|Home | About | Journals | Submit | Contact Us | Français|
This study compares the self-assessments of 67 late stage HIV/AIDS patients regarding their symptomatology, sense of self-worth and several other aspects of their health care situation, to assessments of that situation provided by their informal caregivers. As part of a dyadic study of care preferences, patients and caregivers independently completed nearly identical versions of the Palliative Care Outcome Scale (POS), a short ten-item measure of the patient’s current health, social, and psychological status. Participants in the study were recruited from inpatient and outpatient services at an urban medical center. Substantial or moderate agreement, assessed by weighted kappa, was found between patient and caregiver assessments with regard to only four items assessing physical or emotional states of the patient (pain, other symptoms, anxiety, and life worthwhile). Fair or slight agreement was found for the six remaining items including those assessing patient’s sense of self-worth, family/friends’ anxiety, interactions with family/friends, and practical matters. Mean ratings of self-worth were significantly different for patients and caregivers. These findings underscore the clinical need to assess patient care outcomes directly, and they suggest the importance of facilitating more effective communication about relevant health issues among seriously ill patients, caregivers, and health care providers.
Patients are generally considered the best source of information about their symptoms, states, and opinions, given that this information is often subjective and personal (1). However, cognitive impairment, emotional distress, respondent burden, and fatigue may prevent seriously ill patients from participating in research or decision-making about their care. Additionally, it may become ethically inappropriate to ask patients to complete a research interview, particularly if they are seriously ill (2).
Prior research has investigated the agreement of patient reports with reports from a variety of potential proxies, including: health care professionals involved in the patient’s care, the patient’s friends or relatives, and the patient’s caregiver (3). However, in recent years, given an increasing recognition of the importance of family and friends as an important component of the context within which patient care and treatment decision-making most often occurs, interest is now centered on family caregivers as an alternative source of relevant information regarding seriously ill patients who are unable to participate in clinical or research interviews (4-6). In some instances, such as when patients have infrequent or brief contact with medical professionals or the patient has been legally found incapable of making his or her own health care decisions, family proxies may be the only possible source of information upon which treatment decision-making for the patient can be made. It has been hypothesized that family caregivers represent a unique class of proxy respondents for seriously impaired patients since they, by virtue of their care provision, have often had ample opportunity to observe symptoms and discuss well-being with the patient (1). However, it has also been noted that the quality of family caregivers reports may vary, given that not all caregivers and patients are equally adept at or comfortable with discussing the complex and difficult issues associated with serious illness and potential death; and, caregivers may have their own unique perceptions or beliefs regarding the patient’s health and medical care that the patient does not necessarily share (1).
Although family proxies may be readily available, the value of information they supply to healthcare professionals serving the seriously ill is dependent upon the degree to which the caregiver’s reports accurately reflect the patient’s perceptions and wishes. There are an increasing number of studies that have investigated the accuracy of family caregiver reports as a proxy for patient self-reports. Individual studies have reported on patient-caregiver congruence with regard to a range of attributes of the caregiving situation for different patient populations (e.g., discharge plans of patients with congestive heart failure and adult onset diabetes (3), pain in dementia patients (7), and functional abilities of Alzheimer’s patients (8)), however, the vast majority of congruence studies have described patient-caregiver agreement with regard to the patient’s experience of pain, other symptomatology, and/or quality of life (QOL) in samples of cancer patients (9-13), often with advanced disease or a terminal diagnosis (14-18).
Several patterns have emerged in these cancer studies. For example, in a review of studies assessing agreement between terminal cancer patients and their family caregivers, Tang and McCorkle (1) concluded that most studies found patient-caregiver dyads to agree moderately well (kappa >0.60) on patient QOL. Sneeuw and colleagues found that although exact agreement was not high, 75% to 95% of patient and caregiver responses were identical or within one category when an item had 4 or 5 response categories (19). Similarly, Brown and colleagues found similar mean scores for 39% to 68% of patients and caregivers to be within 10 points of each other on 5 measures assessing various dimensions of QOL that had been transformed to range in value from 0 to 100 (9). Previous reports have also noted that family caregivers tend to hold more negative opinions of the patient’s QOL than does the patient they are caring for (20).
