Abstract
Background:
Prostate-specific antigen (PSA) testing for early detection of prostate cancer is low-value when it is not indicated by guidelines and the harms outweigh the benefits. In this retrospective cohort study, we identify provider and patient factors associated with PSA testing, particularly in situations where testing would be low-value.
Methods:
We used electronic health record data from 2011 to 2018 representing 1,738,021 health system encounters in the United States. Using logistic generalized estimating equation models, we examined patient factors (age, comorbid illness, family history, race and prior PSA results), provider factors (gender, specialty, graduation year and medical school rank), and overall time trends associated with PSA testing in low-value and appropriate settings.
Results:
Comorbid illness (odds ratio (OR) 0.0 for 3+ conditions vs none) and no prior PSA testing (OR 0.2) were associated with a lower likelihood of PSA testing in low-value situations, while family history of prostate cancer (OR 1.6) and high prior PSA test results (OR 2.2 for PSA >6 vs 0 – 1) were associated with a greater likelihood. Men aged 55 – 65 years were at greatest risk for PSA testing in low-value situations.
The provider factor associated with PSA testing in low-value situations was specialty, with urologists being most likely (OR 2.3 versus advanced practice providers). Internal medicine physicians were more likely to perform PSA testing during low-value situations (OR 1.3 versus advanced practice providers) but much more likely to order a PSA test where appropriate (OR 2.2). All PSA testing decreased since 2011.
Conclusion:
We identified several patient and provider factors associated with PSA testing in low-value settings. Some aspects suggest attention to relevant factors for PSA testing in low-value settings (e.g. comorbid illness), while others may encourage PSA testing in low-value settings (e.g. family history). The greatest likelihood of PSA testing in low-value settings is among men within the age range most commonly recommended by guidelines.
Keywords: PSA testing, Low-value screening
1.1. Introduction
Professional societies provide differing guidelines for PSA-based screening for early detection of prostate cancer. Differences are based on age, family history, ethnicity, life expectancy and PSA levels from prior screening1-5. The United States Preventative Service Task Force (USPSTF) currently rates PSA prostate cancer screening as grade C (selectively offer or provide to individual patients based on professional judgment and patient preferences) for men aged 55- 69 years2. The American Urological Association supports screening in men 55 – 69 years old after shared decision making with their physician but gives screening for those under 55 and over 70 a grade C, recommending screening among these men only in selected circumstances1. Given the variety of recommendations and their grading it can be difficult for providers to navigate the guidelines, thereby contributing to variation in PSA screening6,7.
Low-value PSA testing, or testing in situations where potential harms outweigh the anticipated benefits, accounts for substantial PSA testing in certain groups8. Low-value PSA testing can be categorized broadly as testing in men 1) too young to benefit from screening; 2) young and without risk factors such as African American race or a family history of breast, ovarian or prostate cancer9,10; 3) with recent low PSA test results11,12; and 4) with life expectancy <10-15 years13,14. Low-value testing in men who are older or with limited life expectancy has been well documented15,16. When considering each of these contributors to low-value PSA testing, up to 50% are performed when it is low-value to do so8.
In this study, we sought to determine provider and patient factors associated with an increased likelihood of undergoing PSA testing in situations where testing would be low-value. We hypothesized that physician factors such as age, specialty and medical school graduation year may influence PSA screening habits. Similarly, we hypothesized that patient factors such as comorbidities, prior testing, race and family history may influence the likelihood of low-value PSA screening.
2.0. Materials and Methods
2.1. Data
This retrospective cohort study used data extracted from the enterprise data warehouse of a large academic health system in the United States from July 1, 2011 through June 30, 2018. We included all outpatient encounters with the health system for men over the age of 18 without a diagnosis of prostate cancer. Patient level factors of interest included age, race, ethnicity, family history of prostate cancer or breast/ovarian cancer, and Charlson Comorbidity Index (CCI). For the analyses, race/ethnicity were grouped as six categories: i) white, non-Hispanic; ii) African, non-Hispanic; iii) Asian, non-Hispanic; iv) Hispanic; v) other, non-Hispanic; and vi) unknown. CCI values were grouped as four levels: 0, 1, 2, and 3+. Prior PSA values were grouped into six levels: [0, 1], (1, 2], (2, 3], (3, 6], greater than 6, and no prior PSA.
Provider factors included years since graduation from medical/graduate school, rank of medical school, gender, and specialty. For the analyses, years since graduation from medical/graduate school were indicated by the graduation year and categorized as: 1970s or before, 1980s, 1990s, 2000s, 2010s, and missing. Medical school rank was grouped as: rank 1-42, rank 43-83, unranked and missing. Provider gender was considered male or female and specialty was grouped into eight categories: advanced practice provider, family medicine, general surgery, internal medicine, medicine subspecialty, surgical subspecialty, urology, and other.
