Solid orange line indicates cumulative distribution; horizontal dotted line indicates the mean value for the 50th percentile. Primary care physicians are ranked from those with the lowest percentage of patients who underwent PSA screening to the highest. Only PCPs with at least 20 male patients 75 or older in their panels are included to produce estimates of PSA screening rates with reliability greater than 0.80. Error bars indicate the 95% CIs of the estimates, derived from a multilevel model including all the variables listed in the Table. Dark error bars indicate PCPs whose PSA screening rates were significantly different from the mean rate for all PCPs. Compared with the mean rates, 314 PCPs had significantly lower rates and 474 had significantly higher rates.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Jaramillo E, Tan A, Yang L, Kuo Y, Goodwin JS. Variation Among Primary Care Physicians in Prostate-Specific Antigen Screening of Older Men. JAMA. 2013;310(15):1622–1624. doi:10.1001/jama.2013.277514
No organization recommends prostate-specific antigen (PSA) screening in men older than 75 years. Nevertheless, testing rates remain high.1,2 We hypothesized that primary care physicians (PCPs) would vary substantially in PSA screening rates and that much of the variance in whether an older man received a PSA test would depend on which PCP he saw.
Using complete Medicare Part A and B data for Texas, we selected PCPs whose patient panels included at least 20 men 75 years or older without a prior diagnosis of prostate cancer. Primary care physicians were identified as generalist physicians who saw a man on 3 or more occasions in 2009.3 Patients enrolled in health maintenance organizations (approximately 25% of men ≥ 75 years) were not included because of incomplete data on testing and diagnoses. We assessed screening PSA tests ordered by any physician, or restricted to those ordered by the patient’s PCP, using the algorithm developed by Walter et al1 (Table, footnote c). We then conducted a multilevel, multivariable logistic regression analysis controlling for the patient characteristics listed in the Table. We estimated the PSA screening rate in 2010 for men 75 or older, adjusted for patient characteristics, for each PCP. We also calculated the intraclass correlation coefficient (ICC) at the PCP level. This study was approved by the University of Texas Medical Branch institutional review board. SAS version 9.2 (SAS Institute Inc) was used for all analyses.
Our sample included 1963 PCPs whose patient panels included at least 20 men 75 or older (61 351 men). Overall, 41.1% of the men received PSA screening and 28.8% received PSA screening ordered by their PCPs (Table). Both rates declined with patient age. There were small differences in rates of testing by patient ethnicity, markers of socioeconomic status, and location (urban vs rural).
The Figure presents a cumulative distribution of estimated PSA screening rates for each of the 1963 PCPs, showing only the rates for PSA tests ordered by the PCP and adjusted for all the characteristics in the Table in the multilevel model. In all, 474 PCPs (24.2%) had rates significantly greater than the mean, with a mean rate of 49.8% (95% CI, 48.8%-50.8%), whereas 314 PCPs (16.0%) had significantly lower rates, with a mean rate of 6.1% (95% CI, 5.9%-6.3%).
In the model assessing PSA screening ordered by the patient’s PCP, the ICC was 0.27, indicating that 27% of the variance in whether a man received PSA screening was explained by which PCP he saw. Specific patient characteristics (eg, age, comorbidity) explained only 3.7% of the variance in whether a patient received a PSA test ordered by his PCP.
The high variability among PCPs in PSA screening, with a 10-fold difference in rates between the highest and lowest deciles of PCPs, has not been found in other studies of PCP behavior. For example, using similar methodology and data sources for Texas PCPs, ICCs were 0.10 and 0.09 for receipt of mammography and colorectal cancer screening, respectively, compared with 0.27 for PSA screening.4,5 With PSA screening, which PCP a man saw explained approximately 7 times more of the variance in PSA screening than did the measurable patient characteristics.
Limitations of this study include the accuracy of identifying the PCP,3 exclusion of patients in health maintenance organizations, lack of information on patient preference, and the use of data from a single year in Texas. Southern states tend to have higher utilization rates for a number of tests and procedures. We assessed only screening PSA tests, not tests ordered to evaluate symptoms, but some symptoms may not have been coded. The rate of all PSA testing, including testing in men with symptoms, was greater than 50% in men 75 years or older in 2010 in Texas.
The high variability among PCPs in ordering PSA screening for older men requires additional study to understand its causes. It has been suggested that overtesting rates be included as quality measures of PCPs.6 Medicare data can be used to generate such measures.
Corresponding Author: James S. Goodwin, MD, Sealy Center on Aging, University of Texas Medical Branch, 301 University Blvd, Galveston, TX 77555-0177 (email@example.com).
Author Contributions: Dr Goodwin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Jaramillo, Kuo, Goodwin.
Acquisition of data: Tan, Yang.
Analysis and interpretation of data: Tan, Yang, Kuo, Goodwin.
Drafting of the manuscript: Jaramillo, Tan, Goodwin.
Critical revision of the manuscript for important intellectual content: Yang, Kuo, Goodwin.
Statistical analysis: Tan, Yang, Kuo.
Obtained funding: Goodwin.
Administrative, technical, or material support: Goodwin.
Study supervision: Kuo, Goodwin.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This research was supported by Comparative Effectiveness Research on Cancer in Texas (CERCIT) grant RP101207, funded by The Cancer Prevention Research Institute of Texas; grant K05CA134923 from the National Institutes of Health; grant R24H5022134 from the Agency for Healthcare Research and Quality; and University of Texas Medical Branch Clinical and Translational Science Award UL1TR00007.
Role of the Sponsors: The study sponsors had no role in the design and conduct of the study; the collection, management, analysis, and interpretation of the data; the preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.