Should we screen aging physicians for cognitive decline?

Abstract

Objectives

To synthesize evidence relevant for informed decisions concerning cognitive testing of older physicians.

Methods

Relevant literature was systematically searched in Medline, EMBASE, PsycInfo, and ERIC, with key findings abstracted and synthesized.

Results

Cognitive abilities of physicians may decline in an age range where they are still practicing. Physician competence and clinical performance may also decline with age. Cognitive scores are lower in physicians referred for assessment because of competency or performance concerns. Many physicians do not accurately self-assess and continue to practice despite declining quality of care; however, perceived cognitive decline, although not an accurate indicator of ability, may accelerate physicians’ decision to retire. Physicians are reluctant to report colleagues’ cognitive problems. Several issues should be considered in implementing cognitive screening. Most cognitive assessment tools lack normative data for physicians. Scientific evidence linking cognitive test results with physician performance is limited. There is no known level of cognitive decline at which a doctor is no longer fit to practice. Finally, relevant domains of cognitive ability vary across medical specialties.

Conclusion

Physician cognitive decline may impact clinical performance. If cognitive assessment of older physicians is to be implemented, it should consider challenges of cognitive test result interpretation.

Keywords:

• Physician
• cognitive aging
• competence
• clinical performance
• fitness for practice
• cognitive assessment

Introduction

Aging affects the entire human body, and changes in cognitive functioning are among the most noticeable features of growing old (Cohen et al., 2019). Decline in cognitive functions associated with aging cannot always be attributed to pathological processes in the brain, such as neurodegeneration in Alzheimer’s disease. For example, most older adults process information slower than they did when they were younger (Institute of Medicine, 2015). Normal cognitive aging is ‘a process of gradual, ongoing, yet highly variable changes in cognitive functions that occur as people get older. Cognitive aging is a lifelong process. It is not a disease or a quantifiable level of function’ (Institute of Medicine, 2015) (p. 2). While evidence suggests that neurodegeneration leading to Alzheimer’s disease spans decades (Braak et al., 2006), the boundary between normal age-associated cognitive changes and the initial phase of a neurodegenerative disorder is not well understood. Distinguishing between normal and pathological changes may be challenging (Institute of Medicine, 2015). Transition from normal cognition to mild cognitive impairment has not been well studied, though evidence supports several health and lifestyle factors as predictors (Cherbuin et al., 2009).

Many cognitive functions decline in older adults, although there is great variability between individuals. Areas of decline include speed of information processing, selective attention, divided attention (multitasking), working memory, episodic memory, prospective memory, source memory (the ability to remember contextual aspects of an experience), executive functioning, reasoning ability, and spatial ability (Cohen et al., 2019; Glisky, 2007; Institute of Medicine, 2015). In contrast, other functions remain relatively stable over the lifespan, including sustained attention (or vigilance), semantic long-term memory (the ability to store general knowledge about the world acquired over a lifetime), and wisdom (a multidimensional construct) (Cohen et al., 2019; Glisky, 2007; Institute of Medicine, 2015). Procedural memory (remembering how to perform an activity) and language functions are only partially affected in old age. Older adults are good at performing automatic or well-learned procedures and can learn new procedures; however, they may be slower when performing familiar tasks and may learn new procedures at a slower rate than younger adults (Glisky, 2007; Institute of Medicine, 2015). There is evidence that discourse skills improve with age; vocabulary and word comprehension remain stable, whereas language production skills appear to decline (Glisky, 2007; Institute of Medicine, 2015).

Despite high education levels and careers that require constant and involved mental aptitude, physicians are not immune to age-related changes in cognitive function. Many physicians continue to practice beyond age 65, and there exists concern that cognitive impairment may undermine patient safety (Dellinger et al., 2017). For instance, the Quality in Australian Health Care Study reported that failure of cognitive function was the second most frequent cause of errors in the delivery of healthcare that led to adverse events in patients (Wilson et al., 1999).

The current review aimed to synthesize evidence on cognitive decline in aging physicians and its relation to clinical competency. The results of the review will serve to guide policy on possible implementation of cognitive screening of physicians. The review focused on scientific evidence; ethical, legal, and financial considerations were beyond the scope of the review.

Methods

Literature was searched in Medline, Embase, PsycInfo, and Education Resources Information Center (ERIC) in July-August 2022. The search strategy was developed for Medline (See Appendix, Tables A1 and A2) and then adapted for other databases. Bibliographies of review articles were inspected for additional references. References identified from all sources underwent two levels of screening: title and abstract (level 1) and assessment of full text (level 2) based on pre-defined eligibility criteria. Eligible were systematic reviews and original studies that included older physicians of any specialty, and objective measures of cognitive functioning or clinical competence or performance.

Table 1. Studies of the relationship between physician age or years in practice and competence or performance.

Reference	Physicians’ characteristics	Outcome	Association with age^1
B. R. Anderson et al. (2017)	Surgeons Years since graduating from medical school: 9-55 (median 25)	Major morbidity, mortality, and a composite outcome after congenital heart surgery	Negative with major morbidity and composite outcome No association with patient mortality
G. M. Anderson et al. (1997)	General practitioners; family practitioners; specialists Age: <45 to 65+ years	Inappropriate prescribing	Negative
Brown et al. (2016)	Primary care physicians Years since completing residency: 1-46 (mean 20)	Diabetes glucose control	No association
Gershon et al. (2014)	Primary care physicians Age: <35 to >65 years	Pulmonary function testing for diagnosis of COPD	Negative
Goulet et al. (2013)	Family practitioners Age: <40 to ≥70 years.	Composite score for three components: clinical investigation, diagnostic accuracy, appropriateness of treatment and follow-up	Negative
Grace et al. (2014)	Physicians of various specialties Age: 32 to 84 (mean 53.1) years.	Competence assessment tailored to practice area and specialty	Negative
Marco et al. (2018)	Emergency medicine residency–trained physicians Age: 35 to 72 (mean 45) years	Score on the American Board of Emergency Medicine ConCert examination	Negative
Markar et al. (2018)	Surgeons Age: <30 to ≈70 years	90-day all cause mortality; 5-year all cause mortality; 5-year disease-specific (esophageal cancer) mortality after esophageal cancer resection	Negative
Mehrotra et al. (2018)	Endoscopists Years in practice since residency: ≤9 to 51	Adenoma detection rate	Negative
Meier et al. (2020)	General practitioners Age: mean 52 (SD 9.3) years	Quality of diabetes care	No association
Meltzer et al. (2017)	Surgeons (vascular) Years since medical school graduation: 3 to >30	Outcomes of abdominal aortic aneurysm repair: 30-day mortality and 1-year mortality.	No association
O’Neill et al. (2000)	Surgeons Age: 30 to ≥65 years	In-hospital mortality; in-hospital overall bad outcome (mortality or morbidity) following carotid endarterectomy	Negative with mortality No association with the rate of overall bad outcome
Pentzek et al. (2009)	General practitioners Age: mean 51.13 (SD 8.05) years	Dementia knowledge	Negative
Poulin de Courval et al. (1996)	Family physicians Age: 25 to >65 years	Knowledge and skills with respect to dementia	Negative
Prystowsky (2005)	Surgeons Years since certification by the American Board of Medical Specialties: <5 to >30	Mortality, morbidity, and hospital length of stay after alimentary tract surgery	Negative with hospital length of stay^2 No association with mortality and morbidity
Reid et al. (2010)	Physicians of various specialties Years in practice: <10 to ≥50	Measures from RAND’s Quality Assessment Tools: scores on 124 indicators of quality of care for 22 acute and chronic conditions; preventive care	Negative with acute care and male-specific care. No association with chronic care, female-specific care, and all measures Positive association with preventive care.
Satkunasivam et al. (2020)	Surgeons Age: 29–80 years	30-day postoperative outcomes of 25 procedures: deaths, complications, hospital readmission in the 30 days after surgery; a composite outcome; hospital length of stay	Positive with most outcomes. No significant association with hospital length of stay after emergent surgery in the continuous analysis and negative association in dichotomized analysis (surgeons’ age ≥ 65 vs. <65 years).
Smith-Bindman et al. (2005)	Physicians who interpret mammograms (radiologists?) Age: <40 to ≥70 years	Accuracy of screening mammogram interpretation: sensitivity, specificity, overall accuracy	Positive with specificity and overall accuracy. No association with sensitivity
Tsugawa et al. (2018)	Surgeons Age: <40 to ≥60 (mean 50.5, median 50.0) years	Mortality during hospital admission or within 30 days of non-elective surgeries	Positive
Tsugawa et al. (2017)	Hospitalist physicians Age: <40 to ≥60 years	Mortality within 30 days of admission; 30-day readmission rates; costs of care	Negative with 30-day mortality No association with 30-day readmission and costs of care
Waljee et al. (2006)	Surgeons Age: <40 to >61 years	Mortality before discharge from hospital or within 30 days of the surgery	Negative with mortality after pancreatectomy, coronary artery bypass grafting, carotid endarterectomy. No association with mortality after elective repair of an abdominal aortic aneurysm; aortic valve replacement; lung resection; esophagectomy; cystectomy
Wenghofer et al. (2009)	General/family practitioners Age: mean 51.0 (SD 9.91) years	Acute care, chronic care, continuity of care, well care, managing patient records	No association
Weycker and Jensen (2000)	All non-osteopathic physicians Age: <40 to 60+ years	Medical malpractice	Positive

¹An association between physicians’ age and their competence/performance is described as negative if competence/performance declined with rising age; positive if superior performance was seen among older physicians.

