INTRODUCTION

Deciding whether it is worthwhile to advise older adults (aged ≥ 65 years) to continue or stop cancer screening is not straightforward for primary care practitioners (PCPs). At a population level, there is evidence that the benefit/harm trade-off of cancer screening becomes less favorable in older adults. The decision also becomes increasingly complex given competing mortality risks, increasing risks of harm from the screening cascade (follow-up testing, false positive results) and harms from overdiagnosis.1 Moreover, the chance of benefiting at an individual level does not depend on age, but rather the life expectancy and health status of a person, suggesting screening decisions should be individualized.2

Cancer screening recommendations for older adults vary by jurisdiction and cancer type. Many countries implement national breast, bowel, and cervical screening programs that target participation up to a certain age (e.g., 75 years), but it is unclear what role PCPs play in decision-making beyond these ages. In the USA, both researchers and guidelines from the U.S. Preventive Services Task Force have increasingly advocated for individualized screening decisions based on patient health, comorbidities, and prior screening history,1,2,3,4 in which PCPs play a key role. For example, many guidelines recommend against screening older adults with less than 10 years’ life expectancy.1 However, many older adults with limited life expectancy continue to screen5 and have limited knowledge of the potential harms of doing so.6

Understanding more about the evidence on the factors that influence PCPs’ decisions about continuing or stopping cancer screening in older adults is needed to inform the design of interventions to individualize screening for older adults and to identify areas for further research in this area. The aim of this systematic review was to examine the factors influencing PCPs’ recommendations for breast, cervical, prostate, and colorectal cancer screening for older adults. These cancer screening types were chosen given the international relevance of the problem of overscreening and low-value screening for all four types.5,7,8,9

METHODS

Protocol and Registration

The protocol for this review was registered with PROSPERO (CRD42021268219) and reporting is guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist.10

Search Strategy

A comprehensive search strategy previously developed in consultation with an academic librarian to explore patient-reported factors associated with older adults’ cancer screening decision-making11 was adapted by adding terms related to PCPs (see Supplementary Table I). Searches were conducted in Medline, Pre-Medline, EMBASE, PsycINFO, and CINAHL on 16th July 2021. After removing duplicates, two researchers (JS, RD) independently screened titles and abstracts for inclusion/exclusion using an eligibility checklist and disagreements were resolved via discussion. A backward citation search and forward citation search were conducted once decisions were finalized (14th July 2022).

Inclusion Criteria

Studies were included if they:

  1. a.

    Empirically assessed clinician-reported factors influencing primary care practitioners’ (PCPs) recommendations for cancer screening for older adults, defined as either (1) adults who are aged ≥ 65 years or (2) adults with limited (< 10-year) life expectancy.

  2. b.

    Included PCPs (e.g., general practitioners, family medicine practitioners, primary care internists, primary care nurse practitioners, physician assistants) providing care to older adults or a sub-group of older adults.

  3. c.

    Assessed breast (mammography), cervical (Pap smear or human papillomavirus (HPV) test), prostate (prostate-specific antigen (PSA) testing), or colorectal (faecal occult blood testing (FOBT), faecal immunochemical test (FIT), or colonoscopy) cancer screening decisions in asymptomatic older patients; and

  4. d.

    Were quantitative, qualitative, or mixed-methods peer-reviewed publications, or theses.

Studies were excluded if they:

  1. a.

    Analyzed factors associated with retrospective screening, or assessed decision-making regarding diagnostic or surveillance tests; or

  2. b.

    Were reviews, editorials, commentaries, research letters, intervention studies or protocols.

Quality Appraisal

Two researchers (JS, KG) independently assessed the quality of included studies using the Joanna Briggs Institute critical appraisal tools, which are checklists that can be used to determine the sources of bias in a study’s design, conduct, and analysis (Prevalence Studies Checklist for survey studies and Qualitative Research Checklist for interview and focus group studies).12 Different aspects of the methods and reporting of results in included studies were scored individually and then combined to determine an overall assessment of risk of bias (low, moderate, or high). See Supplementary Table II for more details.

