How can good compliance be attained and maintained

The QL data are both multi-dimensional and longitudinal which makes analysis difficult, but the problem is greatly increased if compliance is poor and there is much missing data. When data are 'missing at random,' methods of imputation are possible, but most QL data will be 'not missing at random,' usually due to patients failing to complete forms due to their deteriorating health. Such data cannot normally be imputed, and so every effort should be made to collect as much data as possible.

In QL assessment the strictest definition of compliance is the number of patients who have completed every question of a QL questionnaire at the required timepoint, divided by the number who could have (i.e. were alive at that timepoint). In addition, the definition of 'at the required timepoint' is open to interpretation and most groups now give a time window around the timepoint (e.g. ±7 days) which is considered acceptable. Generally the chosen 'width' of the window will be a reflection of the frequency of form administration, thus if questionnaires are being completed at 3-weekly intervals a window of ±7 days maybe appropriate, whereas 6-monthly forms might have a window of ±2 months. Sometimes an asymmetrical window might be used, for instance at baseline where completion should be before randomization, here acceptable times may be —7 to 0 days from randomization.

It is also important to differentiate between questionnaire compliance and individual question (item) compliance (the proportion of items completed on a returned questionnaire).

Questionnaire compliance

Despite differing definitions of questionnaire compliance, many groups have reported similar and disappointingly low levels of compliance in cancer trials. Overall rates are often quoted as being around 70 per cent, but this usually reflects a higher rate at baseline (around 85 per cent) which falls consistently over time. Indeed, long-term compliance rates, at two or more years, can often be as low as 25 per cent [14,76-79].

When good compliance rates have been achieved it has usually been as a result of a well-planned administrative operation. Thus:

♦ Earl et al. [11] reported an overall compliance rate of 87 per cent in the use of daily diary cards (438 cards completed out of a possible 506) and this was attributed to the fact that QL was carried out in only one centre with a research nurse assigned solely for this purpose.

♦ Langendijk et al. [80] attributed their 90 per cent compliance to the appointment of a specific individual (not the treating clinician) to collect QL data, and to limiting the frequency of administration.

♦ Wisloff etal. [81] achieved an overall compliance rate of over 83 per cent in their trial of patients with multiple myeloma using the EORTC QLQ-C30. They attribute this to the fact that patients confirmed their willingness to take part by mailing the completed baseline form to the study secretariat and that all subsequent communication relating to QL took place between the patient and the secretariat.

Some of the best rates of QL compliance have been reported by the National Cancer Institute of Canada who reported a compliance rate of 95 per cent in three breast cancer trials [82]. However they acknowledge that it remains unclear whether similar success can be obtained with different questionnaires, in different types of trials, in different institutions, and during long-term follow-up. To achieve this impressive level, the NCIC used a level of resource that may not be available for all trials. In one study nurses called the patients at home on the appropriate day to remind them to complete the questionnaire and specific measures were instituted prior to trial activation to ensure maximum compliance in the completion and return of QL questionnaires. These included:

♦ QL assessment was incorporated as a specific trial objective,

♦ rationale for QL data collection was included in the protocol,

♦ specific instructions for administration of questionnaires were included in the protocol,

♦ data collection forms were modified to remind data managers to administer questionnaires,

♦ specific reporting schedules were provided,

♦ successful completion of the baseline QL questionnaire was defined as an eligibility criterion,

♦ computer-based reminders were provided in advance of due dates,

♦ pre-trial workshops for nurses and data managers were implemented,

♦ participating centres were given regular feedback on QL via letters and newsletters.

There is no doubt that given careful planning and provided adequate resources are available it is possible to achieve high compliance rates, although of course the patient group may affect what is possible.

The central trials office needs to be effective, efficient and open to new ideas to improve compliance. Ganz et al. [83] have suggested the use of scannable questionnaires to facilitate the rapid and accurate transference of data onto the computer database, and increased use of flow sheets, study calendars, patient tracking cards, training videos and identification or flagging of patient medical records can all help with compliance [84].

Item compliance

Item compliance is generally much less of a problem, although there are some consistent problems surrounding questions about sexual problems, and sometimes poor questionnaire design means that whole pages are missed because patients did not appreciate there were further questions on the back of the form or on another sheet. This emphasises the need for completed questionnaires to be checked before being returned to the data centre.

Patient compliance

Although several groups have noted that poor compliance in multicentre trials is more often related to a lack of commitment in particular centres, poor patient performance status can have a major effect.

In a series of MRC lung cancer trials, QL questionnaires were completed by 92 per cent of the patients whom the clinician assessed as having good performance status, but this fell dramatically to only 31 per cent of those assessed as very poor.

The problem with poor compliance, however, is that its cause is not always obvious. Cox et al. [85] suggested that as well as the very sick patients not completing forms, perhaps surprisingly 'those experiencing fewer problems may not be so diligent in returning questionnaires.' Indeed Brorsson et al. [86] telephoned non-responders to a mailed-out questionnaire and reported that the vast majority reported considerably better QL than the responders.

Poor compliance in the completion of forms may also result from patients' failure to attend at designated times and patients being perceived as too ill or too distressed to be approached. However, frank refusal is uncommon. It is also of interest that patients can feel very upset if requests for questionnaires to be completed are discontinued. The reason for non-compliance is less likely to be patient refusal, and more likely to be staff reticence about handing questionnaires to patients they perceive to be too sick to complete them.

Centre compliance

Organizational problems within institutions may account for a large part of the failure to collect data. These include:

♦ unavailability of forms,

♦ priority given to other matters,

♦ failure to check, collect and return forms,

♦ staff changes and absences.

A number of studies have investigated the administration of questionnaires. Hopwood et al. [87] reported on a survey of centres participating in three trials conducted by the MRC to assess the administration of QL questionnaires. The aim was to see what lessons could be learned regarding the standardization of procedures and improving compliance. Interesting findings included the facts that:

♦ The baseline questionnaire tended to be given out after and the consultation with the doctor and randomization (and therefore after the patient knew their treatment allocation) but before treatment. Thereafter there was no consistency as to the timing of administration in relation to the consultation with the doctor.

♦ Forms were often not given out when the patient was perceived as too ill or unsuitable because they had just received bad news. However, the majority of centres reported that patients' reactions to completing questionnaires were favourable, and that patients took them seriously and were impressed with the interest being shown in how they felt. Only one centre mentioned patient refusal as a problem.

♦ In seventeen of the thrity-six centres surveyed there had been changes in procedures or staff during the course of the trial. The most frequently mentioned staff-related problems were: staff being on holiday or not available for other reasons, staff being unfamiliar with procedures, poor compliance with trial procedures in peripheral clinics, unavailability of questionnaires, and staff forgetting to hand out, check, or collect the questionnaires.

♦ Many centres were affected by the lack of privacy and quietness for patients completing their forms.

One of the trials had specific funding for dedicated local trials staff, the others had not. The only differences between the specific-funded trial and non-funded trials appeared to be that a variety of staff (doctors, nurses, or radiographers) administered the questionnaires in the non-funded trials, whereas only research nurses did so in the funded trial. Surprisingly, in all other aspects the extra training and resourcing in the funded trial appeared to have no impact on the organizational problems.

Although it is often difficult to find out why QL forms are not completed and returned, questions about QL completion included on the clinical forms can be helpful, especially if the reasons are made explicit, i.e. not just 'patient refused' but the reason why they refused.

0 0

Post a comment