medical education in review

Prevalence of depression amongst medical students: a

meta-analysis
Rohan Puthran,1 Melvyn W B Zhang,2 Wilson W Tam3 & Roger C Ho1

CONTEXT Medical schools are known to be 28.0% (95% confidence interval [CI] 24.2–
stressful environments for students and hence 32.1%). Female, Year 1, postgraduate and Mid-
medical students have been believed to experi- dle Eastern medical students were more likely
ence greater incidences of depression than to be depressed, but the differences were not
others. We evaluated the global prevalence of statistically significant. By year of study, Year 1
depression amongst medical students, as well as students had the highest rates of depression at
epidemiological, psychological, educational 33.5% (95% CI 25.2–43.1%); rates of depres-
and social factors in order to identify high-risk sion then gradually decreased to reach 20.5%
groups that may require targeted interventions. (95% CI 13.2–30.5%) at Year 5. This trend
represented a significant decline (B = 0.324,
METHODS A systematic search was conducted p = 0.005). There was no significant difference
in online databases for cross-sectional studies in prevalences of depression between medical
examining prevalences of depression among and non-medical students. The overall mean
medical students. Studies were included only if frequency of suicide ideation was 5.8% (95%
they had used standardised and validated ques- CI 4.0–8.3%), but the mean proportion of
tionnaires to evaluate the prevalence of depressed medical students who sought treat-
depression in a group of medical students. ment was only 12.9% (95% CI 8.1–19.8%).
Random-effects models were used to calculate
the aggregate prevalence and pooled odds CONCLUSIONS Depression affects almost
ratios (ORs). Meta-regression was carried out one-third of medical students globally but
when heterogeneity was high. treatment rates are relatively low. The current
findings suggest that medical schools and
RESULTS Findings for a total of 62 728 medi- health authorities should offer early detection
cal students and 1845 non-medical students and prevention programmes, and interven-
were pooled across 77 studies and examined. tions for depression amongst medical students
Our analyses demonstrated a global prevalence before graduation.
of depression amongst medical students of

Medical Education 2016: 50: 456–468

doi: 10.1111/medu.12962

Discuss ideas arising from the article at

www.mededuc.com discuss.

1
Department of Psychological Medicine, Yong Loo Lin School of Correspondence: Rohan Puthran, Department of Psychological
Medicine, National University of Singapore, Singapore, Singapore Medicine, Yong Loo Lin School of Medicine, National University
2
Department of General and Forensic Psychiatry, Institute of of Singapore, Singapore 117597, Singapore. Tel: 00 65 9424
Mental Health, Singapore, Singapore 8149; E-mail: rohan.ri@gmail.com
3
Alice Lee Centre for Nursing Studies, Yong Loo Lin School of
Medicine, National University of Singapore, Singapore, Singapore

456 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468
 Depression and medical students