Data Extraction and Synthesis

Previous systematic reviews conducted by the research team and their colleagues were used to inform the development of a standardized data extraction template (Supplementary Table III).11,13 After piloting, two authors (JS, KG) independently extracted data and resolved any disagreements through discussion. Disagreements were evident for the categorization of factors, for example, whether fear of what the patient thought was a “patient psychosocial” or “clinician psychosocial” factor. These discrepancies were solved through input of all other co-authors. The data was narratively summarized. JS conducted a preliminary synthesis, which was continually refined throughout the writing process through input from all co-authors.

RESULTS

After removal of duplicates, 1921 titles and abstracts were screened for inclusion or exclusion, and 57 full texts were reviewed, of which 25 met inclusion criteria (Fig. 1). Five additional records were included: one identified in our review on patient-reported factors,14 three from forward citation searching,15,16,17 and one from backward citation searching.18 This led to the inclusion of 30 studies (20 quantitative, nine qualitative, and one mixed-methods).

Figure 1
figure 1

Flow diagram of included studies.

Full size image

Characteristics of included studies are summarized in Table 1. Eight studies included family and internal medicine specialties only;22,23,24,25,30,35,41,42 21 included a combination of family/internal medicine and geriatricians, urologists, oncologists, and obstetrics and gynecology providers;14,15,16,17,18,19,20,21,26,27,28,29,31,32,33,34,36,37,38,39,43 and one included resident physicians.40 Most did not differentiate results across specialty types, except where reported. Thirteen studies examined breast screening;14,15,17,18,23,26,29,30,32,39,42,43,44 six examined colorectal;21,24,25,31,34,40 five examined prostate;16,36,37,38,41 one examined cervical;19 three examined a combination of breast, colorectal, and prostate;20,27,33 one examined breast and cervical;22 and one examined prostate and colorectal.45

Table 1 Characteristics of Included Studies
Full size table

Eight quantitative studies examined factors associated with recommendations to screen in vignettes.21,22,23,29,34,35,40,43 Six quantitative studies assessed factors associated with self-reported recommendations to screen that were inappropriate (according to guidelines19,26,28,32,36,37 or terminal illness15). Another quantitative study assessed factors associated with recommending colorectal screening or seeking patient input first.24 Other quantitative studies examined factors associated with actual breast screening above 75 years using provider-level data,18 PCP views on what factors should be considered when ordering and stopping PSA screening,38,41 and concerns around recommending breast screening for women ≥ 75 years (promoting uptake).42 Three studies qualitatively explored factors PCPs consider in screening decisions for older adults (≥ 75 years25,30, ≥ 80 years39; one used vignettes25). Five qualitative studies examined decision-making about discontinuing screening or inappropriate screening16,17,20,27,33 and one assessed colorectal screening test options.31 One study examined PCP views on the extension of national breast screening programs to older women using mixed-methods.14

Nine of 20 cross-sectional studies had low risk of bias,15,19,21,22,23,24,26,28,29 one had low-moderate risk,32 six had moderate risk18,34,35,36,37,38 and four had high risk40,41,42,43 (Supplementary Table II). Concerns were sampling (11/20 studies), inadequate sample sizes (14/20 studies), unclear degree of coverage of identified sample (18/20 studies), and measure validity (13/20 studies). Five of nine qualitative studies had low risk of bias,16,17,20,25,27 three had low-moderate risk,30,31,33 and one had moderate risk.39 The mixed-methods study had moderate-high risk of bias (assessed using cross-sectional and qualitative checklist).14 Concerns included no philosophical perspective statement (6/9 studies) and limited acknowledgement of the researcher’s location culturally or theoretically (7/9 studies), and their influence on the research (9/9 studies).

