Clinical guidelines recommend against routine prostate-specific antigen (PSA) screening for older men and for those with lower life expectancies. The authors of this report examined providers' decision-making regarding discontinuing PSA screening.
A survey of primary providers from a large, university-affiliated primary care practice was administered. Providers were asked about their current screening practices, factors that influenced their decision to discontinue screening, and barriers to discontinuing screening. Bivariate and multivariable logistic regression analyses were used to examine whether taking age and/or life expectancy into account and barriers to discontinuing were associated with clinician characteristics and practice styles.
One hundred twenty-five of 141 providers (88.7%) participated in the survey. Over half (59.3%) took both age and life expectancy into account, whereas 12.2% did not consider either in their decisions to discontinue PSA screening. Providers varied in the age at which they typically stopped screening patients, and the majority (66.4%) reported difficulty in assessing life expectancy. Taking patient age and life expectancy into account was not associated with provider characteristics or practice styles. The most frequently cited barriers to discontinuing PSA screening were patient expectation (74.4%) and time constraints (66.4%). Black providers were significantly less likely than nonblack providers to endorse barriers related to time constraints and clinical uncertainty, although these results were limited by the small sample size of black providers.
Since 2008, US Preventive Services Task Force recommendations have advised against routine prostate-specific antigen (PSA) screening for prostate cancer in men aged ≥75 years. The American Cancer Society and American Urological Association recommend incorporating life-expectancy into decisions regarding PSA screening. Draft recommendations released by the US Preventive Services Task Force in October 2011 discourage routine PSA screening for all men.
The recommendations suggest that the potential harms associated with PSA screening in older men and in those with limited life expectancies may outweigh the potential benefits.1, 2 Patients who undergo a biopsy for an elevated PSA may experience psychological distress3 and infectious complications.4 The diagnosis and treatment of prostate cancer may lead to lower health-related quality of life5, 6; may cause urinary, bowel, and erectile dysfunction7; and may result in high financial costs.8-11 A reduction in prostate cancer-specific mortality from PSA screening has not been demonstrated in older men.12
Nonetheless, screening among older men and for those with limited life expectancy remains high: An estimated 57% of men ages 75 to 79 years13 and approximately 33% of men with low life expectancies received screening within the past year.14 A physician's recommendation for PSA screening has been identified as the strongest predictor of whether a patient will undergo screening.13, 15 However, little is known about which factors are important to providers in discontinuing prostate cancer screening. Previous research on colorectal cancer has indicated that decisions to stop screening are complex and involve an assessment of patients' clinical factors and functional status.16, 17
We sought to examine the extent to which providers report taking age and life expectancy into account when ordering PSA screening and to determine the barriers that clinicians face when discontinuing PSA screening. We also examined whether taking age and life expectancy into account varied by provider clinical training, race/ethnicity, and other factors.
MATERIALS AND METHODS
Johns Hopkins Community Physicians (JHCP) is a university-affiliated practice comprised of 26 outpatient sites in 11 counties in Maryland. In 2010, approximately 40,000 men aged ≥40 years who were eligible for prostate cancer screening attended a JHCP site. The patient population of the clinics is diverse—27% of those eligible for PSA screening were nonwhite, and 3 outpatient sites are located in medically underserved areas of the metropolitan area of Baltimore.
The survey was developed in conjunction with the leadership of JHCP and was pretested by primary care providers practicing in the city of Baltimore who were not affiliated with JHCP.
The survey was distributed at a yearly provider meeting. In total, 141 individuals who provide primary care for adult male patients attended the meeting and were eligible to participate. The survey was self-administered and took approximately 10 minutes to complete. Providers were compensated $10 for their participation.
Age and Prostate-Specific Antigen Screening
On the survey, providers were told, “Practice guidelines vary about what age to stop PSA screening,” and they were asked whether there was an age at which they typically stopped screening an otherwise healthy man. Those who responded affirmatively were asked at what age they stopped screening.
Role of Life Expectancy
Providers were asked how strongly they agreed or disagreed with the statement, “I frequently take life expectancy into account when deciding not to order PSA screening.” Providers who agreed or strongly agreed were considered to take life expectancy into account. Then, they were asked how easy or difficult it was for them to estimate a patient's 10-year life expectancy in clinical practice by choosing 1 of the following responses: “very easy,” “somewhat easy,” “neither easy nor difficult,” “somewhat difficult,” or “very difficult.”
