Associations between symptom severity and treatment burden in people living with HIV

Abstract

Aim:

To examine the association between symptoms severity and treatment burden in people living with HIV.

Design:

Correlational, secondary analysis of data from participants diagnosed with HIV enrolled in a descriptive, cross-sectional study examining physical activity patterns.

Methods:

We analysed data from 103 men and women using self-report data collected between March 2016 - February 2017. Our primary statistical analyses consisted of explanatory multivariate modelling with individual PROMIS-29 scores representing symptom severity and treatment burden measured using the Treatment Burden Questionnaire-13.

Results:

Greater symptom severity was associated with higher levels of cumulative treatment burden as well as higher levels of task-specific medication and physical activity burden. Multivariate regression analyses revealed that fatigue was a risk factor of cumulative treatment burden as well as task-specific medication and physical activity treatment burden. Effect sizes of multivariate models ranged from small (0.11) to medium (0.16). Additionally, post hoc analyses showed strong correlations between fatigue and other measured symptoms.

Conclusion:

sFindings support extant treatment burden literature, including the importance of addressing symptom severity in conjunction with treatment burden screening in the clinical setting. Results also suggest clinical interventions focused on the reduction of fatigue could reduce treatment burden in people living with HIV. Strong correlations between fatigue and other symptoms indicate the potential for reducing fatigue by addressing other highly clustered symptoms, such as depression.

Impact:

People living with HIV exhibiting higher levels of fatigue are at high risk for treatment burden and poorer self-management adherence. Clinicians should consider incorporating symptom and treatment burden assessments when developing, tailoring and modifying interventions to improve self-management of HIV and other co-morbid conditions.

1 |. INTRODUCTION

Sub-optimal self-management adherence, specifically adherence to medication (Günthard et al., 2016; Schaecher, 2013), physical activity (Brien, Tynan, Nixon, & Glazier, 2016) and diet (Fields-Gardner et al., 2010), are associated with poor health outcomes for the 39.7 million people living with HIV (PLWH) worldwide. Mitigation of risk factors associated with sub-optimal adherence is essential in improving health outcomes in this population. Treatment burden, the burden of adhering to a self-management regimen, is a risk factor associated with poor medication and physical activity adherence in PLWH (Sav et al., 2013; Schreiner, Perazzo, Rn, Daly, & Webel, 2019; Tran et al., 2012). Identification of antecedent correlates associated with treatment burden could provide healthcare providers with targets for interventions aimed at decreasing treatment burden and improving health outcomes in PLWH. Symptom severity is one such antecedent correlate of treatment burden. Higher levels of symptom severity, such as fatigue, are associated with higher levels of treatment burden (Schreiner et al., 2019). PLWH have a high prevalence of symptoms (Iribarren et al., 2018; Webel et al., 2019; Wilson et al., 2016) and may be at greater risk of treatment burden due to these symptoms, though this relationship has not been previously explored.

2 |. BACKGROUND

Many PLWH experience a multiple distressing symptoms related to the diagnosis of HIV or other co-morbid conditions (Iribarren et al., 2018; Webel et al., 2019; Wilson et al., 2016). Symptoms such as pain, fatigue and depression have a negative effect on the health and quality of life of these individuals (Balderson et al., 2013; Schnall et al., 2017). These symptoms are also associated with sub-optimal self-management adherence and subsequent poor health outcomes, such as medication non-adherence and low levels of physical activity (Langebeek et al., 2014; Vancampfort et al., 2018).

An emerging concept in self-management adherence literature is treatment burden, which is the burden associated with the “work” a patient must undertake to manage one or more chronic conditions (Boyd et al., 2014; Sav et al., 2013; Tran et al., 2012). Specific sources of treatment burden, such as burden related to daily medication taking, are associated with poor self-management adherence (Sav et al., 2013; Schreiner et al., 2019; Tran et al., 2012). For instance a person diagnosed with HIV may experience high medication burden due to a complex medication regimen and may not adhere to these necessary actions. Treatment burden related to medication administration and physical activity are risk factors for poorer medication and physical activity adherence in PLWH, though this explanatory relationship was not found between dietary treatment burden and diet adherence (Schreiner et al., 2019). Schreiner et al. (2019) indicated this non-significant relationship between dietary treatment burden and dietary adherence, measured via Healthy Eating Index 2010 scores, was due to the inclusion of other lifestyle habits (i.e. smoking and consumption of alcohol) in addition to eating a healthy diet.