Findings from the included studies are summarized in Figure 2. It highlights the factors reported as influential by three or more quantitative or qualitative studies. Furthermore, findings are narratively synthesized below according to patient demographic and health characteristics, patient psycho-social factors, clinician characteristics, clinician psycho-social factors, and health system factors, with the most commonly reported factor summarized first. Influential factors are only summarized in-text if reported by two or more studies. However, all the findings from the quantitative studies (including effect sizes where available) are shown in Table 2, and for the qualitative studies in Table 3 (quotes in Supplementary Table III) and interactions between factors in Table 4.

Figure 2
figure 2

Individual studies reporting influence of factors on PCPs’ recommendations about cancer screening for older adults; further displacement from the center indicates a greater number of individual studies reporting factor as influential; range = 0–9; factors were only included if reported by ≥ 3 studies (across qualitative and quantitative).

Full size image
Table 2 Summary of Quantitative Studies Examining Factors Influencing Primary Care Practitioners’ Cancer Screening Recommendations
Full size table
Table 3 Summary of Qualitative Studies Examining Factors Influencing Primary Care Practitioners’ Cancer Screening Recommendations
Full size table
Table 4 Summary of Studies That Examined Interactions Between Factors Influencing Primary Care Practitioners’ Cancer Screening Recommendations
Full size table

Patient Demographic Characteristics

Patient Age

Twelve studies (four of good quality) reported that PCPs’ screening recommendations for older adults were influenced by their age.14,20,22,25,27,31,34,35,37,38,39,41 Quantitative studies using vignettes and linear regression models found greater likelihood of recommending screening to 80-year-olds compared to 90-year-olds;22 to 75-year-olds compared to 85-year-olds;34 and to 70-year-olds compared to 85- or 90-year-olds.14,35 However, these studies had moderate34,35 to high risk of bias14 and the one study with low risk of bias was conducted in 2006.22 In descriptive and qualitative studies, PCPs reported using specific age cutoffs in deciding whether to screen or not. For example, 67.5% reported that they discontinued prostate screening based on age38 and “I’ll just do 75 for everything.”20

Patient Health Characteristics

Life Expectancy

Fifteen studies (four of good quality) found that PCPs considered life expectancy an important factor in screening recommendations, including being less likely to recommend screening for individuals with limited life expectancy as presented in vignettes (< 5 years23,34,41 or < 2 years14,24,40).14,17,25,30,31,33,38,39,43 In qualitative studies, PCPs determined clearly limited life expectancy by considering comorbidities, frailty, and cognitive decline, but also reported difficulty or uncertainty about these estimations38 or disagreed with using life expectancy to guide decision-making. 33,38 When life expectancy was 5–10 years in vignettes, PCPs were more likely to defer to the patient’s preference for colorectal screening.24,40

Health Status, Comorbidities and Functional Status

Five quantitative studies, including three of good quality, found that PCPs were less likely to offer screening (or more likely to recommend against24) to those in poor health compared to good health22,40 or to those with more severe comorbidity.21,34 PCPs also described comorbidity as an important factor in decision-making in six studies (two of good quality) 14,27,31,33,42,43 and one study of poor quality found 68% felt that breast screening should be offered to healthy women regardless of their age.14 Five qualitative studies, including two of good quality, highlighted how PCPs assessed health status globally or based on “gut feeling” (e.g., considering comorbidities, functional status, life expectancy) when making screening recommendations.14,25,27,30,31 One quantitative study of lower quality and one qualitative study of good quality also found functional status and ability to tolerate further tests/treatment were influential.27,42

Cancer Risk

One quantitative study of low quality41 and two qualitative studies of moderate quality30,46 reported that risk due to family or personal history of cancer influenced PCP recommendations for breast and prostate screening.

