Health-Related Quality-of-Life Assessments and Patient-Physician Communication:  A Randomized Controlled Trial FREE

In recent years, interest has been expressed in the use of health-related quality-of-life (HRQL) assessments in daily clinical practice as an aid to detect physical or psychosocial problems that otherwise might be overlooked, monitor disease and treatment, and improve the delivery of care.^1- 4 Although the literature enumerates the putative benefits of routine assessment of patients' HRQL in clinical practice,^3- 6 relatively few empirical studies have investigated the effect of such a procedure.

The results of feasibility studies are consistent and encouraging. Administration of self-reported HRQL questionnaires in outpatient clinic settings requires only a modest investment in material and personnel and is acceptable to both patients and staff.^7- 11 The randomized studies^7,9,12- 19 that have investigated the value of routine HRQL assessments for patient management and outcomes have yielded less consistent and generally less favorable results. Although several studies have reported improved detection of patients' problems (eg, depression),^14,18- 19 others have not.^12,20 No studies have found a salutary effect on patient satisfaction or health status.

There are several possible explanations for the relative paucity of positive findings. In some studies, the HRQL assessments were made at fixed intervals, regardless of whether they coincided with specific medical visits^12,17 or were made at a single time point, thus precluding the possibility of monitoring changes in patients' HRQL.^{7,9,16,18- 19} None of the studies provided summaries of the patients' responses to the HRQL questionnaires to the patients themselves. Most notable, however, is the lack of attention paid to the rigorous evaluation of the effect of HRQL assessments on patient-physician communication. Although in several of the studies the participating physicians and/or patients reported that the availability of the HRQL data facilitated communication,^9,12- 13,15 such self-reports need to be confirmed by direct behavioral measures.²¹

Given that effective communication is an early step in the care process,²² it seems only logical that patient-physician communication be included as the most proximal outcome in studies that evaluate the effect of standardized HRQL assessments in daily clinical practice. If no effect on communication is found, it is unlikely that any effect on more distal outcomes, such as patient satisfaction or HRQL, will be observed.

The current study was undertaken to investigate the potential value of providing oncologists and their patients with timely, structured feedback on the patients' HRQL during palliative chemotherapy treatment. Our primary hypothesis was that such HRQL assessments would facilitate the frequency with which HRQL issues were discussed during the consultations. Additional outcomes included physicians' awareness of their patients' HRQL, clinical management activities, patients' and physicians' satisfaction with their medical interactions, and patients' self-reported HRQL over time.

The study was conducted at the Antoni van Leeuwenhoek Hospital, a specialized cancer treatment center in Amsterdam, the Netherlands. The physician sample consisted of the oncologists working in the Department of Medical Oncology. The patient sample was a consecutive series of patients who were receiving outpatient palliative chemotherapy. Patients were excluded if they lacked basic proficiency in the Dutch language, were younger than 18 years, or were participating in a concurrent HRQL study. Eligible patients were invited to participate in the study after they had received 2 cycles of chemotherapy. They were provided with both oral and written information about the study prior to obtaining their informed consent. The institutional review board of the hospital approved the study.

The study used a longitudinal, randomized, crossover design. The physicians were initially assigned, at random, to either the intervention or control condition. For each physician, a minimum of 10 consecutive patients were recruited into the study. The first study medical visit served as a baseline assessment for both groups. The intervention was introduced at the second study visit and continued through the fourth study visit.

Midway through the study, the physicians originally assigned to the control arm of the study were switched to the intervention arm and those originally assigned to the intervention arm to the control arm. A second cohort of at least 10 patients per physician was recruited and followed-up in a manner identical to that in the first study period. A buffer period of 2 months was introduced before starting recruitment of the second cohort of patients. This crossover design was deemed the most appropriate means of testing the effect of the intervention, given the relatively limited number of participating physicians. With each physician serving as his/her own control, the potential influence of between-physician differences in sociodemographics, professional experience, and attitudes and behavior on key study outcomes could be neutralized. The sample size was established to provide 80% power to detect a moderate between-group difference (effect size = 0.40) in the frequency with which patients' problems were discussed during the medical consultations (α = .05).

All patients had standard follow-up visits with their physician before the start of each cycle of chemotherapy. Patients in the intervention arm were asked to complete a standardized HRQL questionnaire, the 30-item European Organization for Research and Treatment of Cancer, Quality of Life Questionnaire-Core 30 (QLQ-C30) (version 3.0), in the waiting room immediately before each visit. The QLQ-C30 is organized into 5 functional scales (physical, role, cognitive, emotional, and social), 3 symptom scales (fatigue, pain, and nausea or vomiting), a global HRQL scale, and a number of single items that assess additional symptoms. The questions are posed in terms of the previous week. For most items, a 4-point Likert-type response scale is used (ranging from "not at all" to "very much"). Following standard European Organization for Research on the Treatment of Cancer procedures, all scores were linearly converted to a 0 to 100 scale, with higher scores indicating a higher level of functioning and more severe symptoms.²³

The patients' responses to the QLQ-C30 were optically scanned into a desktop computer, scored, and printed as a graphic summary profile (Figure 1). Copies of the summary were given to the patient and physician immediately before consultation. A copy was also placed in the medical records. At the 2 subsequent outpatient visits, the QLQ-C30 summary included both the patients' current scores and those elicited at the previous visit(s).