Symptoms, such as fatigue, are antecedent correlates of treatment burden in older adults diagnosed with multi-morbid conditions (Schreiner, Schreiner, & Daly, 2018). This study demonstrated higher symptom severity was a risk factor for higher levels of cumulative treatment burden. However, this relationship is unexplored in PLWH, despite the high incidence of symptoms and the impact of treatment burden on self-management adherence in this population.

Examining the relationship between highly prevalent symptoms and treatment burden in PLWH can provide insight into the specific symptoms associated with increasing treatment burden. We used Shippee, Shah, May, Mair, and Montori (2012) Cumulative Complexity Model to inform the design of our study. The Cumulative Complexity Model posits chronic condition-associated symptoms negatively affect a patient’s capacity to effectively self-manage these conditions. Disease severity and related symptoms can contribute to increased treatment burden and subsequent poor self-management adherence. By examining these relationships, clinicians can focus palliative efforts, targeting specific-associated symptoms, thereby reducing treatment burden and potentially improving self-management in PLWH.

3 |. THE STUDY

3.1 |. Aim

To examine the association between symptoms severity and treatment burden in people living with HIV. We hypothesized that symptom severity would be positively associated with cumulative, medication and physical activity-related treatment burden, controlling for antecedent correlates of treatment burden in PLWH [i.e. number of multiple chronic conditions (MCC) and years since diagnosed with HIV] and that it would not be related to diet treatment burden.

3.2 |. Design

For the purposes of this study, we conducted a secondary analysis of a descriptive, cross-sectional study examining the physical activity patterns of individuals diagnosed with HIV.

3.3 |. Ethical considerations

The Institutional Review Board at the University Hospitals, Cleveland Medical Center approved all study procedures (UHCMC IRB 11-15-17) prior to enrolment of the participants.

3.4 |. Participants

We analysed data from 103 men and women with HIV enrolled in the Active Living and HIV study (Webel, 2018). Sensitivity analysis conducted in G*Power 3.1 (Faul, Erdfelder, Buchner, & Lang, 2009) based on our sample size of 103 participants, α = 0.05 and β = 0.8, allowed us to detect statistical significance in a multivariate regression model consisting of nine independent variables at a medium effect size of 0.17.

Convenient sampling was used to recruit participants into the Active Living and HIV Study. Participants enrolled in the study were selected from a repository of individuals who previously participated in HIV-related studies and consented to contact for future studies.

Eligibility criteria for participants were: (a) age ≥ 18 years and (b) confirmed HIV (HIV + ELISA with confirmatory PCR or Western blot). Exclusion criteria were as follows: (a) a medical contraindication for exercise determined by the American Heart Association criteria or inability to be physically active without an assistive device (i.e. wheelchair, walker, or cane); (b) unable to understand English or Spanish; or (c) expected to move out of the area or planned to become pregnant within 12 months. Additional eligibility criteria for our secondary analyses consisted of complete data from the Patient-Reported Outcomes Measurement Information System (PROMIS)-29 survey and the Treatment Burden Questionnaire-13. We excluded 13 Active Living and HIV participants for our secondary analysis because of incomplete data (e.g. incomplete or missing Treatment Burden Questionnaire-13 survey).

3.5 |. Data collection

Data collection was conducted at a research-intensive institution with ties to a tertiary medical centre between March 2016 - February 2017. Data analysed in our secondary analysis were collected using self-report questionnaires administered by study staff in electronic format via an institutional REDCap (Research Electronic Data Capture) database.

3.6 |. Validity, reliability and rigor

3.6.1 |. Demographics

We collected univariate measures of age, sex, race, education level and monthly income. We included other known antecedent correlates of treatment burden in our analyses: the number of multiple chronic conditions (Sav et al., 2016; Schreiner et al., 2019; Schreiner & Daly, 2018), as defined by the Centers for Medicare and Medicaid Services (CMS, 2019) and the number of years since participants’ HIV diagnoses (Schreiner et al., 2019).