Screening History

Screening history influenced PCPs in two studies whereby history of normal cervical screening reduced likelihood of offering screening to older women (good quality)22 and longer time since last colonoscopy (> 5 years vs < 5 years) increased likelihood of recommending colonoscopy.34

Patient Psycho-social Factors

Patient Personal Preference

Eight quantitative studies (2 of good quality) and nine qualitative studies (4 of good quality) found that PCPs considered patient preference (also conceptualized across studies as patient’s request, expectation, autonomy, satisfaction, experiences, or anxiety), as influential in screening recommendations.14,16,17,19,24,25,27,30,31,33,35,37,38,39,41,42,43

Clinician Characteristics

Influence on Recommendations

In three lower quality studies, clinician age influenced recommendations, such that older PCPs were less likely than younger PCPs to recommend colorectal screening for 85-year-olds34 but more likely to recommend prostate screening for men ≥ 75 years.36,37 There were mixed findings for the impact of gender and specialty, as two studies (one of good quality) reported no impact for colorectal screening.22,34 However, other studies of lower quality found female PCPs in England were twice as likely to be undecided about breast screening compared to males (who were more likely to recommend against)14 and family medicine PCPs were more likely to recommend prostate screening for men ≥ 75 years in the USA than medical and radiation oncology specialty providers.36,37

Influence on Inappropriate Recommendations (Based on < 5-Year Life Expectancy or Moderate-Severe Comorbidity Including Non-small cell Lung Cancer or Ischemic Cardiomyopathy with Dyspnea)

All studies that assessed influence of clinician characteristics on inappropriate screening recommendations were of good quality. Internal medicine and family medicine doctors were less likely to recommend inappropriate colorectal21 or breast screening29 compared to obstetrics and gynecology physicians. Females were more likely to recommend inappropriate colorectal21 or breast screening22 than males (but no difference for cervical).22

Clinician Psycho-social Factors

Attitudes and Perceived Benefits/Harms

PCPs considered direct and downstream harms for screening recommendations in four studies, including three of good quality (i.e., overdiagnosis, overtreatment, false positives).20,25,26,33 However, two qualitative studies (one of good quality) highlighted differences in PCP recognition of screening harms for older adults, whereby some PCPs did not believe overscreening in older adults occurred, felt it was acceptable, or disagreed with how it is defined based on limited life expectancy.17,33 Similarly, a small study found that PCPs recommend inappropriate prostate screening despite knowing that harms may outweigh benefits.16 PCPs were more likely to recommend screening when concerned about missing cancers in three good quality studies,19,20,26 and when the test was perceived as accurate/effective.20,31 Uncertainty led to higher breast screening rates and recommendations in two more recent studies (one of good quality)18,26 or deferring to patient preference rather than provide a recommendation in one study from 2008.40 Positive attitudes towards screening were associated with recommending screening in two studies (one of good quality).23,35 PCPs reported decisions about screening based on a gut feeling, not necessarily conscious or deliberate,27 given there were so many factors to consider.25

Social Factors

Three studies (one of good quality) found that PCPs reported inappropriately screening due to difficulty discussing risks and benefits or stopping screening.14,19,39 Anecdotal experiences with patients27 or their own social networks32 also influenced PCP decision-making. Fear of negative patient appraisal and liability concerns led PCPs to continue inappropriate cervical screening19 or recommend breast screening to older women.26 PCPs also reported patient-PCP relationship as an important factor in three qualitative studies.25,30,39

Health System Factors

Guidelines/Recommendations

Four studies (two of good quality) reported that guidelines were influential in PCPs’ screening recommendations including US Preventive Services Task Force (USPSTF),29,30,35 American Cancer Society,30 and American College of Obstetricians and Gynecologists15 guidelines. However, three qualitative studies highlighted negative views towards guidelines, including concerns of inequality,14 undermining patient-centered care,33 and inadequate evidence to support best practice.30 In older studies, PCPs also reported that better guidelines were needed.39,42 In recent studies, PCPs expressed desire for decision support (e.g., patient hand-outs, electronic medical record (EMR) notifications with guideline recommendation updates).30,43

Other Health System Factors

In three studies (one of good quality), PCPs reported limited time to provide justification for a recommendation to stop screening.19,38,39 Clinical reminders28,35 and EMR alerts27 influenced recommendations in three studies (two of good quality). In two poorer quality studies, PCPs did not consider stool tests and breast screening for older adults worthwhile due to limited cost-effectiveness and unnecessary burden on the health system14,31 and costs for the patient was also reason for not recommending screening in two other poorer quality studies.30,42 PCPs quality metrics including scores on how well they are screening patients also led them to encouraging screening up to 70 or 75 years, regardless of other factors.30,31

Interaction Between Factors

Table 4 highlights findings from studies that reported the interaction of factors influencing PCPs’ screening recommendations.

