Variations in concerns reported on the patient concerns inventory in patients with head and neck cancer from different health settings across the world

1 INTRODUCTION

The number of cancer cases across the world increased by one third between 2005 and 2015. The main influences were population growth and increasing age.1 In 2018, there will be over 18 million new cancer cases and nearly 10 million cancer deaths.2 Head and neck cancer (HNC) is a global problem and rates continue to climb with the increase in number of oropharyngeal cancers ascribed to human papillomavirus.3

Patient reported outcome measures (PROM) are an established component of cancer outcomes reporting.4 Publications related to PROM reflect international interest with papers from Africa, Asia, Australia/Oceania, Europe, North America, and South America. Health-related quality of life (HRQOL) assessment is embedded in many clinical trials.5 Over 6000 patients with cancer pooled from randomized controlled trials and using the European Organization for Research and Treatment of Cancer (EORTC) QLQ-C30 showed differences between cancer types and the effects of age on HRQOL.6 In a critical review, Aggarwal et al summarized global radiation therapy research between 2001 and 2015, and reported increasing numbers of HRQOL papers from a range of different countries.7

HRQOL evaluation makes a positive difference in clinical practice 8 and research around the expression of concerns during follow-up consultations is ongoing.9 This approach supports patient-centered interventions. 10 Following HNC treatment, there is a large array of potential issues patients wish to discuss. This led to the development of the patient concerns inventory (PCI) as an item prompt list to aid clinical consultations and promote multiprofessional involvement for patients.11 There are subtle differences in items and their frequency reported by HNC patients by site (oral, oropharyngeal, laryngeal, and other) and stage (early, late).12

Most publications relating to the PCI have been UK based and thus far there is a lack of evaluation across a wide variety of healthcare settings. There are likely to be clinical, social, cultural, spiritual, and health economic variances. Recognizing concerns common across centers and reflecting on differences should aid clinicians and their colleagues from multiprofessional backgrounds to consider ways to improve the post-treatment support for patients.

The aim of this project was to collate and contrast PCI responses from a range of different head and neck cancer follow-up clinics around the world. Also, we sought to explore the relationship, if any, between responses to the PCI and overall quality of life (QOL).

2 PATIENTS AND METHODS

Nineteen units from around the globe participated (Table 1) and Aintree University Hospital in Liverpool was the lead co-ordinating site. The intention was to acquire at least 100 patients per site and as close to a consecutive series as possible to minimize selection bias. The sample size was a pragmatic decision based on the likely number of patient responses achievable within a reasonable timeframe. Eligible patients were those following head and neck cancer treatment, aged 18 to 89 years, and attending routine clinic consultations. Patients were free from active cancer (no time limit) and their treatment was with curative intent. Those successfully treated for recurrence and late effects such as osteoradionecrosis were also eligible. Patients were ineligible if they had cognitive impairment, significant psychiatric illness, or with thyroid or skull base cancer. Eligible patients attending several clinics were only included once. Units were aware of patient identity but their submitted data was anonymous. Collaborators approached local medical health boards or University Institutes regarding local ethical approval and sponsorship and receipt of such approval was required centrally. The data collection period was in large part during 2018, when the data was submitted to Aintree. The exception was Aintree itself with consecutive cases from November 2011 to January 2013; this was because Aintree is currently involved in a randomized trial of the PCI.

The PCI as developed at Aintree University Hospital in collaboration with Edge Hll University and consists of 56 specific clinical items (see Figure 5 to see a list of items) and one free-text “others” box, and is an item prompt list which patients select from before their appointment. These items can be grouped within domains 13 of physical and functional well-being (29 items), treatment related (4 items), social care and social well-being (9 items) and psychological, emotional, and spiritual well-being (14 items). The PCI also contains a list of 18 professionals who patients might want to talk with. For this international study, this list was excluded because it is specific to the United Kingdom and job roles and titles do vary between countries. Thus, only a single sheet paper version of the PCI symptom and problem prompt list was created. Also, a single question about QOL was included for analysis in relation to the PCI. The overall QOL question from The University of Washington QOL questionnaire (UW-QOLv4) was chosen and this asks patients to rate their overall QOL during the past 7 days.14 Patients are asked to consider not only physical and mental health, but also other factors, such as family, friends, spirituality, or personal leisure activities important to their enjoyment of life. The response options for the overall QOL question are “outstanding,” “very good,”’ good,’ “fair,” “poor,” and “very poor,” The main reason for including this single question about QOL was that the primary outcome measure in an ongoing multicenter randomized trial9 is the percentage of patients reporting less than good overall QOL. The analyses we present use this dichotomy, that is, where the responses of “fair,” “poor,” and “very poor” are taken as being “less than good.” The PCI and UW-QOLv4 have already been translated into various languages and these were used to create the single page form used in this study. Where there were no translated versions, a standard forward and backward translation process was followed with consensus for any discrepancies.

