How Prostate Cancer Patients are Surveyed may Influence Self-Reported Sexual Function Responses

Abstract

Background The side effects of prostate cancer treatment include decreases in sexual function, hence, the way patient reported outcomes are collected may affect the quantity and quality of responses. Aim To determine the effect that different survey modes (email, telephone, or mail) had on the quantity of missing data and self-reported function following treatment. Methods Men newly diagnosed with prostate cancer and enrolled in the Victorian Prostate Cancer Outcomes Registry formed the study population. The Expanded Prostate Cancer Index Composite (EPIC-26) survey instrument was administered approximately 1 year after their initial treatment. EPIC-26 measures self-reported function in the sexual, urinary, bowel, and hormonal domains. Multivariable regression models were used to examine effects of survey mode, adjusting for age, residence, socioeconomic status, diagnosing institute type, risk group and primary treatment modality. Outcomes The percentage of patients for whom a domain score could not be calculated due to missing responses and the functional score within each domain. Results Registry staff attempted to reach 8,586 men eligible to complete the EPIC-26. Of these, 4,301 (50%) returned the survey via email, 1,882 (22%) completed by telephone, and 197 (2.3%) by mail. 2,206 (26%) were uncontactable or did not respond. Email responders had the highest proportion answering all 26 questions (95% vs 87% by phone and 67% by mail). The sexual function score was unable to be calculated due to missing responses for 1.3% of email responders, 8.8% by phone, and 8.1% by mail. After adjustment for patient and disease factors, phone responders were almost 6 times more likely than email responders to have a missing score in this domain, odds ratio = 5.84 (95% confidence interval: 4.06–8.40). The adjusted mean functional score (out of 100) was higher for those responding by phone than email or mail across all domains. The largest adjusted difference between phone and email was observed in the hormonal domain (mean difference 4.5, 95% confidence interval: 3.5–5.4), exceeding the published minimally important difference for this score. Clinical Implications Studies that ask questions regarding sexual health and use multi-modal data collection methods should be aware that this potentially affects their data and consider adjusting for this factor in their analyses. Strengths and limitations A large study sample utilizing a widely available survey instrument. Patient specific reasons for non-response were not explored. Conclusion Completion mode effects should be considered when analyzing responses to sexual function questions in an older, male population.