Age

One vignette study found that PCPs were less likely to recommend colorectal screening to an 80-year-old with congestive heart failure compared to a healthy patient, but this difference was not evident for patients 65 years old.21 Another study found that the consideration of health status did not change with increasing age.34 The influence of screening history also depended on age, as 53% of PCPs recommended screening for a 75-year-old who had undergone colonoscopy < 5 years ago, but 23% recommended screening for an 80-year-old with the same screening history.34 Patient request was also more likely to influence prostate and colorectal screening recommendations for 70-year-olds compared to 90-year-olds.35

Health

PCPs used average life expectancies of people at that age to guide their colorectal screening decisions for patients in good health, but for patients in poor health, severity of comorbidities and functional status was influential.24 PCPs were also more likely to elicit preferences for a patient in fair health compared to a patient in poor health.24 Reported difficulty estimating life expectancy did not impact whether life expectancy influenced prostate screening decision-making.38

PCP Characteristics

Female PCPs were more likely to recommend prostate screening due to patient anxiety compared to male PCPs41 and non-black providers were more likely than black providers to report that their patients expected them to continue ordering PSA tests, uncomfortable with the uncertainty of discontinuing screening, and that it takes more time to explain rationale for stopping screening compared to just continuing.38 Family and internal medicine PCPs were more likely to be influenced by USPSTF recommendations than obstetrician-gynecologists who were more likely to be influenced by the American Cancer Society or American College of Obstetricians and Gynecologists guidelines.29

DISCUSSION

Our review builds on previous narrative reviews by systematically synthesizing evidence on how patient, clinician, and health system factors influence PCPs’ cancer screening decision-making for older adults. Our findings highlight a complex interplay between patient health factors and clinician psycho-social factors in decision-making for PCPs, and the role of older adults’ screening preferences. US-based research dominates the evidence base, highlighting the need for further research to understand factors influencing PCPs’ decision-making for cancer screening in older adults in non-US countries, including those with national screening programs.

Prior research has posited frameworks for individualizing cancer screening decisions for older adults that consider patient, clinician, and health system factors,2,47 and has summarized the evidence and guidelines for cancer screening in older adults.5,48 Breslau and colleagues developed the Individualized Decisions for Screening (IDS) framework in 2016, providing a guide for the conceptualization, measurement, and implementation of multi-level interventions to improve the quality of older adult’s screening decisions through consideration of patient, clinician, and health system influences.2 Our review provides an up-to-date summary of how PCP decision-making may or may not align with this framework for individualized decision-making.

Regarding patient factors, our review highlights older adults’ preferences particularly influence PCPs’ decision-making, emphasizing the importance of continued efforts to develop and implement tools to support informed choice for screening. Although the gold standard may be that PCPs provide information about the benefits and harms of screening, additional approaches may be helpful such as directly targeting information to older adults, perhaps even before the notion of no longer screening becomes relevant, as previously suggested by Schonberg and colleagues for breast screening.49 Older adults have long heard persuasive screening messaging and some have strong positive attitudes towards screening that may outweigh the influence of factors such as limited life expectancy or a clinician’s recommendation.50 They may need time to consider tailored information, clarify values, and have ongoing conversations with their doctor and family if desired. Researchers have examined how patient decision aids can improve the quality of cancer screening decisions for older adults and reduce overscreening. For example, Schonberg and colleagues developed a decision aid for women aged ≥ 75 years in the USA, which led to more informed screening decisions and a 9% reduction in the number of women choosing to be screened.51 In countries with national screening programs, it may also be useful for information to be provided from screening programs to communicate to individuals before they reach the upper age when screening is no longer recommended. However, further research on the impacts of providing such information outside of US contexts is needed.