3.6.2 |. PROMIS-29

The primary focus of this analysis was to examine symptom severity as an antecedent correlate of treatment burden. We used the PROMIS-29 to assess participants’ symptom severity. The PROMIS-29 is a validated, reliable and widely used measure of symptom severity. It has been used in various populations with Cronbach’s α ranging from 0.85–0.93 (Cella et al., 2010; Kroneke, Yu, Wu, Kean, & Monahan, 2014) and has demonstrated similar psychometric properties (Cronbach’s α = 0.87–0.97) in PWLW (Schnall et al., 2017). Each item is rated on a 1 (low severity) −5 (high severity) scale, except for Physical Function and Ability to Participate in Social Roles, which were reversed coded, with lower scores indicating poorer physical function or ability to participate in social roles and Pain Intensity, which is score from 0 (no pain) - 10 (high pain intensity). We summed total PROMIS-29 scores for each symptom and treated the summed individual symptom score as an independent variable for use in analyses.

3.6.3 |. Treatment burden

Our outcome was treatment burden measured by the Treatment Burden Questionnaire-13 (TBQ-13). The TBQ-13 is a psychometrically sound, 13-item survey measuring the burden associated with the self-management of chronic conditions (Tran et al., 2012). Respondents rate their associated level of burden, ranging from 0 (no burden) −10 (high burden). Total scores range from 0–130, with higher scores indicating higher burden. The TBQ-13 displays strong internal reliability, with a Cronbach’s α of 0.89 (Tran et al., 2012). The internal reliability of the TBQ-13 had not been previously assessed in PLWH, thus we calculated the Cronbach’s α for the instrument in our sample and found strong internal consistency with a Cronbach’s α of 0.9. Cumulative treatment burden scores were calculated by summing all 13 items into a composite score. Medication-specific treatment burden was comprised of the first four TBQ-13 items, with summed composite scores for the subscale ranging from 0–40. Physical activity- and diet-related burden were single items with scores ranging from 0–10.

3.7 |. Data analysis

We analysed study data using SPSS Version 26 (IBM Corp., Armonk, N.Y., USA). We conducted univariate analyses describing the sample percentages, frequencies, means and standard deviations of participant demographics and study variables (Tables 1 & 2). We conducted a post hoc One-way ANOVA test to determine if there were group difference in symptom severity based on the number of MCC. We posited group differences between the reported chronic conditions with symptom severity increasing with greater frequency of chronic conditions. We then tested assumptions prior to running statistical analyses, including testing for influential outliers and linearity associated with regression and found variables met statistical assumptions of all planned analyses. Pearson’s r was used to test relationships between symptoms, other known correlates of treatment burden of PLWH (number of MCC and years since diagnosis of HIV) and treatment burden. Following correlational analyses, we included significant (p = 0.05) correlates of treatment burden as independent variables in our multivariate models. We then ran multivariate regression analyses to determine which symptoms explained treatment burden, while controlling for the number of MCC and years since diagnosis of HIV. Calculated multivariate regression model effect sizes, using Cohen’s (1988) F statistic, ranged from small (0.11) to medium (0.16). Statistical significance for planned analyses was set at an α = 0.05.

TABLE 1.

Sample demographic characteristics

Variable	Frequency	Percent
Sex
Male	67	65%
Female	36	35%
Ethnicity
Black	86	83.5%
White	11	10.7%
American Indian or Native Alaskan	5	4.9%
Education
11th grade or less	23	22.3%
High school or GED	32	31.1%
Some college or technical school training	25	24.3%
2 years of college or technical school training	8	7.8%
College Degree (BA or BS)	12	11.7%
Master’s degree or higher	3	2.9%
Monthly income
No monthly income	6	5.8%
Less than $200	6	5.8%
$200-$399	5	4.9%
$400-$599	1	1.0%
$600-$799	41	39.8%
$800-$999	12	11.7%
More than $1,000	32	31.1%
Variable	Range	Mean	SD
Age	32–71	53.16	7.17
Number of chronic conditions	2–10	3.64	1.75
Years since diagnosis of HIV	3–33	17.01	7.62

TABLE 2.

