Medical Doctors and Dementia: A Longitudinal Study

Abstract

OBJECTIVE:

To examine the association between being a medical doctor (MD) and the risk of incident dementia.

DESIGN:

Cohort study.

SETTING:

Olmsted County, Minnesota.

PARTICIPANTS:

A total of 3460 participants (including 104 MDs), aged 70 years or older, of the population-based Mayo Clinic Study of Aging.

MEASUREMENTS:

Participants were randomly selected from the community and had comprehensive cognitive evaluations at baseline and approximately every 15 months to assess for diagnosis of dementia. For participants who withdrew from the follow-up, dementia diagnosis was also assessed using information available in their medical record. The associations were examined using Cox proportional hazards models, adjusting for sex, education, and apolipoprotein E ε4, using age as the time scale.

RESULTS:

MDs were older (vs “general population”), and most were males (93.3%). MDs without dementia at baseline did not have a significantly different risk for incident dementia (hazard ratio = 1.12; 95% confidence interval = 0.69–1.82; P = .64) compared to the general population.

CONCLUSIONS:

Although the study includes a small number of older, mainly male, MDs, it provides a preliminary insight on cognitive health later in life in MDs, while most previous studies examine the health of younger MDs. Larger longitudinal studies are needed to examine these associations and investigate if associations are modified by sex.

Previous studies suggest that medical doctors (MDs) have a lower mortality rate and healthier lifestyles than the general population, although there is room for improvement.^1–4 Illnesses experienced by MDs are mostly those experienced in the population at large,⁵ but they have a higher occurrence of suicide, burnout, and mental disorders than the general population or other groups.⁶

MDs’ health has an impact on their families, the healthcare system, and the patients.⁷ They face similar barriers as other patients in accessing medical care (eg, fear, embarrass-ment, loss of control, and financial and time costs), but they also have special barriers like knowing the healthcare personnel, fear of loss of confidentiality (especially if experiencing mental health conditions)⁸ and privacy, inability to reverse roles, long working hours that limit access to healthcare for themselves, and feeling it is not culturally appropriate to acknowledge illness to a peer.^9–11 MDs work in environments with high workload, fatigue, and stressors specific for their profession.¹² At the same time, they need to work under excessive cognitive demands as they process quickly large amounts of information for long periods of time,¹² while physician burnout is prevalent in the United States, both in training and in practice.¹³ Medical associations recognize the demands and potential health risks related to this profession and created programs tailored to the needs of MDs.⁴

The study’s overarching hypothesis, based on the available literature, was that MDs, as a professional group, have a mixture of potentially beneficial and detrimental characteristics for their health and specifically for their cognitive health later in life (eg, higher education [potential protective factor] and depression and anxiety [potential risk factors]).^14,15 The objective of the present study was to examine the association between being an MD and the risk for incident dementia in participants, aged 70 years or older, of the population-based Mayo Clinic Study of Aging (MCSA) in Olmsted County, Minnesota (USA).

METHODS

Study Population

The MCSA study design, protocol, and recruitment have been presented previously in detail.^16,17 The Rochester Epidemiology Project (REP)¹⁸ resources were used to enumerate Olmsted County residents, aged 70 to 89 years (initially on October 1, 2004). An age- and sex-stratified random sample of Olmsted County residents was invited to participate in the MCSA.¹⁶ In 2012, recruitment of participants aged 50 to 69 years was initiated. Ongoing recruitment has maintained the study sample. The study was approved by the Institutional Review Boards of the Mayo Clinic and of Olmsted Medical Center. All participants provided written informed consent before participation in the study. There were 3460 MCSA participants (≥70 years old at baseline) without dementia (Figure 1), with available education and occupation information and at least one prospective follow-up (ie, in-person evaluation) or follow-up information available in their medical record. By considering the longest occupation, most recent occupation, name suffix (ie, MD or equivalent), and years of education, we determined there were 104 MDs among the 3460 participants.

Figure 1.

Study flowchart. There were 3460 Mayo Clinic Study of Aging (MCSA) participants (aged ≥70 years at baseline) without dementia, with available education and occupation information and at least one in-person MCSA evaluation or follow-up information available in their medical records. All participants who were evaluated at least in the baseline visit are eligible for medical record review (MRR) for incident dementia at 70 years of age and every 5 years thereafter, if they provided MRR authorization. A total of 228 participants were excluded (3 did not have education data; 147 had only the baseline visit and await their in-person follow-up evaluation; 13 died before they could come in for follow-up; 1 withdrew from the study and withdrew consent for MRR; 43 withdrew from the in-person MCSA evaluation and await MMR; and 21 withdrew from the in-person MCSA evaluation but we could not have MRR for unspecified reasons).

