Clinician Interest in Clinical Decision Support for PSA-based Prostate Cancer Screening

Abstract

Objective:

To evaluate the interest of primary care clinicians in utilizing CDS for PSA screening. Evidence suggests that electronic clinical decision support (CDS) may decrease low-value prostate-specific antigen (PSA) testing. However, physician attitudes towards CDS for PSA screening are largely unknown.

Methods:

A survey was sent to 201 primary care clinicians, including both physicians and Advanced Practice Providers (APP), within a large academic health system. Eligible clinicians cared for male patients aged 40-80 years and ordered ≥5 PSA tests in the past year. Respondents were stratified into three groups, appropriate screeners, low-value screeners, or rare-screeners, based on responses to survey questions assessing PSA screening practices. The degree of interest in electronic CDS was determined via a composite Likert score comprising relevant survey items.

Results:

Survey response rate was 29% (59/201) consisting of 85% MD/DO and 15% APP respondents. All clinicians surveyed were interested in CDS (p<0.001) without significant difference between screener groups. Clinicians agreed most uniformly that CDS be evidence-based. Clinicians disagreed on whether CDS would decrease professional discretion over patient decisions.

Conclusions:

Primary care clinicians are interested in CDS for PSA screening regardless of their current screening practices. Prioritizing CDS features that clinicians value, such as ensuring CDS recommendations are evidence-based, may increase the likelihood of successful implementation, whereas perceived threat to autonomy may be a hinderance to utilization.

1. INTRODUCTION

The United States may spend up to $28 billion annually on low-value screening, testing, or procedures, with a significant portion consisting of high-volume, low-cost services such as PSA testing.[1]^,[2] Despite utilization of the PSA test for over 30 years, poor PSA screening practices are prevalent with evidence suggesting that a significant portion of current PSA testing is considered low-value.[3] Guidelines indicate that low-value testing includes screening men younger than age 40, men with recent low PSA test results, or men with limited life expectancy.[3] While the underlying causes of low-value PSA testing practices are not well understood, the question of who and when to screen is controversial. Some experts advocate that PSA screening should be abandoned because the harms outweigh the benefits.[4] Others argue that benefits of screening outweigh the harms if testing is done in the most appropriate population of men at the ideal time.[5]

Optimizing the use of PSA-based prostate cancer screening to maximize benefit and minimize harm is challenging, particularly for primary care clinicians who are navigating an ever-changing landscape. Electronic clinical decision support (CDS) for PSA screening is a potential vehicle to achieve this balance. A core principal of successful CDS is that it is computerized, integrating with the electronic health record (EHR), providing recommendations for clinicians to deliver guideline-concordant care. Patient-specific, EHR-integrated, recommendations facilitated by CDS could mitigate low-value testing practices, particularly when updated screening guidelines are slow to be implemented. Effective elements of CDS have been studied,[6,7] and while the exact formula for successful CDS implementation is elusive with many frameworks and theories available,[6] consensus and buy-in from physician and organizational stakeholders are critical factors.[8,9] Despite the importance of clinician consensus and input for successful CDS implementation, clinicians’ current beliefs about CDS for PSA screening remain largely unknown.

In the current work, we sought to understand clinician attitudes towards CDS-assisted decision making for PSA-based prostate cancer screening using a survey of primary care physician and Advanced Practice Provider (APP) clinicians. We hypothesized that clinicians would be interested in CDS for PSA screening but may have concerns regarding autonomy and workflow integration.

2. MATERIALS and METHODS

2.1. Study Population and Measures

A survey was distributed to primary care clinicians within a large academic health system. Responses were analyzed to evaluate clinician interest in utilizing CDS for PSA screening. Clinicians identified included both physicians and APPs within the University of Utah Health System who cared for any male patients aged 40-80 years and ordered at least 5 PSA tests in the past year (November 2016 – November 2017). A second query of the survey was sent if no response was received within one month.

Survey items assessed which aspects of CDS would be perceived as helpful, how a CDS would impact practice patterns, and the most important features of an effective CDS. Individual scores were generated for each Likert-style survey question, with composite scores calculated by adding individual Likert scores. The complete survey instrument is available in the Supplement. Survey design, administration, and data capture was performed using Research Electronic Data Capture (REDCap) software.