Univariate statistics describing symptom severity and treatment burden scores

Variable	Range	Mean	Standard deviation
PROMIS physical function	8–20	16.99	3.65
PROMIS anxiety	4–20	7.8	3.96
PROMIS depressive symptoms	4–20	7.35	3.93
PROMIS fatigue	4–20	8.85	4.11
PROMIS sleep disturbance	4–20	10.64	4.44
PROMIS social roles and activities	4–20	15.72	3.99
PROMIS pain interference	4–20	8.69	4.76
Diet treatment burden	0–10	1.99	2.74
Physical activity treatment burden	0–10	1.97	2.85
Total treatment burden score	0–119	22.99	24.49
Medication treatment burden	0–35	6.66	8.43

Abbreviations: PROMIS, patient-reported outcomes measurement information system.

4 |. RESULTS

Our sample was primarily male, African American, had a high school education, with a mean age of 53 years. In addition to HIV, the most prevalent chronic conditions in our sample were hypertension (55.5%), asthma (28.2%), arthritis (26.2%), diabetes (18.4%) and hepatitis B/C (14.6%) and hyperlipidaemia (14.6%).

Post hoc one-way ANOVA analysis examined mean differences in total symptom severity, calculated by summing all PROMIS-29 scores, between categories of MCC, which ranged from 2–10. One-way ANOVA results (F [9, 93] = 1.17; p = 0.33) were not statistically significant, indicating that there was not a difference in symptom severity between groupings based on the number of MCC.

Correlation analyses supported our study hypothesis, although a weak correlation did exist between symptoms and diet-related burden. We found symptoms of fatigue (r = 0.24; p < 0.01) and sleep (r = 0.22; p = 0.01) were positively correlated with cumulative treatment burden. Poorer physical functioning (r = −0.27; p < 0.01), increased depression (r = 0.16; p = 0.05), increased fatigue (r = 0.33; p < 0.001), greater sleep disturbance (r = 0.22; p = .01), worsening ability to participate in social roles and activities (r = −0.2; p = 0.02) and increased pain interference (r = 0.25; p < 0.01) and pain intensity (r = 28; p < 0.01) were correlated with physical activity treatment burden. Worsening physical functioning (r = −0.2; p = 0.02) and ability to participate in social roles (r = −0.18; p = 0.03), increased depression (r = 0.16; p = 0.02), fatigue (r = 0.31; p = 0.001), sleep disturbance (r = 0.21; p = 0.02) and pain (r = 0.16; p = 0.05) were correlated with the summed mediation treated burden scores. Only increased sleep disturbance (r = 0.17; p = 0.05) correlated with diet-related treatment burden.

Post hoc analysis of correlations between symptoms and control variables are listed in Table 3. We focused on fatigue as a primary variable because it was the symptom explanatory of treatment burden in three of our multivariate models. As the severity of fatigue increased, so did the severity of the other symptoms. Fatigue was weakly correlated with the number of MCC and did not correlate with years since diagnosis with HIV.

TABLE 3.

Pearson’s correlation between symptom severity, years since diagnosis of HIV and number of chronic conditions

Variable	1	2	3	4	5	6	7	8	9	10
1. PROMIS physical function	—
2. PROMIS anxiety	−0.28; p = 0.002	—
3. PROMIS Depressive Symptoms	−0.34; p < 0.001	0.81; p < 0.001	—
4. PROMIS Fatigue	−0.61; p < 0.001	0.52; p < 0.001	0.63; p < 0.001	—
5. PROMIS Sleep Disturbance	−0.32; p < 0.001	0.39; p < 0.001	0.46; p < 0.001	0.46; p < 0.001	—
6. PROMIS Social Role and Participation	0.63; p < 0.001	−0.43; p < 0.001	−0.56; p < 0.001	−0.68; p < 0.001	−0.41; p < 0.001	—
7. PROMIS Pain Interference	−0.67; p < 0.001	0.27; p < 0.01	0.37; p < 0.001	0.59; p < 0.001	0.41; p < 0.001	−0.59; p < 0.001	—
8. PROMIS Pain Intensity	−0.78; p < 0.001	0.29; p < 0.01	0.40; p < 0.001	0.61: p < 0.001	0.41; p < 0.001	−0.65; p < 0.001	0.94; p < 0.001	—
9. Years Since Diagnosed with HIV	−0.12; p = 0.11	0.01; p = 0.45	−0.03; p = 0.38	0.04; p = 0.35	0.00; p = 0.5	−0.08; p = 0.28	−0.06; p = 0.29	−0.02; p = 0.41	—
10. Number of MCC	−0.33; p < 0.001	0.13; p = 0.10	0.21; p = 0.02	0.19; p = 0.03	0.21; p = 0.02	−0.15; p = 0.06	0.18; p = 0.03	0.23; p = 0.01	0.14; p = 0.07	—

Note: MCC, multiple chronic conditions; PROMIS, patient-reported outcomes measurement information system.