This in turn translates to higher rates of suicide

INTRODUCTION ideation amongst medical students, for which one
large multi-institution study has reported a preva-
Medical schools around the globe have long been lence of 11.2%.14 It is well established that the rate
considered stressful environments for students of physician suicide is high, with male and female
entering tertiary education. Previous research has physicians having suicide rates that are 40% and
identified long hours of study, and the impositions 130%, respectively, higher than those of their coun-
of emotional burden,1 high workload2 and consider- terparts in the general population.15 This is believed
able financial pressure as the principal stressors. to have roots in medical school and may reflect the
Hence, it is unsurprising that medical students have effects of untreated psychological symptoms in med-
been found to show a significantly higher preva- ical students.16,17 Another important reason for the
lence of depressive symptoms than the general pop- high suicide risk amongst physicians refers to their
ulation.3 A qualitative review yielded no definitive greater knowledge of and easier access to means of
conclusions on the prevalences of depression in suicide.18
medical students compared with non-medical stu-
dents, and different studies included in the review Depression amongst medical students deserves spe-
showed very different results.4 The study did not cial attention; it not only affects students’ lives in
include a meta-analysis and hence made no statisti- terms of academic performance, dropout and pro-
cal comparison.4 fessional development, but may also have repercus-
sions on patient care in the long run: depressed
Some studies report gender differences in preva- and anxious medical students have been reported
lences of depression among medical students,3,5,6 to be more cynical, less empathetic and less willing
whereas others do not.7–9 The association between to care for chronically ill patients.19
year of study in medical school and the preva-
lence of depression is as yet unclear, with many This paper aims to evaluate the global prevalence of
studies reporting different results. James et al.10 depression amongst medical students, and the epi-
reported no difference in median depression demiological, psychological, educational and social
scores between graduate-entry and undergraduate- factors of relevance, in order to identify high-risk
entry students. In terms of geographic differences, groups that may require targeted interventions. We
Steptoe et al.11 showed that amongst university hypothesised that the prevalence of depression in
students in general, those with the highest level medical students would be higher than it is among
of depressive symptoms were from Asian coun- non-medical students as a result of the unique stres-
tries, and postulated that this may be attributable sors in medical school, and that female and Asian
in part to lower perceptions of a sense of control medical students would face the highest prevalences
in these students. To our knowledge, no similar of depressive symptoms based on the aforemen-
comparison has yet been made amongst medical tioned studies. We also hypothesised that postgradu-
students. ate medical students would be subject to a higher
prevalence of depressive symptoms as a result of
Medical students who have completed psychiatry their older age, greater likelihood of being indebted
postings are expected to have better knowledge for bachelor studies, and more likely exposure to
about depression and recommended treatments. additional psychosocial stressors such as marriage
However, studies have found that undertreatment and parenthood.20
of depression remains significant amongst medical
students, with low treatment rates of < 25%
amongst depressed students.12,13 Medical students METHODS
have been found to have higher rates of depres-
sion than the general population, but similar rates Search strategy
of seeking treatment, which suggests that depres-
sion may be undertreated in this group.12 Com- The online databases PubMed, MEDLINE,
mon reasons cited in these studies for this failure PsycINFO, EMBASE and Science Direct were
to seek treatment include fear of a negative impact searched from inception to April 2015. The search
on the student’s academic record and future terms used are summarised as follows: depress* and
career, lack of time, the stigma associated with medical student; depress* and medical undergradu-
using psychiatric health services and fear of ate; depress* and undergraduate medical education;
unwanted intervention. mood disorder and medical student; mood disorder

ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468 457
R Puthran et al

and medical undergraduate; mood disorder and were further evaluated based on their study design.
undergraduate medical education; affective disorder Articles were screened against the inclusion and
and medical student; affective disorder and medical exclusion criteria, as previously defined, by both of
undergraduate, and affective disorder and under- these authors independently. Any disagreements
graduate medical education. In order to ensure that with regard to the selection of articles were resolved
all relevant articles from the inception of the data- through discussion with the last author (RCH).
bases to date were extracted, no date or language Prior to further statistical analysis, all articles were
limitations were applied in the search strategy. In randomised by the blinding of the title of the arti-
addition to these electronic searches, the references cle, the authors and the name of the publishing
of the final selection of papers were manually journal. Data were extracted. The process involved
searched for further material. in this systematic selection is detailed in Fig. 1. This
procedure was based on that for the meta-analysis of
Inclusion criteria observational studies in epidemiology (MOOSE).21

Cross-sectional studies that evaluated the prevalence Statistical methods

of depression amongst medical students were
included for the analysis. Studies were included only Data extraction and critical appraisal
if they had: (i) assessed a group of medical students;
(ii) analysed the prevalence of depression; (iii) Data extracted from the selected articles were then
determined the prevalence of depression using stan- reviewed by the second and last authors, and docu-
dardised validated instruments and questionnaires, mented on a standardised form for: (i) details of
and (iv) provided adequate information for the publication (such as last name of first author, year
authors to compute the aggregated prevalence of and geographic locale of publication); (ii) the num-
depression in the selected cohort of medical ber of medical students (and non-medical students
students. Cross-sectional studies that reported a if applicable); (iii) undergraduate and postgraduate
comparative analysis between medical and non- medical curricula (entry to a postgraduate medical
medical students were included only if they curriculum requires the candidate to hold a bache-
reported comparative prevalence rates and statistics. lor degree prior to admission to medical school);
(iv) prevalences of depression overall and in sub-
Exclusion criteria groups within the sample if available; (v) descrip-
tions of the instruments used to assess for
Selected studies were excluded if they: (i) failed to depressive symptoms; (vi) overall prevalence of sui-
provide any aggregate prevalence of depressive dis- cide ideation in the medical student group and
order in the medical student group; (ii) had treatment rates within depressed medical students if
included nursing, dentistry or pharmacy students as available, and (vii) demographics of the study sub-
part of the medical student group; (iii) had consid- jects. This information is reflected in Table S1. The
ered only a subgroup of medical students with a quality of the 77 studies was appraised using an
specific characteristic not studied in this paper, or adaptation of the Newcastle–Ottawa cohort scale for
(iv) were not accessible for full review and provided cross-sectional studies,22 which takes into account
insufficient information in their abstracts to calcu- the selection of participants, comparability of out-
late aggregate prevalence. Studies for which full come groups and assessment of outcome measures
texts were written in languages other than English (Table S2). We also checked whether the provision
that were not translatable by the present authors of ethical approval was properly defined in these
were excluded. In addition, all interventional stud- papers in order to further review the quality of the
ies, case reports, case series and commentaries were research process; 58 of 77 papers satisfied this crite-
excluded. rion and a comparison analysis was performed to
look for any bias.
Selection of articles
Statistical analysis
Utilising the search strategy, all articles were
screened initially based on their titles, as well as Comprehensive Meta-Analysis (CMA) Version 2.0
their abstracts. The initial selection and screening (Biostat, Inc., Englewood, NJ, USA) was used to per-
of the articles were conducted by two authors (RP form all statistical analyses. Random-effects models
and MWBZ) independently. After the exclusion of were used to derive the summary odds ratio (OR).
articles that were irrelevant, the remaining articles The random-effects model was used because it

