Rogers SN, Lowe D. Screening for Dysfunction to Promote Multidisciplinary Intervention by Using the University of Washington Quality of Life Questionnaire. Arch Otolaryngol Head Neck Surg. 2009;135(4):369-375. doi:10.1001/archoto.2009.7
To investigate the potential of the University of Washington Quality of Life Questionnaire (UW-QOL) in routine clinics as a quick screening tool for possible dysfunction in patients after treatment of head and neck cancer.
Regional Maxillofacial Unit, Aintree University Hospitals National Health Service Foundation Trust Liverpool, a National Health Service teaching hospital.
Consecutive disease-free patients with oral or oropharyngeal squamous cell carcinoma, who had undergone primary surgery with or without adjuvant radiotherapy, for whom UW-QOL version 4 data from 2000 to May 2006 were available in our research database; and consecutive patients from previous studies (4 postal surveys of disease-free patients with oral or oropharyngeal squamous cell carcinoma, 1 clinic-based study that targeted speech and swallowing in patients with oropharyngeal disease, 1 that evaluated shoulder function in patients with various diagnoses, and 1 that recruited patients without cancer attending a general dental practice).
Main Outcome Measures
Cutoff strategies for further evaluation/intervention derived from studies using the UW-QOL in parallel with 13 other established questionnaires. Effects of preferred cutoffs on trigger variation were assessed with the use of all available UW-QOL version 4 data (615 patients).
Trigger rates for further intervention fell between 9% (recreation and speech) and 16% (swallowing). Eighty-one percent of patients with free-flap surgery and adjuvant therapy for T3 or T4 tumors met the trigger criteria at around 2 years, with 42% meeting the trigger on 3 or more domains.
The fourth version of the UW-QOL is suitable for routine screening in clinical practice.
Health-related quality of life evaluation after head and neck cancer treatment is recognized as an important patient-reported outcome measure. An ever-increasing number of articles have been published on this issue.1 Validated questionnaires have emerged that allow clinicians a better insight into the factors that predict health-related quality-of-life outcomes.2,3 They are useful as primary or secondary outcomes in clinical trials. When presented in a simple way, questionnaires can also provide additional information for patient, caregiver, and tumor board members.4 They help provide feedback from the patient perspective in a structured way and encourage the opportunity for further evaluation and intervention for patients who are doing poorly.
A range of issues are affected by head and neck cancer, and it can be difficult to identify patients with problems in the clinical setting. This is made more problematic by the busy nature of outpatient clinics, where there are often considerable time pressures. Another hurdle is that patients tend to have low self-esteem and are reluctant to complain.5 Many patients do not wish to be a burden and to take up time in the clinic on such matters. Hence, some patient problems will inevitably go unrecognized.6 Questionnaires have a role in addressing this issue, and the use of touch-screen technology makes it achievable in routine practice.7
There is no one criterion standard questionnaire for quality of life.8 One that is often used in clinical practice in the United Kingdom is the University of Washington Quality of Life Questionnaire (UW-QOL),9 partly because of its simplicity. The fourth version of the UW-QOL10 has 4 components: 12 domains (pain, appearance, activity, recreation, swallowing, chewing, speech, shoulder, taste, saliva, mood, and anxiety), an importance-rating scale, general quality-of-life questions, and a free-text section where patients can add their own comments. It would be just as appropriate to use other validated head and neck questionnaires to help screen patients in the clinic, but we have a track record for using the UW-QOL questionnaire over several years and it forms the basis of the touch-screen patient-derived outcome tool used in the Liverpool clinic.
The aims of this study were to collate data collected from several previous studies that used a variety of questionnaires in combination with the UW-QOL between 2000 and 2006; to postulate appropriate cutoffs in severity for UW-QOL domains as triggers for further evaluation and intervention; and to estimate how many patients the cutoffs would identify at various times in the cancer journey and in various clinical subgroups.
The 12 domains of the UW-QOL version 4 are scaled from 0 (worst) to 100 (best) according to the hierarchy of response. Patients are also asked which 3 domains have been most important to them in the past 7 days. We have used the UW-QOL questionnaire since 1995 (version 4 since 2000) and have maintained a research database of all UW-QOL questionnaires completed by the 1992 to 2005 cohort of patients with oral or oropharyngeal squamous cell carcinoma whose primary treatment was surgery with or without adjuvant radiotherapy.
