform approaches to clinical decision-making, treatment, and counseling across a wide range of conditions. Providers who intentionally incorporate condition interactions into clinical discussions can facilitate shared decision-making and optimize outcomes not only for individual diseases but for the interactions as well. Inquiry into perceived interactions may elicit patient insight into the multiple resources they draw on to manage these interactions and contribute to their understanding of their overall health. More specifically, recognizing the perceived interactions may be a key step in identifying and addressing previously unrecognized barriers and facilitators to treatment adherence. For example, when a patient reports that chronic pain impedes his or her ability to perform desired self-care activities for diabetes, it could provide both the explanation for suboptimal diabetes control and, by suggesting diabetes as a shared target, help establish a concrete treatment goal (as recommended by guidelines) [24] for any pain prescription. The taxonomy can also be used to highlight chasms between care team and patient perspectives, for example if open-ended inquiry reveals that a patient perceives there is unclear (or no) interaction between their various comorbid conditions, yet the provider feels the MCC and/or recommended treatments have important or obvious interplay, this might highlight opportunity for further exploration, education, and care coordination.

The results of our study resonate with previous qualitative research describing patient experiences with interactions among specific conditions. For example, one qualitative study of patients with diabetes and depression found that some individuals did not see these issues as related, while others described a relationship in which management could be either integrated or conflicting. [18] We found that this spectrum of experiences and the different bidirectional relationships between one condition and a second condition’s treatment were not unique to comorbid diabetes/depression, validating these patterns in patients with an array of different physical and mental health conditions. The exploratory results that patients who are younger, female, and have a greater number of conditions are more likely to perceive condition interactions may reflect unique condition combinations that are common in these groups, or may relate to other characteristics or circumstances; these findings present an opportunity for further study.

Our findings of common cross-cutting themes highlight additional nuances beyond the category and direction of interactions. While experiences with condition interactions are often negative and can exacerbate physical or mental health, in some cases the presence of a comorbidity can have positive consequences, leading to improved medication adherence or positive lifestyle changes. For example, a discussion about the positive impact of lifestyle change on multiple health issues that a patient already identifies as interacting, including cardiometabolic conditions, chronic pain, and depression, may be more likely to motivate a patient.

The taxonomy that emerged from this study may be a valuable tool to guide clinical care and shared decision-making. Understanding a patient’s perception of comorbid condition interactions may also be an important part of understanding the patient’s explanatory model (and recognizing where it diverges from the medical provider’s perspective), and the patient’s broader sense of their health state. For example, one patient in the study described how concerns about prostate cancer recurrence exacerbated his anxiety and depression. By querying this patient about how his physical conditions influence his mental health issues, a clinician could better understand, address, and potentially alleviate the patient’s anxiety. These discussions may also reveal differences in how patients and their clinicians view certain condition interactions, and facilitate discussions about appropriate person-centered plans for treatment or surveillance. Person-centered care enables patients and providers to work together and encourages communication and holistic care. The provider’s ability to connect with their patients using a person-centered approach can ultimately improve the patient-provider relationship and support shared decision-making[25]. Future research should study in more depth the role that contextual factors (e.g., social support, housing situation), continuity of care, and the patient-provider relationship could play in helping patients manage multimorbidity; our taxonomy can provide a framework in which to explore these factors’ influence on patient-perceived condition interactions.

Some limitations to this study should be noted. Although this study describes an analysis of qualitative data from an open-ended question, it is not a conventional qualitative study, and therefore lacks methodological elements that are central to a qualitative research study, such as a qualitative sampling strategy and concurrent data collection and data analysis processes. Our inclusion of the VA as one site resulted in a population that was predominantly male and with limited racial and ethnic heterogeneity, which may have influenced rates of certain conditions and condition combinations in our sample. Analysis of survey data enabled us to include qualitative data from a larger population, but precluded follow-up of ambiguous responses and verification to ensure our interpretation was correct. We mitigated this by sending a follow up survey to a sample of respondents to ensure that the synthesized qualitative findings resonated with them.

4.2. Conclusion

The proposed taxonomy presents an opportunity to break away from a single condition-focused paradigm and incorporate into clinical care the common condition interactions that patients with multimorbidity experience.

4.3. Practice implications

As medical care and research increasingly incorporate person-centered outcomes and shared decision-making, this work highlights the need to recognize that a target condition and its treatment may not be viewed in isolation from other treatments and conditions by patients with multimorbidity. Explicit discussions about interrelated symptoms and condition-treatment interactions may offer opportunities to better support patient self-management and shared decision-making.

Abbreviations:

MCCs

multiple chronic conditions

VA

veterans affairs

HSR&D

health services research & development

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.pec.2019.11.020.