Effect of different forms of information produced for cancer patients on their use of the information, social support, and anxiety: randomised trial

Participants

Location and timing of the study—Patients were recruited at the Beatson Oncology Centre between November 2001 and March 2002. The centre provides specialised non-surgical cancer treatment for patients throughout western Scotland. Recruitment started at the main site, but logistical problems resulted in the study moving to the newly opened Tom Wheldon Centre (a “satellite” centre of the Beatson Oncology Centre at Gartnavel Hospital) recruiting patients there between February 2002 and March 2003.

Sample size—The target sample was 400 patients, giving 80% power to find a difference of 40-27% in improved anxiety between (for example) 200 given personalised information and 200 given general information only. (This was based on conservative estimates of changes in anxiety from our previous work but not taking into account possible interactions between factors).

Patient selection—A total of 874 patients identified from outpatient appointment diaries as starting radiotherapy treatment for breast, prostate, cervical, or laryngeal cancer were registered with the study, and their medical records were reviewed (figure). Of these, 604 patients were sent a letter explaining the study, and 270 were excluded. Reasons for exclusion included receiving palliative care, severe pain or symptoms causing distress, having cancer at other sites, having no spoken English, receiving treatment for psychological or psychiatric problems, visual or mental handicap, and case notes being unavailable, ambiguous, or illegible.

Patient flow through study

• Download figure
• Open in new tab
• Download powerpoint

Recruitment

On each patient's attendance at the hospital a researcher (JP) further explained the study and sought written consent. She excluded 41 (6.8%) patients at recruitment—20 who did not give informed consent, 14 who were “missed” (that is, not contacted when they attended) for logistical reasons, and seven who were excluded at the end of the study when the target sample size was reached.

Patients refusing—Of the 563 patients invited to take part, 400 (71%) consented. Of the 163 who refused, 44 gave no reason, 25 said they were not interested or could not be bothered, 21 said they did not want any information, 16 said they had enough information, 18 were in a hurry or had someone waiting, 13 were feeling ill or could not hear well, 12 did not like filling forms, nine were already in a trial and did not want to take part in another, and five gave other reasons. The 29% of patients who refused to take part were more likely to live in deprived areas (36% of patients in deprivation categories 5-7 refused compared with 25% of those in categories 1-4;χ² = 8.5; 1df; P = 0.003).

Patients recruited—The 400 patients recruited comprised 275 (69%) women and 125 men with ages ranging from 28 to 82 years with mean age 59 (median 61). Of these, 348 completed follow-up. Patients had had their cancer diagnosed between five and 312 weeks before recruitment. Two thirds (262, 68%) of the patients had breast cancer, and just under a third (118, 31%) had prostate cancer. One woman had cervical cancer, and two men and one woman had laryngeal cancer. The breast cancer patients had been diagnosed for a shorter time on average than had the prostate cancer patients (19 weeks v 43 weeks, P < 0.001).

Recruitment data—At recruitment, patients were asked to complete a questionnaire at home (see appendix 5 on bmj.com for details) and to bring it to their next appointment. This included questions on their previous computer use, information preferences5 6 and coping style,7 household, and confidant (the person they were most likely to discuss their cancer with); Helgeson's social support questionnaire (HSSQ),12 and the hospital anxiety and depression scale (HADS).13

Randomisation

All patients were offered booklets based on information from the website of CancerBACUP, a cancer information and support service for patients (www.cancerbacup.org.uk/). Before recruitment, we prepared 400 randomisation sealed envelopes in 10 blocks of 40 using random number tables. We randomised patients to eight groups defined by the three binary factors under study: (a) half received personalised information that included data from their medical records, whereas half had only general information from CancerBACUP for their cancer; (b) half chose information interactively by selecting it with a computer at the oncology centre, and half had a larger volume of material in booklets that were produced automatically; and (c) half had additional anxiety management advice, and half did not. An error in the randomisation envelopes resulted in the misallocation of one patient, with one of the eight groups having 51 patients and another only 49 (see figre).

Interventions

Choices made by patients in our previous study11 guided the information we offered. The patients given booklets that were produced automatically (containing 40-47 sections) did not use the oncology centre computer. Patients provided with general information only received the booklets Understanding Radio-therapy, Diet and the Cancer Patient, and the appropriate cancer-specific booklet (such as Understanding Breast Cancer). Patients provided with general information and who selected information interactively could choose sections from the above three booklets and from three further CancerBACUP booklets (Cancer and Complementary Therapies, Feeling Better Controlling Pain, and Sexuality and Cancer). These six booklets had a total of 78 sections, and the patients were allowed to choose up to 10 sections from a menu. This relatively limited choice was decided partly for pragmatic reasons (such as time available for users to select topics) but mainly in order to test the difference between a small but carefully selected subset of information and a larger more general document. Patients given anxiety management advice had an extra few pages with self help advice based on work in cognitive behaviour therapy for anxiety.