To determine attitudes and beliefs about stopping PSA testing in patients who previously had been receiving routine screening, we asked providers the extent to which they agreed or disagreed with a list of statements (see Table 2). The statements were developed based on previous survey instruments,18 qualitative research,16, 19 and in pretesting with providers in Baltimore. Responses were dichotomized into agree and strongly agree versus neither agree nor disagree, disagree, and strongly disagree.
Table 1. Demographic and Practice Style Characteristics of Johns Hopkins Community Physicians Survey Respondents
No. of Respondents (%)
Abbreviations: PSA, prostate-specific antigen.
Total numbers for individual variables may be less than 125 because of missing responses.
Frequency they typically recommended PSA screening
On a yearly basis
Less than yearly
Practice Styles With Prostate-Specific Antigen Screening
Provider patterns of PSA screening were based on a single item that was developed by Linder and colleagues: “Over the past year, which best described your approach to prostate cancer screening with a healthy, 50-year-old man who has no other risk factors and is otherwise a candidate for screening?”20 This was accompanied by the following response options: “I generally ordered a PSA test without discussing the possible harms and benefits with the patient”; “I generally discussed the possible harms and benefits of screening with the patient, and then recommended the test”; “I generally discussed the possible harms and benefits of screening with the patient, and then let him decide whether or not to have the test”; “I generally discussed the possible harms and benefits of screening with the patient, and then recommended against the test”; and “I generally did not order the PSA test nor discussed the possible harms and benefits with the patient.”
Providers were asked how frequently, over the past year, they recommended screening in men who opted for screening with the following response options: “yearly,” “every other year,” or “other.” They also were asked whether they felt comfortable or uncomfortable with their ability to answer patients' questions about PSA screening with responses ranging from very comfortable to very uncomfortable.
Demographic and Clinical Characteristics
The survey instrument included items on provider training (internal medicine, family medicine, internal medicine/pediatrics, and other), years since residency graduation (<5 years, 5 years to <10 years, and ≥10 years), sex, and race/ethnicity.
Descriptive analyses quantified responses on questions related to the use of age and life expectancy in screening decisions and perceived barriers to screening. Bivariate analyses tested the association between 1) provider use of age and/or life expectancy with perceived barriers to screening, 2) provider use of age and/or life expectancy in screening with clinician characteristics (sex, training, race/ethnicity, and years since residency) and practice styles (patterns of PSA screening, comfort with PSA discussions, and frequency of PSA screening), and 3) perceived barriers to screening with clinician characteristics and practice styles. Multivariable logistic regression models were constructed that simultaneously adjusted for all clinician characteristics and practice styles. Separate models were built for each of the main outcomes: use of age and/or life expectancy and perceived barriers to screening. Rates of missingness for each variable were <3%, and observations with missing values were excluded from analyses. All analyses were conducted using the Stata software package (version 10.0; Stata Corporation, College Station, Tex). This study was approved by the Johns Hopkins University School of Medicine Institutional Review Board.
The response rate was 88.7% (125 of 141 providers). The demographic characteristics of the participants are listed in Table 1. Overall, 62.9% of the providers were women, and 11.5% were black. Approximately half were trained in internal medicine, and the majority (67.7%) had finished residency more than 10 years previously. With regard to screening, 17.2% of participants said they ordered PSA screening without discussing it with their patients, and 33.6% said they discussed PSA screening and typically recommended it. When providers ordered PSA screening, they overwhelmingly (90%) ordered it on an annual basis. Providers typically felt comfortable answering patients' questions about PSA screening (89%).
Table 2. Number and Percentage of Providers Who Agreed or Strongly Agreed With Barriers to Discontinuing Prostate-Specific Antigen Screening and the Odds of Agreement by Provider Racea
No. of Respondents (%)
OR for Agreement With Barrier: Black Providers Compared With Non-Black Providers (95% CI)b
Agreement/disagreement was on a 5-point scale, and missing observations were excluded from the analysis.
The analyses were simultaneously adjusted for clinician characteristics (sex, training, race/ethnicity, and years since residency) and practice styles (patterns of PSA screening, comfort with PSA discussions, and frequency of PSA screening). Bivariate analyses are not presented because of small cell sizes and concerns over confidentiality.