Clinical Evaluation and Diagnostic Assessment

In MCSA, participants undergo a face-to-face evaluation by a nurse or a study coordinator, a physician, and neuropsychological testing by a psychometrist, at baseline and approximately every 15 months thereafter. The nurse or study coordinator collects demographic information, asks questions about memory, and administers the Clinical Dementia Rating scale¹⁹ and the Functional Activities Questionnaire²⁰ to an informant. The physician’s evaluation includes a review of medical history, administration of the Short Test of Mental Status,²¹ and a neurological examination. Nine neuropsychological tests, administered by a psychometrist, are used to assess four cognitive domains (ie, memory, language, attention/executive, and visuospatial skills). At a weekly held conference, the nurse or study coordinator, the physician, and a neuropsychologist discuss all the information for each participant and make the final diagnosis (ie, cognitively unimpaired [CU], mild cognitive impairment [MCI], or dementia) by consensus.^16,17 Individuals who perform in the normative range and do not meet criteria for MCI²² or dementia²³ are classified as CU.

Medical Record Review

We were concerned that MDs might have a different pattern of dropping out of the in-person follow-ups than the “general population”; thus, we also included data on incident dementia diagnosis derived from the review of the medical records of participants who withdrew from the MCSA in-person follow-up. Most MCSA participants who withdraw from the in-person evaluation do not withdraw their permission for MCSA to review their medical records (Figure 1).

Details of the medical record review process have been described previously.²⁴ In summary, for participants aged 70 years or older who withdrew from the MCSA follow-up visits, experienced nurse abstractors review their medical records (every 5 years) to assess an incident dementia diagnosis. Medical record review includes cognitive changes and difficulties, neuropsychological tests, physician diagnosis of dementia, and dementia diagnosis by a neurologist. A dementia diagnosis is based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, criteria.²³ Dementia onset was considered the first date on which symptoms compatible with a dementia diagnosis were recorded in the medical record. Ambiguous or conflicting information was reviewed in consultation with a study neurologist (D.S.K.).²⁴ It is expected that most dementia cases ascertained in person and not by medical record review would be mild.²⁵

Of 3460 participants, 460 (13.3%) withdrew after the baseline in-person MCSA evaluation (Figure 1). For these participants, we obtained incident dementia information by reviewing their medical records. For 1523 (44%) participants, we had data from the baseline evaluation and in-person follow-up visit(s); however, at some point, they withdrew from the in-person evaluations and we supplemented information for incident dementia from the medical record review. For 1477 (42.7%) participants, we had both baseline and follow-up in-person evaluation visits in MCSA without the need to do an additional medical record review at the time this analysis was conducted. The median follow-up time added for those with medical record review, beyond what is captured in the MCSA in-person follow-up visits, was 2.47 (95% confidence interval [CI] = 2.30–2.65) years.

Covariates

Covariates were selected a priori due to their association with cognitive impairment and dementia based on previous background knowledge.^14,15 The Beck Depression Inventory (BDI-II)²⁶ and the Beck Anxiety Inventory (BAI)²⁷ were used to assess symptoms of depression and anxiety at base-line. Apolipoprotein E (APOE) ε4 genotype was assessed from a blood draw at baseline assessment. Using the REP medical records–linkage system, comorbidities were abstracted from participants’ medical records. The chronic disease burden was assessed from a weighted Charlson Comorbidity Index (CCI) score²⁸ based on electronic diagnosis codes (Hospital International Classification of Diseases Adapted, International Classification of Diseases, Ninth Revision, and International Classification of Diseases, Tenth Revision) using the REP resources.

Statistical Analysis

Participant characteristics were compared using Kruskal-Wallis tests or χ² tests as appropriate. In all analyses, we treat “being an MD” (vs general population; ie, including all participants but MDs) as an independent variable. The hazard ratio (HR) of MDs vs “general population” for incident dementia (as assessed by an in-person evaluation or medical record review) was obtained from Cox proportional hazards models, adjusted for sex, years of education, and APOE ε4 carrier status, with age as the time scale (model 1). Separate models adjusting additionally for BDI-II, BAI (model 2), and CCI (model 3) were built. We utilized a standard two-sided .05 alpha level to determine statistical significance. Analyses were performed using the SAS statistical software version 9.4 (SAS Institute) and R version 3.4.2 (R Foundation for Statistical Computing).

RESULTS

Characteristics of Participants

There were 104 MDs among 3460 MCSA participants without dementia at baseline (Table 1) with available follow-up, by either an in-person evaluation or medical record review. MDs were older, most were males (93.3%), had more years of education, and a higher CCI (Table 1). Most of the study population was white (n = 3414 [98.7%]) and not-Hispanic or Latino (n = 3440 [99.4%]). The overall median (95% CI) follow-up time was 7.84 (7.68–7.93) years.

Table 1.

