The instrument
The DBMA is a self-report questionnaire in which the patient must rate a number of medical conditions, if present. The original version of the DBMA included 23 common medical conditions [3]. The instrument was later modified, and the number of conditions was reduced to 21 [8]. The instrument used for this study was the latter version with an additional item exploring depression/anxiety. In the original article of Bayliss and colleagues [3], depression was assessed as a separate outcome measure but due to its importance in patients with multimorbidity [9] we decided to include it as the 22nd item. For each condition present, the patient assesses the degree to which the condition limits his/her daily activities on a five-point descriptive scale in which the first level, "not at all", has a weight of 1, and the fifth level, "a lot", has a weight of 5; all other conditions are scored zero. The questionnaire also allows patients to add medical conditions not included in the list and to score them in the same way. The total score is the sum of the limitation from all conditions, including those added by the patient. The questionnaire is written in a simple language understandable to patients using short sentences. The English version of the questionnaire is shown in more detail in an additional file (see Additional file 1).
Translation
The original version of the DBMA [8] was translated in French Canadian following a procedure inspired by Vallerand [10] and Hébert [11]. Although forward-backward translation is a method commonly used for translating questionnaires in the field of patient-reported outcomes, given the simple nature of the questions we estimated that this procedure was not necessary. A bilingual translator translated the original version into French. A panel of experts examined both versions and made revisions to further adapt it to Quebec French. The translated version (DBMA-Fv) was then submitted to a panel of experts (three physicians and one nurse). The panel verified that the disease list of the instrument faithfully reflected the English version, and that the language was well adapted to the one used by patients seen in primary care. If any discrepancy was found, modifications and specifications were made to the questionnaire following expert recommendations.
Cognitive interview
The DBMA-Fv was first assessed during cognitive interviews to make sure all items were clearly written, without ambiguity, and in a language that could easily be understood by the target population. To achieve this, two of the authors (MF and CH) recruited a convenience sample, as recommended by Dillman [12], from their consulting patients. The sample included 10 patients (7 women) aged from 19 to 79 years (mean ± SD, 63.3 ± 16.9) suffering from various chronic diseases, in a clinical practice. Patients agreeing to participate provided written consent. At each interview, an observer was also present. The interviewer and the observer were research assistants trained by one of the authors of the study (MEP). At the time of the interview, participants were asked to read the questions of the DBMA-Fv out loud and to express any thoughts or doubts they had, or to highlight ambiguities they may find while answering the questions. The interviewer led the activity while the observer had the task to note all the questions and comments of the participants. The maximum duration of interviews was 30 minutes.
After each interview, the interviewer and the observer met three authors of the study (MEP, MF and CH) in order to discuss the comments expressed by the participant. Questions identified as unclear were clarified by the research team for the next participant and so on until no further change was required. The questionnaire, modified according to participant input, was then considered to be in its final format and applied in the validation study. The final version of the DBMA-Fv validated in this study is shown in more detail in an additional file (see Additional file 2).
Validation study
The patients were recruited by a research assistant during consecutive consultation periods from the waiting room of the Family Medicine Unit (FMU) of a regional health centre (Centre de santé et de services sociaux de Chicoutimi) in Saguenay (Quebec) Canada. We aimed to recruit a sample of 100 participants to test the instrument [12]. Patients solicited were asked to provide written consent and to complete a short questionnaire to determine eligibility. Participants had to be at least 18 years old, patients at the FMU for more than two years and able to read and write in French. Pregnant women, patients with an unstable acute condition, or having an uncontrolled psychiatric disease, a cognitive disorder, or unable to provide informed consent were not included in the study. Eligible patients were asked to complete the DBMA-Fv questionnaire without assistance while they were in the waiting room of the FMU (T1). Filling out the questionnaire took no more than 15 minutes. Participants agreed to complete the same questionnaire sent to their homes two weeks later (T2) and provided consent to the research team to access their medical records.
A second copy of the questionnaire along with a letter inviting participants to answer was sent to those who did not return the questionnaire one week after the date of the first mailing. Participants who did not answer the questionnaire one month after their recruitment were called up by the research coordinator of the study to point out the importance of their participation or to accept their withdrawal from the study.
A trained research nurse reviewed the medical records and completed a data extraction grid, including the list of chronic diseases and past medical history of each participant. This data was used to complete, first, a DBMA-Fv based on chart review and also to score the CIRS, another multimorbidity index [4], to assess concomitant validity. The measurement properties of the CIRS and the validity of this method for scoring the CIRS have been described elsewhere [1, 13, 14]. Briefly, the CIRS uses a scoring system that encompasses 14 anatomical domains (cardiac, vascular, hematological, respiratory, ophthalmologic-otorhinolaryngologic, upper gastrointestinal, lower gastrointestinal, hepatic-pancreatic, renal, genitourinary, musculoskeletal-tegument, neurological, endocrine-metabolic-breast, and psychiatric) and assigns a value from 0 (no impairment of the organ or system) to 4 (extremely severe impairment that is a life-threatening condition) to determine a severity score for each domain. In the case of multiple conditions affecting a particular domain, the condition with the highest score determines the score given to the domain. The global score is the sum of each domain's score.
Data analysis
Missing data were analyzed by number of incomplete questionnaires and by unanswered items in the questionnaire. Questionnaires with one missing value or more were considered incomplete. To evaluate the test-retest reliability, we calculated the intraclass correlation coefficient (ICC) of the total score of the DBMA-Fv. After checking for normality of the distributions, the Pearson correlation was used to measure the concomitant validity of the DBMA-Fv compared to the CIRS. The 95% confidence interval (CI 95%) for the correlation coefficient of the DBMA-Fv vs. CIRS relationship at T1 and T2 was calculated using Fisher's Z transformation [15]. Analyses of sensitivity and specificity with a 95% CI were carried out for each medical condition included in the DBMA-Fv when numbers were sufficient. The 'gold standard' used to calculate sensitivity and specificity was the information contained in the medical records. These analyses compared the questionnaires completed by patients to those completed by the research nurse using patient charts. Since the DBMA was originally developed as a mail survey, we used the results of the mail questionnaire at T2 to assess diagnostic validity. Conditions found in an insufficient number of subjects (five or less) to provide a good estimation of the sensitivity were not included. Ill-defined diagnoses (back pain, stomach problem, colon problem, poor circulation) unlikely to be found in the chart review were also excluded from the estimation of sensitivity.
The SPSS 16.0 Software was used for data analysis. Questionnaires with missing values were excluded from analyses of test-retest reliability and concomitant validity. Approval for this project was obtained from the ethics committee of the Centre de santé et de services sociaux de Chicoutimi in February 2009.