There is not at present a suitable disease-specific health-related quality of life instrument for uninvestigated dyspepsia and functional (non-ulcer) dyspepsia.
There is not at present a suitable disease-specific health-related quality of life instrument for uninvestigated dyspepsia and functional (non-ulcer) dyspepsia.
To develop a new multi-dimensional disease-specific instrument.
The Nepean Dyspepsia Index (NDI) was designed to measure impairment of a subject’s ability to engage in relevant aspects of their life and also their enjoyment of these aspects; in addition, the individual importance of each aspect is assessed. A 42-item quality of life measure was developed and tested, both in out-patients presenting to general practice with upper gastrointestinal complaints (n = 113) and in a randomly chosen population-based sample (n = 347).
Adequate face and content validity was documented by an expert panel. Factor analysis identified four clinically relevant subscales: interference with activities of daily living, work, enjoyment of life and emotional well-being; lack of knowledge and control over the illness; disturbance to eating or drinking; and disturbance to sleep because of dyspepsia. These scales had high internal consistency. Both symptoms and the quality of life scores discriminated dyspepsia from health.
The Nepean Dyspepsia Index is a reliable and valid disease-specific index for dyspepsia, measuring symptoms and health-related quality of life.
Quality of life is now recognized as an important health outcome, and is frequently measured in clinical and epidemiological studies.1–3 In clinical trials, quality of life is often used as a secondary outcome to symptom relief, while in health surveys, quality of life information provides valuable insights into the burden of illness amongst diseased populations.1, 2
Quality of life is particularly useful as an outcome measure in studies of disease that have no obvious biological or clinical markers. In these conditions, symptom control becomes a treatment priority, and in the absence of objective clinical criteria, treatment efficacy should be evaluated by its impact on symptoms as well as on patient well-being and functioning.3–5 Quality of life generally correlates well with symptom severity scores in untreated patients,3 and improvements in quality of life consequent to effective treatment have been reported.3
Existing quality of life measures range from totally generic to disease-specific.3 Generic measures tend to contain a large number of items, and many of these may be irrelevant to specific illnesses.3, 4 These items cannot be expected to change in response to treatment. Further, generic instruments may be insensitive to change on relevant items because of the ‘noise’ of those many items which do not change. Consequently, while generic instruments are viewed as adequate for single-point comparisons between subjects, the use of disease-specific quality of life instruments is currently recommended to monitor change within patients.3, 4
Reduced quality of life has been documented in patients with functional dyspepsia.6–8 However, the studies have used generic quality of life instruments, including (but not limited to) the Medical Outcomes Survey (MOS) short form,6, 9 and the Psychological General Well-Being Index.7, 8 Indeed, there have been few attempts to develop a disease-specific quality of life instrument for functional dyspepsia, despite widespread acceptance that quality of life is an important medical outcome in this patient group,10 as it is in irritable bowel syndrome.11, 12
The impact of an illness on quality of life varies depending on what area is affected and how important that area is to the individual.3 Weighting of items by the individual patient, according to their relative importance, may therefore be of value and could represent a more accurate method of measuring the impact of a condition on quality of life. However, there are no quality of life instruments in gastroenterology that have tested this approach.
We set out to develop a new, disease-specific instrument for measuring quality of life both in patients with uninvestigated dyspepsia and in patients with functional (non-ulcer) dyspepsia. We aimed to create a self-report measure that would discriminate dyspepsia from health and could be used to assess changes in symptoms and quality of life. We wanted the instrument to be simple to administer and score, and to be suitable for clinical research in general practice, specialty practice and the community, including clinical trials. Our aim was also to create an instrument that included the option of incorporating into the scoring each subject’s own evaluation of the importance of the affected activities. In this report, we describe the development of the new instrument, including its internal reliability and validity in out-patients and a population sample.