²“… When it comes to LOS, the newer surgeons tend to get their patients out faster. … that might be a bias of older more experienced surgeons who might hang onto their patients a little bit longer …” (Prystowsky, 2005) (p. 502).

Table 2. Studies of cognitive functioning of physicians referred for assessment due to clinical competency or performance concerns.

	Population	Cognitive function measure	Clinical performance or competence measure	Findings
Del Bene and Brandt (2020)	Referred group 30 physicians of different specialties referred for neuropsychological evaluations. 33% had known neurological disorders, 20% had known psychiatric disorders, 27% had suspected cognitive disorder. Many not currently practicing. Mean age 64.27 (SD 12.53) years. Comparison group 39 currently practicing community urologists with no neurological, psychiatric, or cognitive disorders. Mean age 48.51 (SD 8.71) years.	9 variables from 7 cognitive and motor tests	Referral for assessment due to quality-of-care deficiencies, or concern of healthcare professionals about physician neurocognitive wellbeing	The referred group was significantly older and performed below the comparison group on the test battery as a whole and on each variable^1. Impaired test score defined as <5th percentile of the control physicians’ performance. Suggested cut-off to identify impaired physicians: impaired score on >5 of the 9 variables. At least one of the impaired scores must be on a motor-free test. Using this cut-off, 14 referred physicians categorized as impaired and 16 as ambiguous. Neurocognitive scores of impaired physicians were inferior to those of ambiguous physicians. Impaired physicians were older, more often suspected of having MCI, dementia, or Parkinson’s disease, and more likely discontinued practicing.
Korinek et al. (2009)	Competency group 267 physicians of different specialties referred for competency evaluation Age 30-80, mean 51.5 (SD 9.1) years Comparison group 68 physicians of similar age with no competency concerns Age 31-74, mean 49.4 (SD 12.5) years	MicroCog scores for 5 specific cognitive domains and 3 summary scores	Referral for competency evaluations; no details provided	Scores of the competency group significantly lower than those of the control group on 3 summary measures and 4 cognitive domains. 24% in the competency group vs. 0% in the comparison group (p < 0.001) had cognitive difficulty defined as a cognitive proficiency score >1SD below the mean, or any two index scores >1 SD below the mean.
Perry and Crean (2005)	Study Group 148 physicians of different specialties whose primary language was English referred to the Physician Assessment and Clinical Education (PACE) program. Comparison group Normative reference samples.	24 subsets derived from four neuropsychological test batteries	Referral to PACE due to various infractions	The referred group performed in the average range on most neuropsychological measures. 61% scored 1SD below the normative reference sample on three or more neuropsychological measures. 14 (9%) scored in the average range or greater on all measures. The referred physicians performed lower than expected on tests of sequential processing, attention, logical analysis, eye–hand coordination, verbal and non-verbal learning.
Turnbull et al. (2000)	Study group 27 physicians referred to the Physician Review Program (PREP) Comparison group Internal comparisons by level of performance in the PREP	Neuropsychological screening battery to assess problem solving (verbal and visual/spatial), concept formation, reasoning, learning, memory, attention, complex mental tracking, and verbal productivity (fluency)	Performance in the PREP program.	Of 8 physicians who performed well in the PREP assessment, 7 had no, minimal, or mild cognitive impairment. Of 19 physicians who scored poorly at PREP, 6 had moderate or severe cognitive impairment.
Turnbull et al. (2006)	Study group 45 physicians referred to the PREP Comparison group Internal comparisons by level of performance in the PREP	See Turnbull et al. (2000)	Performance in the PREP program.	Of 14 physicians scoring well at PREP, 12 had no, minimal, or mild cognitive impairment using age-adjusted reference norms. Of 31 physicians scoring poorly at PREP, 12 (38%) had moderate to severe cognitive impairment using age-adjusted reference norms, and 17 (55%) using age-independent reference norms. The best predictor of the PREP score was the age-independent global cognitive score (r = 0.57).

¹The authors acknowledged possible incompatibility of the referred and the control group. All physicians in the control group were currently practicing urologists, whereas the referred group included physicians of various specialties, both active and retired.

Additional eligibility criteria were applied to studies that examined associations between physicians’ age and their clinical competence or performance. These studies had to include two or more age categories older than ∼50 years and at least one category younger than 50 years. Studies with age measured as a continuous variable needed to include physicians aged from less than ∼50 years to ∼70 years or older. Additionally, analyses had to be adjusted for potential confounders, such as characteristics of the physician (other than age), the patients, and the clinical practice.

Results

Does aging affect physician cognition and the ability to practice competently and safely?

A PRISMA flow diagram (Moher et al., 2009) detailing the study selection process is shown in Figure 1.

Figure 1. PRISMA flow diagram for studies of age-related changes in physicians’ cognition or their clinical performance/competence.

Physician age and cognitive functioning

Four studies were identified, of which three included subsets of the same study population. Additional details on these studies can be found in Appendix (Table A3).

Bieliauskas et al. (2008) used the Cambridge Neuropsychological Test Automated Battery (CANTAB) to assess 359 surgeons aged 45–86 years. Visual sustained attention and stress tolerance in the rapid visual information processing (RVIP) task, reaction time (RTI task), and visual learning and memory in the paired associates learning (PAL) task declined with increasing physician age; the decline was most pronounced in age groups over 70 years, although some functions showed an earlier decline, from ages 60. When surgeons were retested one to six years after initial testing, an increase in mean latency of response on the RVIP task was observed with no significant changes in accuracy. There was a decrease in number of errors per trial on the PAL task, which the authors explained by test familiarity or the practice effect (Bieliauskas et al., 2008).

Boom-Saad et al. (2008) analyzed data on 308 practicing surgeons from the same dataset, where surgeons were divided into two age groups: 139 mid-career surgeons (45–60 years) and 169 senior surgeons (61–75 years). Their performance on the three tasks was compared with normative age-matched controls and also with the performance of 20 to 35-year-old surgical trainees. The performance of senior surgeons on all three neuropsychological tests was inferior to that of mid-career surgeons and of surgical trainees. Senior surgeons outperformed the age-matched normative control only on one of the two parameters of the RTI test, whereas the performance of mid-career surgeons and surgical trainees was significantly better than that of age-matched normative controls on both the RTI and RVIP tests.

Drag et al. (2010) studied the cognitive status of the same surgeons in relation to their retirement status. Their analysis included data on 168 senior surgeons (60+ years) and 126 younger surgeons (45–59 years). Of the senior surgeons, 36% had retired, 33% were planning to retire within 5 years, and 30% had no imminent retirement plans. It was found that 55% of all senior surgeons, 49% of senior surgeons with no imminent plans to retire, 72% of senior surgeons planning to retire within the next 5 years, and 45% of retired surgeons performed within the range of the younger surgeons on all three cognitive tasks. No senior surgeon performed below the younger surgeons on all three tasks. This study demonstrated that many senior surgeons performed at or near the level of younger surgeons on cognitive tasks, suggesting that older age does not necessarily affect cognition.

Using the computerized cognitive test, MicroCog, Powell and Whitla (1994) assessed cognition of 1,002 physicians and 581 other subjects (referred to as normal subjects). The physicians ranged from 25 to 92 years of age, and normal subjects from 25 to 83 years. The MicroCog test measures reactivity, attention, numeric recall, verbal memory, visuospatial facility (the individual’s capacity to identify visual and spatial relationships among objects), reasoning, and mental calculation. Physician total MicroCog scores gradually declined between 30 and 60 years of age, with a more rapid decline after 60 years of age. Among the normal subjects, the total MicroCog score declined more rapidly after age 50; in the two decades after age 60, their scores were similar to those of physicians ten years older. The gap between the two groups narrowed after age 75; however, at all ages, the normal subjects scored lower than the physicians. As to specific cognitive functions, similar patterns were seen for the physicians and the normal subjects. Visuospatial scores showed the steepest decline, followed by reasoning scores (decline in reasoning scores of physicians occurred about a decade later than the decline in visuospatial scores), and then verbal memory. More resistant to aging were calculation skills (math scores) and attention. Specific abilities, as well as global cognitive functioning, showed a relatively mild decline until age 70–74 years and a more marked decline at 75–79 years. No significant age-related changes were seen for reaction times, which was not consistent with most of the literature. The discrepancy may be explained by either relatively less demanding tasks used in this study or by the older study participants being not representative of their age group (Powell & Whitla, 1994). This study also demonstrated a significant variability in cognitive performance among physicians aged 65 years and older, with a substantial number of older physicians functioning at a level close to that of younger subjects.

