RESUMO
BACKGROUND: To understand our performance with respect to the collection and reporting of patient-reported outcome (PRO) measure (PROM) data, we examined the protocol content, data completeness and publication of PROs from interventional trials conducted at the Royal Marsden NHS Foundation Trust (RM) and explored factors associated with data missingness and PRO publication. DESIGN: From local records, we identified closed, intervention trials sponsored by RM that opened after 1995 and collected PROMs as primary, secondary or exploratory outcomes. Protocol data were extracted by two researchers and scored against the SPIRIT-PRO (PRO protocol content checklist; score 0-100, higher scores indicate better completeness). For studies with locally held datasets, the information team summarized for each study, PRO completion defined as the number of expected (as per protocol) PRO measurements versus the number of actual (i.e. completed) PRO measurements captured in the study data set. Relevant publications were identified by searching three online databases and chief investigator request. Data were extracted and each publication scored against the CONSORT-PRO (PRO manuscript content checklist; scored as SPIRIT-PRO above). Descriptive statistics are presented with exploratory comparisons of point estimates and 95% confidence intervals. RESULTS: Twenty-six of 65 studies were included in the review. Nineteen studies had accessible datasets and 18 studies published at least one article. Fourteen studies published PRO results. Most studies had a clinical (rather than PRO) primary outcome (16/26). Across all studies, responses in respect of 35 of 69 PROMs were published. Trial protocols scored on average 46.7 (range 7.1-92.9) on the SPIRIT-PRO. Among studies with accessible data, half (10/19) had less than 25% missing measurements. Publications scored on average 80.9 (range 36-100%) on the CONSORT-PRO. Studies that published PRO results had somewhat fewer missing measurements (19% [7-32%] vs 60% [- 26 to 146%]). For individual PROMs within studies, missing measurements were lower for those that were published (17% [10-24%] vs 41% [18-63%]). Studies with higher SPIRIT-PRO scores and PROs as primary endpoints (13% [4-22%] vs 39% [10-58%]) had fewer missing measurements. CONCLUSIONS: Missing data may affect publication of PROs. Extent of inclusion of SPIRIT-PRO protocol items and PROs as primary endpoints may improve data completeness. Preliminary evidence from the study suggests a future larger study examining the relationship between PRO completion and publication is warranted.
RESUMO
Importance: The stratification of indeterminate lung nodules is a growing problem, but the burden of lung nodules on healthcare services is not well-described. Manual service evaluation and research cohort curation can be time-consuming and potentially improved by automation. Objective: To automate lung nodule identification in a tertiary cancer centre. Methods: This retrospective cohort study used Electronic Healthcare Records to identify CT reports generated between 31st October 2011 and 24th July 2020. A structured query language/natural language processing tool was developed to classify reports according to lung nodule status. Performance was externally validated. Sentences were used to train machine-learning classifiers to predict concerning nodule features in 2,000 patients. Results: 14,586 patients with lung nodules were identified. The cancer types most commonly associated with lung nodules were lung (39%), neuro-endocrine (38%), skin (35%), colorectal (33%) and sarcoma (33%). Lung nodule patients had a greater proportion of metastatic diagnoses (45 vs. 23%, p < 0.001), a higher mean post-baseline scan number (6.56 vs. 1.93, p < 0.001), and a shorter mean scan interval (4.1 vs. 5.9 months, p < 0.001) than those without nodules. Inter-observer agreement for sentence classification was 0.94 internally and 0.98 externally. Sensitivity and specificity for nodule identification were 93 and 99% internally, and 100 and 100% at external validation, respectively. A linear-support vector machine model predicted concerning sentence features with 94% accuracy. Conclusion: We have developed and validated an accurate tool for automated lung nodule identification that is valuable for service evaluation and research data acquisition.