Before the start of the intervention period, each physician received a single, half-hour educational session on how to interpret the QLQ-C30 summary scores. Patients in the intervention group received a similar explanation in a pamphlet mailed to their home. If desired, a research assistant provided further explanation of the summary. No specific guidelines were provided for the use of the summary during the medical consultations.

Patients' Sociodemographic and Clinical Characteristics. At the time of the baseline visit, the patients completed a brief sociodemographic questionnaire, and the physicians rated their patients' performance status using the Eastern Cooperative Oncology Group scale.²⁴ Clinical information was extracted from the medical records.

Patient-Physician Communication. All of the medical consultations were audiotaped, transcribed, and content analyzed with the aid of a checklist to determine whether the HRQL topics included in the QLQ-C30 were discussed and to record the total length of the consultations. Coding was performed directly from audiotape by 3 trained raters (S.B.D. and L.D.V.W.) who were blinded to group assignment. All raters coded a random sample of 15% of the audiotapes to assess interrater reliability. A high level of agreement was reached (mean, 95%). A composite communication score (range, 0-12) was calculated by summing all HRQL-related topics that were discussed. This composite score served as the primary study outcome.

Physicians' Awareness of Patients' HRQL. At the first and fourth study visits, the physicians and patients were asked to complete the Dartmouth Primary Care Cooperative Information Functional Health Assessment (COOP) and the World Organisation Project of National Colleges and Academics (WONCA) charts.²⁵ These charts assess physical fitness, feelings, daily and social activities, pain, and overall health. An additional chart that assessed fatigue was also included. A 5-point response scale is used, with 1 representing the best and 5 representing the worst level of functioning. The charts were used to (1) determine the physicians' awareness of their patients HRQL and (2) identify patients with serious HRQL impairments.

Patient Management. Medical records and the audiotapes were used to abstract notations and comments relating to HRQL-related patient management, including prescription of medication, ordering of tests, referrals to other health care practitioners, and counseling. A composite patient management score was calculated by summing all HRQL-related actions taken by the physicians per patient.

Patient and Physician Satisfaction. Following the first and the fourth study visits, patients completed the 5-item Patient Satisfaction Questionnaire C, which assessed satisfaction with how their needs were addressed, their active involvement during the visit, patient-physician interaction, and information and emotional support received.²⁶ An overall satisfaction score was calculated by averaging the responses to the 5 questions. At the same time points, the physicians were asked to rate their global satisfaction by means of a single question: "How satisfied were you with the communication with your patient during this visit?"

Patients' Self-Reported HRQL. At the first and fourth visits, the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36)^27- 28 was administered to all patients. The SF-36 is organized into 8 multi-item scales. The questions referring to general health perceptions were omitted because they had been shown to be upsetting to patients with advanced cancer.²⁸

Patient and Physician Evaluation of the Intervention. Following the fourth study visit, patients in the intervention group completed a questionnaire and a brief telephone interview regarding their experience with the intervention. Similarly, all of the participating physicians underwent a semistructured interview to obtain their views.

All statistical analyses relating to the effectiveness of the intervention were performed on an intention-to-treat basis and were based on data obtained from the fourth study visit. Patients' sex, the study period (precrossover vs postcrossover), and, when possible, baseline (ie, first visit) values were used as covariates. Stepwise linear regression analysis was used to test for between-group differences in the mean scores on the composite communication scale, and forward, unconditional logistic regression analysis was used to compare the percentage of consultations in the intervention with the control group in which the specific HRQL topics were discussed.

Physicians' awareness of patients' HRQL problems was assessed by calculating the percentage of exact and global agreement between patient and physician ratings for the COOP/WONCA charts. Exact agreement was defined as identical patient and physician ratings, whereas global agreement was defined as agreement within 1 response category in either direction.²⁹ Logistic regression analyses were performed for the total sample and for the subgroup of patients experiencing moderate-to-severe problems (rating of 3 to 5 on the COOP/WONCA charts) at the fourth visit. Additionally, within-group change over time in physician awareness of patients with moderate-to-severe problems was examined. For these latter analyses, no formal statistical testing could be performed because the patients with moderate-to-severe problems at the fourth visit were not necessarily the same patients as those at the first visit. Rather, a within-group shift over time of at least 10% of cases in which agreement was taken as an indication of improved physician recognition of patients' problems.