Low-value encounters were defined as an encounter meeting any of the following 5 criteria: 1) men younger than 40 years; 2) men younger than 45 years without risk factors (i.e. African American Race or family history of prostate, breast, or ovarian cancer); 3) men older than 75 years; or 4) men with a life expectancy less than 10 years. Life expectancy was calculated using a previously validated methodology combining age and number of comorbidities17. Life expectancy was deemed less than 10 years for men aged 70–75 with 2+ comorbidities, 60–69 with 3+ comorbidities, 50–59 with 4+ comorbidities, or age 45–49 with 5+ comorbidities. The final criterion was based upon prior testing, with the encounter considered low-value if prior PSA testing was conducted within 24 months and was <1 ng/mL or conducted within 12 months and was between 1-4 ng/mL18. Appropriate encounters were defined as encounters that did not meet any of the previous 5 criteria.
2.2. Analysis
Each encounter was assessed for whether it represented an encounter with the health system during which it would be low-value to conduct a PSA test and models were stratified by low-value/appropriate encounter. Our outcome was whether a PSA was ordered during each encounter to assess patient and provider factors associated with PSA testing.
Descriptive statistics of patient and provider factors were stratified by encounters in which PSA testing would be low-value and encounters in which PSA testing would be appropriate. The frequency (N) and percentage (%) were calculated for categorical variables. Mean and standard deviation (SD) were calculated for continuous variables. The associations of patient and provider factors on the odds of ordering a PSA test were assessed in the context of multivariable logistic generalized estimating equations (GEE) models which included the all of our patient and provider factors simultaneously, for the overall set of patient encounters as well as stratified by potentially low-value and appropriate situations. In the context of these adjusted models, each effect estimate can be interpreted as the impact of the variable in question on the response (odds of PSA testing) given that all of the other included variables remain fixed. Spline effects were included for patient age and encounter year. Robust sandwich estimators of variance were applied to compute standard errors, and corresponding P values and 95% confidence intervals, accounting for within patient dependencies19. Separate models were fit to estimate the associations between the variables of interest and PSA testing within the two settings. Statistical significance of the setting-wise differential impacts was determined in an overall model for PSA testing, which included interactions between setting (low-value/appropriate) and the factors of interest.
P values less than 0.05 were considered as statistically significant. All the analyses were conducted in R (version 3.5.1), with the package ‘geepack’ used for GEE modeling.
3.0. Results
The original data pull consisting of all encounters with male patients aged at least 18 years from July 1, 2011 through June 30, 2018 contained data on 1,804,413 encounters. Of these, 66,392 (3.7%) were excluded due to missing National Provider Identifier or provider specialty, leaving 1,738,021 encounters in the analytic dataset. Missingness in any of the remaining included patient and provider variables was encoded as a separate category, "Missing", and is summarized in Table 1. This study included 1,738,021 encounters from 197,584 individual men. A PSA test would have been appropriate to order in 36.8% of these encounters. As expected due to the definition of a low-value encounter, patients seen in an encounter when a PSA order would be low-value were more likely to be younger than 40 or older than 70, tended to have more medical problems, were more likely to be a racial minority, and were less likely to have a concerning family history or to have undergone prior PSA testing (Table 1). Providers caring for patients during encounters when PSA testing would be low-value were more likely to be female, an advanced practice clinician, or graduated from medical school more recently or from an unranked program.
Table 1.
Baseline characteristics stratified by appropriateness of encounters for PSA testing
| All | Low-Value | Appropriate | P-value | |
|---|---|---|---|---|
| Patient Factors | 1738021 (100%) | 1099179 (63.2%) | 638842 (36.8%) | |
| Age | < 0.001 | |||
| Mean (SD) | 51.3 (±18.0) | 47.3 (±20.8) | 58.2 (±8.1) | < 0.001 |
| Groups | ||||
| <40 | 512,427 (29.5%) | 512,427 (46.6%) | 0 (0.0%) | |
| 40-49 | 265,710 (15.3%) | 151,078 (13.7%) | 114,632 (17.9%) | |
| 50-59 | 326,378 (18.8%) | 75,085 (6.8%) | 251,293 (39.3%) | |
| 60-69 | 335,987 (19.3%) | 125,014 (11.4%) | 210,973 (33.0%) | |
| 70-79 | 209,686 (12.1%) | 147,742 (13.4%) | 61,944 (9.7%) | |