Characteristics of Study Participants (Aged ≥70 Years) Without Dementia at Baseline

Characteristics	General Population (N = 3356)a	MDs (N = 104)	Total (N = 3460)	P Valueb
Age, mean (SD), y	78.57 (5.40)	79.95 (5.36)	78.61 (5.40)	.010
Age groups (per decade), y				.015
70–79	1983 (59.1)	49 (47.1)	2032 (58.7)
≥80	1373 (40.9)	55 (52.9)	1428 (41.3)
Sex, male	1673 (49.9)	97 (93.3)	1770 (51.2)	<.001
APOE ε4 positivec	862 (26.3)	29 (28.2)	891 (26.3)	.668
Education, mean (SD), y	13.760 (2.75)	19.99 (0.10)	13.95 (2.91)	<.001
Cognitive status (baseline)				.998
Cognitively unimpaired	2840 (84.6)	88 (84.6)	2928 (84.6)
Mild cognitive impairment	516 (15.4)	16 (15.4)	532 (15.4)
BDI-II score (baseline), mean (SD)d	5.10 (4.81)	4.09 (4.33)	5.07 (4.801)	.020
BDI-II, total ≥13	263 (8.0)	6 (6.1)	269 (8.0)	.476
BAI total (baseline), mean (SD)e	3.01 (4.14)	1.93 (2.76)	2.98 (4.11)	.015
Charlson Comorbidity Index, mean (SD)	3.70 (3.246)	4.07 (2.92)	3.71 (3.24)	.052
History of depressionf	1140 (34.0)	26 (25.0)	1166 (33.7)	.056
History of alcohol use problemsg	131 (3.9)	4 (3.9)	135 (3.9)	.975
History of coronary artery diseasef	1336 (39.8)	54 (51.9)	1390 (40.2)	.013
History of strokef	196 (5.8)	6 (5.8)	202 (5.8)	.975
History of congestive heart failureh	382 (11.4)	10 (9.6)	392 (11.3)	.574
History of atrial fibrillationf	554 (16.5)	31 (29.8)	585 (16.9)	<.001
History of hypertensionf	2629 (78.4)	70 (67.3)	2699 (78.0)	.007

Note. Data are given as number (percentage) unless otherwise stated.

Abbreviations: APOE, apolipoprotein E; BAI, Beck Anxiety Inventory; BDI-II, Beck Depression Inventory; MD, medical doctor.

^aAll participants, except MDs.

^bKruskal-Wallis or χ² test.

^cA total of 75 missing.

^dA total of 86 missing.

^eA total of 11 missing.

^fA total of one missing.

^gA total of 35 missing.

^hA total of two missing.

We examined whether differences in characteristics were partly due to differences in the sex and age distributions between the two comparison groups. When we adjusted the analysis for sex and age at baseline, the difference in the history of coronary artery disease (P = .69), as well as CCI (P = .54), lost their statistical significance, while the difference in the history of atrial fibrillation was border-line statistically significant (P = .05).

Overall, 608 participants developed dementia during follow-up (587 [17.5%] in the general population and 21 [20.2%] in MDs). A total of 282 cases of dementia were captured during the in-person MCSA evaluations (14 cases were in MDs [14.9%] and 268 were in the general population [9.2%]); 326 cases were ascertained through the medical record review (7 were in MDs [11.9%] and 319 were in the general population [16.6%]), possibly suggesting that MDs were more likely to stay in the study while they decline to dementia whereas individuals from the general population tended to drop out probably more before dementia.

Association Between Being an MD and Dementia, Including All Data

In participants without dementia at baseline, the HR for incident dementia was modestly increased without reaching statistical significance, suggesting that MDs did not have a significantly different risk for incident dementia (vs the general population; HR = 1.12; 95% CI = 0.69–1.82; P = .51, adjusting for sex, education, and APOE ε4, using age as the time scale; Table 2 and Figure 2). Adjusting additionally for BDI-II and BAI scores, as well as CCI, did not change estimates appreciably. Empirically, we observed a lower rate of dementia in MDs younger than 80 years (Figure 2). Inspection of the data showed there were few MD participants younger than 75 years with dementia; thus, estimation of the risk of dementia for MDs younger than 75 years was imprecise.

Table 2.

HR of MD Associated With Dementia

Outcome: Dementia	MDs vs General Population
	Events/N	HR (95% CI)a	P Value
Model 1	600/3385	1.12 (0.69–1.82)	.636
Model 2	578/3298	1.22 (0.75–2.00)	.426
Model 3	578/3298	1.17 (0.72–1.92)	.527

Abbreviations: CI, confidence interval; HR, hazard ratio; MD, medical doctor; N, total number of participants for whom data are included in the model.