All version 4 questionnaires from January 1, 2000, to May 31, 2006, were analyzed to identify patients with significant deficits in their quality of life who were, as such, potential candidates for intervention. For each UW-QOL domain, the domain scores were cross-tabulated with domain importance to derive risk groups. Our intention was to identify a group in which the trigger would select around 5% to 10% of the patients with the worst responses and a second group in which the trigger would select the next 5% to 10%. Other subgroups were identified to take the total at-risk selection to 25% to 35%.
We then analyzed data from a series of previous studies.10- 18 We calculated summary statistics for other measures collected concurrently with the UW-QOL for each of the risk groups identified for the UW-QOL domains. The other measures were the Center for Epidemiology Studies Depression Scale, Derriford Appearance Scale, European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Head and Neck, EQ-5D EuroQol, Hospital Anxiety and Depression Scale, Liverpool Oral Rehabilitation Questionnaire version 3, M. D. Anderson Dysphagia Inventory, Neck Dissection Impairment Index, Shoulder Disability Questionnaire, Swallowing (Dysphagia) Specific Quality of Life, Voice Handicap Index, Voice-Related Quality of Life Measure, and Xerostomia-Related Quality of Life Questionnaire. These results formed the evidence base from which we then derived clinically relevant UW-QOL cutoff criteria in each domain as our triggers for intervention. We also took the strength of the wording of UW-QOL domain levels into consideration. Our aim was to derive domain criteria that would trigger intervention in no more than 1 in 5 patients. We recognize the subjectivity in our preferred criteria, and our results are presented as far as possible to allow others to vary their criteria for triggering more or fewer patients.
Using our preferred criteria, we then revisited the total UW-QOL database to explore the variation in trigger rates between main clinical groups at around 2 years after primary surgery.
All of the studies we included in this report recruited consecutive patients. Four were postal surveys of disease-free patients with oral or oropharyngeal squamous cell carcinoma treated by primary surgery with or without adjuvant radiotherapy. Another study was clinic based and investigated speech and swallowing issues, specifically studying patients with oropharyngeal tumors. Another clinic-based study measured shoulder function and included a wider range of diagnoses. Another study recruited subjects without cancer attending a general dental practice.
Ethical approval for each study was given by the Sefton Research Ethics Committee.
Details of patient populations, response rate, relevant measures, and aims for each study are summarized in Table 1. The research database identified 758 patients in a consecutive 1992 through 2005 cohort having had primary surgery with or without adjuvant radiotherapy after a diagnosis of squamous cell oral or oropharyngeal cancer. Of these patients, 143 had died before January 2000, the time from which version 4 of the UW-QOL was used. From January 2000 to May 2006, 79% of patients (487 of 615) completed at least 1 UW-QOL questionnaire (median, 3; interquartile range, 1-5; range, 1-9). There were 1615 questionnaires in all, obtained at a median time from surgery of 36 months (interquartile range, 15-65 months; 10th-90th percentile, 5-101 months).
The database of 1615 UW-QOL questionnaires was used to ascertain the severity of response. For each UW-QOL domain, the domain scores were cross-tabulated with domain importance to derive the most at-risk patient groups. Application of the criteria described in the “Methods” section to each domain produced up to 4 risk subgroups (A through D) to reflect gradations in risk, with subgroup A having the worst UW-QOL responses (Table 2). For example, a UW-QOL pain score of either 0 or 25 defines the worst at-risk group for pain (subgroup A); a UW-QOL pain score of 50 together with the patient choosing pain as being an important issue defines subgroup B; a UW-QOL pain score of 50 together with the patient not choosing pain as an important issue defines subgroup C; and a UW-QOL pain score of 75 together with pain being important defines subgroup D. The other possible scores, 75, not important and 100, indicate less risk and are not shown in the table.
These UW-QOL groupings were then used to compare patient scores on other relevant measures collected concurrently (Table 3 and Table 4) since 2000. Table 3 and Table 4 indicate the ability of the UW-QOL to distinguish between patients in regard to these other measures and were used to decide our most preferred trigger for each domain. We aimed to derive trigger criteria that would select not more than 1 in 5 patients. For example, in regard to swallowing, from variation in the M. D. Anderson Dysphagia Inventory and Swallowing (Dysphagia) Specific Quality of Life results between UW-QOL subgroups shown in Table 4, we decided to include UW-QOL swallowing subgroups A and B as trigger criteria. Table 2 indicates that subgroup A is defined by a UW-QOL swallowing score of 0 and subgroup B by a score of 30, and that by choosing subgroups A and B we might expect 16% of UW-QOL questionnaires (5% in subgroup A, 11% in subgroup B) to be captured by these criteria. In practice, the rates for our preferred triggers ranged between 9% (recreation and speech) and 16% (swallowing) of questionnaires.