In summary, studies reviewed in this section demonstrated a decline in cognitive functioning of older physicians detectable at about 60 years, which is 10 years later than in the general population. This is likely due to physicians’ higher educational level and intellectually demanding occupation, factors that increase and maintain cognitive reserve (Stern et al., 2020). Cognitive reserve poses a possible issue for interpretation of results, e.g. appropriateness of using age-matched general population data for comparison (Turnbull et al., 2006). There was a substantial inter-individual variability in older physicians’ cognitive functioning. It should be noted that, because all these studies are cross-sectional, a causal association cannot be inferred between the observed differences in test scores and physician age. These differences in cognitive performance may also be related to other unaccounted differences between the age groups. Although cross-sectional data are generally adequate to answer the question about age-related cognitive decline, they may overestimate both the magnitude of the decline and how early the decline occurs (Powell & Whitla, 1994).

Physicians’ age and ability to practice competently and safely

Two related concepts were considered: competence and performance. Competence reflects how physicians perform in a controlled environment, specifically in testing situations such as a written examination or Objective Structured Clinical Examination. Physicians acquire competence through medical school and residency/fellowship training, and on passing qualifying medical licensure examinations. Performance represents how physicians apply their attitudes, knowledge, and skills in a multidimension practice context that is impacted by many factors. Competence is necessary but not sufficient for high-quality physician performance (Kain et al., 2019; Rethans et al., 2002). There is some uncertainty regarding the informativeness of competence assessments noted in an assessment center for performance in clinical practice (American Medical Association, 2015), as demonstrated by studies showing inconsistent correlations between competency test results and parameters of physician performance in actual practice (Rethans et al., 2002).

Twenty-three original studies that meet the eligibility criteria and report on results of original research are summarized in Table 1 and detailed in the Appendix (Table A4). Three systematic literature reviews were also identified and summarized below.

One of the three systematic reviews (Jung et al., 2022) included meta-analysis of 10 retrospective cohort studies of postoperative outcomes with significantly heterogeneous results. Postoperative mortality among patients of older surgeons (older than 50, 55 or 60 years in different studies) was 14% higher compared to patients of middle-aged surgeons and 23% higher compared to patients of young-aged surgeons (younger than 40–45 years in most studies), though the latter difference was not statistically significant. Postoperative major morbidity among patients did not differ across surgeon age groups. Morbidity after minor (small organ) or major (large organ) surgeries also did not differ by age group.

Another systematic review (Choudhry et al., 2005) included 59 studies on the relationships between physicians’ age or years in practice and their medical knowledge, guideline adherence, patient mortality, and other quality of care outcomes. These studies, reporting on 62 groups of outcomes, were classified into six groups based on the observed association. A decline in performance for all assessed outcomes with increasing physician age or years in practice was reported in 32 of 62 evaluations (52%); heterogeneous results (only some showing decline) were reported in 13 evaluation (21%); 2 studies (3%) reported an initial increase in performance with growing experience, a peak, and a subsequent decline; 13 (21%) reported no association; one study (2%) reported improved performance for some outcomes but not for others, and one reported improved performance with age for all outcomes assessed. These results did not change substantially when analyses were restricted to objective outcome measures (use of chart audits or administrative databases vs. self-reports) or to studies adjusted for other known predictors of health care quality, such as patient comorbidity, physician volume or specialization. Although this review was based on heterogeneous studies, the results suggest that older physicians may be at risk for providing lower-quality care (Choudhry et al., 2005).

Studies reviewed by Eva (2002) indicated that analytic capabilities declined with age, whereas nonanalytic processing remained intact and became increasingly dominant. Older physicians tended to rely more on their prior experience and to a lesser extent critically incorporated novel conflicting information. When facing a new case, older physicians were more influenced by information encountered earlier in the case. While experienced physicians may have more accurate first impressions, a failure to engage analytic processing in the decision-making process may result in medical errors. This review noted strong inter-individual differences that tended to increase with age. Although, on average, the performance of older physicians tended to be inferior, many older physicians performed at or above the level of their younger colleagues (Eva, 2002).

Although cognitive aging is a risk factor for deficits in physicians’ clinical practice, it should not be interpreted in isolation (Medical Board of Australia, 2017). Older physicians’ training occurred decades in the past and may not be updated regularly (Choudhry et al., 2005). For example, Day et al. (1988) demonstrated that, on recertification examinations, older physicians who were further out of training, performed less well compared to more recently trained physicians on items related to new or changing knowledge; at the same time, they performed as well as younger examinees on items testing stable knowledge. Choudhry et al. (2005) noted that older physicians may be less accepting to practice innovations.

Physician performance may be influenced by individual characteristics (such as medical school and certification status), organisational factors (for example, solo practice, support systems, hours worked, practice volume, appointment lengths), and systems level factors (such as community size and physician-to-population ratio) (American Medical Association, 2015; Kain et al., 2019; Wenghofer et al., 2009). Aging physicians who collaborate with their colleagues, engage in evidence-based discussions, receive performance feedback, and who use computerized reminder systems, may have advantage over those who practice in isolation or in more traditional settings (Choudhry et al., 2005).

Patient factors, such as disease acuity and complexity, should also be taken into consideration (American Medical Association, 2015). Samuels and Ropper (2005), for instance, suggested that older doctors might tend to treat older and sicker patients.

Age-related physiological changes, such as reductions in manual dexterity, hearing, and visual acuity, may affect physician clinical competence and performance (American Medical Association, 2015; Moutier et al., 2013). Although intact cognition is important for optimal performance of surgeons, good eyesight and hand dexterity are also essential (Fortunato & Menkes, 2019; Yule et al., 2006).

Lack of valid tools for measuring physician competence or practice performance, and the variable nature of physician practices, are among the challenges faced when assessing practicing physicians (American Medical Association, 2015). Quality of care in one domain may not predict quality of care in another. Therefore, caution is recommended when interpreting results of physician assessments based on narrow measures of performance (Parkerton et al., 2003; Yen & Thakkar, 2019). Parkerton et al. (2003) assessed primary care physicians on four measures of performance and found that most physicians (57.4%) ranked in the highest tertile for at least one measure and in the lowest tertile for at least one other measure.

Reported improvements in outcomes for patients of older surgeons may be partly explained by self-selection of highly skilled surgeons to continue performing surgical procedures, whereas lower skilled surgeons may decide to quit clinical practice (Satkunasivam et al., 2020; Tsugawa et al., 2018). Also, volume and complexity of surgical cases may decrease with increasing surgeon age (Bieliauskas et al., 2008).

Postoperative mortality, as a measure of surgery outcome, is affected by technical quality of the operation and other aspects of care, including postoperative inpatient care (R. J. Campbell et al., 2019). Postoperative mortality is specialty-specific and of greater relevance in some surgical specialties than others (Maruthappu et al., 2014).

In summary, although many studies demonstrated age-related decline in physicians’ clinical performance or competence, other studies detected no relationship with age, and still others reported superior performance of older physicians compared to younger counterparts. Although cognitive aging is a risk factor for deficits in physicians’ clinical practice, numerous other factors may contribute to underperformance of older physicians.

What is the association between physician cognition and clinical performance?

Studies relevant to this question include measures of both cognitive function and clinical performance or competence. A PRISMA flow diagram (Moher et al., 2009) detailing the study selection process is shown in Figure 2.

Figure 2. PRISMA diagram for studies of a relationship between physician cognition and clinical performance or competence.

Five cross-sectional studies were identified in this review (Table 2). They demonstrate that, on average, cognitive functioning of physicians with problematic clinical performance or competence is inferior to that of physicians without such problems, or below the normative reference data. Turnbull et al. (2006) found that a cognitive score was a strong predictor of performance in a standardized competency assessment test. Overall, these studies suggest that lower cognitive functioning may be a contributing factor to physician competence and performance problems.

Is there a cognitive threshold at which a physician is no longer fit to practice medicine?

Serra et al. (2007) defined fitness for work as a capacity to work without risk to self or others. Physician fitness for practice (professionalism) includes multiple components, such as knowledge and skills, ethical aspects and values, effective relationships and communication with patients and colleagues, responsibility for their own health and well-being and that of their colleagues (General Medical Council (UK), 2021; Royal College of Physicians and Surgeons of Canada, n.d.).