| 80+ | 87,833 (5.1%) | 87,833 (8.0%) | 0 (0.0%) | |
| Year | < 0.001 | |||
| 2011 | 44,798 (2.6%) | 27,540 (2.5%) | 17,258 (2.7%) | |
| 2012 | 131,205 (7.5%) | 83,027 (7.6%) | 48,178 (7.5%) | |
| 2013 | 200,093 (11.5%) | 123,989 (11.3%) | 76,104 (11.9%) | |
| 2014 | 250,066 (14.4%) | 157,705 (14.3%) | 92,361 (14.5%) | |
| 2015 | 292,025 (16.8%) | 185,835 (16.9%) | 106,190 (16.6%) | |
| 2016 | 320,057 (18.4%) | 203,150 (18.5%) | 116,907 (18.3%) | |
| 2017 | 302,762 (17.4%) | 192,691 (17.5%) | 110,071 (17.2%) | |
| 2018 | 197,015 (11.3%) | 125,242 (11.4%) | 71,773 (11.2%) | |
| CCI | < 0.001 | |||
| 0 | 821,006 (47.2%) | 475,287 (43.2%) | 345,719 (54.1%) | |
| 1 | 287,043 (16.5%) | 122,273 (11.1%) | 164,770 (25.8%) | |
| 2 | 174,221 (10.0%) | 82,964 (7.5%) | 91,257 (14.3%) | |
| 3+ | 455,751 (26.2%) | 418,655 (38.1%) | 37,096 (5.8%) | |
| Family hx of PSA | < 0.001 | |||
| No | 1,664,001 (95.7%) | 1,058,883 (96.3%) | 605,118 (94.7%) | |
| Yes | 74,020 (4.3%) | 40,296 (3.7%) | 33,724 (5.3%) | |
| Family hx of BROCA | < 0.001 | |||
| No | 1,652,756 (95.1%) | 1,046,291 (95.2%) | 606,465 (94.9%) | |
| Yes | 85,265 (4.9%) | 52,888 (4.8%) | 32,377 (5.1%) | |
| Race/Ethnicity | < 0.001 | |||
| White, non-Hisp | 1,374,053 (79.1%) | 854,180 (77.7%) | 519,873 (81.4%) | |
| African, non-Hisp | 34,235 (2.0%) | 23,924 (2.2%) | 10,311 (1.6%) | |
| Asian, non-Hisp | 40,895 (2.4%) | 28,098 (2.6%) | 12,797 (2.0%) | |
| Hispanic | 158,679 (9.1%) | 110,301 (10.0%) | 48,378 (7.6%) | |
| Other, non-Hisp | 76,113 (4.4%) | 48,966 (4.5%) | 27,147 (4.2%) | |
| Unknown | 54,046 (3.1%) | 33,710 (3.1%) | 20,336 (3.2%) | |
| Prior PSA Testing | < 0.001 | |||
| [0,1] | 157,494 (9.1%) | 81,702 (7.4%) | 75,792 (11.9%) | |
| (1,2] | 78,614 (4.5%) | 39,244 (3.6%) | 39,370 (6.2%) | |
| (2,3] | 34,015 (2.0%) | 18,397 (1.7%) | 15,618 (2.4%) | |
| (3,6] | 33,535 (1.9%) | 19,351 (1.8%) | 14,184 (2.2%) | |
| (6,inf) | 12,787 (0.7%) | 7,684 (0.7%) | 5,103 (0.8%) | |
| No prior test | 1,421,576 (81.8%) | 932,801 (84.9%) | 488,775 (76.5%) | |
| Provider Factors | ||||
| Gender | < 0.001 | |||
| Male | 1,163,253 (66.9%) | 727,774 (66.2%) | 435,479 (68.2%) | |
| Female | 574,768 (33.1%) | 371,405 (33.8%) | 203,363 (31.8%) | |
| Primary specialty | < 0.001 | |||
| Advanced Practice Clinician | 361,727 (20.8%) | 241,796 (22.0%) | 119,931 (18.8%) | |
| Family Medicine | 246,137 (14.1%) | 153,775 (14.0%) | 92,362 (14.4%) | |
| General Surgery | 2,422 (0.1%) | 1,685 (0.2%) | 737 (0.1%) | |
| Internal Medicine | 176,962 (10.2%) | 109,052 (9.9%) | 67,910 (10.6%) | |
| Medicine Subspecialty | 512,284 (29.4%) | 324,472 (29.5%) | 187,812 (29.4%) | |
| Surgical Subspecialty | 203,598 (11.7%) | 118,335 (10.8%) | 85,263 (13.3%) | |
| Urology | 31,260 (1.8%) | 19,752 (1.8%) | 11,508 (1.8%) | |
| Other | 203,631 (11.7%) | 130,312 (11.9%) | 73,319 (11.5%) | |
| Graduation year | ||||
| <1980 | 114,099 (6.6%) | 67,626 (6.2%) | 46,473 (7.3%) | |
| 1980’s | 295,579 (17.0%) | 182,240 (16.6%) | 113,339 (17.7%) | |
| 1990’s | 303,148 (17.4%) | 188,925 (17.2%) | 114,223 (17.9%) | |
| 2000’s | 626,796 (36.1%) | 394,650 (35.9%) | 232,146 (36.3%) | |
| 2010’s | 205,075 (11.8%) | 139,091 (12.7%) | 65,984 (10.3%) | |
| Missing | 193,324 (11.1%) | 126,647 (11.5%) | 66,677 (10.4%) | < 0.001 |
| School rank | ||||
| Rank 1-42 | 574,155 (33.0%) | 351,684 (32.0%) | 222,471 (34.8%) | < 0.001 |
| Rank 43-83 | 219,676 (12.6%) | 135,848 (12.3%) | 83,828 (13.1%) | |
| Unranked | 756,127 (43.5%) | 488,059 (44.4%) | 268,068 (42.0%) | |
| Missing | 188,063 (10.8%) | 123,588 (11.2%) | 64,475 (10.1%) |
P-values were calculated to compare distributions between low-value and appropriate groups
Wilcoxon rank sum tests were applied for continuous variables, and chi-squared tests were applied for categorical variables
3.1. Patient Factors
Logistic GEE models were then employed to identify factors associated with whether a PSA test was ordered during each encounter. Table 2 provides PSA testing rates across patient and provider factors, as well as multivariable adjusted effects estimates (odds ratios, 95% confidence intervals) for PSA testing overall and stratified by potentially low-value and potentially appropriate settings with a test of factor by potentially low-value vs. appropriate setting interaction. Overall, greater comorbid illness was associated with a lower likelihood of PSA testing. Having a family history of prostate cancer or breast or ovarian cancer were both associated with an increased likelihood of testing (odds ratio (OR) 1.8 and 1.2, respectively). Racial and ethnic minority status were each associated with a lower likelihood of testing. Being seen in an encounter when no prior PSA test (OR 0.7) was previously performed or when prior testing results were low (OR 0.9) was associated with a lower likelihood of undergoing a PSA test. Whereas prior PSA testing with results greater than 3 mg/mL was associated with greater odds of PSA testing.