The conceptual basis of the Nepean Dyspepsia Index (NDI) was developed by the Sydney research team. Initial items were developed through consultation with patients and with local gastroenterology and psychology staff. These items were then pre-tested amongst patient populations (n = 55) in four countries (Australia, Germany, Italy and the Netherlands) using focus groups and face-to-face interviews to identify a comprehensive set of relevant disease-specific items. To ensure face and content validity, the NDI was reviewed by an international panel of experts in gastroenterology (G. N. J. Tytgat, V. Stanghellini, G. Holtmann, M. Verlinden, N. J. Talley) and items were carefully selected. A revised version of the NDI was then created; this incorporated 42 items which assessed quality of life across 17 key aspects and also a symptom checklist that measured the frequency, intensity and bothersomeness of 15 upper gastrointestinal symptoms over the prior 2 weeks. A 2-week period was chosen to minimize recall bias yet cover an adequate period for measurement of usual activities and symptom status.3
The inclusion of a symptom checklist provides the instrument with an internal standard against which to compare the NDI’s quality of life measurement. The underlying philosophy is that quality of life is affected by the severity of symptoms. Hence a change in quality of life ought to parallel a change in symptom severity.
An important aspect of the philosophy underlying the NDI was that it be adaptive to the differing priorities and individual value systems of dyspepsia patients. This was achieved by allowing individuals to apply differing weights (equivalent to priorities) to common groups of items or themes (based on key aspects of life) in the NDI, such as the impact of dyspepsia on daily living, sleep or sex. These weights were incorporated into the NDI scoring system that was developed and tested as described below.
The revised version of the NDI underwent a validated translation from Australian English to French, Dutch, Italian, German, Spanish and American English, to allow for a later multinational evaluation study. The purpose of this initial evaluation study was to assess the psychometric properties of the NDI (including internal reliability and validity), to shorten the questionnaire, and to develop and test the NDI scoring system.
Two groups of subjects were evaluated: (i) consecutive patients presenting to general practice with dyspepsia or reflux and (ii) community subjects identified as having or not having dyspepsia by a screening questionnaire.
Ten general practitioners in Sydney and eight general practitioners in Melbourne recruited consecutive patients presenting to their practices with dyspepsia or reflux (n = 113). Dyspepsia was defined as upper abdominal pain or discomfort over the preceding 3-month period. The population has been described in more detail elsewhere.13 Patients aged 18 years presenting with dyspepsia were considered eligible for inclusion. Patients with a history of gastrointestinal bleeding or major upper gastrointestinal surgery, patients presenting with other serious conditions such as drug and alcohol abuse, mental illness or dementia were excluded, as were pregnant or lactating women. Patients presented to the research centre all completed both the Bowel Symptom Questionnaire (BSQ), a validated measure of gastrointestinal complaints in the prior year,14 and the Nepean Dyspepsia Index.
A total of 1000 adult subjects (aged ≥ 18 years) were randomly chosen from the electoral rolls of the Greater Wellington region, New Zealand, with an eligible population of 212 383. A brief screening questionnaire, derived from the BSQ, investigating dyspepsia and reflux symptoms during the previous 12 months14 was mailed to the sample. Forty-eight subjects had moved, and of the 952 eligible subjects, 778 returned a questionnaire giving a response rate of 82%. Subjects were asked if they would be prepared to be involved in further studies, and 435 agreed in writing. These subjects were mailed more detailed questionnaires, and 347 (80%) responded. The New Zealand population sample completed the Nepean Dyspepsia Index, full BSQ and a generic quality of life measure, the SF-36.9, 14
This questionnaire consisted of 42 questions, measuring health-related quality of life structured around 17 common themes, namely medications, doctor visits, and control over and attitude to disease (5 items), finance (1 item), eating/drinking (3 items), sleep (2 items), work/study (2 items), daily tasks (2 items), partner relationship (2 items), family (2 items), friends/other social (2 items), sex (2 items), recreation (2 items), religious/spiritual (2 items), emotions (5 items), self-confidence (2 items), thinking/concentrating (1 item), cognition (2 items) and energy and wellness (2 items). These domains represent areas considered by patients and the expert panel as the most likely to be affected by dyspepsia.
For most of the themes, impact of the illness was considered to occur in two dimensions: interference with a subject’s ability to perform or engage in an aspect of life (e.g. a reduced ability to spend time with friends because of dyspepsia); and interference with their enjoyment of that aspect of life (e.g. impaired enjoyment of time spent with friends because of dyspepsia). These were measured by 5-point Likert scales, from not at all to extremely. While clearly for some groups of items these dimensions were not appropriate (e.g. emotions), and others overlap considerably (e.g. sleep, partner relationships), they still represented potentially separable concepts and were tested in this study.
