To the best of our knowledge, this is the first published report of the development and validation of a questionnaire for estimation of the breast cancer total, patient and provider intervals and the correlated factors with delays. Due to the association between delay and prognosis for breast cancer
, it is important to quantify it and identify correlated factors. This is especially relevant in developing countries, where the majority of patients are diagnosed in advanced stages, as in Mexico
The instrument was developed, modified and validated using standardised test-construction methods
[28, 34, 40]. The results of the current study show that the instrument has good face validity, comprehensibility, patient acceptance, and content validity. It has acceptable internal consistency in most dimensions (considering the social nature of most of the items) and very good reliability for most items
Although the two modules of the questionnaire were intertwined to facilitate its administration for this study, they would be easy to separate if someone wished to measure only the time intervals in another context. Nonetheless, we found that this intertwining of modules facilitated the questionnaire’s administration in our population, easing the flow of the interview.
Reliability was high for both total and patient intervals. In agreement with other studies, this shows that patients tend to recall the precise time when they first discovered their symptoms
[12, 31] or at least the month and year of this discovery
. Nevertheless, the provider interval did not show very good consistency. Because the test-retest measurements were taken with a 3-month separation, the variation in the estimation of this interval may be explained by recall bias. Apparently, memory was more affected in our participants in relation to the first medical consultation than in relation to the beginning of the problem. It is likely that the passage of time makes it harder for patients to recall dates and events that occurred prior to their admission, especially as medical consultations might blend with in-hospital consultations. This process could be even more challenging for patients who have already started chemotherapy (as was the case of most of our retest participants) because common secondary effects of chemotherapy include memory loss and difficulty concentrating
. These findings suggest that to minimise recall bias when assessing delay, the patients should be interviewed as early as possible, as has been suggested in previous studies
Recall bias also seems to explain the poor external consistency scores of some other questionnaire items, including symptoms present when the patient first arrived at INCAN, the diagnosis offered by the first doctor consulted, the tests requested by this first doctor, and the time that passed from identification of the problem until the patient told someone about the problem. The remaining poor test-retest consistency scores could be explained by changes in these items over time. The first interview took place before the cancer diagnosis, while the second interview took place after treatment had begun. Items with poor consistency that could have changed over time include use of alternative medicine, knowing a person with cancer and knowledge of recommended breast cancer screening practices.
In regard to the length of the intervals reported in our findings, the median total interval was very prolonged for our study population (median: 234.5 days) and the main delays seem to be presenting within the provider interval (median: 151 days). This is similar to findings of studies that have been done in other Latin American countries. For example, another Mexican study that was done with a small sample of breast cancer patients (n=32) with public health insurance reported a mean total interval of 8.4 months
. A Colombian study of 1106 breast cancer patient files reported a median time from the first medical consultation to diagnosis confirmation (diagnosis interval) of 91 days
 and a Brazilian study with 104 patients reported a median diagnosis interval of 6.5 months
Another matter that we would like to briefly discuss is related to whether a self-complete tool could have provided a more objective way of administering the questionnaire. Such a tool could certainly reduce inter-observer variability and response bias
. It has been shown, however, that self-complete tools are difficult for undereducated populations (like our study population) to use
[36, 37]. This strategy would have most likely yielded a high rate of incomplete and inadequately filled-out questionnaires.
One limitation of our study is that external consistency measures are not available for some items because these items were identified as relevant only after the second pilot study. Because these items are “patient perceptions of barriers within the provider interval”, we hypothesise that they are likely to have reliabilities similar to those found for “perceptions of barriers within the patient interval” (moderate; between 0.4 and 0.75).
One more limitation was the impossibility of constructing scales for the social network of support for medical attention and health service utilisation. Another limitation was the lack of correlation between most of the social network items and the delay intervals. Nevertheless, we think that it is still premature to decide which of these items should be kept or discarded before the instrument is tested in a greater and more heterogeneous sample with higher levels of education, formal employment and higher socioeconomic status.
A disadvantage of the instrument is that its application is time-consuming, especially for older participants and those with low levels of education. On the other hand, it has high acceptability for patients, especially if it is administered when they first arrive at the hospital, before they see the breast specialist.