Development and validation of a new bariatric-specific health-related quality of life instrument ''bariatric and obesity-specific survey (BOSS)''RM Tayyem, JM Atkinson, CR Martin
Department of General Surgery, The Ayr University Hospital, University of the West of Scotland, Scotland, United Kingdom
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0022-3859.143952
Source of Support: None, Conflict of Interest: None
Background: There is a lack of a psychometrically robust bariatric-specific health-related quality of life (HRQOL) tool. Aim: A single centre, prospective study to develop and validate a new bariatric-specific 81-item self-report HRQOL instrument called the Bariatric and Obesity-Specific Survey (BOSS). Materials and Methods: Data were collected from 236 participants. Of these 83 patients were under consideration for bariatric surgery (Group 1) 68 patients had undergone a bariatric procedure (Group 2) and 85 normal healthy participants (Group 3). At baseline, participants were asked to complete the Short Form Health survey (SF-36), Hospital Anxiety and Depression (HADS) scale, Moorehead-Ardelt Quality of life Questionnaire (M-A QoLQ II), and a demographic data sheet. Two weeks following the completion of these five questionnaires, participants were asked to complete BOSS once more along with a feedback sheet. Results: Exploratory factor analysis revealed a multidimensional instrument consisting of 42 items distributed over six domains that addressed various HRQOL aspects and dimensions pertinent to bariatric surgery, and relevant to morbidly obese patients. Further psychometric analysis showed that BOSS has adequate internal consistency reliability (Cronbach α = 0.970), test re-test reliability (ICC = 0.926), construct validity, criterion validity, face validity and acceptability. Conclusion: BOSS thus appears to be a valid and reliable multidimensional instrument that provides a clinically useful and relevant measure to assess HRQOL in patients undergoing bariatric surgery.
Keywords: Bariatric-specific instrument, bariatric surgery, factor analysis, health related quality of life, reliability and validity
A wide range of research activities have contributed to improvement in the clinical management of obesity and particularly bariatric surgery. Traditionally, the most important criterion for success of bariatric surgery is considered weight loss. This has however been challenged and a multidimensional concept of success has been proposed.  One particular aspect that needs further attention is the translation of the research observations of the impact of bariatric surgery on health-related quality of life (HRQOL) into evidence-based medicine (EBM).  Therefore, it is of paramount importance to have an accurate measurement of HRQOL in the field of bariatric surgery, not only for research purposes but also for use in regular clinical practice.  Few studies have reported HRQOL using instruments with psychometrically established properties in morbidly obese patients.  A systematic review analyzed the content validity of HRQOL instruments that are currently used in bariatric surgery and showed lack of satisfactory HRQOL instruments in this rapidly growing field. 
Tayyem et al. (2011) proposed a conceptual framework comprising 20 domains pertinent to bariatric surgery and morbid obesity which can help create a new bariatric-specific HRQOL instrument with sound content validity.  Once validated, the new instrument will enable clinicians to assess the full impact of obesity and bariatric surgery on patients' HRQOL. The study aimed to construct a new self-report questionnaire that can measure with reliability and validity the impact of obesity and bariatric surgery on HRQOL. The new bariatric-specific instrument, is entitled Bariatric and Obesity-Specific Survey (BOSS).
A variety of approaches and methods were used for the initial stage of item generation including literature review, other generic and disease-specific instruments, patient involvement and suggestions, and discussion with health care professionals. This led to the creation of a preliminary item pool of approximately 200 draft items covering all the main domains affected by obesity and bariatric surgery as informed by the conceptual framework.  A multidisciplinary team of various specialties which are involved in the management of the condition including an experienced bariatric surgeon, two university professors with interest in mental health, and wide expertise in HRQOL instrument development and validation, a bariatric specialist nurse, a bariatric dietician, a bariatric psychologist and a bariatric research fellow participated in selecting the items.
A preliminary questionnaire was piloted on a small group of bariatric patients (12 patients) who were excluded from the later part of the study for determination of validity of the instrument. Each item in the preliminary pool was marked according to its readability, relevance to obesity and bariatric surgery, and impact on HRQOL. The highest scoring items representing each domain were selected. The feedback and comments from the expert panel and pilot group were used to modify ambiguous and unclear items, reject irrelevant items, and add new items that resulted in selecting and aggregating 81 items into ten domains to cover the entire framework domains which contributed to the instrument content and face validity.
Gradation of responses
The authors used a five-point Likert scale (zero-four) to record and grade participants' responses to each item. In order to prevent automatic selection of the first or last choice without reading the item statement, the order of responses within the questionnaire was changed. An 81-item instrument was then formatted into a self-report questionnaire which included written instruction to direct participants completing the questionnaire. Instructions asked participants to carefully read each question, and to give their best answer for each item. This working version of BOSS was called BOSS-81.