To date, other investigations have found the level of concordance between reports of the patient and their potential proxy to vary, based on the properties of the domain assessed. Research suggests that some patient characteristics such as psychological symptoms or levels of pain are inaccurately reported by potential proxies as these attributes are subjective judgments about internal states (4, 21). Several studies (10, 16) have also concluded that comparability increases for observable constructs such as levels of functioning or physical symptoms (22). Therefore, it might be expected that patient-caregiver agreement varies with the degree of visibility of the attribute being assessed.
Congruence between patient and potential proxy reporting has also been found to vary based on attributes of the caregiver. For instance, caregivers’ perceptions of the patients’ status may be affected by their own pain and depressive symptoms (23) or caregiver burden (24). In a study of advanced cancer patients and their informal caregivers in the UK, Higginson and Gao (25) found that higher caregiver burden was significantly associated with disagreement on the “feeling anxious,” “life worthwhile,” and “felt good” items of the Palliative Care Outcomes Scale.
Prior research also suggests that cancer caregivers tend to rate patients’ psychological and physical health poorer than do patients themselves. In specific, earlier studies have found caregivers’ to overestimate patients’ experience of psychological symptoms (16), pain and/or other physical symptoms (12,15,18). Poorer ratings of patients’ health (26) and functioning (18) by caregivers have also been reported. On the other hand, Higginson and Gao (25) found substantial chance-corrected agreement (weighted kappa) between patients and caregivers with regard to reports of pain; moderate agreement in reports of other symptoms, sharing feelings, patient anxiety, family anxiety; fair agreement with regard to reports of wasted time, patients feeling that their life was worthwhile, and patients feeling good about themselves; and only slight agreement with regard to reports of help with practical problems and the information provided. In that study, caregivers also reported more problems than patients with regard to: information given, sharing feelings, patients feeling that their life was worthwhile, and patients feeling good about themselves.
The generalizeability of findings from studies of cancer patients and their caregivers to other patient populations is unknown. Virtually nothing is known about agreement of HIV/AIDS patients’ reports regarding their health and care with those provided by their family caregivers. As with cancer, attention to issues of seriously ill patients remains a necessary and important aspect of HIV/AIDS care and service delivery (27). Even in the era of HAART, HIV/AIDS continues to be a serious, progressive, incurable, and potentially life-threatening illness. HIV/AIDS patients with advanced disease continue to live with not only a risk of dying from the disease itself, but increasingly also from significant co-morbidities such as hepatitis and a variety of malignancies. Like their counterparts living with cancer, patients with advanced HIV/AIDS and their informal caregivers must continually cope with health care issues such as symptom management and threats posed by their illness to their psychological well being (28,29). Despite such similarities with regard to the caregiving situation, important sociodemographic differences between the HIV/AIDS and cancer patient populations associated with risk of infection with HIV, e.g., race/ethnicity, sexual orientation, history of intravenous drug use, age, suggest a need to further investigate patient-caregiver agreement within the context of HIV.
The aim of this report is to describe agreement of patient and caregiver assessments of ten important domains (physical, psychological, and spiritual) of the caregiving situation in a sample of seriously ill HIV/AIDS patients and their family/friend caregivers. Differences in the means of patient and caregiver responses will also be assessed. Based on results of prior cancer studies, we hypothesized that such agreement will be higher for items assessing more overt or observable dimensions of the patient’s health or care (e.g., visible physical symptoms, receipt of information, and personal affairs) than for items assessing dimensions that are either more subjective or internal (e.g., pain, anxiety, and feeling life is worthwhile). It is also predicted that there will be few, if any, systematic differences between patients and caregivers on mean scores. The findings will inform healthcare professionals working with HIV/AIDS patients with regard to areas of the caregiving situation for which patients and family caregivers are most likely to agree as to whether or not there is an issue that needs attention. This information may also be useful for the development of programs to enhance caregivers’ skills as advocates on behalf of patients, by maximizing the likelihood of a shared understanding with regard to the current impact of the illness on the patient and his support system, as well as the patient’s long term goals with regard to care and treatment.