Each of the 17 themes was weighted by the individual subject at the end of the NDI, according to its self-rated importance in determining the overall quality of their life, on a 5-point Likert scale.
A validated questionnaire was used to identify subjects with dyspepsia and to classify these persons into dyspepsia subgroups. The BSQ addresses gastrointestinal symptoms in the past year and lifetime, socio-demographic items and healthcare seeking, and serves to identify subjects with dyspepsia (defined as pain or discomfort centred in the upper abdomen).14 It is derived from the validated Bowel Disease Questionnaire.15 Subgroups of dyspepsia (ulcer-like, dysmotility-like, reflux-like and unspecified) were identified based on standard criteria as defined below.6
Category 1—Ulcer-like dyspepsia. Subjects with upper abdominal pain or discomfort and at least three of the following symptoms: (a) pain or discomfort often (>25% of the time) relieved by food; (b) pain or discomfort often relieved by antacids; (c) pain or discomfort often before meals or when hungry; (d) periodic pain or discomfort (periods of at least 1 month without pain, with periods in between of weeks to months when there is pain); and/or (e) night pain or discomfort (waking the subject from sleep).
Category 2—Dysmotility-like dyspepsia. Subjects with upper abdominal pain or discomfort and at least three or more of the following symptoms suggesting slow gastric emptying or upper intestinal dysmotility: (a) nausea or vomiting once a month or more; (b) abdominal bloating or visible distension, often; (c) early satiety; (d) pain or discomfort often aggravated by food or milk; (e) pain or discomfort often after meals; and/or (f) pain or discomfort often relieved by belching.
Category 3—Reflux-like dyspepsia. Subjects with upper abdominal pain or discomfort associated with one or more of the following symptoms: (a) heartburn at least once a week; and/or (b) acid regurgitation once a week or more.
Category 4—Non-specific dyspepsia. Subjects with upper abdominal pain or discomfort that did not fit into categories 1, 2 or 3.
Differences among these groups in their responses to the NDI were assessed as part of the analyses.
This is a widely accepted self-administered health-related quality of life generic measure from the Medical Outcomes Study.9 Eight multi-item dimensions measure functional status and well-being. It has been very extensively tested for reliability and validity, 9, 16 and has been used in previous studies on dyspepsia.6
The actual subscale (domain) structure was assessed through factor analysis of the observed responses. Only subjects meeting the criteria for dyspepsia (see above) were included in this analysis as they represent the instrument’s target population. Comparisons were made between the observed domain (factor structure) and the groups of items theorized in the NDI’s construction. A close concordance between the observed and theorized structure would support the theorized constructs. The factor analysis was performed twice, once on the entire collection of items and again on an item-reduced set. The reduced item analysis was necessary because an initial exploratory analysis of the data revealed that many items were answered ‘not at all’ or ‘not applicable’ by a majority of subjects who met the criteria for dyspepsia. After expert review, we viewed either of these responses as indicating a ‘nil’ impact of dyspepsia on a subject’s usual functioning. The finding of numerous ‘nil’ responses has two immediate consequences. One is that it suggests these items are generally not relevant to dyspeptics. The second is that items with little variance will interfere with a factor analysis because there cannot be covariance without variance. Hence the factor structure of the instrument was examined twice, once including all items and then only considering items where the rate of ‘not at all’ or ‘not applicable’ responses was set a priori at < 60%. Although arguably an arbitrary cut-off, 60% is sufficiently severe to unquestionably warrant exclusion. Only the reduced item factor structure is reported here because the alternative was not tenable. An eigenvalue of 1.0 or more was used as the criterion for factor formation and Varimax rotation was employed. Only rotated factor loadings are displayed in the results.
The subscales in the quality of life part were scored for analysis as described below.
Weights were applied to all individual items defined a priori for pre-specified groups of items, based on the self-rated importance of the group. For example, the sleep disturbance questions (ability to sleep and quality of sleep) comprise a group. Suppose a subject responds moderately important (3 out of 5) to the question: ‘How important in determining the quality of your life is sleep?’ In this case both sleep items are weighted by this value. A weight was applied to every individual item within its group by multiplication of the item score and the weight to form a new, weighted item score.