2.2. Data Analysis

Respondents were categorized into two groups—screeners and rare-screeners—based on their beliefs and practices surrounding PSA screening. Screeners were defined as those who routinely recommended PSA screening and believed the benefits of screening outweighed the limitations, or if they indicated that the benefits of ordering a PSA test should be evaluated on a case-by-case basis (see supplement, survey question 1). Rare-screeners were defined as those who generally did not recommend PSA screening and believed the limitations of screening outweighed the benefits (see supplement, survey question 1). Screeners were further categorized into two groups, appropriate screeners and low-value screeners, by adherence to age-based PSA screening guidelines such as the percentage of patients screened at a given age and start and stop PSA screening ages of 45 and 75 years old respectively (see supplement, survey question 2). All other screeners who did not meet these criteria were categorized as low-value screeners (Figure 1).

Figure 1. Surveyed Clinicians by Screener Type.

Flowchart showing categorization of surveyed clinicians into three groups: appropriate screeners, low-value screeners, and rare-screeners.

CDS interest was quantified and represented by a CDS interest score, calculated using Likert scores between 0 - 4 assigned to each question in sections 9 and 10 of the survey. Section 9 included questions that assessed if CDS for PSA screening would be helpful in deciding when and when not to screen, when to repeat screening, and when to refer to a urologist. Section 10 contained questions that assessed whether CDS for PSA screening would affect job ease, delivery of guideline concordant care, and physician autonomy. The Likert scores obtained from the survey questions in sections 9 and 10 were summed to determine a composite Likert score that demonstrated CDS interest. A baseline CDS interest score, called neutral CDS interest, was generated by summing a neutral Likert score of 2 for each question in sections 9 and 10 for a total score of 16. Under section 10, the Likert scale for the question addressing professional discretion over patient care decisions was inverted for consistency. The primary outcome was CDS interest amongst clinicians stratified by screener type (rare screener, appropriate screener, low-value screener). We also sought to investigate the factors clinicians thought to be most important in designing an effective CDS system for PSA testing.

2.3. Statistical Analysis

We compared CDS interest and PSA knowledge scores of screeners and rare screeners as well as appropriate and low-value screeners using two sample t-test. The CDS interest scores for all screeners, appropriate screeners, low-value screeners, and rare screeners were compared to a neutral CDS interest score using a one sample t-test. All analysis was done using R (version 4.1.0) and p-value of 0.05 was considered statistically significant.

3. RESULTS

We identified 201 clinicians fitting our criteria. After sending initial and follow-up invitations, a total of 59 clinicians fully completed the survey yielding a 29% (59/201) response rate. A majority (85%, 50/59) were physicians (MD/DO), followed by physician assistants (12%; 7/59) and advance practice registered nurses (3%, 2/59) (Table). Respondents worked in various disciplines including family medicine (39%, 23/59), internal medicine (general, subspecialty, and med/peds; 46%, 27/59), and geriatrics (14%, 8/59). Of all respondents, 49% (29/59) were classified as appropriate screeners, 34% (20/59) as low-value screeners, and 17% (10/59) as rare screeners. No significant differences were observed between screener types in regards to clinical degree, department affiliation, or era of training completion.

Table:

Demographics Stratified by Screener Type

	Screener Type
	All	Appropriate	Low Value	Rare
Degree
MD	48	24	16	8
DO	2	2	0	0
PA	7	3	3	1
APRN	2	0	1	1
Department
Family Medicine	23	14	4	5
IM- primary	11	6	4	1
IM- subspecialty	14	5	7	2
IM/pediatrics	2	0	1	1
Geriatrics	8	5	3	0
Other	1	0	0	1
Completion of Training
before 1990	14	8	5	1
1990-1999	11	5	4	2
2000-2009	18	8	6	4
2010 or later	16	9	4	3
NA	1	0	1	0

All three screener types—rare screeners, appropriate screeners, and low-value screeners—displayed above-neutral CDS interest scores (Figure 2). Low-value screeners trended toward higher CDS interest scores, however this difference was not statistically significant. Agreement among respondents was universally high regarding the most helpful aspects of implementing CDS for PSA screening. Over 85% of respondents either agreed or strongly agreed that a PSA CDS system would be helpful for deciding when to screen, when not to screen, when repeat tests should be ordered, and when to refer the patient to a urologist (Figure 3a.).