Del Bene and Brandt (2020) sought to establish an approach to identify neuropsychologically impaired physicians. Seven cognitive and motor tests were administered, and nine key variables were derived. An impaired test score was defined as being below the 5^th percentile of control physician performance. An impaired score on more than five of the nine cognitive and neuromotor variables was proposed as a cut-off to identify impaired physicians. Del Bene and Brandt (2020) noted that, without a ‘gold standard’ for impairment, it was not possible to determine the perfect neuropsychological threshold; the threshold could differ for physicians in different specialties. The authors recommended additional research to develop specialty-specific neuropsychological criteria for impairment and noted that the cross-sectional retrospective design was a limitation of this study. To derive a useful predictor of cognitive impairment, a prospective longitudinal study of currently practicing physicians would be required (Del Bene & Brandt, 2020).

No prospective longitudinal data on cognitive functions in physicians over time, correlating them with clinical performance or competence, were identified for this review.

Overall, the literature to date provides no data to determine a firm cognitive cutoff for identifying physicians unfit to practice.

Other considerations regarding cognitive screening of older physicians

Considerations in support of cognitive screening

Physicians may not be good at assessing their own clinical performance and cognitive skills. For example, Davis et al. (2006) systematically reviewed 17 studies conducted in the United Kingdom, Canada, United States, Australia, or New Zealand and comparing physician self-assessments with objective assessments of knowledge, skill, or performance. These studies included 20 comparisons between self- and external assessments, of which seven found agreement, whereas 13 demonstrated little, no, or even an inverse association between self-assessment measures and external indicators, suggesting a limited ability of physicians to accurately self-assess. Bieliauskas et al. (2008) administered a battery of computerized neuropsychological tests to 359 older surgeons, of whom 294 surgeons also completed a self-report survey asking about perceived cognitive difficulties. It was found that there was no notable relationship between surgeon perception of cognitive difficulties and their performance on cognitive tests. Subjective cognitive problems were significantly related to surgeon retirement status or declared decision to retire in the next five years, whereas no notable relationship was seen between objective cognitive status and the declared retirement decision. These findings suggest that perceived cognitive decline, although it might not be an accurate reflection of true abilities, may play a role in a surgeon’s decision to retire (Bieliauskas et al., 2008).

Although it is their duty to report their own and their colleagues’ cognitive problems, physicians are generally reluctant to do so (Fortunato & Menkes, 2019). In a national survey among physicians in the United States (E. G. Campbell et al., 2007), 96% of respondents agreed that physicians should report significantly impaired or incompetent colleagues. However, 45% of physicians who had personal knowledge of an impaired or incompetent colleague indicated that at least once in the last three years they had not reported that colleague. Forty-six percent of physicians with personal knowledge of a serious medical error did not report that error in the last three years. In another national survey conducted in the United States (DesRoches et al., 2010), 64% of surveyed physicians agreed it was their duty to report impaired or incompetent colleagues. Of the 309 physicians who personally knew a colleague incompetent to practice, 105 (33%) did not report this colleague to the relevant authority. Weenink et al. (2015) conducted a survey of legally regulated healthcare professions in the Netherlands: dentists, midwives, nurses, pharmacists, physicians, physiotherapists, psychologists, and psychotherapists. Although ∼31% of respondents had an experience with an impaired or incompetent colleague, only ∼69% acted upon this experience. The definition of action was broad and included talking to the colleague or discussing with other colleagues. Only one-fifth of participants reported the impaired/incompetent colleague to relevant authorities: the board of the organisation (11.7%), the professional association (3.0%) or the Health Care Inspectorate (6.0%) (Weenink et al., 2015).

Overall, evidence suggests that physicians may not be good at assessing their own clinical performance and cognitive skills and are generally reluctant to report impaired or incompetent colleagues.

Potential challenges with cognitive screening of physicians

Outcomes of most currently available cognitive assessment tools are difficult to interpret due to lack of normative data for physicians (Garrett et al., (2021). The American Medical Association (2015) (p. 7) stated:

… it is uncertain whether and how physician results should be compared to the general population and whether their results should be age-matched for interpretation purposes. … The nature of physician decisions, in terms of their difficulty, acuity and gravity, suggests that even minor changes in cognitive function may be impactful in patient care situations. … Results for cognitive testing that are interpreted as normal based on comparison to an age-matched, non-physician population could potentially represent a significant decline in highly intelligent individuals….

Although some cognitive assessment tools include normative data for physicians, there are limitations associated with inter-individual comparisons in using of normative data (Gaudet & Del Bene, 2022). Baseline cognitive testing would offset these limitations by comparing an individual’s current cognitive performance with prior performance; if longitudinal testing is done, reliable change indices should be used to determine whether the observed cognitive changes are clinically meaningful, or represent normal intra-individual variability (Gaudet & Del Bene, 2022).

There are no standards on the level of cognitive decline at which a doctor is no longer fit to practice (Australian Medical Association, 2017; Lee & Weston, 2012; Medical Board of Australia, 2017). Prospective longitudinal studies correlating physician functioning in specific cognitive domains with key measures of clinical performance or competence in different medical fields would be needed to establish such standards in the face of normal cognitive aging. Such standards, however, would not necessarily apply to a physician compromised by a diagnosed Alzheimer’s disease or after having suffered a cerebral infarction. Lack of scientific evidence linking specific cognitive deficits with physician performance has been acknowledged by others (American Medical Association, 2015; Armstrong & Reynolds, 2020; Medical Board of Australia, 2017; Sataloff et al., 2020; Steffany, 2022).

Sets of neurocognitive abilities required for different medical specialties are diverse, reflecting the diversity of skills required and the level of responsibility that each specialty involves (Garrett et al., 2021; Gaudet & Del Bene, 2022). For instance, psychomotor impairment may weigh more heavily for surgeons, critical care, or emergency medicine specialists (Garrett et al., 2021); changes in processing speed may be particularly relevant for intensivists who rely on rapid decisions and responses (Skowronski & Peisah, 2012); visuospatial abilities are important for radiologists (Gaudet & Del Bene, 2022). Even if cognitive cut-off scores were available, these may differ by physician specialty area (Garrett et al., 2021; Gaudet & Del Bene, 2022; Katz, 2016; Saver, 2020). This in turn may require determination of appropriate assessments and tools, including development of new tests (Fortunato & Menkes, 2019).

Ethnic, cultural, and language factors may also affect performance on cognitive testing, especially for foreign-born physicians or international medical graduates, and must be considered to avoid inaccurate interpretation of test scores. For example, bilingual people are disadvantaged on verbal fluency tasks in their non-dominant language, and their lower-than-expected performance on such tasks may not be due to cognitive dysfunction but rather due to cultural or language differences (Gaudet & Del Bene, 2022).

Conclusion

This review provides some evidence suggesting that cognitive testing of late-career physicians may be useful. Physicians, despite their high education levels and careers that require constant and involved mental aptitude, are not immune to age-related changes in cognitive function. On average, cognitive abilities of physicians decline with increasing age, although this occurs about a decade later than in the general population. Physician competence and performance may also decline with age, though study results varied. Numerous factors, including measures of competence and performance, affect interpretation of results. Cognitive test scores are lower in physicians referred for assessment because of clinical performance or competence concerns than in control physicians with no such concerns, which suggests that lower cognitive functioning may contribute to physician performance problems. Cognitive functioning and clinical performance among older physicians are highly variable, and physicians may not be able to accurately self-assess. And yet, physicians were frequently reported to be reluctant to report their colleagues’ cognitive problems. So, some physicians may continue to practice despite a significant decline in the quality of the care they provide. For others, perceived cognitive decline, although it may not reflect true abilities, can accelerate their decision to retire.

The periods covered in some of the studies of physicians are 10 to 20 years old. Evidence from older studies should be considered cautiously, especially with changing specialities and demographics (AAMC, 2023), as well as struggles with meeting changing demands (Gupta et al., 2017).

Complex factors require contemplation in deciding to implement routine cognitive assessment of older physicians. Few cognitive assessment tools include normative data for physicians. The ability to competently practice medicine is driven by complex cognitive aptitudes that may vary across specialties, and specialty-specific cognitive assessments linked to impaired clinical performance are lacking. Evidence to date is not sufficient to comment on differing performance issues across medical specialties tied to cognitive functioning with age. No longitudinal studies were identified that would correlate cognitive function of physicians with their clinical performance, nor were there any that determined a level of cognitive decline that would be sufficient to render a physician unfit for practice. Cultural and language factors can also affect performance on cognitive testing, which need to be considered as many practicing physicians are graduates from international medical schools.

Beyond scientific evidence, implementation of routine cognitive screening to improve patient safety involves practical elements (e.g. valid and reliable cognitive assessment instruments for physicians), ethical considerations and legal factors (e.g. privacy and human rights concerns). These considerations were beyond the scope of the review.

Overall, the results of this review provide some support both for and against routine cognitive testing of older physicians, which will need to be carefully weighed by regulators, such as provincial Colleges of Physicians and Surgeons in Canada, in deciding whether to implement a policy requiring age-based cognitive assessment of physicians. Further research and investigation into the actual cognitive testing of older physicians and correlation to physician performance and patient outcomes is warranted.