Multivariate analyses supported our hypotheses. In examining symptom severity as a risk factor of cumulative treatment burden, while controlling for the number of MCC and years since diagnosis of HIV (Table 4), we found that fatigue, the number of MCC and years since diagnosis of HIV explained 10% of cumulative treatment burden’s variance (adjusted r² = 0.10; F [9, 93] 2.24; p = 0.03). Fatigue (standardized beta coefficient = 0.30; p = 0.05), total number of MCC (standardized beta coefficients = 0.27; p = 0.01) and years since diagnosed with HIV (standardized beta coefficient = −0.21; p = 0.03) were risk factors for cumulative treatment burden.

TABLE 4.

Results of multivariate regression model with cumulative treatment burden as the outcome

Variable	Unstandardized beta coefficient	Standardized beta coefficient	t	p
PROMIS physical function	0.30	0.04	0.25	0.81
PROMIS anxiety	−0.42	−0.07	−0.41	0.68
PROMIS depression	−0.70	−0.11	−0.59	0.56
PROMIS fatigue	1.84	0.30	1.99	0.05
PROMIS sleep disturbance	0.84	0.15	1.35	0.18
PROMIS satisfaction with social roles	−0.50	−0.08	−0.54	0.59
PROMIS pain interference	−1.69	−0.08	−0.29	0.78
PROMIS pain intensity	−0.48	−0.09	−0.27	0.79
Years since diagnosis with HIV	−0.69	−0.21	−2.19	0.03
Total number of chronic conditions	3.42	0.27	2.55	0.01

Note: Dependent variable = cumulative treatment burden.

Our second multivariate regression model examining symptom severity as a risk factor of physical activity-related treatment burden controlling for the number of MCC and years since diagnosis of HIV (Table 5) explained 14% of the variance of physical activity treatment burden (adjusted r² = 0.14; F [9, 93] = 2.77; p < 0.01). Fatigue (standardized beta coefficient = 0.31; p = .03) and years diagnosed with HIV (standardized beta coefficients = −0.23; p = 0.02) were risk factors for physical activity treatment burden.

TABLE 5.

Results of multivariate regression model with physical activity treatment burden as the outcome

Variable	Unstandardized beta coefficient	Standardized beta coefficient	t	p
PROMIS physical function	−0.13	−0.17	−0.99	0.33
PROMIS anxiety	−0.17	−0.24	−1.49	0.14
PROMIS depression	0.06	0.08	0.44	0.66
PROMIS fatigue	0.23	0.31	2.16	0.03
PROMIS sleep disturbance	0.07	0.11	0.95	0.34
PROMIS satisfaction with social roles	0.04	0.06	0.42	0.68
PROMIS pain interference	0.11	0.05	0.17	0.87
PROMIS pain intensity	−0.08	−0.13	−0.39	0.70
Years since diagnosis with HIV	−0.09	−0.23	−2.42	0.02
Total number of chronic conditions	0.18	0.12	1.16	0.25

Note: Dependent variable = physical activity treatment burden.

Our third multivariate regression model with symptom severity as a risk factor of medication-related treatment burden, controlling for the number of MCC and years since diagnosis of HIV (Table 6), explained 13% of medication-related treatment burden [adjusted r² = 0.13; F(9,93) = 2.74; p < 0.01]. Fatigue (standardized beta coefficient = 0.38; p = 0.01) and years since diagnosed with HIV (standardized beta coefficient = −0.2; p = .04) were risk factors for total medication treatment burden scores. Our final multivariate regression model with examining symptom severity as a risk factor for diet-related treatment burden controlling for the number of MCC and years since diagnosis of HIV (Table 7) was non-significant (p = .16).

TABLE 6.

