The Rotterdam Elderly Pain Observation Scale (REPOS) is reliable and valid for non-communicative end-of-life patients

Several studies reported that 45% to 70% of patients with incurable cancer, either admitted to a hospice or staying elsewhere, suffer moderate to severe pain [1‐3]. Forty-five per cent of those patients receive opioids, which are the strongest analgesics [2, 4]. Pain levels should be accurately assessed to guide pain treatment.

Patients’ self-report of pain is considered the “gold standard” for pain assessment [5]. However, in the terminal phase of life, patients may not be able to self-report pain. To illustrate this, 68% to 83% of patients had cognitive failure [6, 7] and 90% to 98% of patients were drowsy or unresponsive prior to death [8‐10]. Proxy pain assessment by a nurse was needed, therefore, in 90% of patients in a palliative care unit at the day of death [9]. Assessment of suffering, including pain, is challenging anyway in the terminal phase, especially in sedated patients [11].

Proxy assessment often underestimates the pain level, with consequent risk of under treatment [12, 13]. The risk might be lower if attention is paid to well-defined behaviour indicative for pain. Therefore, application of an observation scale that includes such behaviours (and has been validated for the target group) could be more beneficial for non-communicative patients.

We previously developed the Rotterdam Elderly Pain Observation Scale (REPOS, see Additional files 1, 2 and 3) in a study among nursing home residents with speech limitations caused by various disorders. The REPOS is available in Dutch and English [14, 15]. Dutch palliative guidelines recommended the REPOS for specific non-communicative patient groups [16], such as persons with dementia and intellectual disability. It proved a valid tool to measure pain in the nursing home population, including those who could not communicate [14]. However, the REPOS was not yet validated for palliative patients and needed to be re-validated for this population.

Beside the REPOS several other observational pain assessment tools have been developed for specific non-communicative patients groups [17‐19], but to our best of knowledge only one of these has been validated for end-of-life patients, the Multidimensional Objective Pain Assessment Tool (MOPAT) [20]. This tool, published in 2011, was developed for hospice patients who are unable to self-report pain. It was tested in a small sample of 28 alert patients and 30 non-communicative patients and showed good internal consistency and sensitivity to change after a pain-reducing intervention [20]. A disadvantage of the MOPAT, however, is that it includes blood pressure and heart rate measurements, which are often stopped at the end of life, as recommended in the Liverpool Care Pathway for the dying patients [21].

In the Netherlands, the REPOS is increasingly adopted in nursing homes and institutions for intellectually disabled or non-communicative patients [15]. Hospice patients, however, may have other characteristics. They often suffer from advanced cancer (> 90% of patients) and are mostly bedridden. Self-report is not possible due to their illness state (comatose, delirium or adverse effects of medication), in contrast to nursing home patients who more often have dementia. In addition, end-of-life patients are in another emotional state and may be extremely anxious, facing death. All these aspects may influence experiences or expressions of pain [17].

The findings from this study show that the REPOS has promising psychometric properties to assess pain in non-communicative end-of-life patients; including adequate internal consistency, sufficient concurrent validity and good sensitivity to change after a pain-reducing intervention.

A variety of observational pain scales have been developed for other settings where non-communicative patients are treated, including intensive care units [35] and nursing homes [19]. Only one, the MOPAT [20], has been validated for non-communicative end-of-life patients, albeit preliminary and without establishing a cut-off score. The MOPAT was published after start of our study (2010). We therefore could not use it, although it would have been interesting to compare the two scales.

The overall internal consistency of the REPOS in the present study was adequate as judged from the Cronbach’s alpha coefficient of 0.73 [31, 32]. Although the item-total correlations of 4 items (panicky, fearful look, moaning and moving body parts) were below 0.30, we chose to retain those items. These low values could be related to infrequent occurrence of those behaviours in terminally ill patients as they receive high doses of strong analgesics and/or sedatives. For example, the ability to fully react with all body parts is often diminished in the end-of-life stage. In other non-communicative patients who use lower doses analgesics those behaviours might be more obvious. In observations of nursing home residents during a possible painful situation, i.e. washing and clothing or during physiotherapy, when manipulation irritates already affected tissue (i.e. arthritis) all 10 behaviours were seen. Deletion of those 4 items could therefore create a risk of underestimating pain in a broader group of non-communicative patients. The major reason for retaining these items is that the scale should be applicable in other settings with non-communicative patients as well.