^aHR (95% CI) from Cox proportional hazards models using age as the time scale; model 1 was adjusted for sex, education, and apolipoprotein E ε4 status; model 2 was additionally adjusted for the Beck Depression Inventory and the Beck Anxiety Inventory scores; and model 3 was additionally adjusted for the Charlson Comorbidity Index; reference group for all comparisons was the “general population” (ie, including all participants but MDs).

Figure 2.

Survival curves for outcome of incident dementia, adjusted for sex, education, and apolipoprotein E ε4 status, with age as the time scale. The y axis presents the proportion of participants without dementia. MD indicates medical doctor.

DISCUSSION

MDs in our study population (≥70 years) did not have a significantly different risk for dementia compared with the general population. Although the study includes a small number of older, mostly white male, MDs, it provides a preliminary insight on cognitive health later in life, while most previous studies examine the health of younger MDs. Larger longitudinal studies are needed to examine these associations and investigate if associations are modified by sex.

Issues concerning MD health have been under inspection for many years⁷ but not specifically for cognitive impairment later in life. Although previous studies support that MDs have lower mortality rates and healthier lifestyles than the general population,^1–3 burnout in MDs has been prevalent, at least in recent years,¹³ with higher rates in females and younger physicians. MDs in the present study are older (mean [SD] = 79.95 [5.36] years and 93.3% retired at baseline) than MD groups described in previous research reports (eg, medical students, residents, and active medical professionals), and we can hypothesize that their professional careers had many of the characteristics described in the literature, although they could have missed more recent changes. We know also that risk factors for dementia span the lifetime, including the prevention or management of many common chronic conditions (eg, car-diovascular diseases) with the potential to reduce or delay dementia cases by a third.²⁹ MDs have access to this information and apply it to their practice but not always to themselves,^4,5,9 although MDs with healthy habits are more likely to counsel patients about prevention and for the patients to practice these preventive health practices.^30,31

Programs tailored to MDs’ specific needs as patients are developing. Physicians, for example, consider useful rapid referral access, confidential telephone consultation, comprehensive care provided in a short time, and advice on work-life balance.⁹ Professional medical associations and health authorities in many countries, including the United States, have taken measures to optimize physicians’ healthcare (eg, MDs who specialize in having MDs as their patients, psychological services, services for medical students and trainees, and conferences on physician health).³² Medical schools have introduced wellness programs promoting healthy nutrition, exercise, emotional health, and community involvement, to establish healthy habits in hopes that physicians will maintain these habits during their medical careers.⁴ However, healthcare access for MDs is more complex than just offering a general practitioner¹¹; this might be the first step.

Identified barriers to healthcare access for MDs, for the most part, are modifiable, requiring the collaboration of policy makers, employers, and MDs.⁷ Poor health in MDs has been associated with long working hours, shift work, and high job demands,⁶ which make access to medical care difficult for MDs; confidentiality of the information and cultural issues, like the pressure on MDs to be healthy and control their health, self-treatment encouraged by colleagues, and the belief that it is not culturally appropriate to acknowledge illness in a peer are additional barriers.¹¹ Systemic barriers (eg, long hours, cultural issues, and lack of encouragement by employer)⁷ rather than personal ones seem to be more important hindrances for healthcare access for MDs.¹¹ A common result of barriers could be self-diagnosis and self-treatment,⁷ while often colleagues mutually neglect themselves and are not sure of whether they should interfere.³³

The study is not without limitations. Although MDs have a higher occurrence of suicide, burnout, and mental disorders than the general population or other groups, in our study population they were less likely to have a history of depression and had lower levels of anxiety (as assessed by the BAI total score); this might suggest that MD participants later in life have different characteristics than nonparticipant MDs or than MDs earlier in their carrier, or that MDs in our study have different characteristics than MDs in previous reports and other study populations. Additional studies in diverse populations would be desirable. On the other hand, the disease burden of MDs might have been underestimated due to the barriers to access healthcare (eg, resulting in a lower frequency of depression history). Additionally, medical record review might have missed mild dementia cases that are not yet under care. The small number of MDs included in this population-based study was not unexpected as it is estimated that there were 275 active licensed physicians per 100 000 population in 2010³⁴; however, with only 104 MDs, the study might not have had enough statistical power to detect a difference in the dementia incidence between the groups, and future studies with a higher sample size are necessary. The present study population was comprised mainly of white (98.7%), not-Hispanic or Latino (99.4%) participants, and further studies in more heterogeneous populations would be necessary. Most of the 70 year and older MDs (93.3% of them retired) in our study were male, with only seven female MDs. The proportion of female physicians has been increasing over time (eg, increased from 7.1% to 15.3% between 1970 and 1986),³⁵ while females comprised recently one-third of all licensed physicians.³⁴ Thus, future studies are needed to examine if associations are modified by sex.

In conclusion, MDs in our study population did not have a significantly different risk for dementia compared with the general population. Further investigations in larger diverse study populations are needed to validate these findings.