The study sample was randomly drawn from the 20% Medicare claims files for the period August 1, 2008, through December 31, 2008. Medicare beneficiaries who met the following criteria were preliminarily eligible: (1) an inpatient claim for radical prostatectomy (International Classification of Diseases, Ninth Revision 9 [ICD-9] SX code of 605 in any position), (2) a prostate cancer diagnosis during the admission when the prostatectomy was performed (ICD-9 diagnosis codes 185, 1850, 2365, 2395, 2334, 19882, V1046, or V1045), (3) a surgeon's claim for the procedure (Current Procedural Terminology [CPT] codes 55810, 55812, 55815, 55840, 55842, 55845, 55866, 55899, or 55899), (4) at least 66 years old at the time of surgery (to have 12 months of preoperative claims available), (5) no health maintenance organization participation during 2008, and (6) lived in the United States. Exclusion criteria were having died before selection or residence in a nursing home. A list including coded identification numbers of potential participants was sent to the Center for Medicare and Medicaid Services (CMS), which removed decedents according to Social Security records and returned names and addresses to the Center for Survey Research in Boston, MA.
Selected patients were first sent a letter signed by the CMS Privacy Officer, indicating they had been chosen for a survey of surgical decision making. The letter explained that participation was voluntary and would not affect their Medicare benefits. Potential participants were asked to call a toll-free number to decline.
Two weeks later, participants who had not called were sent a cover letter, a questionnaire, a postage-paid return envelope, and a $5.00 cash incentive. A postcard reminder was sent 2 weeks later. Nonresponders subsequently received another packet, excluding the payment. Nonresponders were also called to ensure they had received the materials, answer any questions, and encourage participation. The first survey was mailed on November 12, 2009, and the last survey received on March 31, 2010. All materials were provided in English and Spanish.
Survey instrument development included cognitive interviews and usability testing with members of the target population. Items were cognitively tested to learn whether most respondents understood them as intended and whether their answers were good reflections of what they had to say. Focus groups were conducted to test instrument usability and to ensure the suitability of the materials included in the mailings.
The relevant sections of the questionnaire for this study asked participants to confirm that they had surgery for prostate cancer in the month and year specified in the claims. Participants were then asked, “In laparoscopic surgery, several small openings are made to do the operation instead of one big incision. Was this prostate surgery done as laparoscopic surgery?” If the answer was yes, they were asked, “During this prostate surgery, did the surgeon use a robot to help with the operation?” For these questions, “I'm not sure” was one of the responses offered.
For the analyses in this article, the two key outcome questions were “Since this prostate surgery, how much of a problem have you had with leaking or dripping urine?” and “Since this prostate surgery, how much of a problem have you had with sexual functioning, such as problems with erections?” Possible responses were “No problem,” “A very small problem,” “A small problem,” “A moderate problem,” and “A big problem.”
Given the interest in minimizing respondent burden, the survey focused on the cardinal issue of adverse effect bother rather than frequency.18
These two questions were taken from the Prostate Cancer Outcomes Study (PCOS),19
although slightly modified on the basis of cognitive testing. The response frame for both questions was identical with that of the PCOS. These two questions from PCOS were in turn adapted from an earlier survey conducted by our group among Medicare beneficiaries undergoing radical prostatectomy.20
Similar bother items with the same response options are included in the University of California at Los Angeles Prostate Cancer Index.21
The survey finished with demographic questions covering age, self-rated overall and mental health (both rated as excellent, very good, good, fair, poor), education, marital status, and race/ethnicity.
Analytic techniques included descriptive statistics and simple correlations. The Pearson χ2
test was used to compare the distributions of categorical variables, and the Wilcoxon rank sum test was used to compare distributions of follow-up time after surgery. To adjust for potential confounders, logistic regression was used with the adverse effect variables dichotomized a priori (an adverse effect was coded positively if the response was “A moderate problem” or “A big problem,” and coded negatively for “No problem,” “A very small problem,” or “A small problem.”). Surgical procedure (RALRP v
ORRP), as well as potential demographic confounders, were included in the models as independent variables. These primary analyses were confirmed by using ordinal regression with the same dependent and independent variables.22
Analyses were conducted by using the Predictive Analytics Software Statistics 18 analytic package. The study was approved by the institutional review boards at the University of Massachusetts (Boston) and Dartmouth Medical School.