Acknowledgments

The authors would like to thank Natalie Jensen for help identifying relevant literature and for conducting environmental scans during the project.

Disclosure statement

The authors state that there are no competing interests to declare.

Additional information

Funding

This work was supported by funding from the College of Physicians and Surgeons of Alberta.

Appendix

Table A1. Medline search for literature on physician’s age as a determinant of cognitive functioning and clinical performance or competence.

#	Search terms
1	(Aging/ or Cognitive Aging/ or Aged/ or Age Factors/) and exp Physicians/
2	Ag?ing doctor*.mp.
3	Ag?ing physician*.mp.
4	Ag?ing clinician*.mp.
5	(Ag?ing adj2 practitioner*).mp.
6	Aged doctor*.mp.
7	Aged physician*.mp.
8	Aged clinician*.mp.
9	(Aged adj2 practitioner*).mp.
10	Older doctor*.mp.
11	Older physician*.mp.
12	Older clinician*.mp.
13	(Older adj2 practitioner*).mp.
14	Elderly doctor*.mp.
15	Elderly physician*.mp.
16	Elderly clinician*.mp.
17	(Elderly adj2 practitioner*).mp.
18	Senior doctor*.mp.
19	Senior physician*.mp.
20	Senior clinician*.mp.
21	(Senior adj2 practitioner*).mp.
22	(Pre-retirement adj3 doctor*).mp.
23	(Pre-retirement adj3 physician*).mp.
24	(Pre-retirement adj3 clinician*).mp.
25	(Pre-retirement adj3 practitioner*).mp.
26	(Preretirement adj3 doctor*).mp.
27	(Preretirement adj3 physician*).mp.
28	(Preretirement adj3 clinician*).mp.
29	(Preretirement adj3 practitioner*).mp.
30	(Late-career adj3 doctor*).mp.
31	(Late-career adj3 physician*).mp.
32	(Late-career adj3 clinician*).mp.
33	(Late-career adj3 practitioner*).mp.
34	1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27 or 28 or 29 or 30 or 31 or 32 or 33
35	(Mandatory adj4 assess*).mp.
36	(Mandatory adj4 test*).mp.
37	(Mandatory adj4 evaluat*).mp.
38	(Mandatory adj4 screen*).mp.
39	(Compulsory adj4 assess*).mp.
40	(Compulsory adj4 test*).mp.
41	(Compulsory adj4 evaluat*).mp.
42	(Compulsory adj4 screen*).mp.
43	(Obligatory adj4 assess*).mp.
44	(Obligatory adj4 test*).mp.
45	(Obligatory adj4 evaluat*).mp.
46	(Obligatory adj4 screen*).mp.
47	(Competenc* adj4 assess*).mp.
48	(Competenc* adj4 test*).mp.
49	(Competenc* adj4 screen*).mp.
50	(Performance* adj4 assess*).mp.
51	(Performance* adj4 test*).mp.
52	(Performance* adj4 evaluat*).mp.
53	(Performance* adj4 screen*).mp.
54	(Cognit* adj4 assess*).mp.
55	(Cognit* adj4 test*).mp.
56	(Cognit* adj4 evaluat*).mp.
57	(Cognit* adj4 screen*).mp.
58	Cognition Disorders/di [Diagnosis]
59	Cognitive Dysfunction/di [Diagnosis]
60	exp Professional Competence/st, sn [Standards, Statistics & Numerical Data]
61	exp Professional Impairment/
62	Malpractice/pc, st, sn [Prevention & Control, Standards, Statistics & Numerical Data]
63	Work Performance/
64	‘Quality of Health Care’/st, sn [Standards, Statistics & Numerical Data]
65	Outcome Assessment, Health Care/st, sn [Standards, Statistics & Numerical Data]
66	Medical Errors/sn [Statistics & Numerical Data]
67	35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43 or 44 or 45 or 46 or 47 or 48 or 49 or 50 or 51 or 52 or 53 or 54 or 55 or 56 or 57 or 58 or 59 or 60 or 61 or 62 or 63 or 64 or 65 or 66
68	34 and 67

Table A2. Medline search for literature on a possible cognitive threshold at which a physician is no longer fit to practice medicine.

#	Search terms
1	(Aging/ or Cognitive Aging/ or Aged/) and exp Physicians/
2	Ag?ing doctor*.mp.
3	Ag?ing physician*.mp.
4	Ag?ing clinician*.mp.
5	(Ag?ing adj2 practitioner*).mp.
6	Aged doctor*.mp.
7	Aged physician*.mp.
8	Aged clinician*.mp.
9	(Aged adj2 practitioner*).mp.
10	Older doctor*.mp.
11	Older physician*.mp.
12	Older clinician*.mp.
13	(Older adj2 practitioner*).mp.
14	Elderly doctor*.mp.
15	Elderly physician*.mp.
16	Elderly clinician*.mp.
17	(Elderly adj2 practitioner*).mp.
18	Senior doctor*.mp.
19	Senior physician*.mp.
20	Senior clinician*.mp.
21	(Senior adj2 practitioner*).mp.
22	(Pre-retirement adj3 doctor*).mp.
23	(Pre-retirement adj3 physician*).mp.
24	(Pre-retirement adj3 clinician*).mp.
25	(Pre-retirement adj3 practitioner*).mp.
26	(Preretirement adj3 doctor*).mp.
27	(Preretirement adj3 physician*).mp.
28	(Preretirement adj3 clinician*).mp.
29	(Preretirement adj3 practitioner*).mp.
30	(Late-career adj3 doctor*).mp.
31	(Late-career adj3 physician*).mp.
32	(Late-career adj3 clinician*).mp.
33	(Late-career adj3 practitioner*).mp.
34	1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27 or 28 or 29 or 30 or 31 or 32 or 33
35	Declin*.mp.
36	Disorder*.mp.
37	Impair*.mp.
38	Dysfunction*.mp.
39	Deteriorat*.mp.
40	Decrement*.mp.
41	Deficit*.mp.
42	Dement*.mp.
43	Loss*.mp.
44	Cognition Disorders/
45	Cognitive Dysfunction/
46	Memory Disorders/
47	Neurocognitive Disorders/
48	Dementia/
49	35 or 36 or 37 or 38 or 39 or 40 or 41 or 42 or 43 or 44 or 45 or 46 or 47 or 48
50	Threshold*.mp.
51	Cutoff.mp.
52	Cut-off.mp.
53	(Fit* adj4 practi#e).mp.
54	(Fit* adj4 dut*).mp.
55	(Fit* adj4 responsibilit*).mp.
56	(Fit* adj4 job).mp.
57	(Fit* adj4 work*).mp.
58	(Fit* adj4 perform*).mp.
59	(Unfit adj4 practi#e).mp.
60	(Unfit adj4 dut*).mp.
61	(Unfit adj4 responsibilit*).mp.
62	(Unfit adj4 job).mp.
63	(Unfit adj4 work*).mp.
64	(Unfit adj4 perform*).mp.
65	(Ability adj4 practi#e).mp.
66	(Ability adj4 dut*).mp.
67	(Ability adj4 responsibilit*).mp.
68	(Ability adj4 job).mp.
69	(Ability adj4 work*).mp.
70	(Ability adj4 perform*).mp.
71	(Inability adj4 practi#e).mp.
72	(Inability adj4 dut*).mp.
73	(Inability adj4 responsibilit*).mp.
74	(Inability adj4 job).mp.
75	(Inability adj4 work*).mp.
76	(Inability adj4 perform*).mp.
77	(Able adj4 practi#e).mp.
78	(Able adj4 dut*).mp.
79	(Able adj4 responsibilit*).mp.
80	(Able adj4 job).mp.
81	(Able adj4 work*).mp.
82	(Able adj4 perform*).mp.
83	(Unable adj4 practi#e).mp.
84	(Unable adj4 dut*).mp.
85	(Unable adj4 responsibilit*).mp.
86	(Unable adj4 job).mp.
87	(Unable adj4 work*).mp.
88	(Unable adj4 perform*).mp.
89	(Capable adj4 practi#e).mp.
90	(Capable adj4 dut*).mp.
91	(Capable adj4 responsibilit*).mp.
92	(Capable adj4 job).mp.
93	(Capable adj4 work*).mp.
94	(Capable adj4 perform*).mp.
95	(Incapable adj4 practi#e).mp.
96	(Incapable adj4 dut*).mp.
97	(Incapable adj4 responsibilit*).mp.
98	(Incapable adj4 job).mp.
99	(Incapable adj4 work*).mp.
100	(Incapable adj4 perform*).mp.
101	Competen*.mp.
102	exp Professional Competence/
103	(Professional* adj2 impair*).mp.
104	exp Professional Impairment/
105	Malpractice.mp.
106	Malpractice/
107	Dyscompeten*.mp.
108	(Work adj2 perform*).mp.
109	Work Performance/
110	(Health care adj2 quality).mp.
111	‘Quality of Health Care’/st, sn [Standards, Statistics & Numerical Data]
112	Outcome Assessment, Health Care/st, sn [Standards, Statistics & Numerical Data]
113	Medical Errors/lj, pc [Legislation & Jurisprudence, Prevention & Control]
114	50 or 51 or 52 or 53 or 54 or 55 or 56 or 57 or 58 or 59 or 60 or 61 or 62 or 63 or 64 or 65 or 66 or 67 or 68 or 69 or 70 or 71 or 72 or 73 or 74 or 75 or 76 or 77 or 78 or 79 or 80 or 81 or 82 or 83 or 84 or 85 or 86 or 87 or 88 or 89 or 90 or 91 or 92 or 93 or 94 or 95 or 96 or 97 or 98 or 99 or 100 or 101 or 102 or 103 or 104 or 105 or 106 or 107 or 108 or 109 or 110 or 111 or 112 or 113
115	34 and 49 and 114

Table A3. Summary of studies of the relationship between physicians’ age and their cognitive abilities.