Results of multivariate regression model with physical activity treatment burden as the outcome

Variable	Unstandardized beta coefficient	Standardized beta coefficient	t	p
PROMIS physical function	−0.13	−0.17	−0.99	0.33
PROMIS anxiety	−0.17	−0.24	−1.49	0.14
PROMIS depression	0.06	0.08	0.44	0.66
PROMIS fatigue	0.23	0.33	2.16	0.03
PROMIS sleep disturbance	0.07	0.11	0.95	0.34
PROMIS satisfaction with social roles	0.04	0.06	0.42	0.68
PROMIS pain interference	0.11	0.05	0.17	0.87
PROMIS pain intensity	−0.08	−0.13	−0.39	0.70
Years since diagnosis with HIV	−0.09	−0.23	−2.42	0.02
Total number of chronic conditions	0.18	0.12	1.16	0.25

Note: Dependent variable = physical activity treatment burden.

TABLE 7.

Results of multivariate regression model with dietary treatment burden as the outcome

Variable	Unstandardized beta coefficient	Standardized beta coefficient	t	p
PROMIS physical function	−0.11	−0.15	−0.80	0.43
PROMIS anxiety	−0.16	−0.23	−1.34	0.19
PROMIS depression	0.02	0.03	0.18	0.86
PROMIS fatigue	0.09	0.13	0.81	0.42
PROMIS sleep disturbance	0.14	0.22	1.88	0.06
PROMIS satisfaction with social roles	0.00	0.00	0.02	0.98
PROMIS pain interference	−0.68	−0.30	−1.00	0.32
PROMIS pain intensity	0.00	0.00	0.00	1.00
Years since diagnosis with HIV	−0.03	−0.08	−0.82	0.42
Total number of chronic conditions	0.29	0.20	1.88	0.06

Note: Dependent variable = dietary treatment burden; PROMIS = patient-reported outcomes measurement information system.

5 |. DISCUSSION

Our results demonstrated that high levels of symptom severity are associated with higher levels of cumulative treatment burden levels and task-specific (medication and physical activity) burden in PLWH. Worse symptoms were correlated with higher levels of treatment burden, though we only found a weak correlation between worsening sleep disturbance and higher levels of diet-related treatment burden, with the small effect size (0.17) most likely explained by random error. Our results indicating that higher levels of fatigue is a risk factor for greater treatment burden supports Shippee et al.’s (2012) hypothesis of symptom severity as an antecedent risk factor of treatment burden.

Our results also support Schreiner et al. (2018) finding of the relationship between symptom severity and treatment burden in the older adult population diagnosed with multi-morbid conditions. Both analyses found fatigue as the only symptom explaining treatment burden variance in a multivariate analysis, emphasizing the impact of fatigue across populations. The importance of this finding is further supported by extant HIV symptom literature, which demonstrates fatigue is prevalent and distressful within PLWH. Wilson et al. (2016) found fatigue as the third most reported symptom (47% percent of the sample) using the HIV Symptom Index Scale in a sample of 1945 men and women diagnosed with HIV. Eighty-four percent of these participants also rated fatigue as bothersome and 31% of participants rated fatigue within the ‘Bothers Me a Lot’ category. In conjunction, these findings identify fatigue as a prevalent risk factor for increasing treatment burden and subsequent poor health outcomes within PLWH.

Fatigue associated with chronic conditions is often difficult to treat. Certain medications may be beneficial in the palliative treatment of fatigue, but there is no current standard pharmacological treatment for chronic condition-associated fatigue (Mucke et al., 2016). Some non-pharmacologic interventions, such as yoga (Evans et al., 2017) and resistance training (Hagestrom et al., 2016) have been shown to improve fatigue in various chronic conditions.

Despite the difficulty in treating fatigue, these findings strengthen the argument for adding symptom screening (Schreiner et al., 2018) when evaluating treatment burden (Dobler et al., 2018; Eton et al., 2017; Sav et al., 2016) in populations diagnosed with chronic conditions, including HIV. In particular, assessment of fatigue can identify persons at high risk for self-management non-adherence.

Our findings also suggest PLWH who are earlier in their diagnosis of HIV are more susceptible to higher levels of cumulative, medication and physical activity-related treatment burden, independent of symptom severity and the number of chronic conditions. This finding is especially important given that our sample consisted primarily of African American males and in light of previously reported prevalence of sub-optimal medication adherence among younger African American males (Zanoni & Mayer, 2014). Adherence to an anti-retroviral medication regimen is the most essential component to preventing complications from HIV (Benator et al., 2015; Voisin, Quinn, Kim, & Schneider, 2017). Further research into how being earlier in the HIV disease trajectory increases treatment burden and undermines optimal anti-retroviral adherence in younger African American males would provide better insight into how to achieve anti-retroviral adherence in this sub-population.

