Impact and learnings

The outputs planned by the OCAFP were subject to a series of data-related challenges. By no means comprehensive, the list below highlights some of the main data challenges faced during the OCAFP.

Data completeness

Across successive projects, poor data completeness on key data items was likely to have impacted the interpretation of results.

Performance status

As noted earlier in the report, separate efforts were undertaken to isolate the effect of geographic variations in clinical practice on the treatment pathway and short-term mortality. To do so reliably requires adjustment for all factors which were associated with these outcomes and that may have been distributed unevenly between geographies. However, data completeness of performance status at diagnosis was particularly poor, negating its inclusion into the adjustment model. Presented estimates of geographic variation are therefore subject to residual confounding, and this should be borne in mind during interpretation.

Comorbidities

Multivariable-adjusted models account for differences in patient comorbidity via adjustment for their Charlson comorbidity index, which assigns scores to a series of pre-defined comorbid conditions recorded during hospital admissions. Although this variable had high completeness, the ability to identify comorbid conditions was dependent solely on recording within the inpatient setting, potentially biasing scores downward in instances where conditions were predominantly managed and documented in a primary care rather than a secondary care setting. Due to the lack of sensitivity of this methodology in capturing comorbidity, almost four-fifths of patients in some reports were scored as presenting without comorbidity – a figure at odds with the age profile of ovarian cancer patients in England and population-based research publications elsewhere.

Residual disease

Multivariable-adjusted models account for differences in patient comorbidity via adjustment for their Charlson comorbidity index, which assigns scores to a series of pre-defined comorbid conditions recorded during hospital admissions. Although this variable had high completeness, the ability to identify comorbid conditions was dependent solely on recording within the inpatient setting, potentially biasing scores downward in instances where conditions were predominantly managed and documented in a primary care rather than a secondary care setting. Due to the lack of sensitivity of this methodology in capturing comorbidity, almost four-fifths of patients in some reports were scored as presenting without comorbidity – a figure at odds with the age profile of ovarian cancer patients in England and population-based research publications elsewhere.

Data availability

Alongside the issue of data completeness, the OCAFP was challenged by the lack of data availability for certain data items of interest.

Route to diagnosis

The route to diagnosis can be an important determinant of a patient’s treatment pathway, with individuals who present in an emergency setting likely exhibiting more advanced disease than those referred for assessment as suspected cancer cases for from primary care. Unfortunately, adjustment for this important factor was not possible for some of the OCAFP analyses as the data item was unavailable for all years of diagnosis.

Timely cancer registration

Due to the impact of the Covid-19 pandemic on cancer registration, reports produced as part of the OCAFP have a greater lag that would typically have been the case between the most recent year of included diagnoses and the date of publication.

Data accuracy

The OCAFP was not funded in the manner of a national audit to include resources for data cleaning and pursuit of missing data items. The project was therefore reliant on the accuracy of routinely captured data. The feasibility aspect of the project included an assessment where possible of the validity of the data, which were all captured through existing processes during routine management of ovarian cancer cases in England. The project to assess surgical radicality highlighted a lack of data accuracy for HES surgical procedure coding. Whilst there is generally more scrutiny and oversight of routine cancer registration data capture which formed the basis of the profile, treatment and short-term mortality reports, hidden data inaccuracies may have also impacted the analyses published in these reports.

The OCAFP resulted in the release of new ovarian cancer data that:

• provides a more accurate and up-to-date picture of ovarian cancer diagnosis and survival than had been available previously.
• demonstrated statistically significant geographical variation in ovarian cancer treatment.
• brought attention to previously unknown inequalities, such as higher rates of short-term mortality according to factors like deprivation and whether diagnosis occurred at a hospital housing a specialist gynaecological cancer centre.