Reference	Study location (Country)	Participants (Number, age, gender, specialty)	Assessments (Functions tested and tests used)	Results
Bieliauskas et al. (2008)	USA	N = 359 Age: 45 to 86 (mean 61.4) years 330 men and 29 women Surgeons	CANTAB Visual sustained attention and stress tolerance in the rapid visual information processing (RVIP) task Reaction time (RTI task) Visual learning and memory in the paired associates learning (PAL) task	RVIP task Mean response latency increased with increasing age (p = 0.002). Age groups 70-74 and 75+ years were significantly (p < 0.05) slower than all younger age groups. The probability of a hit decreased with increasing age (p = 0.01) and was lower in age groups 65 to 69, 70 to 74, and 75+ years compared to the age group 45-49 years (p < 0.05). Age had no significant effect on the probability of a false alarm (p = 0.64). RTI task Five-choice reaction time increased with increasing age (p = 0.0001). Increases in age groups 60-64, 65-69, 70-74, 75+ years vs. 45-49 years (p < 0.05); 65-69, 70-74, and 75+ years vs. 50-54 years (p < 0.05); 70-74 and 75+ years vs. 60-64 years (p < 0.05); 75+ vs. 65-69 and vs. 70-74 years (p < 0.05). Five-choice movement time increased with age (p = 0.0001). Increases in age groups 60-64, 65-69, 70-74, 75+ years vs. 45-49 years (p < 0.05); 75+ tears vs. 50-54 years (p < 0.05); 60-64, 70-74, and 75+ years vs. 55-59 years (p < 0.05); 75+ years vs. 60-64 years (p < 0.05). PAL task Number of stages where the correct responses were made on the first trial decreased with increasing age (p = 0.0001). Decreases in 75+ age group vs. 45-49, 50-54, 55-59, and 60-64 years (p < 0.05). Mean number of errors in learning all stimulus locations increased with age (p = 0.0001). Increases in 60-64, 65-69, 70-74, and 75+ years vs. 45-49 and 50-54 years (p < 0.05); 65-69, 70-74 and 75+ years vs. 55-59 years (p < 0.05); 70-74 and 75+ years vs. 60-64 and 65-69 years (p < 0.05). Retest one to six years after the initial testing Increase in mean latency of response on the RVIP task (p = 0.001) with no significant changes in the accuracy of responding (the probability of a hit (p = 0.12) and the probability of a false alarm (p = 0.35). Decrease in mean number of errors per trial on the PAL task (p < 0.05). No significant changes on the other tasks
Boom-Saad et al. (2008)	USA	Surgeons N = 308 [subset of data from Bieliauskas et al. (2008)] Age: 45-75 years Women: 1.8% (61-75-year-old group); 18% (45--60-year-old group) Medical students (4^th year) entering a surgical field N = 21 Women: 28% Age: 20-35 years	See Bieliauskas et al. (2008)	RVIP task Stimulus response time (s): 401 ± 16, 477 ± 7, 508 ± 7 for students, 45-60-year-old and 61-75-year-old surgeons, respectively (p < 0.001) Accuracy (% correct hits): 89 ± 16, 74 ± 16, 66 ± 19 for students, 45-60-year-old and 61-75-year-old surgeons, respectively (p < 0.001) RTI task Efficiency of motion (s): 320 ± 9, 384 ± 7, 426 ± 8 for students, 45-60-year-old and 61-75-year-old surgeons, respectively (p < 0.001) Response time (s): 313 ± 7, 340 ± 3, 358 ± 4 for students, 45-60-year-old and 61-75-year-old surgeons, respectively (p < 0.001) PAL task Stages completed first trial: 4.0 ± 0.15, 2.9 ± 0.7, 2.7 ± 0.07 for students, 45-60-year-old and 61-75-year-old surgeons, respectively (p < 0.001) Total adjusted errors: 3.1 ± 0.65, 16.5 ± 0.9, 30.0 ± 1.9 (p < 0.001) Comparison with age-matched normative data [Normative data were available for RVIP and RTI] RTI, movement time: significantly shorter in all study groups than in age-matched controls (p < 0.05). RTI, total reaction time: in student and 45-60-year-old surgeon group significantly shorter than in respective controls (p < 0.05); in 61-75-year-old surgeons shorter but not significantly than in controls (p = 0.08). RVIP task (accuracy -% correct hits): significantly greater in students and 45-60-year-old surgeons than in respective controls; not different from control in 61-75-year-old surgeons.
Drag et al. (2010)	USA	[Subset of data from Bieliauskas et al. (2008)] Senior surgeons N = 168 Age: 60+ years 166 men, 2 women 36% retired, 33% planning to retire within 5 years, 30% had no imminent retirement plans Younger surgeons N = 126 Age: 45-59 years 105 men, 21 women	See Bieliauskas et al. (2008)	Z-scores of senior surgeons - the number of standard deviations (SD) that a raw score of senior surgeons deviated from the mean score of younger surgeons [Z-score of below −1.5 (i.e. more than 1.5 SD below the younger group score) was considered significantly lower than the score of the younger surgeon group] PAL: −1.50, −0.80, −1.95, −1.43 for surgeons with no imminent retirement plans, those planning to retire within 5 years, retired surgeons, and all senior surgeons, respectively. RVP: −0.41, −0.26, −0.75, −0.48 for surgeons with no imminent retirement plans, those planning to retire within 5 years, retired surgeons, and all senior surgeons, respectively. RTI: −0.62, −0.63, −0.86, −0.71 for surgeons with no imminent retirement plans, those planning to retire within 5 years, retired surgeons, and all senior surgeons, respectively. Effect of age on cognitive performance among the 108 practicing senior surgeons Age was negatively correlated with z-scores on the PAL (r=-0.29, p = 0.002), the RVP (r=-0.25, p = 0.01), and the RTI (r=-0.19, p = 0.05). Percentage of senior surgeons who perform within the range of the younger surgeons’ performance 55% of all senior surgeons, 49% of senior surgeons with no imminent plans to retire, 72% of senior surgeons planning to retire within the next 5 years, and 45% of retired surgeons performed within the range of the younger surgeons on all three cognitive tasks. No senior surgeon performed below the younger surgeons on all three tasks.
Powell and Whitla (1994)	USA	Physicians [referred to as MDs] N = 1002 (740 practicing, 222 retired) Age: 25-92 years 930 men, 72 women Specialty not reported Non-physicians [referred to as ‘Normal’ subjects] N = 581 Age: 25-83 years 173 men, 408 women	MicroCog Reaction time Visuospatial abilities (Tictac subtest) Reasoning scores (Analogies subtest) Verbal memory - delayed recall (Stories subtest) Calculation skills (Math) Attention (Wordlist 1, 2, Alphabet, Numbers Forward)	Analysis of raw scores MDs Gradual decline between 30 and 60 years of age (a 60-year-old MD scored only 8% lower than a 30-year-old doctor). More rapid decline after age 60 years; the differences with the youngest group almost doubled with each decade (14% at age 70 years and 26% at 75+ years) Decline was steepest in visuospatial memory (correct responses on Tictac dropped by 24% between ages 30 and 50 years, by 10% between ages 50 and 60 years, and by 39% between 65-74 and 75+ years. Decline in reasoning scores (Analogies subtest) was second steepest (occurred about a decade later than the decline in visuospatial scores). Decline in verbal memory (Stories subtest) was in the third place. Calculation skills (Math scores) and attention (Wordlist) were more resistant to aging. ‘Normal’ subjects Gradual decline between 30 and 50 years of age More rapid decline after age 50 years In the two decades after age 60, scores of ‘Normal’ subjects were approximately in the same range as those of MDs ten years older. The gap between ‘Normal’ subjects and MDs narrowed after age 75; however, at all ages, ‘Normal’ subjects scored lower than the MDs. For specific cognitive functions, patterns were similar to those of MDs [Although the decline in visuospatial abilities in ‘Normal’ subjects was also steepest, it was slower than in MDs, likely due to lower initial scores.] Effect size (ES) correlation analysis for MDs 10-year age groupings Visuospatial faculty, reasoning, and verbal memory were more affected by aging than other functions. Flattening in the magnitude of decline at ages 50-60 compared with the decades on either side of this age span (suggesting slowing of decline in late mid-life). 5-year age groupings (more in-depth investigation of the timing of decline) The downward sloping of global cognitive functioning and specific abilities relatively mild until 70-74 years, with a remarkable increase at 75-79 years. Reaction time No significant findings for reaction times with either grouping (10- or 5-year intervals) [The authors noted that this finding was not consistent with most of the literature and explained the discrepancy by either relatively less demanding task used, or by older participants in their study being not representative of their age group. A more likely explanation, in the authors’ view, was that older subjects performed as well as younger ones on a simple reaction time test, but they responded much slower if the task requires both accuracy and a quick response.]