Post hoc analyses also suggest that treatment of other symptoms highly associated with fatigue could lead to a reduction in treatment burden and a subsequent improvement in cumulative, medication or physical activity adherence. The phenomenon of symptom clustering, or the occurrence of two or more concurrent symptoms, is prevalent in PLWH (Miaskowski et al., 2017; Moens et al., 2015; Namisango et al., 2015; Wilson et al., 2016). The literature suggests treatment of one symptom may alleviate other co-existing symptoms (Kwekkeboom et al., 2012). Based on this literature, treating a participant’s depression, which was one of the most highly correlated symptoms with fatigue in our sample, might reduce a participant’s overall level of fatigue, thereby decreasing treatment burden. Conversely, the treatment of one symptom can exacerbate or create issues related to another symptom. For instance the pharmacological (e.g. anti-depressive management) treatment of depression can create issues with decreased mobility due to increased fatigue, anxiety and worsening sleep quality.

Our post hoc analyses also demonstrated no difference in symptom severity between groups based on the number of chronic conditions, indicating most symptoms relate to the diagnosis of HIV and not the subsequent diagnosis of other co-morbid conditions. We acknowledge the need for further testing to validate this finding, though it emphasizes the centrality of HIV self-management to reduce symptom severity in this population.

5.1 |. Limitations

Our study was not without limitations. Our small sample size limited our ability to detect statistical significance within our planned analyses. A prospective study designed around a larger sample would yield an improved understanding of how symptom severity affects treatment burden in PLWH.

Secondary data analyses limited our ability to use multiple symptom scales, such as the HIV Symptom Index (Wilson et al., 2016) to improve the validity and reliability of results. Using a specific HIV-symptom scale in comparison to the PROMIS-29 measure would provide additional validation of results when comparing mean differences in symptom severity by the number of MCC.

Given the large proportion of African-American participants in the study sample, we were unable to address the importance of HIV-related social stigma on symptom severity and treatment burden within the study model. HIV-related social stigma is prevalent in the HIV population, especially in the African-American community. HIV-related social stigma is associated with increased depressive symptomology and decrease treatment adherence in the African-American HIV population (Quinn et al., 2017; Williamson, Mahmood, Kuhn, & Thames, 2017). Further studies measuring the association of symptom severity and treatment burden should consider controlling for the influence of HIV-related stigma, especially in a majority male, African-American sample. We were unable to test whether certain symptoms, such as pain or depression, were antecedent to fatigue due to the study design. In the future, a longitudinal approach employing a latent growth curve statistical analysis could help determine the temporal ordering of these symptoms.

The secondary analysis design of this study also limited our ability to compensate for potential sampling bias associated with single site recruitment in the primary study. Single-site sampling can contribute to homogeneity of the sample; thus results may lack generalizability to the general population of PLWH.

As previously noted in the literature, the diet-related treatment burden question lacks specificity (Schreiner et al., 2019). The question not only inquires about the respondent’s diet but also about smoking cessation and alcohol intake, not referring only to eating a healthy diet. Rewording the question, removing the language about burden related to lifestyle changes (e.g. smoking cessation and alcohol consumption), could improve the accuracy of results when testing for the association between symptoms and diet-related treatment burden in PLWH.

6 |. CONCLUSION

The findings of this study are applicable to the clinical setting. Fatigue in PLWH is prevalent, distressful and puts patients at risk of increased treatment burden. Screening of symptom severity in conjunction with treatment burden allows clinicians, including nurses and nurse practitioners, to identify patients most at risk for higher levels of treatment burden. This screening is also an essential step in identifying and decreasing barriers to effective implementation and sustainability of any prescribed self-management intervention. Clinicians could reduce treatment burden by targeting fatigue directly or via treatment of other highly correlated symptoms, such as depression. Further replication of these findings are needed in a robust, heterogeneous sample of individuals diagnosed with chronic conditions before a recommendation implementing symptom and treatment burden screening in the clinical setting can be endorsed as a standard of care.