Units collected categorical clinical and demographic details on each patient for age (<55, 55-64, 65-74, ≥75), sex (men, women), clinical stage (early stage 1-2 or late stage 3-4), site (oral, oropharynx, larynx, other), surgery (yes, no), radiotherapy (yes, no), chemotherapy (yes, no), and months from primary diagnosis (<12, 12-23, 24-59, ≥60). Each unit entered data into a preprepared excel worksheet which was submitted centrally via secure email for collation into a single SPSS (Version 25) dataset.

3 RESULTS

Nineteen units participated with a median (IQR) of 100 (89-140) cases submitted, range 50 to 204, in total, data were submitted on 2136 patients, and 70% (1488/2135) were males. Twenty-four percent (505/2132) were aged under 55 years, with 32% (686) 55 to 64 years, 32% (672) 65 to 74 years, and 13% (269) 75 years or older. Just over half (55%, 1157/2115) had “late” stages 3 to 4 tumors, and tumor location for 48% (1025/2129) was oral, 20% (424) oropharynx, 20% (419) larynx, and 12% (261) “other.” About one-third (36%, 763/2122) were treated by surgery alone, 23% (482) by radiotherapy and/or chemotherapy without surgery, while 41% (877) had surgery combined with radiotherapy and/or chemotherapy. Specifically, 77% (1642/2131) received surgery, 62% (1326/2130) radiotherapy, and 32% (689/2128) chemotherapy. One quarter (27%, 560/2073) were within 12 months of diagnosis, 20% (416) within 12 to 23 months, 30% (620) within 24 to 59 months, and 23% (477) at 60 months or later.

The median (IQR) total number of PCI items selected was 5 (2-10), range 0 to 48 items, mean 6.95 and 48% (1026/2136) with six or more PCI items overall. Corresponding results for each PCI domain were: Physical and functional well-being 3 (1-6), 0-28, 4.31, and 47% (1013/2136) with four or more items; Treatment related 0 (0-1), 0-4, 0.38, and 31% (665/2136) with one or more items; Social care and social well-being 0 (0-1), 0-9, 0.75, and 41% (869/2136) with one or more items; Psychological, emotional, and spiritual well-being 1 (0-2), 0-13, 1.50, and 36% (773/2136) with two or more items. There was considerable variation between units in the number of items their patients selected (Figure 1), and a fourfold difference in the mean total number selected (Figure 2). However, there was considerable variation in unit case-mix (Table 2). Figure 3 presents funnel plots showing unit variation both before and after case-mix adjustment in the number of patients selecting six or more PCI items overall. Case-mix adjustment had minimal impact on unit variation implying there remain stronger systematic unit differences. Figure 4 shows case-mix adjusted funnel plots for PCI domains and though most units follow the pattern delineated by the funnel some units had a tendency for their patients to select more items across all domains, and others to select fewer. One unit was excluded from all adjusted analyses because no data was submitted for one of the case-mix variables (Table 2).

Binary regression assessed the association of each case-mix variable with the likelihood of endorsing more than the median number of PCI items, after adjustment for other case-mix variables as independent predictors and adjustment for unit clustering. Separate models evaluated each PCI score (ie, total score, physical function, treatment-related issues, social care/social well-being, and psychological/emotional/spiritual well-being; Table 3). The first column of results shows univariate case-mix variable variation in the percentage of patients selecting ≥6 PCI items overall and greater percentages were observed for females, patients with later stage tumors, patients having radiotherapy and/or chemotherapy, and patients within 12 months of diagnosis. Overall and across domains, the adjusted risk ratios suggest a consistency for females to be more likely to select more items, particularly psychological emotional and spiritual items, though some of the confidence intervals about these risk ratios did include the possibility of no added risk (ie, a risk ratio of 1.00). Similarly, there was also consistency observed in risk ratios greater than 1.00 for patients treated with radiotherapy and/or chemotherapy, and patients within the first 12 months since diagnosis, though again many of the confidence intervals included the possibility of no added risk. There generally seemed little association with tumor site while tumor stage seemed more specifically relevant to physical functioning.

Figure 5 shows the PCI items selected by patients over the whole sample, the most common of which were “fear of the cancer returning” (39%) and “dry mouth” (37%). Other items selected by 20% to 29% of patients were “chewing/eating,” “swallowing,” “speech/voice/being understood,” “dental health/teeth,” “fatigue/tiredness,” “salivation,” “pain in the head and neck,” “cancer treatment,” and “mouth opening,” “Dry mouth” was in the top five items selected for 17 of the 19 units and in the top 10 items selected for 18 units (Table 4) and “fear of the cancer returning” was in the top 10 selected items for all 19 units. In all, across 19 units there were 20 different PCI items that made their way into the top five items selected and 35 items into the top 10 selected. Other free-text items were few in number (2%, 51/2136); three were related to work, three to itchy skin, and three to bad breath, whereas the remainder were a disparate collection of nonspecific issues, such as lip ulcer, tremor, postoperative hair growth in mouth, dimension of tracheostomy stoma, loneliness, facial numbness, cramps, aching joints, CT findings, and blocked tear duct.