Stepwise linear regression analysis was used to test for between-group differences on the measures of patient management, patient and physician satisfaction, and patients' self-reported HRQL as measured by the SF-36. Additionally, between-group comparisons were made of the percentage of patients whose SF-36 scores improved by at least 0.5 SD unit from the first to the fourth visit. All statistical tests were 2-sided, with the significance level set at P = .05. All analyses were performed using SPSS version 10.1 (SPSS Inc, Chicago, Ill).

Of the 12 physicians asked to participate in the study, 10 agreed to do so. The 2 who declined objected to having their consultations audiotaped. Four of the participating physicians were women. Their mean age was 44 years (range, 35-53 years), with an average of 11 years of working experience in oncology (range, 2-24 years).

Between June 1996 and June 1998, 382 patients were invited to participate in the study, of whom 273 agreed (71% response rate). Of the 109 nonrespondents, 50 declined to participate because of poor physical or emotional condition, 43 indicated insufficient interest or lack of time, and 16 had objections to the audiotaping. A nonrespondent analysis indicated that patients who declined to participate were significantly less well educated than those who agreed to participate (P<.001). No statistically significant differences were found for other background characteristics. During the study, 31 patients in the intervention group and 28 in the control group were lost to follow-up: 33% died, 30% changed physicians, and 37% transferred to another hospital. There were no statistically significant differences between groups in reasons for loss to follow-up (Figure 2).

The background characteristics of the remaining 214 patients are presented in Table 1. The intervention and control groups were well-balanced on variables except primary diagnosis, with the control group having proportionally more breast cancer patients than the intervention group (62% vs 41%, P = .03).

The mean (SD) composite communication score at the fourth visit was 4.5 (2.3) in the intervention group and 3.7 (1.9) in the control group (P = .01; effect size = .38). Ten of the 12 HRQL issues were discussed more frequently in the intervention group compared with the control group. These differences reached statistically significant levels for social functioning, fatigue, and dyspnea (Table 2).

There were no statistically significant between-group differences at the fourth visit in exact or global physician-patient agreement in ratings on the COOP/WONCA charts. Similar results were obtained when limiting the analysis to those patients who reported moderate-to-severe problems on one or more of the COOP/WONCA charts. In this later case, the only statistically significant group difference was in ratings of social functioning (83% agreement in the intervention group vs 65% in the control group; P = .05). However, within the intervention group, an increase over time of at least 10% in physician recognition of moderate-to-severe problems was observed for daily activities, feelings, social activities, pain, and fatigue. Within the control group, this was the case only for daily activities and pain (Table 3).

The between-group difference in the mean number of HRQL-related patient management actions taken per patient was negligible (0.6 for the intervention group and 0.5 for the control group). No statistically significant between-group differences were observed in the prescription of medications, ordering of tests, or referrals to other health care practitioners. However, a significantly greater percentage of patients in the intervention group (23%) received counseling from their physician on how to manage their health problems than the control group (16%) (P = .05).

Overall patient satisfaction with the fourth medical visit was high in both groups. At the individual item level, the only statistically significant between-group difference was with regard to the degree of emotional support received (mean [SD], 4.3 [.72] vs 4.0 [.89] for the intervention and control groups, respectively; P = .05). The level of physician satisfaction with the medical encounter was similarly high in the intervention and control conditions (mean score, 4.5 in both groups).

There were no statistically significant between-group differences at the fourth visit for any of the SF-36 scales (Table 4). However, a significantly greater percentage of patients in the intervention group compared with the control group exhibited improvement over time (defined as a 0.5-SD unit or greater change) in mental health (43% vs 30%; P = .04) and role functioning (22% vs 11%; P = .05).

No statistically significant between-group difference was found in the mean duration of the visits (mean [SD], 19.8 [6.2] minutes for the intervention group and 20.4 [6.8] minutes for the control groups). Ninety-seven percent of the patients in the intervention group reported that the HRQL summary profile provided an accurate picture of their functioning and well-being, 57% reported that it was used explicitly during the medical visits, 79% believed that the summary enhanced their physician's awareness of their health problems, and 87% believed that it would be useful to introduce the intervention as a standard part of the outpatient clinic procedure. Approximately 25% of the patients indicated that they had discussed the HRQL summary profile with family members or close friends, and 6% had shared it with their family physician.

All physicians reported that the summary profile provided a useful, overall impression of their patients' symptom experience and functional health and indicated that it facilitated communication, especially with regard to psychosocial topics and "unexpected" symptoms (eg, sleep disturbances). Although all of the physicians indicated that they would like to continue use of the HRQL summary profile in their daily practice, several suggested that the information could be more tailored to the problems of specific patient groups (eg, site of pain and use of pain medication for patients with bone metastases).