Scoring of items
Items were scored zero-four with the negatively scored items reverse-scored at the time of instrument scoring so that a high score (e.g. four) always represented a better level of HRQOL, and a low score (e.g. zero) always represented a worse level of HRQOL. The item scores were then standardized to a range of 0 to 100%; e.g. score 0 is standardized into zero%, score 1 is standardized into 25%, etc. Related items, which were verified through factor analysis and reliability analysis, were used to create composite scores (domain score). Composite scores were calculated using unit-weighted approach by computing the mean score of the items representing a particular domain. If an item is missing, then the composite score of the domain is the mean of the remaining items.
Adults between the ages of 18 and 65 years were divided into three groups. Group-one were patients with BMI ≥35 referred to the bariatric clinic for consideration of surgery. Group-two were patients who underwent bariatric surgery. Group-three were normal healthy participants from the hospital staff with any BMI. Participants were excluded if they were unable to give a written consent or who did not have English as their first language. The IRB approved the study.
This being a pilot study, no formal sample size calculations were made. However, the effects of bariatric surgery on HRQOL are very large and as such, significant results would be expected in all parameters. Kline suggested that a 200-sample is an adequate sample size for factor analysis.  Therefore, the authors aimed at recruiting 200 participants as this number is felt to be achievable.
Psychometric evaluation and statistical analysis
Questionnaires dissemination started in the beginning of August 2010 and data collection was terminated at the end of October 2010. In addition to BOSS-81, participants were asked to complete Short Form Health Survey (SF-36), , Moorehead-Ardelt Quality of life Questionnaire II (M-A QOLQ II), ,, Hospital Anxiety and Depression (HAD) Scale ,, and a participant demographic sheet. Two weeks after completing the first set of questionnaires, participants were asked to complete BOSS-81 once again, in addition to a feedback sheet. SPSS software version 16 (SPSS Inc., Chicago, IL, USA) was used to conduct a series of exploratory factor analysis (EFA) and statistical analysis. Correlations were conducted using Pearson Product-Moment correlation. Pearson correlation coefficient r >0.4 was considered satisfactory. The two-tailed test of significance P-value < 0.05 was considered statistically significant. ANOVA was used for score comparison according to age, weight and body mass index (BMI). Other variables were compared using Pearson chi-square test. All analyses were done at 5% significance.
BOSS-81 was administered to a total of 300 adults. Of these, 236 participants responded to the questionnaire (78.7%). Of the 236, 215 participants completed all the 81 items (91.1%). The bariatric surgery group (Group 2) included all four main bariatric procedures consisting of: Laparoscopic gastric bypass, laparoscopic gastric band, laparoscopic gastric sleeve and endoscopic gastric balloon. Of the 236 participants who completed the first set of questionnaires, 116 participants completed and returned BOSS-81 once more 2 weeks later. [Table 1] shows the demographic and clinical characteristics of the three groups. A total of 112 participants provided feedback regarding the questionnaire which showed good face validity of BOSS.
Missed responses for a single item ranged from 0 to 11 (0-4.7%); therefore, no items were discarded. Missed items from a single participant questionnaire ranged from 0 to 8 (0-9.9%) which meant that all participants' questionnaires were included in the EFA. A Kaiser-Myer-Olkin measure of sampling adequacy was 0.954 (well above the minimum of 0.5). Bartlett's test of sphericity was statistically significant (P < 0.001) which confirmed the suitability of data for EFA.
Principal component analysis
A series of EFA using principal component analysis with direct Oblimin rotation was conducted on all items of BOSS-81 because it was anticipated that the extracted factors would be correlated. Missed items were excluded list-wise. Item factor loading was set at coefficient level ≥ 0.4 to suppress small coefficients and to identify the most parsimonious interpretation of the factor structure. Items that did not load on any of the extracted factors and items that did not belong to the rest of the items of a particular factor were deleted. Extracted factors with less than three loaded items and the non-interpretable factors were deleted. The emerging pattern matrix showed well differentiation of 42 items under six extracted factors.
Cronbach's alpha, Pearson's correlation and domain scores
Cronbach Alpha α > 0.7 is considered satisfactory for internal consistency reliability to ensure the coherence of the instrument. [Table 2] shows that Cronbach alpha ranged from 0.785 to 0.975 which exceeded Kline criteria for acceptable instrument internal reliability.  [Table 3] shows that Pearson r two-tailed correlation coefficients for BOSS-42 items were positively correlated and statistically significant and intraclass correlation coefficient analysis was satisfactory as well. Pearson r also revealed a satisfactory convergent validity of BOSS-42 with the theoretically relevant instruments. BOSS-42 domains were positively and significantly correlated with the domains of SF-36 (r = 0.418-0.786, P < 0.001), HAD scale (r = 0.376-0.773, P < 0.001) and M-A QoLQ II (r = 0.387-0.844, P < 0.001). BOSS-42 domain scores correlated weakly and insignificantly with participant feedback scores (r = 0.009-0.127, P = 0.200-0.925). [Table 4] demonstrates statistically significant difference in the score of BOSS-42 domains between the three groups. BOSS-42 domains correlated negatively and significantly with weight (r = −0.291- −0.646, P < 0.001) and BMI (r = −0.303- −0.740, P < 0.001).