Table A4. Summary of studies of the relationship between physicians age (or years in practice as a surrogate for age) and their competence/performance.

Reference	Study location	Physicians’ characteristics	Type of outcome	Assessments	Covariates	Association with age
Authors (year)	Country	Specialty, age/years in practice	Competence or Performance	Measure of competence/performance	Potential confounders adjusted for	Negative (decline with age) Positive (superior performance of older physicians) No association
B. R. Anderson et al. (2017)	USA	Surgeons Years since graduating from medical school: 9 to 55 (median 25)	Performance	Patients’ major morbidity, mortality, and a composite outcome (major morbidity or mortality) after congenital heart surgery	Patient’s characteristics (sex, prematurity, weight (for neonates and infants), year of surgery, noncardiac abnormalities or syndromes/chromosomal abnormalities, previous cardiothoracic operation, and preoperative mechanical ventilation, circulatory support, persistent shock, renal failure, any other preoperative risk factor) STAT Mortality Category for the primary operative procedure. Surgeon and center volume	Negative association with patients’ major morbidity and composite outcome. The odds of poorer patient’s outcomes increased for physicians 35+ years in practice (≈60-80-year-old) vs. <15 years in practice (<40-year-old) No association with patients’ mortality
G. M. Anderson et al. (1997)	Canada	General practitioners; family practitioners; specialists Age from <45 to 65+ years	Performance	Inappropriate prescribing	Physician’s specialty (specialist, family practitioner, general practitioner) Volume of prescriptions	Negative association. Physicians aged 45-64 years and 65+ years were more likely to prescribe potentially inappropriate drugs compared to physicians aged <45 years.
Brown et al. (2016)	USA	Primary care physicians Years since completing residency: 1-46 (mean 20)	Performance	Diabetes glucose control	Patient’s characteristics (demographics, BMI, blood pressure, biomarkers, diabetes severity) Physician’s characteristics (internal medicine as primary specialty; number of patients with diabetes, proportion of patients with diabetes taking insulin)	No association
Gershon et al. (2014)	Canada	Primary care physicians Age: from <35 to >65 years	Performance	Pulmonary function testing (PFT) for diagnosis of COPD	Patient’s characteristics: socio-demographic; rural/urban residence, long-term care, immigration status, year of COPD diagnosis, previous hospitalizations/ED visits; specialist care (yes/no); primary care physician visits in previous year; flu short in previous year; coexisting medical conditions. Physician’s characteristics: gender, foreign medical graduate (yes/no)	Negative association. Physicians in all older age groups (36-50, 51-65 and >65 years) were less likely to order PFT compared to <35 year-old physicians.
Goulet et al. (2013)	Canada	Family practitioners Age: from <40 to ≥70 years.	Performance	A composite score for three quality components: clinical investigation, diagnostic accuracy, appropriateness of treatment and follow-up	Physician’s characteristics (participation in continuing professional development, sex, country of graduation, primary place of practice, completion of a residency in family medicine)	Negative association. In a stepwise multiple linear regression analysis, physicians’ age was a significant predictor of inferior quality of practice: β=-0.15; p = 0.03
Grace et al. (2014)	USA	Physicians of various specialties Age: from 32 to 84 (mean 53.1) years.	Competence	Competence assessment tailored to physician’s practice area and specialty	Physician’s characteristics: gender, years in practice, medical school, degree type, board certification, years of postgraduate training, training specialty Practice characteristics: practice-to-training match, solo practice, rural practice, practice status Referral source (board, hospital, self, other), previous board action affecting license status, Drug Enforcement Administration revocation	Negative association. Increase in the odds of inferior assessment outcomes for each year of increasing age OR = 1.07 (95% CI: 1.03-1.10); p < 0.01
Marco et al. (2018)	USA	Emergency medicine residency–trained physicians Age: 35 to 72 (mean 45 years)	Competence	Score on the American Board of Emergency Medicine (ABEM) ConCert examination	Physician’s characteristics (number of years since residency graduation, diplomate status at time of examination, gender, ethnicity)	Negative association: gradual decline over the age range
Markar et al. (2018)	Sweden	Surgeons Age: from <30 to ≈70 years [based on figures]	Performance	Patients’ 90-day all cause mortality; 5-year all cause mortality; 5-year disease-specific (esophageal cancer) mortality after esophageal cancer resection	Patient’s characteristics: demographics, comorbidity, tumor stage and histology, use of neoadjuvant therapy Surgeon volume of esophagectomies, calendar period of surgery	Negative association. Deterioration in patients’ mortality rates in the highest surgeons’ age category (> =56 years)
Mehrotra et al. (2018)	USA	Endoscopists Years in practice since residency: from ≤9 to 51	Performance	Adenoma detection rate (ADR)	Patients’ characteristics: age, sex, and colonoscopy indication Physician’s characteristics: gender, primary specialty (gastroenterology or other); volume of colonoscopies performed	Negative association. ADR lower in the highest category of years in practice (27-51 years) relative to the lowest category (≤9 years).
Meier et al. (2020)	Switzerland	General practitioners Age: mean 52 (SD 9.3) years	Performance	Quality of diabetes care (process indicators, outcome indicators and cumulative indicators)	Patient’s characteristics (age, sex, number of comorbidities, number of consultations, number of consulted GPs) Physician’s characteristics (gender, physician’s network participation) Practice characteristics (location, type)	No association
Meltzer et al. (2017)	USA	Surgeons (vascular) Years since medical school graduation: 3 to >30 [based on Figure 2]	Performance	Outcomes of abdominal aortic aneurysm repair: 30-day mortality and 1-year mortality.	Patients’ characteristics: age, gender, race, insurance status, comorbidities Surgeon’s annual volume and facility’s annual volume	No association
O’Neill et al. (2000)	USA	Surgeons Age: 30 to ≥65 years	Performance	In-hospital mortality; in-hospital overall bad outcome (mortality or morbidity) following carotid endarterectomy	Patient’s characteristics (age, gender, source of admission, carotid endarterectomy procedure, payer, Medis-Groups severity score Surgeon’s characteristics (gender; specialty, years of experience, board certified, surgical volume)	Negative association with patients’ mortality (p = 0.04). Mortality rate 0.61%, 0.59%, 0.68%, 1.22% and 2.53% among patients of surgeons aged 30-39, 40-49, 50-59, 60-64 and 65+ years, respectively. No significant association with the rate of overall bad outcome: 4.46%, 3.53%, 3.58%, 4.40% and 5.49% among patients of surgeons aged 30-39, 40-49, 50-59, 60-64 and 65+ years, respectively.
Pentzek et al. (2009)	Germany	General practitioners Age: mean 51.13 (SD 8.05) years	Competence	Dementia knowledge assessed using a questionnaire with 37 multiple-choice questions.	Physician’s characteristics (gender, years of practice, location of surgery -rural vs. urban, estimated prevalence of dementia, personal views on dementia). An additional covariate (care facility knowledge score) was used in the analysis of the teaching GP sub-sample.	Negative association with the knowledge test score. Stepwise linear regression analysis: β=-0.255; p = 0.003. Analysis in the sub-sample of teaching physicians: β=-0.246, p = 0.005
Poulin de Courval et al. (1996)	Canada	Family physicians Age: from 25 to >65 years	Competence	Knowledge and skills with respect to dementia assessed using a questionnaire with 47 multiple-choice questions	Physician’s characteristics (gender, percentage of time spent with patients aged 65 or over, years in practice, certification from the College of Family Physicians of Canada – yes or no) Practice characteristics (solo office, group or local community service center, short-term hospital, long-term hospital, method of remuneration, practice settings - urban, semi-urban, rural)	Negative association. Gradual decline over age intervals 32-44, 45-54, 55-64 and > =65 years relative to 25-34 years
Prystowsky (2005)	USA	Surgeons Years since certification by the American Board of Medical Specialties: from <5 to >30	Performance	Patients’ mortality, morbidity, and hospital length of stay (LOS) after alimentary tract surgery	Patient’s characteristics (age, admission type - emergency or nonemergency, severity of illness) Surgeon volume	Negative association with hospital length of stay^1 No association with patient’s mortality and morbidity