My patients expect me to continue getting yearly PSA tests
It takes more time to explain why I am not screening than to just continue screening
By not ordering a PSA, it puts me at risk for malpractice
I am uncomfortable with the uncertainty if I discontinue screening
Patients have been told by other physicians to get a PSA test, and I am reluctant to go against their advice
If I stop PSA testing in my older patients, they will think I am trying to cut costs
If I stop PSA testing, my patients will think I am “giving up” on them
Age to Discontinue Prostate-Specific Antigen Screening
Nearly one-third of providers (32.5%) said that they did not have an age at which they typically stop recommending PSA screening. Among the 67.5% who said they discontinued screening based on patient age, 26.8% discontinued screening at age ≤70 years, 52.4% discontinued screening at age 75 years, and 20.7% discontinued screening at age ≥80 years.
Incorporating Life Expectancy
The majority of providers agreed or strongly agreed that they typically took life expectancy into account when ordering PSA screening (76.8%). Providers who stated that they had an age at which they typically discontinued screening also were significantly more likely to take life expectancy into account (Fig. 1). However, 66.4% of providers described it as very difficult or somewhat difficult to estimate life expectancy in clinical practice. Providers who had difficulty estimating life expectancy were not significantly less likely to take life expectancy into account compared with those who did not report difficulty. In this setting, the majority of providers (69.1%) believed it would be helpful to have a clinical decision support tool that would help predict mortality.
Barriers to Discontinuing Prostate-Specific Antigen Screening
Providers endorsed multiple barriers to stopping PSA testing among men who previously had been receiving PSA testing (Table 2). The most frequently cited reason was, “My patients expect me to continue getting yearly PSA tests” (74.4% of providers agreed) followed by, “It takes more time to explain why I'm not screening than to just continue screening” (66.4%). Agreement with the barriers did not differ among providers who did and did not take age/life expectancy into account.
Characteristics Associated With Prostate-Specific Antigen Screening
We examined whether there was variation in the use of age and/or life expectancy by clinician characteristics and practice styles. In bivariate analyses, black providers were significantly less likely to have an age at which they discontinued screening compared with nonblack providers (percentages not reported because of small cell sizes and concerns over confidentiality; P = .044). This result was not statistically significant after simultaneously adjusting for all clinician characteristics and practice styles. Taking life expectancy into account was not associated with provider race, other clinical characteristics, or practice styles in bivariate or multivariable logistic regression analyses.
Similarly, perceived barriers to discontinuing PSA screening varied by provider race but did not vary with other clinician characteristics or practice styles. In multivariable logistic regression analyses (Table 2), black providers were significantly less likely than nonblack providers to agree with the following barriers: “It takes more time to explain why I'm not screening than to just continue screening” (odds ratio [OR], 0.18; 95% confidence interval [CI], 0.05-0.72); “I am uncomfortable with the uncertainty if I discontinue screening” (OR, 0.16; 95% CI, 0.03-0.81); and, “My patients expect me to continue getting yearly PSA tests” (OR, 0.25; 95%CI, 0.07-0.92). Results of analyses remained consistent after excluding providers who generally did not order PSA screening (ie, those who discouraged PSA screening after a discussion or those who did not discuss it and did not order it; N = 7).
Age and life expectancy were used frequently, but not universally, by primary care providers in a university-affiliated primary care network when deciding whether to discontinue prostate cancer screening. The majority of these providers said they incorporated both factors into their recommendations. It will be necessary to assist this group of providers with tools that may more accurately determine life expectancy in clinical practice. A minority of providers did not consider either age or life expectancy in their screening decisions. For these providers, it may be necessary to focus on education regarding current recommendations. Across both groups of providers, understanding and addressing barriers regarding patient expectations, time constraints, and malpractice fears are critical.
Even among JHCP providers who said they used life expectancy in screening decisions, many said they found it challenging to estimate it in clinical practice, and the majority of providers believed that having a tool to estimate life expectancy at the point of care would be beneficial. Provider's difficulty with estimating life expectancy has been documented in studies on colon cancer screening,16, 17 and prognostic indices for older adults require additional testing and validation.21 It has been demonstrated that providing physicians with life expectancy estimates from life tables changes screening recommendations using clinical vignettes17; however, further research is necessary to test not only the most appropriate tool to calculate life expectancy during the clinical encounter but also whether giving providers such a tool would change decisions regarding PSA screening.