Table 2.
Adjusted Odds Ratios (OR) for patient and provider factors associated with PSA order
| All Encounters | Low-Value | Appropriate | |||||
|---|---|---|---|---|---|---|---|
| PSA Not Ordered |
PSA Ordered |
OR (CI) | P- Value1 |
OR (CI) | OR (CI) | P- Value2 |
|
| Patient Factors | |||||||
| CCI | |||||||
| 0 | 801,078 (97.6%) | 19,928 (2.4%) | 1 | <0.001 | 1 | 1 | <0.001 |
| 1 | 278,762 (97.1%) | 8,281 (2.9%) | 0.8 (0.73, 0.80) | 0.9 (0.81, 0.96) | 0.8 (0.77, 0.84) | ||
| 2 | 169,373 (97.2%) | 4,848 (2.8%) | 0.7 (0.65, 0.73) | 0.3 (0.26, 0.33) | 0.7 (0.68, 0.77) | ||
| 3+ | 446,526 (98.0%) | 9,225 (2.0%) | 0.5 (0.46, 0.50) | 0.0 (0.04, 0.05) | 0.6 (0.54, 0.66) | ||
| Family hx of prostate ca | |||||||
| No | 1,625,904 (97.7%) | 38,097 (2.3%) | 1 | <0.001 | 1 | 1 | 0.01 |
| Yes | 69,835 (94.3%) | 4,185 (5.7%) | 1.8 (1.73, 1.96) | 1.6 (1.40, 1.73) | 2.2 (2.03, 2.37) | ||
| Family hx of BROCA | |||||||
| No | 1,613,361 (97.6%) | 39,395 (2.4%) | 1 | 0.04 | 1 | 1 | 0.26 |
| Yes | 82,378 (96.6%) | 2,887 (3.4%) | 1.2 (1.11, 1.27) | 1.1 (1.00, 1.24) | 1.3 (1.18, 1.40) | ||
| Race/Ethnicity | |||||||
| White | 1,339,135 (97.5%) | 34,918 (2.5%) | 1 | <0.001 | 1 | 1 | <0.001 |
| African | 33,569 (98.1%) | 666 (1.9%) | 0.8 (0.68, 0.92) | 0.7 (0.59, 0.92) | 0.8 (0.64, 0.99) | ||
| Asian | 39,881 (97.5%) | 1,014 (2.5%) | 1.0 (0.94, 1.17) | 0.8 (0.71, 0.98) | 1.2 (1.06, 1.39) | ||
| Hispanic | 155,591 (98.1%) | 3,088 (1.9%) | 0.8 (0.80, 0.90) | 0.8 (0.71, 0.85) | 1.0 (0.92, 1.06) | ||
| Other | 74,593 (98.0%) | 1,520 (2.0%) | 0.8 (0.72, 0.86) | 0.8 (0.69, 0.93) | 0.8 (0.74, 0.92) | ||
| Unknown | 52,970 (98.0%) | 1,076 (2.0%) | 0.6 (0.53, 0.67) | 0.7 (0.57, 0.80) | 0.6 (0.49, 0.64) | ||
| Prior PSA test value | |||||||
| [<1] | 146,928 (93.3%) | 10,566 (6.7%) | 1 | <0.001 | 1 | 1 | <0.001 |
| (1,2] | 73,444 (93.4%) | 5,170 (6.6%) | 0.9 (0.80, 0.94) | 0.8 (0.71, 0.88) | 1.2 (1.00, 1.39) | ||
| (2,3] | 31,755 (93.4%) | 2,260 (6.6%) | 0.9 (0.81, 1.01) | 0.8 (0.65, 0.87) | 1.3 (1.08, 1.68) | ||
| (3,6] | 30,286 (90.3%) | 3,249 (9.7%) | 1.3 (1.17, 1.46) | 1.0 (0.84, 1.12) | 2.0 (1.53, 2.51) | ||
| (6+) | 10,312 (80.6%) | 2,475 (19.4%) | 2.6 (2.26, 3.11) | 2.2 (1.84, 2.76) | 3.1 (2.12, 4.66) | ||
| No prior test | 1,403,014 (98.7%) | 18,562 (1.3%) | 0.7 (0.62, 0.71) | 0.2 (0.22, 0.27) | 2.3 (1.98, 2.68) | ||
| Provider Factors | |||||||
| Gender | |||||||
| M | 1,133,143 (97.4%) | 30,110 (2.6%) | 1 | 0.02 | 1 | 1 | <0.001 |
| F | 562,596 (97.9%) | 12,172 (2.1%) | 0.8 (0.81, 0.86) | 1.0 (0.91, 1.00) | 0.8 (0.73, 0.80) | ||
| Primary specialty | |||||||