Our preferred triggers were used to determine variation in trigger rates between the main clinical groups at around 2 years from surgery (Table 5). Of 615 patients alive in January 2000, 479 had 15 months or more of follow-up (to death or alive at last follow-up) to May 2006. Of these, 386 (81%) had quality-of-life data beyond 15 months; sometimes a patient had several questionnaires that qualified, and so the closest after 24 months was chosen. The median was 29 months and the interquartile range was 23 to 45 months for the 386 patients.
Table 5 highlights considerable differences in trigger rates at around 2 years after surgery according to the clinical characteristics of the patient, particularly adjuvant radiotherapy. A high percentage (81%) of patients who had undergone free-flap surgery with adjuvant therapy for T3 or T4 tumors met the trigger criteria, with 42% meeting criteria on 3 or more domains. At around 2 years there was also considerable variation in trigger rates by age and sex for appearance and by age for mood and anxiety. The trigger rate for appearance was 13% to 20% in females and in males younger than 65 years compared with only 1% (1 of 88) in older males. The trigger rate for mood was 20% (40 of 197) in patients younger than 65 years and 7% (12 of 168) in older patients. For anxiety the trigger rate was 24% (46 of 194) in patients younger than 65 years and 7% (12 of 166) in older patients.
Our preferred triggers were applied to the 372 patients without cancer who attended general dental practice and who were asked UW-QOL questions in relation to their condition rather than to cancer. The trigger rates by domain for these patients without cancer (with trigger rates after 2 years for patients with cancer given in parentheses) were as follows: pain, 18% (12%); appearance, 5% (12%); activity, 9% (13%); recreation, 6% (10%); swallowing, 1% (14%); chewing, 1% (13%); speech, 2% (10%); shoulder, 5% (12%); taste, 0% (12%); saliva, 1% (16%); mood, 12% (15%); and anxiety, 12% (16%).
Patient-reported outcomes are a vital component in our understanding of outcomes after treatment.19 Health-related quality of life is an important facet of patient-reported outcomes. This is the first study of this type, to our knowledge, to postulate cutoff scores by means of the UW-QOL version 4 suitable for screening at-risk patients in a routine clinical setting. We have adopted a systematic approach in the data handling based on previous studies (Table 1). A large number of questionnaires have been brought together for analysis. However, we recognize that this includes relatively few patients with baseline data. This reflects the way the questionnaire has been used in our practice, where the focus has been on posttreatment dysfunction. It has already been recognized20 that the main issues at baseline are pain, mood, and anxiety rather than the aspects of head and neck function. Surveys tend to take place annually, and there is inevitable loss of data at different time points. Nonresponses are liable to represent bias in some ways, but at least our data contain responses from around 80% of the cohort at some stage during their cancer process.
The idea of using a questionnaire to screen patients is not new. Although cutoffs can be calculated for other validated head and neck questionnaires, it is novel to use the UW-QOL version 4. The questionnaire briefly covers the range of common problems associated with cancer and its treatment. The single-item questions are a potentially suitable screening tool because they have been shown to correlate well with specific measures, namely the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Head and Neck (dry mouth),10 EORTC H&N (taste),10 European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Core 30 (emotional),10 Hospital Anxiety and Depression Scale and Center for Epidemiology Studies Depression Scale,11 EQ-5D Euro-Qol,12 Derriford Appearance Scale,13 Xerostomia-Related Quality of Life Questionnaire (S.N.R. and D.L., unpublished data, 2008), M. D. Anderson Dysphagia Inventory and Swallowing (Dysphagia) Specific Quality of Life,14 Voice-Related Quality of Life Measure and Voice Handicap Index,15 Shoulder Disability Questionnaire and Neck Dissection Impairment Index,16 Liverpool Oral Rehabilitation Questionnaire version 3,17 and mouth opening.18 The way the questions are worded in the UW-QOL helps to provide an indicator of outcome that patients, caregivers, and tumor board members can easily understand and apply in clinical practice.4
Applying cutoff criteria allows the patients reporting problems to be identified. This is part of the continued progress toward incorporating health-related quality-of-life data into routine clinical practice. The choice of where to set the cutoff is open to debate. We combined the patient score in each item together with whether it was one of the 3 important issues identified by the patient. The rationale for this was that, if there is mild dysfunction, it still might be appropriate to factor it in if the patient identifies it as an important issue. When looking at suitable cutoffs, we also considered the wording of the questionnaire. Finally, given the range of issues within the questionnaire and the likelihood of patients reporting problems in several of these, particularly after radical treatment for advanced disease, we took a pragmatic approach for clinical practice in not identifying too many patients. Our initial selection of at-risk subgroups A to D given in Table 2 only include one-quarter to one-third of patients for each domain. However, even this selection would result in a large proportion being identified overall. In Tables 3 and 4 we compared our groups A to D with concurrent data from a series of studies to further build up an evidence base from which we could refine our trigger criteria for identifying patients with problems. In practice, the rates for our final preferred triggers ranged between 9% (recreation and speech) and 16% (swallowing) of questionnaires.