Figure 2. CDS Interest by Screener Type.

Box and whisker plot showing which types of screeners were more or less interested in CDS for PSA screening. Screener type was determined by analyzing participant responses to survey questions assessing clinicians’ knowledge of PSA screening and current screening practices. Degree of interest in CDS for PSA screening was measured by analyzing participant responses to questions in sections 9 and 10 in the survey. These questions assessed clinician opinion on the usefulness of CDS for PSA screening. A neutral CDS score (the red line in the figure) is a summation of individual neutral (2 on the Likert scale) responses on all questions in sections 9 and 10. Figure 3 shows above-neutral interest in CDS for PSA screening regardless of screener type.

CDS: clinical decision support

PSA: prostate specific antigen

Figure 3.

a. Survey Section 9 – PSA CDS Would help with:

Graph showing the relative proportions of participant’s responses to questions in section 9 of the survey. Participants could choose to respond with the following: strongly agree, agree undecided, disagree, and strongly disagree.

CDS: clinical decision support

PSA: prostate specific antigen

b. Survey Section 10 - Using a CDS for PSA Screening Would:

Graph showing the relative proportions of participant’s responses to questions in section 10 of the survey. Participants could choose to respond with the following: strongly agree, agree undecided, disagree, and strongly disagree.

CDS: clinical decision support

PSA: prostate specific antigen

c. Survey Section 11: Most Important Features of a PSA CDS

Graph showing the relative proportions of participant’s responses to questions in section 11 of the survey. Participants could choose to respond with the following: strongly agree, agree undecided, disagree, and strongly disagree.

CDS: clinical decision support

PSA: prostate specific antigen

When asked about the impact of using a CDS system for PSA screening, over 85% of respondents either agreed or strongly agreed that implementation would reduce unnecessary tests and improve their ability to provide guideline concordant care (Figure 3b.). While 70% of respondents agreed or strongly agreed that a CDS system would make their job easier, opinions were polarized with more than 10% disagreeing or strongly disagreeing on this point. Respondents showed the lowest level of agreement when asked whether CDS would decrease their professional discretion over patient decisions, with 51% agreeing or strongly agreeing and 34% disagreeing or strongly disagreeing.

In response to questions about the relative importance of various potential CDS features, respondents most strongly agreed that aspects of the CDS system should be evidence-based with 98% agreeing or strongly agreeing (Figure 3c.). Additionally, 88% and 90% agreed or strongly agreed that the CDS system should fit within their existing workflow and provide patient-specific support, respectively. Respondents also agreed that use of the CDS should be optional and that the CDS should automate the recommended action, with 72% and 76% agreeing or strongly agreeing, respectively.

4. DISCUSSION

Our survey of primary care clinicians showed high levels of interest in electronic CDS for PSA screening regardless of current screening behaviors. Low-value screeners demonstrated increased interest in CDS compared to appropriate and rare screeners, albeit this trend did not reach statistical significance. While our study was underpowered to evaluate this relationship, the numerical trend may suggest that clinicians are interested in CDS because selecting the right patient for PSA screening at the right time is a challenge. Respondents agreed that CDS should be evidence-based, but disagreed whether CDS would impact professional discretion over patient decisions. Universal clinician interest in CDS may hint at frustration with current screening practices and a desire for novel strategies to improve them.

Respondents’ universal interest in CDS for PSA screening may suggest belief in CDS as a possible solution to the challenge of PSA-based screening. Electronic CDS has previously shown to improve healthcare processes,[10] including PSA testing in men older than 75, an age cohort in which screening is considered low-value.[11] Additionally, CDS has decreased unnecessary vitamin D testing,[12] improved dyslipidemia treatment,[13] and reduced medication errors.[14] While CDS has shown to improve health processes, data does not support CDS improvement of patient outcomes,[10,15] especially in oncology care.[15]

