Using Cognitive Interviewing to Assess, Improve, and Validate Health-Related Quality of Life Measures in Women Diagnosed with Ovarian Cancer | BMC Women’s Health


Fourteen participants took part in individual telephone or video interviews lasting an average of one hour and fifteen minutes (range of 30 min to one hour 50 min). Interviews for one participant were conducted twice for convenience, and the initial interview that provided pilot data was also incorporated into the dataset. Feedback from the pilot interview helped modify the statement response scale to be used in subsequent interviews. Participants varied by age, occupation and marital status, but all lived in a metropolitan area. Seven participants had been diagnosed more than five years previously and two participants had received their diagnosis within a year of the interview. Of the 14 respondents, five participants were in active treatment due to recurrence of their cancer (Table 2).

Table 2 Description of study participants (n=14)

Draft questionnaire reviewed using CI and CTA

The draft items were categorized into the domains identified in our previous qualitative study. Related to diagnosis and treatment (chemotherapy, surgery and complementary therapies), relationships with family/friends, financial aspects, health services and interactions with health professionals, and coping strategies. Diagnostic and treatment challenges have been documented. Key themes such as physical well-being, emotional well-being, relationships with family/friends, health services and interactions with health professionals, and coping strategies were identified as being linked to life with a diagnosis of CO throughout the clinical course. A 5-point Likert response scale was created for the items.

The draft questionnaire acknowledged all data from the original qualitative study and included a series of questions to gather information about participants’ demographic and cancer histories. The next four sections focused on diagnostic and treatment challenges and included jumping questions that directed participants to the appropriate sections based on the answers given, with the response scale measuring the severity of the challenges. The remaining sections containing items related to HRQoL, satisfaction with services, and coping strategies were provided with a response frequency scale (Table 3).

Table 3 Total number of items per section of the questionnaire before modification based on CI data

Refinement of HRQOL elements

A multi-step analysis process was then used to rank how the items could be changed, using four broad themes. The refinement of HRQOL elements is illustrated by examples below.

Intent and clarity of the article

Responses for several items were inconsistent with the intent of the item, making it impossible to interpret the researcher’s goals. More specifically, for some items, participants said: “I don’t understand what this means” or “I didn’t understand the meaning of the question”. For example, an original item was “I felt frustrated during and/or after receiving treatment”. A majority of participants understood ‘frustrated’ in terms of the treatment received, but the researcher’s intent was to measure ‘frustrated’ in relation to a participant’s activities while undergoing treatment. With feedback from participants, the item was changed to “I felt frustrated at not being able to participate in usual activities during and/or after treatments”.

Label of the article

Difficulties in understanding the meaning of certain items were identified. For example, the statement “I feel valued because I can still contribute to the job market” confused the participant. A particular response was “I don’t work, but I still feel valued. I would put not applicable for that one. But I would say strongly disagree because I feel valued because of the work.” To better structure it, the item was changed to “I felt valued because of the work I can do (home, work)”.

Other items were considered vague by participants and difficulties arose in communicating the researcher’s intent regarding the items to participants. For example, the item “My family/friends reacted unexpectedly to my illness” could have been interpreted as a positive or negative experience. With feedback from participants, the item was changed to “My family/friends reacted unexpectedly (negatively) to my illness”.

Item Relevance and Context

Some items had little relationship to the age of the participants. These items included: “I was bothered by the way my body changed” and “I felt less feminine because of my illness.” In addition, several items specific to emotional well-being were not relevant for participants who were under observation and not receiving treatment. For example, when asked “I felt bad because of the side effects of the treatments I had,” participants responded by saying, “It’s hard because during the chemo, I got felt bad and sick, but since the chemo I’ve had no problems.”

Finally, the complexity of the formulation of certain items made it difficult for the participants to answer with precision. For example, a quirky item was “My family/friends are generally supporting me right now.” However, depending on the participant’s perspective, the support received from family and friends could have been different. Thus, the item was separated into two items. In particular, one participant could not understand the context of the item in the social domain. Initially, the statement was developed as follows: “I felt socially isolated because of my illness. However, the participant could not answer in what context the word “socially” meant. The item was then changed to ‘I had difficulty socializing with people because of my illness (eg, at work, in public)’.

Refinement of items describing contextual factors for HRQOL

The qualitative data set contained additional themes on illness/treatment and financial issues, communication with healthcare professionals, and coping strategies. These themes were developed into three sets of questions reflecting the contextual factors of HRQoL: patient symptoms, satisfaction with health services, and strategies for self-sufficiency and resilience. These elements have been refined using the same multi-step analysis process and examples are shown below.

Label of the article

Some items concerned technical terms that were difficult to understand. For example, concerns have been raised over specific terminologies such as ‘cancer recurrence’, ‘mucositis’, ‘full cycle of chemotherapy’, ‘complementary therapies’. In response, clear definitions were constructed (Supplementary File 3).

Item Relevance and Context

The relevance of specific items depended on age. For example, questions about “sudden onset of menopause” and “inability to have children” had no impact on the well-being of many participants because they had already experienced menopause before receiving their CO diagnosis. Thus, a “not applicable” response column was added to the patient’s symptoms section. As there were some irrelevant items in the questionnaire and after joint agreement of the supervision team, a timeline was provided for each section.

The context of the item “There is a lack of financial assistance with practical support” was unclear, as one participant stated, “I didn’t have to look for that. So, I really don’t know. I didn’t need it but listening to other ladies, I think there is probably a need for it”. However, the researcher’s intention was to find out whether participants had difficulty accessing services or not. The item was modified to read: “There was a lack of practical support offered to me”. In addition, the item “I maintained a feeling of gratitude” seemed out of context to one participant. The item was changed to “I have maintained a sense of gratitude for what I am able to do/achieve”.

Construction of the questionnaire

Some participants mentioned that the type of response format for the HRQoL and well-being items could be improved as responses could have reflected frequency, rather than just agreeing/disagreeing with a statement. For example, participants responded using phrases such as “sometimes” or “sometimes” and found it difficult to simply agree/disagree with items. Thus, an Always/Often/Sometimes/Rarely/Never frequency format was applied to sections that measured aspects of HRQoL and well-being.

In summary, the CI and CTA processes informed substantial item reconstruction to achieve acceptable content validity for this group of participants. The questionnaire as a whole was restructured: of which 67 items were modified, 10 items were condensed and merged into appropriate sections, 66 items were deleted and 15 items were added. The items describing HRQOL have been grouped together, forming an HRQOL instrument named the OVA Cancer Health-Related Quality of Life Scale (OVAQOL). This scale includes items that conceptually contributed to three HRQOL domains: physical well-being (n = 7), emotional well-being (n = 21), and social well-being (n = 10). The rich set of original qualitative data and comprehensive interview and analysis processes informed the development of question sets on demographic characteristics, disease and treatment status, patient satisfaction at the regard to health services and patient resilience (Table 4).

Table 4 Final structure and number of items in the questionnaire with examples of OVAQOL items

Comments are closed.