Subscale scores were computed by summation of each item score within a domain identified by factor analysis. In analyses which incorporate importance weights, items were weighted (as described immediately above) and then summed.
Because the domains identified by factor analysis are composed of varying numbers of items the actual values of the scores can be considered arbitrary. To create a scale whose interpretation is both simple and analogous to well-known scores, such as the SF-36, each subscale was re-scaled to a minimum of zero and a maximum of 100. This was achieved by subtracting the minimum possible score and dividing by the range of possible scores. For each subscale this was achieved as shown in Table 1 1 .
To obtain a scale in which low scores indicate poor quality of life (maximum disease impact) and high scores indicate high quality of life (minimum/nil disease impact) each item’s score was reversed prior to any further calculation. The total quality of life score was then obtained by summation of the four subscales and division by four.
For example, suppose a subject responds with a value of 4 (quite a lot) on both questions in the sleep subscale. The score in this case would be calculated as follows:
The scale reversal means that low values are undesirable while high values indicate a high quality of life. Note that the constants 2 and 10 are the minimum and range of the scale, respectively, while the constant 100 (used twice) makes the scale 0–100 rather than 0–1.
To be interpretable, a subscale must be composed of items which are homogeneous in the sense of measuring a consistent underlying construct. A common index of this notion of internal consistency is Cronbach’s alpha.17 An internal consistency analysis was conducted twice. The first analysis assessed the internal consistency of the observed item structure using all items, while the second used the reduced number of items. Again, only the reduced-item analysis is reported here and only dyspeptic subjects have been included because dyspepsia was the target of the instrument development.
It is expected that a disease-specific measure of quality of life will exhibit two properties: (i) some correlation with another accepted generic quality of life measure, and (ii) a correlation of r 0.5.
The SF-36 was used as the external reference; analysis was by non-parametric Spearman rank correlation between SF-36 and NDI subscales.
Discriminant validity refers to the capacity of the instrument to discriminate between two or more groups of individuals. Discriminant validity applies in the current study because the NDI sought to differentiate between respondents who independently met the criteria for dyspepsia and those who did not. We considered discriminant validity in two parts, the symptom checklist and the health-related quality of life subscales.
The purpose of the symptom checklist was to enable correlation of change in the NDI with change in perceived dyspepsia symptoms. To ensure that the symptoms reported are actually dyspepsia symptoms it was necessary to establish discriminant validity for the total symptom rating. Failure to establish this dimension of validity would cast doubt upon the symptoms being true dyspepsia symptoms as opposed to other, related disorders.
As there is no clear theoretically optimal model for combining the frequency (F), intensity (I) and bothersomeness (B) ratings, we applied four possible models a priori, namely purely additive (F + I + B), purely multiplicative (F x I x B), mixed A [(F + I) × B], and mixed B [F x (I + B)]. This was assessed by comparing the scores between respondents who met the criteria for dyspepsia and those who did not.
Analysis of the quality of life subscales followed the same logical sequence as the symptom checklist to test discriminant validity. The analysis was performed with and without the importance weighting.
Those with and without dyspepsia were compared using the Mann–Whitney test in terms of both the symptom checklist and the quality of life subscales.
All P-values calculated were two-tailed; the alpha level of significance was set at P < 0.05 except when multiple comparisons were performed, when P 0.01 was applied.
Among the out-patient sample, 81% had dyspepsia (mean age 51 years (s.d. = 14), 55% male) and in the population sample, 24% had dyspepsia (mean age 45 years (s.d. = 15), 50% male).
The total symptom score discriminated dyspepsia from health. With all methods the positive scores were all at least four times greater than the negative scores; no method appeared to be clearly superior to the others (Table 2 2 ). All methods applied thus had similar, good discriminant validity. Simple addition of the frequency, severity and bothersomeness items to create a total score is recommended in practice.
All items included. Factor analysis suggested that there were eight factors (subscales). All but one had high internal consistency but these factors mixed together a number of clinically distinct domains and was judged by expert review to be unsuitable. This analysis also suffered from the problem of many items with ‘nil’ frequencies in excess of 60%.