With respect to concurrent validity, a high correlation (0.73) was found – not surprisingly – between REPOS and NRS scores assigned by the same person. The correlation between REPOS and the gold standard (NRS-patient) was only moderate [32], as is seen in other pain scale studies too [36, 37]. This moderate correlation is perhaps explained by patient characteristics. Patients who are unable to report pain with the NRS seem to show more nonverbal behaviour, such as grimace [36, 38]. For these patients their pain would be better reflected by a behavioural score than by a proxy NRS only.

It is not unexpected that the REPOS score and the NRS-proxy score may differ to some extent. A high REPOS score combined with a lower NRS-proxy score, or a so-called false positive score, is typically seen in patients who show “emotional” behaviour not related to pain, but based on anger, fear or agitation [39, 40]. The opposite, a false-negative score, may occur when the attending nurse has observed behaviour not included in the REPOS score, such as muscle tension. Alternatively, the nurse’s NRS score reflects knowledge of relevant characteristics, such as history or medication use, illness and other patient specific characteristics [12, 13, 41].

This study showed that the cut-off score of 3 or higher is applicable for non-communicative end-of-life patients. However, application of a ‘one-fits-all’ cut-off score is debated [42, 43]. A reason suggested in literature is that different underlying conditions cause different types of damage to the brain, and consequently different responsiveness to pain [43‐45]. Based on these arguments one could consider the use of an individualized cut-off score, which has been recommended for other vulnerable patients groups, e.g. young children [46]. However, this approach asks more from the caregivers: it is a dynamic approach requiring evaluation and adjustment at regular times and when indicated. As daily pain assessment itself was shown to be problematic [47], one can wonder if an individualized approach is feasible in a daily care situation [48, 49].

A strength of the present study is that most observations were done within the last two weeks of life and therefore including even those patients near the time of death. In the previous validation study of the REPOS [14], only a small proportion of the population was at the end of life. In addition, communicative patients rated their pain themselves, which enabled comparison between the REPOS and the gold standard of self-report. Also, the sample size in the present study was 100 patients, which far exceeds the minimal number of 50 patients [31, 50].

Some limitations of this study have to be addressed, however. First, this is a single-centre study in mainly advanced cancer patients and care must be exercised in extrapolating the findings to other settings and patients populations, like palliative home care or community-based palliative care. Nevertheless, it seems unlikely that pain behaviour would be different in a different palliative environment. Second, the sensitivity-to-change analysis concerned only a relatively small sample. This limitation is encountered in many other psychometric studies, seeing that researchers often are not available when patients receive additional analgesia and also because nurses may tend to forego reassessment after a pain-reducing intervention [51]. The comparable MOPAT study [20] also included fewer patients than planned, mainly due to logistical reasons. The fact that the observer knew whether the patient received pain medication or not, could be considered a weakness of the sensitivity-to-change analysis. In the ideal situation, observers blinded for this condition apply the REPOS when watching video recordings made before and after an intervention. However, it was felt undesirable to ask relatives’ approval to film their loved ones in the dying phase for research purposes. Still, knowing that the REPOS is sensitive enough to measure small changes after an intervention means that it is suitable for pharmacodynamic studies, which are urgently needed in this palliative patient group [11]. Lastly, in the present study the REPOS was compared with NRS scores, although a comparison with another behaviour pain scale would have strengthened the reliability and validity testing [52]. In future studies the outcomes of different observational scales like the PACSLAC, PAINAD and MOPAT should be compared and the possibility should be explored if blinded observation is possible.

In conclusion, the REPOS seems to meet the criteria for the use of a pain measurement tool in palliative care of the Expert Working Group of the European Association of Palliative Care (EAPC) [53]. That is, the brevity of the scale (10 well-defined behaviors, scored yes/no after a two-minute observation period) and the cut-off score increases clinical utility. The scale was validated for nursing home residents with speech limitations and the present study increases the psychometric knowledge about sensitivity to treatment effect and reliability during palliative care. The fact that the REPOS is available in Dutch as well as English can be seen as an advantage as well. In addition, next to the Dutch palliative guidelines [16], the use of the REPOS for pain assessment in non-communicative patients is recommended in a report on quality indicators in palliative care, published by the Netherlands Institute for Health Services Research [54]. We have demonstrated that the REPOS seems a valid tool for the assessment of pain in non-communicative end-of-life patients. We recommend its use on a daily basis for every non-communicative palliative patient. After a brief training course every professional palliative caregiver will be able to use it in daily practice. A REPOS electronic educational module is available (both in Dutch or English) to guide implementation and training [15].