Along with difficulty estimating life expectancy, many providers reported that patients' expectations were a barrier to discontinuing screening. This is consistent with studies on older patients' expectations about discontinuing cancer screening. Interviews with older adults indicate that 62% of older adults believed their own life expectancy was not important in their decisions to continue cancer screening, and 48% did not want to discuss their life expectancy with their clinician.22 Other studies have indicated that patients with limited life expectancies wanted to discuss this with their physician,23 and Smith and colleagues have advocated that physicians should routinely offer to discuss prognosis with patients who have a limited life expectancy or at least by the time a patient turns aged 85 years.24 Along with providing clinicians with better tools to estimate life expectancy, providers may need training in how to engage in these conversations and adequate time to do so.
The barriers that providers endorsed to discontinuing screening were similar to the barriers observed to engaging in shared decision-making more generally. Davis and colleagues identified competing clinical priorities (95.5%) as the most important barrier to shared decision-making followed by lack of time (80.5%) and level of patient interest (69.9%).25 To the extent that these barriers may be addressed, it may help lead to better shared decision-making and higher rates of discontinuing screening. With their emphasis on patient-centeredness, use of physician extenders, and changes in payment mechanisms, primary care medical homes may provide an important opportunity to mitigate these barriers.26
With the exception of provider race/ethnicity, providers who used age and life expectancy did not have different characteristics (years since residency, sex, training) compared with providers who did not use these factors. Prior research indicated that black physicians were more likely to encourage PSA screening,27, 28 potentially because of the greater burden of prostate cancer among black men. Although they were no less likely to report taking age and life expectancy into account when deciding about PSA testing, black providers were significantly less likely than nonblack providers to report barriers to discontinuing screening. In particular, time constraints and clinical uncertainty were perceived by fewer black providers as significant barriers. The reasons for these differences are uncertain, and it is possible that this variation may be because of underlying differences in patient panels.
The current study has several potential limitations. First, despite the high response rate, nonresponse bias remains a concern. Responders were not significantly different from all providers who attended the retreat with respect to sex, training, and years since residency graduation. Second, questions were based on self-report, which may not accurately correlate with objective measures of screening patterns and patterns of discontinuing. Third, the study was conducted shortly after the release of the US Preventive Services Task Force draft recommendations, which may alter providers' perceptions of PSA screening. The USPSTF recommendation against routine screening in men aged ≥75 years has been in place since 2008. Fourth, the providers all worked within JHCP, potentially limiting the generalizability of this study. These providers, however, represent a range of practice sizes and provide care to a diverse patient population across the state of Maryland. Their practice patterns for PSA screening are comparable to those reported in studies of other providers,20 and the overall rates of PSA screening at JHCP in 2010 (26% of patients aged >40 years) are the same as the rates of PSA screening nationwide.14 Nonetheless, these results should be interpreted in the context of a single, regional health care system, and it would be important to test whether similar associations are present in other health care settings and outside of Maryland. Fifth, we were unable to control for some aspects of practice and training that may affect outcomes (eg, payer type, number of days per week in clinical practice). Finally, additional factors may influence a provider's decision to discontinue PSA screening. Although we did not ask whether factors like prior PSA level, PSA velocity, and family history influence decisions to discontinue screening, 43% of providers stated that they often or always altered the frequency of screening based on a patient's prior PSA level, and 48% did so based on family history. These and other factors may be important when deciding to discontinue screening for an individual patient.29, 30
In summary, we observed that the majority of providers in a university-affiliated primary care practice consider both age and life expectancy when deciding on prostate cancer screening. However, when they do take these factors into account, there is a lack of agreement over the age at which to discontinue testing, and they reported difficulty in assessing life expectancy. The current results suggest that, for some providers, interventions will need to teach clinicians to take age and life expectancy into account when deciding whether to initiate and continue screening. For providers that already consider these factors, the development of tools that help clinicians more accurately and quickly estimate life expectancy may be an important next step. Across both scenarios, addressing the barriers that make it difficult to stop testing—including patient expectation, lack of time, and fear of malpractice litigation—will be necessary to reduce the high rates of PSA screening in men who are least likely to benefit.
We thank Justin Bekelman for his advice on the article and Erin Murphy for her assistance with the survey.
The Maryland Cigarette Restitution Fund at Johns Hopkins provided funding. Dr. Pollack was supported by a career development award from the National Cancer Institute and the Office of Behavioral and Social Sciences Research (K07 CA151910). Dr. Bhavsar was supported by a training grant from the Agency for Healthcare Research and Quality (T32 HS019488).