Domain structure among dyspeptics after items with 60% nil response removed. Due to some data being lost through missing values, 156 of the 172 subjects who met the criteria for dyspepsia were included in the analysis. Items that 60% or more of subjects did not believe affected them at all, or were not applicable to themselves, were removed from this factor analysis. The domain structure among dyspeptics after these items were removed was different to that observed using all items. Four clinically relevant factors (subscales) were identified in the final factor analysis (Table 3 3 ). These were: interference or difficulty with activities of daily living including work, enjoyment of life and emotional well-being (13 items), lack of knowledge and control over the illness (7 items), disturbance to eating or drinking (3 items) and disturbance to sleep (2 items) because of dyspepsia. This structure was substantively different to that theorized in the instrument’s development. Hence we concluded that the original construct was overly complicated and that the actual structure of the impact of quality of life in dyspepsia sufferers is simpler and can be expressed in four distinct constructs.
We found strong internal consistency between the items in each of the four subscales (Table 4 4 ). Weighting according to subject-determined importance scores did not affect internal consistency noticeably (Table 4 4 ).
Patients with dyspepsia had worse scores than patients without the disorder and the difference reached statistical significance in each case (Table 5 , Mann–Whitney test; all P < 0.05). Table 5 5 shows that weighting did not substantively affect the discrimination between dyspeptics and non-dyspeptics. Analysis of items with ‘low positive’ frequencies indicated that it was not necessary to include these items in the scale (data not shown).
Females had a slightly better quality of life than males across the domains of sleep disturbance (P = 0.05) and eating or drinking (P = 0.06) but the differences were not of clinical relevance (Table 6 ). Advancing age was significantly correlated with sleep disturbance (P = 0.005) and total score (P = 0.01) but not with the other subscales. However, the low correlation coefficients (0.16 and 0.14 for sleep disturbance and total score, respectively) indicated that any age effect was negligible.
Table 7 presents the SF-36 scores in subjects with and without dyspepsia. There were modest but not strong correlations between scales on the SF-36 and the NDI subscales (Table 8 8 ). Role-physical and role-emotional functions were moderately (and significantly) correlated with the interference subscale of the NDI. Moderately significant correlations were also identified between knowledge and control on the NDI scale and the SF-36 items physical and emotional role function. The correlation between the SF-36 scales and the disturbance to eating and drinking as measured by the NDI subscale were modest for role-physical functioning, while the correlations between the SF-36 scales and sleep disturbance were modest (but not significant) for bodily pain and social functioning. Overall, there were moderate and significant correlations between the physical and emotional role functions of the SF-36 and the total quality of life score as measured on the NDI (Table 8 8 ). All correlations were < 0.4.
The dyspepsia subgroups as defined overlapped considerably (Table 9 ). The quality of life subscales from the NDI were reasonably similar across the dyspepsia subgroups. There was no statistically significant difference between ulcer-like dyspepsia vs. the other groups combined for each subscale of the NDI. However, in dysmotility-like dyspepsia the knowledge/control scale was slightly better than in the other groups combined (P = 0.03). The reflux-like dyspepsia group had slightly worse scores than the other groups combined on all of the NDI subscales (all P < 0.05).
Although it is not possible to assess the correlation between changes in symptom severity and changes in quality of life score in the current study, it is reasonable to hypothesize that symptom severity and quality of life score ought to be correlated. Correlation of the total quality of life score with the additive model of symptom scoring yielded a Pearson correlation r = − 0.74 (P < 0.001). Hence, the symptom rating correlated strongly but negatively with the quality of life measure. The negative correlation is appropriate because for symptoms, higher scores are worse, while for the quality of life total scale, higher scores indicate a better quality of life.