Our review also highlighted clinician-level barriers to achieving individualized screening decisions for older adults. Contrary to Breslau and colleagues’ suggested framework,2 some PCPs may not make conscious, deliberate screening decisions, may only consider health status in a general sense and life expectancy to varied degrees when making recommendations, and have fears around recommending stopping screening. Other contributors to this challenge are the influence of PCP attitudes and anecdotal experiences on their decision-making, the importance of age-based guidelines, and differences in how they perceive harms such as overscreening. These findings were particularly highlighted in qualitative studies. Continued efforts to develop and implement patient-level interventions such as patient decision aids must also be accompanied by further efforts to support PCPs to provide evidence-based, tailored recommendations to stop screening when it is relevant to do so.

Recommendations to stop screening that incorporate health status (“other health issues should take priority”) may be more acceptable to older adults than recommendations that incorporate life expectancy (“you may not life long enough to benefit”).52,53 However, in our experimental scenario-based study, recommendations incorporating life expectancy resulted in reduced intention to screen compared to recommendations incorporating health status.6 Specific scripts and strategies co-designed with patients and PCPs to discuss stopping screening in various clinical scenarios have also been developed in the USA, including a briefer statement about stopping, a longer script highlighting reasons for stopping, and a shared decision-making script to support discussion about the benefits and harms,54 which could be adapted for use in countries with national screening programs. Further research is needed to test communication strategies that are acceptable to PCPs and effectively support older adults to have a realistic understanding of the benefits and harms of cancer screening, especially for those with strong positive attitudes towards screening.

System-level factors such as time, age-based guidelines, and clinical reminders and electronic medical record alerts are also important influences on PCP recommendations and may form barriers to individualized decision-making. In the USA, it is essential that existing systems are leveraged to better align with the goal of individualized screening decisions. Suggestions include training for PCPs in the use of decision aids, professional organizations expressing support for the use of decision aids, incorporating use of decision aids and life expectancy estimation into electronic medical record alerts,46,51 and pre-clinical visits for PCPs to understand patient preferences and understanding ahead of a discussion. In non-US contexts, further research is needed to understand whether and how individualized decisions for cancer screening in older people can be achieved in the context of national screening programs. The design of organized screening programs is likely an important influence on decision-making about cancer screening, for both younger and older adults.

Limitations

The studies included in our review are limited due to the focus of the literature on the US context, therefore lacking generalizability to other international contexts. Some included studies were also vignette-based, meaning it is difficult to understand whether the factors would have a similar impact in clinical settings. Our review itself was limited as it was not possible to conduct a meta-analysis using any of the quantitative data due to the variability in study designs, measurement, and operationalization of decision-making and influencing factors in the included studies. Extracted data was synthesized narratively, as understanding the relative magnitude of effect sizes for influencing factors was not the aim of this study. This narrative approach allowed for the variation and heterogeneity across studies to be captured, building on the existing narrative reviews on this topic. Strengths of this review were having two independent reviewers conduct screening, data extraction and quality appraisal, and searching studies from grey literature (e.g., theses).

Conclusion

There are a wide range of patient, clinician, and health system factors that influence PCPs decisions and recommendations for cancer screening in older adults. Decision support should continue to be developed and implemented to support informed choice for older adults and assist PCPs to consistently provide evidence-based recommendations, alongside system changes in electronic medical records, professional organizations, and clinician training. Our findings also highlight the importance of supporting informed choice for younger adults who are beginning to screen, so this information does not become difficult for clinicians to communicate to patients when the benefit/risk trade-off becomes less favorable. Further research is also needed in non-US contexts, especially where national screening programs are implemented, to understand the role of PCPs in older adults’ cancer screening decisions.