Reid et al. (2010)	USA	Physicians of various specialties Years in practice: from <10 to ≥50	Performance	Overall performance scores on 124 indicators of quality of care for 22 acute and chronic conditions, as well as preventive care, from RAND’s Quality Assessment Tools.	Physician’s characteristics (gender, board certification status, medical school location -domestic or international, medical school ranking, malpractice claims and disciplinary actions by the board in the past 10 years)	Negative association with acute care and male-specific care. Gradual decline over 10-19, 20-29, 30-39, 40-49 and 50+ years in practice relative to <10 years in practice. No association with chronic care, female-specific care, and all measures Positive association with preventive care. Gradually increasing score over the years in practice categories
Satkunasivam et al. (2020)	Canada	Surgeons Age: 29–80 years	Performance	30-day postoperative outcomes of 25 common procedures performed by surgeons of various subspecialties: deaths; complications; hospital readmission; a composite outcome including the presence of any of death, complications, or hospital readmission in the 30 days after surgery; hospital length of stay	Patient’s characteristics (age, gender, income, rurality, comorbidity - Johns Hopkins Aggregated Diagnosis Group, year of surgery) Surgeon’s characteristics (gender, surgical volume, specialty) Practice setting (academic or community) Stratified analyses by several variables were conducted, including type of surgery (emergent or elective) and case complexity	Positive association with most outcomes. Each 10-year increase in surgeon’s age was associated with decreased odds of the composite outcome (OR = 0.95; 95% CI: 0.92–0.98), death (OR = 0.95; 95% CI: 0.91–0.998), hospital readmission (OR = 0.98; 95% CI: 0.97–0.99), and complications (OR = 0.94; 95% CI: 0.89–0.98) Dichotomized surgeons’ age: patients treated by surgeons aged ≥ 65 years were less likely to have the composite outcome (OR = 0.93; 95% CI: 0.88–0.97) Surgeon’s age ≥65 years was associated with lower odds of the composite outcomes after emergent (OR = 0.97; 95% CI: 0.94–0.99) and elective procedures (OR = 0.96; 95% CI: 0.93–0.99); readmission within 30 d after elective surgery (OR = 0.98 (0.97–0.99); complications within 30 days after emergent (OR = 0.94; 95% CI: 0.91–0.98) and elective surgery (OR = 0.94; 95% CI: 0.89–0.999). No significant differences between older (≥65 years) and younger (<65 years) surgeons in patients’ mortality after emergent or elective surgery, readmission after emergent surgery (ORs 0.96 to 1.00) No significant association with hospital length of stay in the continuous analysis (OR = 1.01; 95% CI: 0.99-1.02 per 10-year increase in surgeon’s age) and a negative association in dichotomized analysis after emergent surgery (OR = 1.01; 95% CI: 1.01-1.02 for older surgeons vs. younger surgeons).
Smith-Bindman et al. (2005)	USA	Physicians who interpret mammograms (radiologists?) Age: from <40 to ≥70 years	Performance	Accuracy of screening mammogram interpretation: sensitivity, specificity, overall accuracy	Patient’s characteristics (age, breast density, whether examination was a first or a subsequent) Physician’s characteristics (annual volume of mammogram interpretations, ratio of screening to diagnostic mammographic interpretations)	Positive association with specificity and overall accuracy. Specificity [the OR (95% CI) for a correct mammogram interpretation by years since receipt of medical degree]: 1.0 (reference); 1.16 (0.88 to 1.54); 1.22 (0.92 to 1.64); 1.18 (0.88 to 1.59); 1.54 (1.14 to 2.08); 1.67 (1.25 to 2.22); 1.59 (1.12 to 2.22) for <10; 10-14; 15-19; 20-24; 25-29; 30-34 and >34 years, respectively. Overall accuracy [the OR (95% CI) for a correct mammogram interpretation by years since receipt of medical degree]: 1.0 (reference); 1.14 (0.73 to 1.78); 1.31 (0.92 to 1.87); 1.14 (0.78 to 1.64); 1.54 (1.05 to 2.26); 1.44 (0.99 to 2.12); 1.20 (0.82 to 1.73) for <10; 10-14; 15-19; 20-24; 25-29; 30-34 and >34 years, respectively No association with sensitivity [the OR (95% CI) for a correct mammogram interpretation by years since receipt of medical degree]: 1.0 (reference); 0.98 (0.68 to 1.43); 1.07 (0.79 to 1.46); 0.96 (0.70 to 1.33); 1.00 (0.72 to 1.40); 0.86 (0.63 to 1.19); 0.76 (0.55 to 1.04) for <10; 10-14; 15-19; 20-24; 25-29; 30-34 and >34 years, respectively
Tsugawa et al. (2018)	USA	Surgeons Age: from <40 to ≥60 (mean 50.5, median 50.0) years	Performance	Operative mortality, defined as death during hospital admission or within 30 days of non-elective surgeries (emergency or urgent admissions and admissions from a trauma center)	Patient’s characteristics (age, gender, race/ethnicity, median household income estimated from residential zip code, Medicaid status, comorbidities) Surgeon’s characteristics (gender, credentials – allopathic (MD) vs. osteopathic (DO), number of non-elective surgeries) Adjustment for hospital where patient was treated by using hospital fixed effects	Positive association Adjusted OR (95% CI) for patients’ mortality vs. surgeons’ age group <40 years (reference category): 0.98 (0.96-1.01), p = 0.25; 0.97 (0.94-0.99), p = 0.02; 0.95 (0.92-0.99), p = 0.02) among patients of surgeons aged 40-49, 50-59 and 60+ years, respectively.
Tsugawa et al. (2017)	USA	Hospitalist physicians Age: from <40 to ≥60 years	Performance	Patients’ mortality within 30 days of admission; 30-day readmission rates; costs of care	Patient’s characteristics: age, gender, race/ethnicity, primary diagnosis, comorbidities, median household income estimated from residential zip code, Medicare/Medicaid coverage, day of the week of admission, and year indicators. Surgeon’s characteristics: gender, medical school (foreign vs. domestic); credentials (allopathic -MD vs. osteopathic -DO) Adjustment for hospital where patient was treated by using hospital fixed effects	Negative association with 30-day mortality. Increased mortality in patients of physicians aged 40-49, 50-59 and 60+ years relative <40 years of age; the increase was greater at ages 60+ years. No association with 30-day readmission and costs of care
Waljee et al. (2006)	USA	Surgeons Age: from <40 to >61 years	Performance	Operative mortality, defined as death before discharge from hospital or within 30 days of the surgery	Patient’s characteristics: age, gender, race, admission acuity - elective or nonelective, pre-existing conditions Surgeon’s characteristics: procedure volume, Hospital volume and teaching status	Negative association with mortality after 3 of 8 surgical procedures analyzed. ORs (95% CI) for patients of 60+ year old surgeons relative to patients of 41-50-year-old surgeons: pancreatectomy (1.67; 1.12–2.49), coronary artery bypass grafting (1.17; 1.05–1.29), carotid endarterectomy (1.21; 1.04 − 1.40). The effect was largely restricted to surgeons with low procedure volumes. No association for 5 of 8 surgical procedures analyzed (elective repair of an abdominal aortic aneurysm; aortic valve replacement; lung resection; esophagectomy; cystectomy)
Wenghofer et al. (2009)	Canada	General/family practitioners Age: mean 51.0 (SD 9.91) years	Performance	Acute care, chronic care, continuity of care, well care, managing patient records	Physician’s factors (age; gender, years in practice, medical school - North American vs. other, CFPC certification, years in current practice, previous peer assessment by the CPSO) Organizational factors: (solo practice, episodic care practice/walk-in clinic, number of clinical and administrative staff, hours worked per week in primary practice, number of patient visits per week in primary practice, active hospital appointment - yes/no, teaching -yes/no, focused practice -yes/no) Systemic factors: access to 911 services - yes/no, estimated minutes for access to emergency medical services, availability of four core diagnostic tests, physician per 1,000 population, northern practice locations -yes/no)	No association
Weycker and Jensen (2000)	USA	All non-osteopathic physicians Age: from <40 to 60+ years	Performance	Medical malpractice	Physician’s characteristics (gender, born US citizen, training credentials, previous litigation history) Practice characteristics (medical research/teaching, direct patient care, solo practice, practice in Wayne County, general practice, medical specialty, surgical specialty)	Positive association: older physicians less likely to be sued for malpractice. ORs from 0.342 to 0.679 (in different models) for 60+ year-old physicians compared to <40-year-old physicians. No significant difference between physicians aged 40–59 and physicians <40 years.

¹When it comes to LOS, the newer surgeons tend to get their patients out faster. … that might be a bias of older more experienced surgeons who might hang onto their patients a little bit longer …” (Prystowsky, 2005) (p. 502)