The patients allocated personalised information that was produced automatically received selected information from the three general booklets plus information from their medical records. The less relevant sections in the three booklets (determined by patient's cancer type, treatment, age, and sex) were omitted, so that patients received slightly shorter booklets than did the patients given general information. Patients allocated personalised information that they chose interactively could select topics from their medical record such as “problem list,” “treatment list,” or “your cancer.” For the patients who chose information interactively, we recorded the sections chosen, whether they required help with the computer, whether they used the computer mouse or the touch screen. We also noted those patients who did not wish to take a booklet. (See appendix 6 on bmj.com for details of the intervention.)

Follow-up data

Patients were sent follow-up questionnaires three months later. The questionnaires included Helgeson's social support questionnaire, the hospital anxiety and depression scale, and questions about the patients' use and opinions of the booklets and their reported understanding of cancer. We sent reminders at three and six weeks later to non-responders.

Statistical analysis

Of the 400 patients we recruited (figure), 384 (96%) completed the intervention and 348 (87%) returned the follow-up questionnaire. Completed data for the hospital anxiety and depression scale were available for 329 (82%) patients. Here we present the results, based on an intention to treat analysis, for the 325 patients with cancer of the breast or prostate, excluding the four patients with cervical and laryngeal cancer for clarity. (Appendix 2 on bmj.com compares the 325 patients included with the 75 excluded from analysis, and appendix 4 presents analysis for all 400 patients, with estimated values for those not completing follow-up.)

We compared both the differences in the anxiety and depression scores at recruitment and at follow-up and in the proportions of patients with scores ≥ 8 (cases or probable cases of anxiety or depression). Patients answered questions about the booklets on Likert scales, and we grouped these according to the modal answer into binary scores. Helgeson's social support questionnaire produces four scores—instrumental, informational, and emotional support (20 = “best”) and “negative interactions” (50 = “worst”).

Analyses to ensure robustness of results included using difference in scores and in scores grouped by tertiles and quartiles. We examined patients' views and use of the booklets and differences in the social support and the anxiety and depression scores between recruitment and three months by means of general linear models and multiple logistic regression as appropriate. To compensate for the problem of finding statistically significant results by chance because of multiple testing, we considered significant results to be important only if they formed a consistent pattern from various analyses.

Baseline characteristics

At recruitment, there was no difference between the intervention groups in terms of anxiety, depression, social support, age, sex, or length of diagnosis (table 1). Of the patients who answered the questions, 326/375 (87%) were satisfied or very satisfied with the cancer information they had already received, 231/373 (62%) had read at least one CancerBACUP booklet. Only 52/382 (14%) had obtained health information themselves from the internet, but 67 (18%) had been given information from the internet by someone else, and 164 (43%) had never used a computer before.

Among the 325 patients with breast or prostate cancer who completed the study, all four of the social support scales were skewed towards having lots of support—median scores were 17 for informational, 20 for emotional, and 20 for instrumental support, and 13 for “negative interactions” (appendix 8 on bmj.com). A third (99, 30%) had anxiety and 43 (13%) had depression (scores of ≥ 8). General linear models used to predict anxiety and depression scores and multiple logistic regression to predict cases or borderline cases of anxiety or depression produced similar results. The negative interactions measure of social support predicted both anxiety and depression at baseline (appendix 7 on bmj.com).

Interactive selection of information

For the 190 patients who selected their cancer information, the average time spent using the computer (including explanation given by the researcher) was 9 minutes (range 2-30). A third required help in using the computer; two thirds chose to use the touch screen, and a third used the mouse. Of the 82 (43%) patients who had not used a computer before, only two chose to use the mouse. The researcher operated the computer for four people.

On average, patients chose eight sections (range 0-10); there was no difference by intervention or other factors. Patients with breast or prostate cancer were similar in the sections they chose (see box). Sections chosen by both men and women and that were available to those given information produced automatically were “healthy eating,” “side effects,” “causes,” “general tips,” and “why prescribed.” Other popular choices for men were “radiotherapy” and “prostate cancer” and for women were “living with surgery” and “feelings.” The last section, from the booklet Cancer and Complementary Therapies, was not issued to those given information produced automatically. The other two booklets available exclusively to those who selected their information (Feeling Better Controlling Pain and Sexuality and Cancer) were not frequently chosen.

Three month follow-up

Patients' opinions of the booklets and perceived understanding—The booklets produced automatically, which were larger than those produced interactively by patients, were more likely to be found useful and to tell the patient something new and less likely to be seen as too limited, but they were also more likely to overwhelm some patients than the booklets produced interactively (table 2). The booklets with personalised information were more likely than those with only general information to tell the patient something new. The patients given automatically produced booklets had higher overall satisfaction scores than those who produced their booklets interactively (54% v 47% with score > 2; P = 0.02 from general linear models). When asked to rate their current understanding of their cancer, 26 (8%) rated it less than they had done at recruitment, 188 (58%) rated it the same, and 110 (34%) rated it better, but there was no difference by any of the intervention factors. One hundred and thirteen patients (35%) made positive comments about the booklets and 38 (12%) made negative comments, and patients with personalised booklets were more likely to mention the relevance of the information than those given only general information (41% v 15%; χ² = 9.3, 1df; P = 0.002).

