GENERAL COMMENTS

This is a well constructed and well presented analysis of the national cancer patient experience survey conducted in 2011/12. The authors aim to interrogate the so-called "London effect" seen in this and subsequent surveys, whereby cancer patients treated in London generally report a poorer experience of care than those treated elsewhere in England, even in large metropolitan areas.

I would like to see the following comments addressed:

1. Introduction

2. The authors set out the 4 hypotheses they consider may underlie the relatively poorer experience of care reported by cancer patients in London. They should consider a fifth possibility, which is that cancer patients in London are more likely to have their care fragmented across multiple hospitals compared to those treated in the more consolidated health care provision elsewhere in England's towns and cities. This could also explain why London's cancer patients report relatively worse care than those in the general hospital inpatient survey in London, as the latter patients would be more likely to be inpatients for conditions only requiring care from a single institution.

3. Do the authors have access to data that could address this fifth possibility? Could they control for it by taking Q 66 (hospital and community staff always worked well together) as a surrogate for those patients who had the most fragmented cancer pathway (although this does not specifically address whether one or more hospitals was involved in the patient's care).

4. Methodology Given that the NCPES was not constructed to test the hypothesis that London hospitals provide poorer care than hospitals elsewhere in England, could they explain more clearly to the general reader, who will not be an expert in statistics, how they controlled for multiple testing? Should there be an explicit "false discovery rate" limit set? I would advise seeking input from a statistician on this point.

5. Results - (page 6 line 2) there is a potential error – the median age is stated as 65 vs 64 years in the text (for Rest of England vs London) but in table 2 it is stated as 66 vs 65 years. Which is correct?

6. Results - paragraph on "Variation in experience adjusted for patient case-mix, page 10), the authors should comment in more detail on the nature of the 8 questions (according to the abstract) whose differential effect disappears for London hospitals once the results are adjusted for case mix, and on the significance of this. From my reading of table 4a, three of these questions are about communication of information (Q's 13, 36, 67), two are about patients having trust in staff and being treated with dignity (Q's 37, 44, 48) and the remaining two are about treatment of pain (Q 50) and overall rating of care (Q 70). The major difference between the case mix of patients treated in London hospitals versus the rest of England is in ethnicity - as shown in table 1, where only 84% of the London cancer patient population who responded to this survey is reported as "white" compared to 97.9% for the rest of England, (the differences in age distribution, gender and case mix seem trivial). This implies that cancer patients from black and ethnic minority groups are disproportionately receiving poorer care in these important areas, not only in London but elsewhere in the country. The authors should comment whether their data shows that the whole country is providing differentially poorer care for BME cancer patients in these areas.

7. Results - paragraph on "London variation for cancer patients and general in-patients, page 11) , the authors should include a comment on whether this difference might be due to differences in the more fragmented nature of a cnacer patient's treatment pathway, usually involving several hospitals under the current configuration of cancer services in London (see comments under “introduction” above).

8. Discussion – second paragraph, the authors should give more detail about what the sources of variation in patient experience between California and the rest of the United States in reference 17 that are relevant to this study.