Purpose: The purpose of this thesis was threefold: 1) To explore the use of text-search methods for identifying rectal cancer patients in large datasets; 2) To examine temporal trends of surgical quality indicators for rectal cancer at a single, tertiary-care institution; 3) To validate the use of administrative codes for identifying rectal cancer patients in population-based datasets.
Methods: 1) A text-search algorithm was developed, validated, and applied to all pathology reports at The Ottawa Hospital (TOH) over a 15-year period. Positive records were confirmed through manual chart review, and a gold-standard cohort of all rectal cancer resections performed at TOH was created. 2) Univariate and multivariate analyses were performed to assess temporal trends and associated factors for four (4) key surgical quality indicators. 3) Previously published methods for identifying rectal cancer resections in population-based datasets were validated using the cohort of patients created in Objective 1 as a gold standard.
Results: 1) The text-search algorithm had a sensitivity and specificity of 100% and 98.4%, respectively. Because of low disease prevalence, positive predictive value (PPV) was 18.6%. 2) The proportion of resections with successful lymph node retrieval improved significantly over the course of the study period. No change was demonstrated for the remaining 3 surgical quality indicators. 3) Previously described methods that utilize procedure codes to identify rectal cancer resections in large administrative datasets had a sensitivity and specificity of 89.5% and 99.9%, respectively, with a PPV of 64.9%.
Conclusions: It is feasible to utilize both procedure codes and text-search methods to identify patients with surgical resections for rectal cancer in administrative datasets. However, these methods are at risk of being inaccurate and resulting cohorts should be validated. Creating large cohorts of rectal cancer patients can be useful for studying a variety of subjects, including surgical quality.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/34300 |
| Date | January 2016 |
| Creators | Musselman, Reilly Patrick |
| Contributors | van Walraven, Carl |
| Publisher | Université d'Ottawa / University of Ottawa |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0022 seconds