| APP | 237,479 (98.2%) | 4,317 (1.8%) | 1 | <0.001 | 1 | 1 | <0.001 |
| Fam. Medicine | 150,211 (97.7%) | 3,564 (2.3%) | 1.3 (1.20, 1.32) | 1.0 (0.96, 1.11) | 1.5 (1.39, 1.56) | ||
| Gen. Surgery | 1,636 (97.1%) | 49 (2.9%) | 0.9 (0.66, 1.18) | 1.0 (0.70, 1.44) | 0.6 (0.39, 1.05) | ||
| Int. Medicine | 104,155 (95.5%) | 4,897 (4.5%) | 1.7 (1.65, 1.81) | 1.3 (1.25, 1.42) | 2.2 (2.05, 2.32) | ||
| Med. Sub. | 321,138 (99.0%) | 3,334 (1.0%) | 0.3 (0.32, 0.36) | 0.4 (0.39, 0.45) | 0.3 (0.28, 0.33) | ||
| Surg. Sub. | 117,651 (99.4%) | 684 (0.6%) | 0.2 (0.17, 0.20) | 0.2 (0.22, 0.28) | 0.2 (0.15, 0.18) | ||
| Urology | 18,285 (92.6%) | 1,467 (7.4%) | 1.9 (1.73, 2.08) | 2.3 (2.03, 2.63) | 1.8 (1.58, 2.02) | ||
| Other | 129,610 (99.5%) | 702 (0.5%) | 0.3 (0.31, 0.36) | 0.3 (0.28, 0.36) | 0.4 (0.36, 0.44) | ||
| Graduation year | |||||||
| <1980 | 111,478 (97.7%) | 2,621 (2.3%) | 1 | <0.001 | 1 | 1 | <0.001 |
| 1980’s | 284,985 (96.4%) | 10,594 (3.6%) | 1.0 (0.96, 1.12) | 0.9 (0.77, 0.96) | 1.2 (1.07, 1.30) | ||
| 1990’s | 292,762 (96.6%) | 10,386 (3.4%) | 1.0 (0.94, 1.11) | 0.8 (0.75, 0.93) | 1.2 (1.10, 1.34) | ||
| 2000’s | 613,949 (98.0%) | 12,847 (2.0%) | 0.7 (0.64, 0.75) | 0.6 (0.55, 0.68) | 0.8 (0.71, 0.86) | ||
| Missing | 201,268 (98.1%) | 3,807 (1.9%) | 0.7 (0.64, 0.76) | 0.7 (0.60, 0.76) | 0.7 (0.66, 0.82) | ||
| Provider’s school rank | |||||||
| Rank 1-42 | 558,026 (97.2%) | 16,129 (2.8%) | 1 | <0.001 | 1 | 1 | <0.001 |
| Rank 43-83 | 213,038 (97.0%) | 6,638 (3.0%) | 1.0 (0.93, 1.02) | 0.8 (0.79, 0.90) | 1.1 (1.00, 1.12) | ||
| Unranked | 738,564 (97.7%) | 17,563 (2.3%) | 0.9 (0.88, 0.94) | 0.9 (0.82, 0.92) | 1.0 (0.93, 1.02) | ||
| Missing | 186,111 (99.0%) | 1,952 (1.0%) | 2.0 (1.30, 3.00) | 1.3 (0.73, 2.16) | 2.8 (1.62, 4.83) | ||
P-values are for whether the factor is significantly associated with PSA testing.
P-values are for interactions between the named factor and appropriate versus low-value testing.
BROCA: Breast or Ovarian cancer; APP- Advanced Practice Provider
To determine which patient and provider factors influenced whether a PSA was ordered during an encounter in which the order was low-value or appropriate, we examined the interaction of appropriateness versus low-value on each of the factors in the models. Several patient factors had a differential impact depending on whether an encounter was low-value or appropriate for PSA testing. For example, the OR for ordering a low-value PSA test was 0.0 for 3+ comorbid conditions versus none compared to the OR for appropriate PSA testing of 0.6 for 3+ comorbid conditions versus none (interaction p<0.001).