As part of a longitudinal investigation of the care preferences and quality of life of patients with advanced HIV/AIDS, 94 patients and their familial/informal caregivers were interviewed between October 2004 and May 2007. Participating patients were accrued from among those receiving either inpatient or HIV/AIDS outpatient clinic services at an urban teaching hospital. Potential subjects were first approached and enrolled by a hospital-based project recruiter. Patients who agreed to participate were asked to identify their primary, informal (i.e., non-paid) caregiver, who was providing them with the most practical and emotional support. For the purposes of this study, familial caregiver was broadly defined to include partner, relative, or close friend. This patient-nominated caregiver was then enrolled into the study. Informed consent was obtained from all patients and caregivers prior to accrual. The data presented are from individual face-to-face interviews conducted by project staff with the patients and their familial caregiver shortly after the dyad’s accrual into the study. To allow for comparisons, analysis was restricted to dyads for which a completed Palliative Care Outcomes Scale (30) was obtained from both the patient and the caregiver (n=91). Although efforts were taken to minimize the interval between the patient and caregiver interviews in order to assure assessments provided by both respondents referred to their experiences over the same interval of time, the length of time between the patient and the caregiver interval varied widely. Consequently, we further restricted the analyses presented to the 67 dyads with patient and caregiver assessments completed within one week of each other. All instruments and protocols received approval from the appropriate Institutional Review Board.
The Palliative Care Outcome Scale (POS) (30) comprises 10 items, each assessing the respondent’s perceptions regarding a different dimension of care or need for patients and their families in a palliative care setting as it was “over the last 3 days.” The scale’s authors designed separate but parallel forms for the patient and care staff to elicit assessments from each source to permit comparisons of patients’ perceptions of their caregiving situation with those of their care providers. As a result, instructions for completion, item content, and the order of items on both forms are identical. The forms were designed to be administered quickly, thereby minimizing respondent and staff burden while maximizing the breadth of information elicited. Eight of the ten items utilize a five-point Likert-like scale and the remaining two items utilize a three-point scale. Scores for respondents’ ratings on all items can range from 0 (indicating no problem) to 4 (indicating a very severe or overwhelming problem). The POS was initially created for use with advanced cancer patients and health care staff in the United Kingdom. However, in consultation with the scale’s author (Higginson), wording of the staff form was modified slightly to make it appropriate for use with the informal/familial caregiver, and the wording of several items on both the patient and staff forms was modified slightly to make it appropriate for use with HIV/AIDS patients in the United States.
The scale’s authors identified item scores of “3” or “4” as indicative of a severe problem in that dimension. For the purposes of the current analyses, we have chosen to further distinguish among those with an item score of “0” (“no problem”) and those with a score of “1” or “2” (“slight/moderate problem”). In addition to grouping responses into more substantively meaningful categories, this data reduction has the advantage of reducing all items to a three-point scale for comparisons.
Seven items on the patient and caregiver form assesses the patient’s state. Comparison of caregivers’ responses to those of patients for these items is equivalent to that of a proxy. Such a comparison with regard to the remaining three items that assess family/friend states (i.e., family anxiety, information and support) is complicated by the relative weight given to the caregiver and other friends’/family members’ feelings.