Dyspepsia is a major health problem in the community,18 yet validated measures of outcome for dyspepsia have up to now been largely unavailable.10 Health-related quality of life is a label that refers to functional status and well-being;1, 2 it should describe how a person feels, and how he or she functions in daily activities in relation to the disease under study. Few comprehensive disease-specific measures of quality of life have been developed for gastrointestinal disease. A number of excellent generic quality of life measures exist but these contain a large number of questions which the expert panel in this study felt were likely to be irrelevant to functional dyspepsia and which failed to measure a number of areas of specific importance in dyspepsia. Furthermore, some items in generic instruments are not appropriate for detecting change over repetitive administration, such as the item in the Sickness Impact Profile ‘I have attempted suicide’.19 Therefore, generic instruments are unlikely to be as useful as a disease-specific measure for quantifying impaired quality of life in dyspepsia and functional dyspepsia.5
The NDI questionnaire described in this study is actually composed of two distinct instruments: a symptom checklist and a disease-specific quality of life measure. The former is made up of a list of common upper gastrointestinal symptoms which respondents rate in terms of three dimensions while the NDI is composed of 42 items (some of which are composite) which include disease impact items and an importance rating. (We have named the instrument after the Nepean river region at the foot of the Blue Mountains in Western Sydney where the Sydney team is currently based. A copy of the Nepean Dyspepsia Index is available from Prof. N. J. Talley.) Another useful outcome measure for dyspepsia, the Glasgow Dyspepsia Severity Index, has been published recently.20 While this instrument adequately assesses global symptom status, it was not designed to comprehensively assess the impact of dyspepsia on health-related quality of life.
A potential omission with all previous quality of life instruments is the lack of weighting of items relating to the relative importance to each individual.3 In individuals, certain activities that really matter to that person are likely to be more affected by an illness. Weighting the importance of domains may therefore theoretically improve the performance of an instrument. This has not previously been specifically tested in functional dyspepsia or other similar chronic conditions. We found that the originally constructed quality of life instrument with 42 questions was over-complicated and could be simplified into four subscales (domains), with a total of 22 items. Applying individual weights from the original groups of items did not improve the already excellent psychometrics of the instrument. While these initial results suggest that it is unnecessary to weight individual importance in order to assess quality of life in dyspepsia, we have not deleted this from the revised instrument as it is conceivable that weighting may increase the sensitivity of the instrument to change. Work is in progress to test this hypothesis in prospective clinical studies.
The four independent subscales identified by the factor analysis appear to measure aspects not captured by symptom measurement or generic quality of life instruments. One major subscale comprised interference or difficulty with activities of daily living or work because of dyspepsia, combined with impaired enjoyment of life and emotional well-being. Psychological distress is documented to be greater in those with dyspepsia who present for medical care,21, 22 but this is only one component of the subscale and would not be captured by standard psychological measures. Another major area identified was lack of knowledge and control over the illness. This seems sensible as dyspepsia is a long-term problem with relapses and remissions.23 The poorly understood pathogenesis, often attributed mistakenly to anxiety or diet,24 is likely to contribute to patients feeling that no one understands their problem and they have no control over it. Disturbance of eating or drinking because of dyspepsia was also recognized as an important area. There is no convincing evidence that eating patterns are different in patients with dyspepsia but this has been poorly studied.25, 26 However, our data clearly suggest that the disorder does have an important impact in this area. Finally, sleep disturbances because of dyspepsia was identified as a relevant area. Although dyspepsia does not commonly wake patients from sleep,27 poor sleep quality has recently been recognized to occur in functional dyspepsia. 28, 29
Our results suggest that quality of life is similarly impaired across different dyspepsia subgroups compared with health applying the NDI subscales except for slightly worse scores in reflux-like dyspepsia. There were no striking gender differences but women did report slightly less sleep disturbance. There was no major effect of age on the NDI subscales. However, it must be acknowledged that the subjects included in this study suffered from uninvestigated dyspepsia. It is therefore unknown whether the psychometric properties of the NDI would be different in patients with documented functional dyspepsia and this is currently being actively investigated. Further work is also under way to evaluate in functional dyspepsia the sensitivity of the NDI to change in the clinical trial setting.
In conclusion, we have developed a useful multi-dimensional disease-specific quality of life measure that was reliable and could discriminate dyspepsia from health. The instrument had adequate face and content validity, and correlated modestly with the SF-36, indicating it provides useful disease-specific information. The responsiveness of the NDI in the clinical trial setting now needs to be established.
This study was supported by an unrestricted research grant from Janssen Research Foundation, Beerse.