Family history of prostate cancer appeared to exert a greater impact towards PSA testing in situations where it would be appropriate than low-value situations (OR 2.2 vs 1.6). Although having a family history of breast or ovarian cancer was associated with PSA testing overall, there was little evidence of a differential impact depending on whether testing was performed during situations when it would be appropriate compared to low-value (interaction p=0.26).
Prior PSA results showed a different pattern in situations where testing would be appropriate as compared to low-value. PSA testing in low value settings were least likely when no prior testing was done. The men at greatest increased likelihood for receiving a PSA test in a low-value setting were those who had a prior high result (OR 2.2). PSA testing in appropriate situations was most likely during encounters with patients who had never had a prior test (OR 2.3) or in those with high previous results (OR 3.1 for prior PSA results 6+).
3.2. Provider Factors
Provider factors were also important in determining whether a PSA test was ordered. Female providers were less likely to order a PSA during situations when it may be appropriate (OR 0.8) but just as likely as male providers to order a PSA during situations when it was low-value (OR 1.0).
Urologists and internal medicine physicians were the most likely to order a PSA, with medical and surgical subspecialists least likely. Urologists were also more likely to order tests in low-value situations (OR 2.3) as compared to situations where testing may be appropriate (OR 1.8), while family medicine and internal medicine providers were more likely to order PSA tests in appropriate situations (OR 1.5 and 2.2 respectively) compared to low-value testing situations (OR 1.0 and 1.3, respectively).
Providers graduating more recently were less likely to order a PSA, with much of this due to a higher likelihood for older providers to preferentially order PSA tests in appropriate situations, while younger providers ordered fewer PSA tests in both low-value and appropriate encounters. We observed minimal impact of school rank on PSA testing behavior.
3.3. Time Trends
Consistent with national data, there was a stark decline in PSA testing after 2011 (Figure 1). There was a pattern of decline in testing for all PSA orders including both low-value and appropriate testing. While low-value PSA testing had an initial decline greater than appropriate testing, the proportion of PSA tests that were low-value then mirrored appropriate testing over time including possible small recent increases in both.
Figure 1.
Odds ratio for PSA testing over time from 2011 through 2018 for all encounters, as well as encounters for which testing is appropriate and low-value (Jan 1, 2012 is the reference).
3.4. Age Effects
Overall, PSA testing was most likely among men age 55 to 65 years old with a similar pattern for appropriate PSA testing (Figure 2). While there was some low-value PSA testing in both older and younger men, the likelihood of low-value PSA testing was greatest among men who are the same ages as men in whom it is appropriate.
Figure 2.
Odds ratio for PSA testing across patient ages for all encounters, as well as encounters for which testing is appropriate and low-value (50 years is the reference age).
4.0. Discussion
This study identified patient and provider factors associated with PSA testing in over 1.7 million encounters with a large health system. Most of these factors had a differential impact based upon whether PSA testing was appropriate or low-value. Understanding these factors may prove beneficial in designing interventions that reduce the provision of low-value PSA screening while maintaining or improving appropriate testing.
The decline in PSA testing we identified in conjunction with the USPSTF grade D recommendation for PSA testing in 2011 is well documented20. While it is interesting to note that this policy change was effective in reducing low-value PSA testing, it was a relatively blunt tool as appropriate PSA testing also dramatically declined. In addition, there appears to be an indiscriminate increase in both low-value and appropriate PSA testing that may be a result of the more recent USPSTF grade C recommendation for PSA testing2.
It is reassuring that certain factors appear to be associated with an expected change in the likelihood for PSA testing in low-value settings. For example, comorbid illness appears to have an important impact, leading to fewer PSA tests. Yet, some providers are responding to known prostate cancer risk factors even when guidelines suggest testing would be low-value. For instance, family history remains a factor increasing the likelihood a provider orders a test in a low-value situation. This may represent scenarios when patients are requesting prostate cancer screening even when it would be low-value.
There is a well-documented increased risk of prostate cancer among African American men, with evidence that biological and social factors increase the likelihood of worse outcomes1-3,21. While African Americans constituted only 2% of the studied encounters, the decreased likelihood for PSA testing among this population warrants further investigation.
The association of prior PSA testing results raise a few interesting issues. Those without any prior PSA test are substantially less likely to undergo PSA testing. This suggests that some patients decided not to check their PSA or some providers do not routinely screen for prostate cancer. Simultaneously, higher previous PSA test results are associated with a greater likelihood of conducting additional PSA tests. We would expect that abnormal PSA results are followed with additional directed screening or further testing, however, it speaks to the potential harms that patients experience when providers order tests in low-value situations given that high prior PSA results lead to an increased likelihood for additional testing among patients in whom testing is low-value. That is, a high PSA result in someone with a limited life expectancy may lead to additional testing even though identifying a cancer is very unlikely to be beneficial, and subsequent workup exposes patients to real harm22-24. As a result, providers must ensure they do not initiate the risky cascade initiated by a single low-value PSA test.