The distribution of patient and caregiver ratings as well as the proportion of dyads for which agreement or disagreement was reported is presented for each item of the POS. Mean scores of patients and caregivers as a group were compared using paired t-tests. Patient-caregiver agreement was further assessed using weighted kappa and Spearman’s rho as tests of concordance. The weighted kappa statistic, a chance corrected measure of agreement particularly appropriate for use with ordinal rankings such as the items comprising the POS, was interpreted using the Landis and Koch guidelines: less than 0.00 indicates poor agreement, 0.00 to 0.20 signifies slight agreement, 0.21 to 0.40 is fair, 0.41 to 0.60 is moderate, 0.61 to 0.80 is substantial, and 0.81 to 1.0 is nearly perfect agreement (31). Spearman’s rho was calculated as a test of a monotonic relationship between patients and caregiver responses. Finally, dyads were grouped on the basis of level of agreement between the patient and the caregiver regarding their assessments of the prevalence and severity of each POS item (complete agreement, agree on existence of problem but disagree on severity, disagree on existence of problem). Although total percentage agreement as a measure of concordance is affected by chance and the prevalence of the state being assessed, measures such as weighted kappa and Spearman’s rho make assumptions regarding the relative “distance” between responses that may or may not be appropriate (32). Detailed data regarding the proportion of dyads for which various types of agreement or disagreement were reported are still necessary to understand the ways in which different kinds of proxy data may be biased. Total agreement percentages will be evaluated using descriptors developed by House and colleagues: ≥70% “acceptable,” ≥80% “adequate,” and ≥90% “good” (33, 34). The minimum criterion for significance on all statistical tests was a P-value ≤ 0.05. The sample size of 67 provides sufficient power (1 — β=0.80) to detect a correlation of 0.337 at α=0.05. All analyses were performed using SPSS version 15.
Commensurate with their advanced disease, over two-thirds of patients (69%) reported having been diagnosed with HIV in 1996 or earlier, i.e., prior to the widespread availability of HAART. Nineteen percent reported being diagnosed with the disease since 1996 but at least five years prior to the interview; 9% reported having been diagnosed 3-5 years prior to the interview and only two patients (3%) reported having been diagnosed within two years of their interview. Most caregivers were the spouse/partner (36%) or friend of the patient (21%) with the remainder the patient’s child (15%), sibling (15%), parent (13%), grandparent (2%), or nephew (1%). About half of the patients lived with their caregiver (51%).
Table 1 presents patient and caregiver demographics. A substantial majority of both patients and caregivers were 35-64 years of age (86% vs. 68%, respectively), although caregivers were more likely than patients to be either younger adults (less than 35 years old) or elderly (65 years or older). Whereas approximately equal proportions of patients were male and female, over two-thirds of caregivers (70%) were female. A majority of both patients and caregivers were either Hispanic (47% and 51%, respectively) or Black Dnon-Hispanic (38% and 40%). About two-fifths of patients (45%) and caregivers (39%) had not completed high school or obtained an equivalent degree. Over three-fourths of patients (77%) were on disability. Nearly one-third of the caregivers (31%) were also on disability and 24% were not employed. Only 27% of caregivers were working at the time of the interview.
Table 2 presents the marginal distribution and the mean score of patients’ and caregivers’ responses for each POS item. A statistically significant difference between patient and caregiver means was found with regard to only one item, self-worth (t=-3.20, P=0.002). While 70% of patients reported no problem with regard to self-worth (i.e., feeling good about themselves), well over half of caregivers (58%) reported that the patient had at least some problem in this area.
Although the mean scores of patients and caregivers did not differ for nine of the 10 items assessed, it is interesting to note that for six items, including all five of the items assessing a physical or emotional state of the patient, at least 10% more caregivers than patients noted some problem. This difference was not evident with regard to family/friends anxiety, the patient’s ability to share his/her feelings with the caregiver/family/friends (support), or two of three items assessing practical matters (wasted time and personal affairs).
Results of additional analyses assessing overall concordance between patient and caregiver reports while controlling for chance are presented in Table 3. Weighted kappas computed for each item, suggest agreement was only “slight” (family/friends anxiety, support, provision of adequate and understandable information, and personal affairs) or “fair” (self-worth and wasted time) for six items. Conversely, “moderate” (other symptoms, anxiety, and life worthwhile) or “substantial” (pain), or was reported for only four items, all assessing a current physical/emotional state of the patient.