Use of the booklets with others—Compared with patients with general information only, patients with personalised information were more likely to show their booklets to their “confidant” (85% v 70%; χ² = 10.1, 1df; P = 0.001), to someone else in the household (32% v 19%; χ² = 6.8, 1df; P = 0.009), and to someone outside the household (33% v 22%; χ² = 4.3, 1df; P = 0.04). There was no difference for the other two intervention factors. Those with personalised information were more likely than those with general information only to think that it helped in discussing their cancer or its treatment (80% v 65%; χ² = 4.2, 1df; P = 0.04).

Changes in social support—Patients' social support scores showed a considerable range of changes from baseline to follow-up. Changes in informational support ranged from - 12 to 12, in emotional support from - 10 to 7, in instrumental support from - 8 to 7, and in “negative interactions” from - 11 to 22. There were some unexpected differences by the intervention factors among patients who had shown their booklets to their confidant (table 3). On the “negative interactions” scale, 42% of patients with personalised information deteriorated, compared with only 24% of those with general information only. Patients who were given anxiety management advice were more likely to have deteriorated on the instrumental support scale than those not given the advice (27% v 13%).

Changes in anxiety and depression—At follow-up, 145 patients (45%) had improved anxiety scores. The percentage improvement ranged from 33% (13/39) in those given automatically produced, personalised information with anxiety management advice to 55% (21/38) in those given automatically produced, general information with no anxiety management advice, but there were no significant differences between the three intervention factors (allowing for interactions between these). There was no difference between mean anxiety scores at baseline (5.9) and follow-up (6.0), nor in the percentage of patients with scores ≥ 8 at baseline (31%) and follow-up (33%). None of the three intervention factors was a significant predictor of change in anxiety in a general linear model (table 4). Patients who had worsening negative interactions had worsening anxiety (P = 0.005). There was no difference between mean depression scores at baseline (3.4) and follow-up (3.4), nor in the percentage of patients with scores ≥ 8 at baseline (14%) and follow-up (13%).

Comparison with previous study

Comparison of the present study (patients recruited in 2001-3) with our previous one (patients recruited 1996-7)11 shows that the catchment population was the same, the rate of refusal to participate was similar (29% v 27%), and the proportion included in analysis of those randomised was similar (81% v 83%). Patients from more deprived areas were more likely not to participate in both studies. Unlike in our previous study, however, we found no overall improvement in patient anxiety in this study. The percentage of patients with anxiety at recruitment was lower (31% v 38%) but did not change. There was no relation between anxiety change and intervention or other factors.

Patients in the present study were generally better informed than five years previously. There had been a major improvement in the hospital environment (with the opening of a new building) and much improved access to information, including increased input from Macmillan nurses in the radiotherapy department. The internet was a source of information for a third of patients in the present study, but its impact should not be overstated; 43% of patients had never used a computer before entering the study. Better information provision might partly explain the lower level of anxiety at baseline, but it cannot explain the absence of the differential effect of personalised versus general information.

Implications of results

One possible explanation for the different result between this and our previous study is that in the previous study all patients were encouraged to write a list of questions to ask their oncolo-gist one week after the intervention.14 Other researchers have shown the possible beneficial effect of prompt sheets and written lists on anxiety.15 Although the use of written lists in our previous study10 was applied to all groups, it may have led to further reduction in anxiety and possibly had a greater effect in those with personalised information. The other explanation is a chance effect in our earlier study.

The single “information intervention” with anxiety management advice in this study did not reduce levels of anxiety. Although this was based on a successful information intervention for patients with generalised anxiety disorder, it had not previously been tested “on its own” with cancer patients. It is perhaps not surprising that such “information only” interventions have limited impact on psychological symptoms. More proactive methods of cognitive behaviour therapy are more likely to be worth while in the management of cancer related anxiety.16

Patients' coping style (according to Miller's classification7) and information preferences did not seem to have a strong effect on the impact of the different types of information. In our patients the type of newspaper read more often predicted reaction than did coping style. Tabloid readers were more likely to think information was too technical, feel overwhelmed, or to have deterioration in perceived informational support.

In this study and our previous one11 four out of five patients showed their booklets to family members, whereas only a third of cancer patients given a patient-held record in a general practice study did so.17 Why there should be such a big difference is not clear.

Conclusions

Our present and previous studies suggest that the claim that patients like personalised information and are more likely to show it to their confidant is robust. However, further research is needed into the effects of sharing information on patients' social support and anxiety.