Differences between providers suggest that some specialties are more likely to order PSA tests in low-value settings; urologists were the most likely. Perhaps this is a reflection of referrals for additional evaluation in patients who had abnormal testing despite being part of a population outside current recommended screening guidelines. It may also be that PSA testing is a default order more commonly for urologists, thereby increasing a patient’s risk for a low-value PSA test simply by being seen in a urology clinic. Financial incentives to work up elevated PSA and treat prostate cancer could further drive this provision. Urologists could be seeing low-value testing candidates who have symptoms concerning for prostate cancer, or who are planning to undergo a procedure for which ruling out prostate cancer is important (e.g. transurethral resection of the prostate for benign prostatic hyperplasia or pre-transplant screening).
Internal medicine doctors were also more likely to order PSA tests with much of that difference driven by testing in appropriate encounters. This may be a reflection of differences in professional guidelines with family practice guidelines more strongly recommending against PSA testing25 while the American College of Physicians3 provides more room for shared decision making around the issue.
That female providers were as likely as male providers to order PSA tests in low-value settings but less likely to order tests when it might be appropriate is interesting. This may speak to the screening practice of various providers and warrants further study.
Much of the literature documenting low-value PSA testing in older men, which leverages Medicare data, misses the findings of this study that the age group of men most likely to undergo low-value PSA testing is 55 – 65 years. These men fall in the age range supported by the USPTF (55-69) and all other professional society guidelines. This emphasizes that a focus on age alone as a factor for determining appropriateness of PSA testing is inadequate.
There are several limitations to consider. First, this study utilized data collected from the electronic health record system. All observations were based upon data collected as part of clinical care or billing, therefore, we are unable to draw conclusions about the thought process of providers or the impact of patient-provider interactions that may have influenced decision-making. Second, retrospective electronic health record data is subject to ascertainment bias, where particular features are more likely to be recorded for some types of patients than others, non-response bias, where the patients represented in the dataset differ from the target population, and potentially recall bias, where data entries are subject to errors in a patient's or provider's memory. Third, while this analysis comes from a large health system, it may not represent factors that impact PSA testing in other regions of the US or the broader international community. Most notably, there were few patients who identified as racial minorities (roughly 20%). Nonetheless, this study offers significant insights into practice patterns around PSA testing, which if proven in a broader context, may serve as the basis of health services interventions to improve the value of care provided.
5.0. Conclusion
Several provider and patient factors appear to be associated with PSA testing in situations where testing may be low-value. Observed differences between providers may reflect specialty differences in training, professional society recommendations and patient referral patterns. Some patient aspects suggest close attention is paid to important factors by providers including the presence of comorbid illness. On the other hand, there are areas for improvement in incorporating factors such as family history of prostate cancer, race/ethnicity, and prior PSA testing. In addition, the observation that the age (55 – 65 years) with the greatest risk for low-value PSA testing overlaps closely with the most common age range for professional guidelines (USPSTF, AUA, ACP) raises concerns about how closely providers discern which patients should not undergo PSA testing in this age range. Finally, while there was a substantial decrease in PSA testing since 2011, the ratio of low-value to appropriate PSA testing has largely stayed the same. Taken together, this work suggests that while providers are paying attention to some factors that increase the harms of PSA testing, there is room for improvement in order to reduce the harms of prostate cancer screening and thereby improve its value.
Highlights.
Low-value PSA testing is most likely among patients with prior PSA values
Men 55 – 65 year old are most likely to receive low-value PSA testing
Urologist are most likely among providers to order low-value PSA testing
Acknowledgments
1. Research reported in this publication utilized the Cancer Biostatistics Shared Resource at Huntsman Cancer Institute at the University of Utah and was supported by the National Cancer Institute of the National Institutes of Health under Award Number P30CA042014. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
2. Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under Award Number K08CA234431. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest: None to declare from any authors.
