National Evidence on the Use of Shared Decision Making in Prostate-specific Antigen Screening

Paul K. J. Han, MD, MA, MPH; Sarah Kobrin, PhD; Nancy Breen, PhD; Djenaba A. Joseph, MD, MPH; Jun Li, MD, PhD; Dominick L. Frosch, PhD; Carrie N. Klabunde, PhD


Ann Fam Med. 2013;11(4):306-314. 

In This Article


Data Source and Study Population

We used data from the 2010 National Health Interview Survey (NHIS), an annual survey of the public conducted by the National Center for Health Statistics, and a principal source of health information on the US civilian, noninstitutionalized population.[29] The NHIS collects information on sociodemographic characteristics, health status, and health care utilization, and in 2010 the NHIS included a Cancer Control Supplement, developed and cosponsored by the National Cancer Institute and the Centers for Disease Control and Prevention. The NHIS obtains a nationally representative sample of US adults using complex random, stratified, clustered, multistage sampling with oversampling of African-American and Hispanic households. In-person interviews are conducted by US Census Bureau interviewers. Methodological details are at

Male respondents were asked whether they ever had a PSA test; those responding yes were asked the date and main reason for their most recent test: (1) part of a routine physical examination/screening test, (2) because of a specific problem, and (3) other. Our analysis included men who reported either testing for routine screening or no past testing. We included men aged 50 to 74 years to reflect the most conservative screening age recommendations of the clinical practice guidelines at the time.[11,12] We excluded men who reported testing for a specific problem, a personal history of prostate cancer, or more than 5 PSA tests in the past 5 years, given that high-frequency testing is almost always undertaken for diagnostic rather than screening purposes.

Conceptual Definitions and Measures of Shared Decision Making

Definitions of shared decision making vary in conceptual breadth and number of essential elements.[30–33] We defined shared decision making in accordance with ACS, AUA, and USPSTF guidelines for PSA screening that existed during the 2010 NHIS; all recommended that patients be provided with key information about the advantages, disadvantages, and uncertainties of PSA screening, and that patients' preferences be considered.[12,13,16] These essential elements of informed decision making[31,32] constitute shared decision making when facilitated by health care clinicians.[14–16] All 3 guidelines in 2010 recommended shared decision making, assigning responsibility to "health care providers,"[16] "clinicians,"[13] or "physicians"[12] to inform and involve patients in screening decisions. Correspondingly, we focused on shared decision making and the extent to which physicians provided key information to patients.

Physician discussion of advantages and disadvantages were measured using 2 items: (1) "Did a doctor ever talk with you about the advantages of the PSA test?" and (2) "Did a doctor ever talk with you about the disadvantages of the PSA test?" Physician discussion of uncertainty was measured by a single item: "Did a doctor ever tell you that some experts disagree about whether men should have PSA tests?" Scientific uncertainty is not the only uncertainty pertaining to medical interventions,[34] and expert disagreement is not the only manifestation of scientific uncertainty. Scientific uncertainty, however, is central to the controversy surrounding PSA screening,[12,13,16,20,23,25,35] and expert disagreement is a final manifestation of this uncertainty. Questions used yes or no response categories and were asked of all respondents including nonscreened men (for screened men, questions were preceded by the phrase, "Before you had the PSA test…). All questions underwent 2 rounds of cognitive testing and modification by expert staff. We used responses to the advantages, disadvantages, and uncertainty items to create a composite variable, extent of shared decision making, with 5 categories ranging from no elements to all elements, reflecting the discrete information elements discussed by physicians.

Sociodemographic and Health-related Variables

The analysis included sociodemographic and health-related factors examined in previous studies:[18,19,36–41] health insurance, presence of usual source of medical care, family history of prostate cancer, self-reported health status, number of chronic diseases. Physician recommendation for PSA screening was measured by the question, "Has a doctor ever recommended that you have a PSA test?"

PSA Screening Intensity

To assess PSA screening intensity, we created a 3-level variable from an item asking men whether they ever had a PSA test; those who answered no constituted the lowest intensity category (none). Remaining men were stratified into low- and high-intensity categories based on the total self-reported number of PSA tests in the past 5 years: men reporting 1 to 3 tests were categorized as low intensity (less than annual), men reporting 4 to 5 tests were categorized as high intensity (approximately annual). This approach underestimates high-intensity screening in younger men who may not have undergone 4 to 5 tests even if screened annually. Additional analysis including men aged 40 to 49 years, however, resulted in no significant differences in the proportions of men in the 3 intensity categories.

Data Analysis

We calculated descriptive statistics of the study population's characteristics, prevalence of shared decision making, and PSA screening intensity. We fit separate multivariable polytomous and binary logistic regression models, adjusting for all sociodemographic and health-related characteristics, for the following dependent variables: (1) PSA screening intensity, (2) discussion of advantages and disadvantages, and (3) discussion of uncertainty. For the first model, we included extent of shared decision making as an independent variable. For the second and third models, examining predictors of shared decision making, discussion of uncertainty, and discussion of advantages and disadvantages, respectively, were included as independent variables to examine associations of different shared decision-making elements with one another. We used the Wald χ2 test to assess statistical significance, using a stringent cutoff (P <.01) to minimize type I error from our large sample size. We used SUDAAN 9.0.1 (RTI International) to adjust for the survey's complex sampling design, using weights provided with the NHIS data file to account for selection probability and nonresponse and to provide estimates representative of the US population.