The implications of how to use the algorithm in a busy clinic need careful thought. For example, with our preferred algorithm (Table 5), 42% of patients with early disease (T1 and T2, oral, no free flap, and no adjuvant radiotherapy) had 1 or more domains that met the trigger criteria compared with 81% with advanced disease (T3 and T4, oral, free flap, and adjuvant radiotherapy). The percentage that met trigger criteria for 3 or more domains was much lower but was still 42% in advanced disease.
Each domain was flagged at between 9% and 16%, which was part of our intention in setting the cutoff. Our final preferred algorithm was influenced by the concurrent questionnaires (Tables 3 and 4). Where we perceived there to be a big jump between groups (A to D) was where we thought it suitable to put a cutoff. We were also mindful to derive domain criteria that would trigger intervention in no more than 1 in 5 patients. In practice, the logical choice of cutoff for most domains seemed to be either at group B or C, but B was preferred to keep the trigger rate to a manageable proportion. The exceptions were chewing and speech, where trigger A was chosen because group B contained too many patients and it was impossible to subdivide group B any further. The choice of B in addition to A would have selected 33% for chewing and 27% for speech. It can be argued that these are important domains and that patients with problems are better identified than missed. It is possible also that these percentages would be less for other head and neck cancer sites.
We recognize in general the subjectivity in our preferred criteria, and our results are presented as far as possible to allow others to vary their criteria for triggering more or fewer patients. We did not start out with any specific hypothesis as to which cutoffs would be useful, and we have let the data dictate our final selections. As such, this can be seen as hypothesis generation, and future research may confirm or refute the suitability of our choice of criteria. These criteria are not fixed and may change in the future as implications of the current preferred criteria are experienced in practice in a busy clinic setting. Even if there is no need to actively intervene, for some patients recognizing and discussing their issues may be beneficial in itself. Our preferred criteria have already been integrated into touch-screen technology as part of a wider patient concerns inventory, and the algorithm triggers form part of an output available to the consultant during clinic consultations. The results of this experience will be published in due course.
The ability to use touch-screen technology to acquire patient-derived data in real time in the clinic has tremendous potential to improve patient care. It is possible to use a questionnaire such as the UW-QOL version 4 to screen patients for dysfunction. To assist in the incorporation of these data into clinical practice, it is feasible to define criteria so that patients and their problem domains can be easily identified. This allows the clinician and members of the head and neck team a chance to discuss these issues with the patient and, where appropriate, arrange further support or intervention.
Correspondence: Simon N. Rogers, FDS, RCS, FRCS, MD, Regional Maxillofacial Unit, Aintree University Hospitals NHS Foundation Trust, Aintree, Liverpool L9 7AL, England (firstname.lastname@example.org).
Submitted for Publication: July 1, 2008; final revision received November 18, 2008; accepted November 24, 2008.
Author Contributions: Dr Rogers and Mr Lowe had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Rogers and Lowe. Acquisition of data: Rogers. Analysis and interpretation of data: Rogers and Lowe. Drafting of the manuscript: Rogers and Lowe. Critical revision of the manuscript for important intellectual content: Rogers and Lowe. Statistical analysis: Lowe. Study supervision: Rogers.
Financial Disclosure: None reported.
Previous Presentation: This study was presented at the Seventh International Conference on Head and Neck Cancer of the American Head and Neck Society; July 20, 2008; San Francisco, California.