Clinicians surveyed in our study disagreed most about the impact of CDS on their professional discretion over patient decisions. Additionally, Less than 75% of clinicians in our study agreed that a CDS would make their job easier. These findings are consistent with a study showing CDS may improve cancer screening practices if the system was easy to use, saved time, and wasn’t perceived as a threat to physician autonomy.[16] Other work has also highlighted clinicians prioritizing a CDS that preserves workflow and the concern that a CDS that may threaten professional autonomy.[9,17–19]

Respondents strongly agreed that a CDS tool for PSA screening should provide evidence-based support. Interpretation of the evidence surrounding PSA screening is controversial. However, it has been suggested that a significant portion of low value PSA testing could be reduced by adhering to major societal guidelines,[3] and that up to 42% of overdiagnosis from PSA comes in men >70 years old.[5] Alerts based on guideline recommendations of who to screen, and who not to screen, could be integrated into the CDS system as an evidence-based support feature.

CDS has clear downsides. For example, CDS can produce alert fatigue, a phenomenon where the increased burden of digital reminders paradoxically produces the opposite desired behavior—ignoring alerts entirely in the EHR.[20] Also, CDS can interrupt workflows, making routine daily tasks harder to manage or complete. Additionally, many CDS implementations fail, wasting time, money, and energy. Even amongst successful CDS implementations, the overall effect size can be minimal, and it is not well understood what separates successful CDS implementations from those that fail entirely or produce minimal impact.[21] However, amongst a multitude of frameworks exploring factors involved in successful CDS implementations,[7] one theory is that the broader social context in which evidence and guidelines exist, coupled with physician attitudes towards such evidence, is crucial for successful CDS implementation and uptake .[8,9,17,22] Our study found that clinicians were interested in using CDS for PSA screening in the context of a current controversial social climate surrounding PSA screening. This finding is a strong first step towards building consensus, clinician buy-in, and ultimately engaging end users.

Despite widespread interest in CDS for PSA screening in our study, not all clinicians surveyed agreed on how a CDS could help, or the most important features of a CDS. Even the best designed CDS system won’t be a “one size fits all” model, making clinician buy-in crucial to the success of any attempt to use CDS to alleviate low-value PSA testing. Previous studies show that uptake and utilization improve when senior clinician leaders champion CDS[17] and consensus is built amongst clinicians, nurses, and administration.[8] Additionally, clinician buy-in during CDS implementation protects against burnout.[23]

Our study has limitations, including that it was limited to primary care clinicians practicing within a single academic health system. While there is limited data comparing CDS deployment in different health systems (i.e. Veteran’s Affairs (VA), academic, or private), implementation of CDS in the VA may be more successful; CDS may impact collections for ancillary services on which the VA system does not rely. Academic health systems are likely to benefit from increased clinician buy-in given incentives surrounding research projects and funding to implement CDS. While private settings are likely to have the most barriers to CDS implementation, community practices may benefit from CDS implementation under ideal circumstances[24].

Less than one third of clinicians responded to our survey resulting in a small sample size. We did not examine how provider gender may have influenced PSA screening practices and beliefs due to constraints from our IRB. Additionally, respondents were mainly MD/DO physicians. Therefore, these results may not apply more broadly to APPs or other clinicians not included in this sample. Another recent study concluded that APPs had more positive views about using CDS systems than physicians,[25] but our limited APP sample size did not allow us to detect the same trend. Finally, our study was underpowered to evaluate the differences in CDS interest when stratified by screener type.

5. CONCLUSION

Despite the emergence of the PSA test three decades ago, current PSA screening practices are poor and often result in low-value care. Evidence suggests that CDS can reduce waste and improve value in health care. To our knowledge, clinician attitudes regarding a CDS system for PSA screening have not been previously assessed. We found that clinicians are interested in using an electronic CDS system for PSA screening regardless of their current PSA screening patterns. This is promising, as clinician interest and consensus may increase uptake and successful implementation of any future CDS system implemented. Furthermore, addressing clinician-specific concerns, such as a system which emphasizes evidence-based recommendations and is sensitive to perceived threats to clinician autonomy, may increase the likelihood of successful implementation.

Highlights.

• Clinicians are interested in clinical decision support for PSA screening

• Low-value screeners trended toward higher interest in clinical decision support

• Clinicians agreed most uniformly that clinical decision support be evidence-based