As hypothesized, the intervention resulted in a significant increase in the frequency with which HRQL issues were discussed. The observed, salutary effect of the intervention on patient-physician communication is particularly encouraging in that the most notable increase was in the discussion of HRQL issues that are less observable (ie, social functioning) or are of a more diffuse and long-term nature (ie, fatigue) and thus are often left unaddressed by health care practitioners.^29- 30

No statistically significant between-group differences were observed in physicians' awareness of the level of functioning or symptom experience of their patients. However, a series of within-group analyses indicated that, over time, physicians in the intervention group improved in their recognition of patients' with moderate or severe problems in 5 of the 7 HRQL domains covered by the COOP and WONCA. Again, it is noteworthy that substantial improvement was observed in the recognition of problems in areas that tend to be underestimated by health care practitioners (eg, emotional and social health, fatigue, and pain).

The intervention had only a modest effect on patient management activities, primarily in terms of increased levels of patient counseling. The only statistically significant difference between the intervention and control groups in patient satisfaction was in the perceived emotional support received from the physician. This is not trivial in that previous research has indicated that patients are significantly more likely to disclose information about their psychosocial functioning when their physician exhibits such supportive behavior.³¹ The failure to detect significant between-group differences in other aspects of patient satisfaction and in physician satisfaction may be because the levels of satisfaction tended to be high, leaving little room for improvement. Such a "ceiling" effect has been reported in other studies.^8,13,32

Most patients exposed to the intervention and all of the physicians reported that the HRQL summary profile was useful in facilitating communication and in enhancing physician awareness of patients' problems and favored continued use of the intervention as a standard part of the outpatient clinic procedure. An additional, unanticipated finding was that approximately 25% of the patients shared the results of the summary profile with family members or with their primary care physician. We consider this to be a positive spin-off, in that previous studies have shown that patients' partners and formal caregivers are often less than fully informed about the patients' HRQL³³ and that problem-oriented feedback (eg, letters and lists) from medical specialists can be a useful means of keeping primary care physicians informed about their patients' health status.^34- 36

The intervention had only a modest impact on patients' self-reported HRQL over time as assessed by the SF-36. This was not unanticipated, given the complex array of factors that can affect patients' health. Nevertheless, the finding that significantly more patients in the intervention than the control group improved by more than half an SD unit on the 2 SF-36 indicators of emotional health is encouraging.

We are cognizant of a number of limitations of our study. First, the results need to be interpreted with some caution because of the relatively large number of statistical tests performed. When correcting for multiple testing using the Hommel procedure,³⁷ the between-group differences noted in the percentage of consultations in which specific HRQL topics were discussed no longer reached conventional levels of statistical significance. Nevertheless, statistically significant between-group differences were observed for the mean composite communication score and 10 of the 12 specific HRQL topics were discussed with greater frequency (in absolute terms) in the intervention than in the control group. These 2 findings suggest that the intervention had the intended, salutary effect on patient-physician communication.

Second, although the patient sample was large, the physician sample was limited. However, we have no reason to believe that the results are in any way atypical of what would be found with a larger group of physicians in that the participating physicians varied substantially in age, sex, and years of work experience.

Third, although the use of a crossover in the study design enabled us to largely neutralize any effect that might be attributed to physicians' background characteristics, it also carried with it the risk of a carryover or contamination effect. In fact, some evidence was found suggesting that the physicians who began in the experimental condition and subsequently crossed over to the control condition may have been sensitized to HRQL issues and may have changed their behavior during the period in which they were no longer exposed explicitly to the intervention. This can be illustrated by the pattern of results pertaining to the discussion of patients' emotional functioning (Figure 3). In the absence of a carryover effect, one would expect that, both before and after the crossover, the percentage of baseline consultations in which patients' emotional functioning was discussed would be similar for the groups that included the same physicians. This was, in fact, the case for those physicians who were first in the control group (36%) and then crossed over to the intervention (41%). However, the percentage of baseline consultations in which emotional functioning was discussed was considerably higher after crossover when physicians were first exposed to the intervention (22%) and then crossed over to the control group (39%) (P = .06). A similar pattern of results was observed when exploring other HRQL outcomes. Importantly, however, any such carryover effect would tend to have a conservative effect, because it would tend to suppress between-group differences in the principal study outcomes.

Finally, the study was conducted in a single hospital that specialized in the treatment of patients with cancer. Future studies are needed to evaluate the efficacy of the intervention in other treatment settings and with larger and more diverse physician and patient samples.

In conclusion, our results support the use of standardized HRQL assessments in the palliative cancer treatment setting as a means of facilitating the discussion of HRQL issues and of heightening physicians' awareness of their patients' problems. Future efforts should be directed at improving the flexibility and precision of HRQL assessments and at linking patients' HRQL (change) scores to specific treatment and care strategies.