References
- 1.Carter HB. American Urological Association (AUA) Guideline on prostate cancer detection: process and rationale. Bju International. 2013;112(5):543–547. [DOI] [PubMed] [Google Scholar]
- 2.Grossman DC, Curry SJ, Owens DK, et al. Screening for Prostate Cancer: US Preventive Services Task Force Recommendation Statement. Jama. 2018;319(18):1901–1913. [DOI] [PubMed] [Google Scholar]
- 3.Qaseem A, Barry MJ, Denberg TD, Owens DK, Shekelle P, Amer Coll P. Screening for Prostate Cancer: A Guidance Statement From the Clinical Guidelines Committee of the American College of Physicians. Annals of Internal Medicine. 2013;158(10):761-+. [DOI] [PubMed] [Google Scholar]
- 4.Carroll PR, Parsons JK, Andriole G, et al. Prostate Cancer Early Detection, Version 2.2016. Journal of the National Comprehensive Cancer Network. 2016;14(5):509–519. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wolf AMD, Wender RC, Etzioni RB, et al. American Cancer Society Guideline for the Early Detection of Prostate Cancer Update 2010. Ca-a Cancer Journal for Clinicians. 2010;60(2):70–98. [DOI] [PubMed] [Google Scholar]
- 6.Jaramillo E, Tan A, Yang L, Kuo Y-F, Goodwin JS. Variation Among Primary Care Physicians in Prostate-Specific Antigen Screening of Older Men. Jama-Journal of the American Medical Association. 2013;310(15):1622–1624. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bynum J, Passow H, Carmichael D, Skinner J. Exnovation of Low Value Care: A Decade of Prostate-Specific Antigen Screening Practices. Journal of the American Geriatrics Society. 2019;67(1):29–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.O'Neil B, Martin C, Kapron A, Flynn M, Kawamoto K, Cooney KA. Defining low-value PSA testing in a large retrospective cohort: Finding common ground between discordant guidelines. Cancer Epidemiology. 2018;56:112–117. [DOI] [PubMed] [Google Scholar]
- 9.Thompson IM, Ankerst DP, Chi C, et al. Assessing prostate cancer risk: Results from the prostate cancer prevention trial. Journal of the National Cancer Institute. 2006;98(8):529–534. [DOI] [PubMed] [Google Scholar]
- 10.Brandt A, Bermejo JL, Sundquist J, Hemminki K. Age-Specific Risk of Incident Prostate Cancer and Risk of Death from Prostate Cancer Defined by the Number of Affected Family Members. European Urology. 2010;58(2):275–280. [DOI] [PubMed] [Google Scholar]
- 11.Aus G, Damber JE, Khatami A, Lilja H, Stranne J, Hugosson J. Individualized screening interval for prostate cancer based on prostate-specific antigen level - Results of a prospective, randomized, population-based study. Archives of Internal Medicine. 2005;165(16):1857–1861. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Yao SL, Lu-Yao G. Interval after prostate specific antigen testing and subsequent risk of incurable prostate cancer. Journal of Urology. 2001;166(3):861–865. [PubMed] [Google Scholar]
- 13.Schroder FH, Hugosson J, Roobol MJ, et al. Screening and prostate cancer mortality: results of the European Randomised Study of Screening for Prostate Cancer (ERSPC) at 13 years of follow-up. Lancet. 2014;384(9959):2027–2035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ilic D, Neuberger MM, Djulbegovic M, Dahm P. Screening for prostate cancer. Cochrane Database of Systematic Reviews. 2013(1). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tang VL, Shi Y, Fung K, et al. Clinician Factors Associated With Prostate-Specific Antigen Screening in Older Veterans With Limited Life Expectancy. Jama Internal Medicine. 2016;176(5):654–661. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kotwal AA, Mohile SG, Dale W. Remaining life expectancy measurement and PSA screening of older men. Journal of Geriatric Oncology. 2012;3(3):196–204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Hawken SR, Auffenberg GB, Miller DC, et al. Calculating life expectancy to inform prostate cancer screening and treatment decisions. BJU Int. 2017;120(1):9–11. [DOI] [PubMed] [Google Scholar]
- 18.Carroll PH, Mohler JL. NCCN Guidelines Updates: Prostate Cancer and Prostate Cancer Early Detection. Journal of the National Comprehensive Cancer Network. 2018;16(5):620–623. [DOI] [PubMed] [Google Scholar]
- 19.Zeger SL, Liang KY, Albert PS. Models for longitudinal data: a generalized estimating equation approach. Biometrics. 1988;44(4):1049–1060. [PubMed] [Google Scholar]
- 20.Misra-Hebert AD, Hu B, Klein EA, et al. Prostate cancer screening practices in a large, integrated health system: 2007-2014. BJU Int. 2017;120(2):257–264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hanson HA, Martin C, O'Neil B, et al. The Relative Importance of Race Compared to Healthcare and Social Factors in Predicting Prostate Cancer Mortality: A Random Forest Approach. In. Vol In press. Journal of Urology 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Borghesi M, Ahmed H, Nam R, et al. Complications After Systematic, Random, and Image-guided Prostate Biopsy. Eur Urol. 2017;71(3):353–365. [DOI] [PubMed] [Google Scholar]
- 23.Brewster DH, Fischbacher CM, Nolan J, Nowell S, Redpath D, Nabi G. Risk of hospitalization and death following prostate biopsy in Scotland. Public Health. 2017;142:102–110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Whitson JM, Murray KS, Thrasher JB. Prostate Biopsy is Associated with an Increased Risk of Erectile Dysfunction. J Urol. 2016;196(1):21–23. [DOI] [PubMed] [Google Scholar]
- 25.Clinical Preventive Service Recommendation: Prostate Cancer. AAFP. Published 2019. Accessed 5/30/2019, 2019. [Google Scholar]