The heterogeneous nature of obesity comorbidities, the complex impact of obesity on HRQOL and the various available bariatric procedures make the task of quantifying HRQOL extremely difficult. This may explain why there are only two instruments that have been specifically developed for the use in bariatric patients including M-A QoLQ and Bariatric Quality of Life Index (BQL).
The currently used HRQOL instruments including the above-mentioned bariatric-specific instruments (M-A QoLQ, BQL) are not psychometrically sound if used on bariatric patients. Previous research showed that they are deficient in terms of content validity.  In addition, researchers raised their concerns as a number of the psychometric properties of BQL were questionable such as the heterogeneity of BQL factor structure and the low internal consistency reliability of part of the questionnaire. Furthermore, the original German version was not translated and validated in the English language.  M-A QoLQ, on the other hand, has some questionable psychometric properties. ,, In addition, it only consists of five items which makes its content validity suffer considerably. ,
The authors have created a new HRQOL to evaluate issues pertinent to obesity and bariatric surgery, and to evaluate the HRQOL postoperatively. This new bariatric-specific HRQOL instrument was called Bariatric and Obesity-Specific Survey (BOSS). The development of BOSS did not only include items related to the aspects or domains that may be affected by obesity and improve with weight loss and resolution of obesity comorbidities following bariatric surgery, but also to include items related to those aspects or domains that may be adversely affected by bariatric surgery in line with the multidimensionality concept of success in bariatric surgery. 
The initial stage of item generation included a variety of approaches and methods to make sure that it has a sound content validity. Item selection began by identifying a conceptual framework of HRQOL domains which might be affected by the condition or intervention under question. A master list of items that came up in previous research pertinent to bariatric surgery was created and then refined. Item selection also utilized related and relevant items from previously constructed HRQOL instruments. Clinicians, other health care providers and patients contributed in generating the item pool by describing the most frequent and relevant complaints of bariatric patients and suggested what the most important effects of obesity and its intervention (Bariatric surgery) are.
The working version of the instrument (BOSS-81) was created in line with the standard techniques used in this field. It consisted of 81 items to cover all the domains relevant to the bariatric patient as suggested by the conceptual framework. 
The satisfactory face validity, high response rate (78.7%) and high completion rate (91.1%) indicate that this instrument is user friendly, easy to use, and highly acceptable.
During EFA, it was ensured that item redundancy was eliminated without compromising the comprehensiveness and the content validity of the questionnaire. EFA reduced the original 81-item instrument by selecting 42 items which loaded on six extracted factors. The final version of BOSS was called BOSS-42. The items loaded on each extracted factor correlated well with one another and measured the same dimension or domain enabling the interpretation the extracted factors based on their item content, namely: Incapacity, Work and Wellbeing, Social Function, Appearance and Health, Eating Patterns and Sexual Health.
BOSS-42 could be seen as consisting of two distinctive but still closely related major subscales: Generic and bariatric-specific subscales. Factors one, two and three were more generic domains that could be relevant to the bariatric and non-bariatric patient. Factors four, five and six were more specific to the bariatric patient.
The use of volunteers in the control group during the development of HRQOL instruments is a standard practice. However, the use of hospital staff could potentially introduce sampling bias compared to population control. On the other hand, the response rate is usually higher among hospital controls compared to the general population. Further, given the pilot nature of the study, the use of hospital staff offers an additional context-specific comparison in that the geographical demographic of the majority of staff in the control group is more likely to be similar to the clinical group, given that the clinical group and the staff group are generally from the same region, an artifact of the lower general level of occupational mobility in this part of the UK. For these reasons, the researchers decided to use hospital staff in the control group.
To conclude, BOSS-42 showed satisfactory factorial structure with excellent internal consistency reliability, test re-test reliability, convergent validity, divergent validity, concurrent validity and face validity in a population of severely and morbidly obese patients who are either under consideration for bariatric surgery or already have had one. In addition, this valid and reliable instrument, which requires only 10 minutes to complete (as BOSS-42 completion time has been measured on a small group of patients), showed good acceptance by its users.
The satisfactory psychometric properties of BOSS-42 make it ready for use in clinical practice by bariatric teams to assess the HRQOL of obese patients referred for bariatric surgery and evaluate their HRQOL postoperatively. It can also be used in research settings as an accurate instrument to assess HRQOL in patients with severe and morbid obesity in bariatric surgery setting. Nevertheless, it is still utilizable in studying obesity in general.
[Table 1], [Table 2], [Table 3], [Table 4]