Overall, QOL was known for 2099 patients and was reported by 4% (94) of patients as “outstanding,” 30% (625) as “very good,” 41% (856) as “good,” 17% (362) as “fair,” 6% (123) as “poor,” and 2% (39) as “very poor,” Thus, overall QOL was less than good for 25% (524/2099) and Figure 6 presents funnel plots showing unit variation both before and after case-mix adjustment. Case-mix adjustment made little difference implying stronger systematic unit differences exist, though these differences were smaller than seen for numbers of PCI items selected. Binary regression assessed the association of each case-mix variable with the percentage of patients reporting less than good QOL after adjustment for other case-mix variables as independent predictors and for unit clustering (Table 5). The adjusted risk ratios suggest females were more likely to report less than good QOL, risk ratio 1.27, 95%CI: 1.05 to 1.53. There were also higher risk ratios observed for late tumor stage and use of radiotherapy/chemotherapy though confidence intervals about these risk ratios do include the possibility of no increased risk. Overall, these case-mix variables were less compelling as predictors of QOL than PCI predictors of QOL, as seen in the results from separate analyses using PCI scores as predictors in which a greater number of PCI items endorsed were associated with poorer QOL (Figure 7). Similar (though less striking) progressive associations were seen within each PCI domain (Figure 8).

Binary regression also assessed the association of each PCI item with less than good QOL, adjusting for unit clustering, and 38 of the 56 were significant at the P < .001 level and all but three (appetite, hearing, and carer) were significant at P < .05. All 56 had risk ratios over 1.00, with 55 over 1.30, median (IQR) risk ratios 1.75 (1.57-1.92) range 1.05 to 2.34. This again supports the number of items selected either overall or by domain being progressive indicators.

4 DISCUSSION

This is the first time that an item prompt list such as the PCI has been used across such a diverse number of units in different healthcare settings. In routine clinical practice, it is feasible to use both the UW-QOLv4 and PCI in digital format with algorithms to identify immediately those patients doing badly and what issues they wish to talk about.11 However, in situations where digital systems for collecting patient reported outcomes are not used, a single sheet paper PCI is a way of alerting clinical teams as to which patients need additional support. A PCI approach in routine follow-up clinics could possibly result in a clinically meaningful and significant difference in QOL, emotional dysfunction, and distress at 1 year and is the subject of ongoing research.9 The primary outcome measure for this ongoing randomized trial is the percentage of patients reporting less than good overall QOL, hence the inclusion of the one item UW-QOLv4 question in our study. The PCI is a condition-specific prompt list and is different to other tools such as the Cancer Survivors' Survey of Needs (CSSN), and Cancer Survivors' Unmet Needs measure (CaSUN).16, 17

There are several limitations to this study. It was nonfunded with self-selecting units and with no attempt to obtain representation from every continent, nor within each country. The intention was for consecutive patients to be approached to reduce selection bias and although reports from collaborators suggest that relatively few patients declined it is a limitation of the study that no log was made of such patients and consequently the accrual rate for the study is unknown. Missing data in selected patients was minimal. Contributing units reported little difficulty in translating the PCI. Some words caused confusion such as “activity” and “regurgitation” and some items may not always have seemed applicable to some healthcare settings, such as having access to financial benefit support, nursing care at home and gastrostomy feeding tubes. Regarding the question of QOL, the term “outstanding” is extreme for some cultures, where “very good” is the best to be expected. In some cultures, it might be considered unacceptable to ask about certain topics such as intimacy and sex; however, using the PCI patients can choose not to identify certain issues and focus on other aspects. If the PCI is to be used more widely it would be appropriate to undertake further cross-cultural translation studies to help frame each item within the context of the individual healthcare setting, which would help determine the utility of using the item prompt list. The cohort were all first-time users of the PCI though in future a longitudinal study of PCI responses would help reveal any temporal differences in items within clinical settings. In terms of data analysis, we looked at variation between units by the case-mix factors used and future studies could include other aspects such as comorbidity, educational level, occupation, and carer support.

5 CONCLUSION

Although there are similarities in the PCI between the 19 units, differences do exist and are larger than what would be expected by case-mix factors alone. It is likely that the PCI reflects subtle differences in priorities across cultures that need to be addressed in order to improve QOL outcomes.

CONFLICT OF INTEREST

The authors declare no potential conflict of interest.