Spearman correlations ranged from 0.118 to 0.610 in a pattern that mirrored results described based on weighted kappas. Correlations were highest for pain, patient anxiety, rating one’s life as worthwhile, and time wasted on appointments. Correlations were lowest for family anxiety and support from family and friends.
Table 4 presents the distribution of dyads by agreement and/or disagreement with regard to the prevalence and severity of problems for each domain assessed by the POS. While the proportion of dyads for which complete agreement was reported regarding both the existence and severity of the problem (A+B+C) ranged from 45%-70%, it was above 60% for only two items (life worthwhile and wasted time) and for only one item did this proportion reach 70%, the cutoff for “acceptable” agreement as defined by House et al. (33). However, agreement between patients and caregivers with regard to the prevalence of several problems (A+B+C+D+E) was substantially higher, with “adequate” agreement noted regarding the prevalence of pain (80%) and “acceptable” agreement regarding the prevalence of a problem for two additional patient states (anxiety (75%) and life worthwhile (73%)), as well as, family anxiety (73%) and wasted time (75%). In those instances where dyads were in agreement with regard to the existence, but not the severity of a problem, the patient was as likely or more likely to describe the problem as severe (E>D). This imbalance was most notable with regard to patient anxiety (17% vs. 6%) and pain (15% vs. 10%).
Given that delivery of appropriate services predicated upon the identification of need, dyads in which one member reported the existence of a problem, but the other did not bear further discussion. Instances such as those reported in column (F) where the caregiver reports a need but the patient does not, suggest that providers’ reliance on proxy reports could result in the delivery of unneeded services and a potential waste of resources. Conversely, instances such as those in column (G) where the caregiver does not see a problem but the patient reports one, may well represent the greatest risk to the well-being of the patient, assuming the patient’s assessment is most accurate. The total proportion of dyads for which one member of the dyad reported at least some problem when the other reported none, ranged from 19%-46%, with similar proportions in (F) and (G) for family anxiety, support and the three items assessing practical matters. However, for three of the five items reflecting patient physical/emotional states (pain, life worthwhile, and self-worth), the proportion of dyads for which the caregiver reported a problem but the patient did not was at least twice that for which the patient reported a problem but the caregiver did not. This discrepancy was most pronounced with regard to reports of the patient having a problem with feelings of self-worth, with 37% of caregivers reporting a problem the patient did not, compared to 9% of patients reporting a problem the caregiver did not.
This analysis was conducted to evaluate the potential effectiveness of relying on informal caregivers as proxy respondents for patients with late-stage HIV/AIDS. Caregiver reports were compared to patient “gold standard” ratings on a variety of items on the Palliative Care Outcome Scale (POS). First, the level of chance-corrected agreement, correlations between patient and caregiver ratings, and proportions of agreement were assessed. Results of these analyses suggest little support for the hypothesis that greater patient-caregiver agreement would be reported for items assessing attributes of the caregiving experience that were visible and observable than that reported for items assessing more subjective or internal experiences of the patient or the caregiver. In fact, among the HIV patients and their caregivers in our sample, “moderate” or “substantial” levels of chance-corrected agreement were reported only for items assessing the patient’s experience of physical and emotional symptomatology. These are states that have generally been viewed as difficult to be assessed by proxy. Perhaps the more substantial agreement found on these items in this study is unique to the HIV/AIDS care setting, which developed within the palliative care and palliative/hospice model of care for patients with a serious and incurable illness. Such care focuses on control or remediation of not only pain and other physical symptoms but also emotional, social, psychological, and spiritual problems (30). It is possible that these states may be more salient for HIV/AIDS patients and they may talk about them more with their caregivers (20). It bears noting, however, that, as evidenced by the distributions of these items in Table 4, there is still substantial disagreement about the presence or absence of even these attributes.
On the other hand, only “slight” agreement was reported for two of the practical matters assessed, including receipt of information, purportedly one of the more observable items. This may be due to the fact that the information item on the POS was not solely about the patient’s state, but also included the family’s needs. Therefore, the information item had an element of self-report for both patients and caregivers and does not exactly parallel the other proxy items in which the caregiver was reporting solely on the patient’s state.
The data presented also provide mixed support for our second hypothesis that there would be few systematic differences at the aggregate level. A statistically significant difference was noted between mean patient and mean caregiver ratings for one item -- patient self-worth. On average, caregivers tended to rate patient self-worth lower than patients rated their own self-worth. In fact, 37% of caregivers stated that the patient had some or severe trouble with self-worth when the patient reported no trouble at all. This finding seems congruent with earlier findings in the cancer literature asserting that caregivers systematically rate patients as having lower QOL (1) than do patients themselves. It is possible that the caregivers in our study had a difficult time imagining their seriously ill relative or friend being able to maintain a sense of self-worth in the context of advanced disease. It is also possible that the patients in our study gave a socially desirable response stating they felt good about themselves.
While aggregate analyses of concordance provided some indication of agreement between patient and caregiver assessments, analyses of dyads with regard to the types of disagreements reported suggest ways in which the use of proxy data might impact provision of care and services to seriously ill HIV/AIDS patients (assuming the patient report of his/her own physical states is more accurate). On the one hand, evidence of a substantial proportion of instances in which the caregiver reports a problem but the patient does not suggests that reliance on proxy data may result in the provision of unnecessary care or services. Consistent with findings from previous studies of cancer patients and their caregivers (26), the caregivers of HIV/AIDS patients described in this report tended to assess the patient as being in poorer health and experiencing greater problems with physical symptoms than did the patient himself. For over a quarter of the sample, caregivers reported a problem with pain, the patient’s sense of self-worth and/or the receipt of adequate, understandable information when the patient reported no such problem. On the other hand, the finding that in those instances where the patient and caregiver agreed on the presence of a problem but not it’s severity, the patient who was more likely to describe the problem as “severe” suggests reliance on proxy data might result in insufficient provision of care or services among those who need it most.
Limitations to this research include the fact that the patient-caregiver sample was a convenience sample and accrued from the patient population of one medical center. However, although the sample is not randomly selected, we believe that it captures many of the characteristics and experiences of terminally ill HIV/AIDS patients and their caregivers in an urban healthcare setting. Furthermore, as with findings of earlier studies investigating the accuracy of caregiver reports as proxies for patient self-reports in other patient populations, the findings presented here may be biased since they necessarily rely on data from patient-caregiver dyads in which the patient was still able to provide an assessment of his or her situation (35). In other words, the accuracy of family caregivers as proxies for those patients for whom reliance on proxy decision-making is most likely mandatory, remains unknown. This methodology is particularly problematic with regard to the assessment of more subjective and time-varying states such as pain and perceived QOL. Finally, the high level of anxiety associated with one’s illness or that of a loved one, particularly when the illness is serious, may have biased either patient or caregiver responses. In instances where the caregiver perceived the patient’s illness as terminal, the caregiver’s assessments of the patient’s health status and care may have been biased due to anticipatory grief. In addition, the caregiver’s ability to accurately report on the patient’s state may also impeded by their own distress at the patient’s illness experience.
The analysis of the POS items at the item level and in the aggregate indicates that patient-caregiver agreement is only fair on the majority of items. This suggests that when circumstances allow, information on health care outcomes is best obtained directly from the patient. The findings from this investigation do not support the use of proxy informants as sole sources of information when HIV/AIDS patients are seriously ill. Although there was little evidence of systematic bias, patient and caregiver responses show substantial disagreement. Indeed, reliance on caregivers as proxy informants may contribute to both unmet need and delivery of unnecessary or unwanted care. Future investigations might examine predictors of non-comparability within the dyad which could in form interventions to improve patient and caregiver agreement on important outcome measures. Conflicted dyads may especially benefit from an intervention designed to assist both patients and caregivers in communicating more effectively about health care issues, to help provide more accurate accounts of the patient’s illness experience.
This research is supported by Grant No. RO1-NR008343 from the National Institute of Nursing Research (V.H. Raveis, Principal Investigator).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.