458 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468
 Depression and medical students

Number of tles retrieved from

online database search = 1395

1140 tles excluded for

irrelevant content

255 abstracts reviewed

Papers were evaluated against
inclusion criteria via abstracts to
determine inclusion for furthur
review
69 abstracts excluded for
- Wrong study design (35)
Wrong sample (33)
Duplicate (1)
186 papers shortlisted

17 papers not accessible

169 papers included for full-text

review 92 papers excluded
- Prevalence not reported (63)
- Duplicates (6)
- Other study designs (22)
77 papers used in analysis - Paper in Polish (1)
- 64 only medical students
- 13 medical versus non-medical
students

Figure 1 Process and outcome of the study selection strategy

assumes variance in effect sizes between studies but performed to examine the impact of moderator
leads to a markedly conservative null hypothesis variables on the study effect size when significant
model,23 and takes into consideration subject-speci- heterogeneity was detected, using a random-effects
fic effects.24–27 All studies reported depression as a model. Publication bias was examined by Egger’s
dichotomous variable (presence versus absence). regression method. In the event that publication
Thus, the aggregate prevalence of depression, its bias was detected, the classic fail-safe test was per-
corresponding p-value and 95% confidence interval formed to establish the potential number of missing
(CI), the Cochran’s Q-statistic and the estimated studies.27 Forest plots for the prevalence of depres-
effect size (t2) were reported for the medical stu- sion overall and within subgroups were made to
dent group. For studies comparing medical with represent the overall estimate, as well as individual
non-medical students, to test the hypothesis that the study estimates.
prevalence of depression was higher in the medical
student group, the summary OR was calculated. For Subgroup analyses
these analyses, the summary OR, its corresponding
p-value, 95% CI and Z-value were reported. The Subgroup analyses were performed to examine the
level of significance was set at 5% and the result was effects of gender and year of study on the preva-
regarded as statistically significant if the p-value was lence of depression in the medical student group.
< 0.05. The I2 statistic was used to assess hetero- Subgroup analyses also examined the effect of geo-
geneity among studies. As a reference, I2 values of graphic region by grouping studies based on the
25% are considered low, those of 50% moderate, United Nations standard geographic region classifi-
and those of 75% high.27 Meta-regression was cation.28 Middle Eastern countries were differenti-

ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468 459
R Puthran et al

Table 1 Results of random-effects meta-regression of demographic moderators for prevalence of depression amongst medical students
for categorical moderators

Moderator Studies used, n Univariate coefficient Z-value p-Value Estimated t2

Age of medical students, years, mean 43 0.009 0.166 0.868 0.558

Proportion of female students, % 70 0.376 0.709 0.478 0.723
Proportion of single students, % 22 0.518 0.658 0.510 0.552

ated from Asian countries based on traditional defi-

nitions. We also used subgroup analysis to investi- Table 2 Results of subgroup analyses of prevalences of
gate differences in prevalence between medical and depression in medical students for categorical moderators
non-medical students.
Pooled
Subgroups prevalence,
RESULTS
comparison Studies, n % 95% CI

Seventy-seven cross-sectional studies covering a total

1 Continent
of 62 728 medical students and 1845 non-medical
Asia 19 30.1 21.3–40.6
students were included in the analysis (Fig. 1). Only
Europe 13 20.0 13.2–29.2
one paper written in Polish was excluded for lan-
guage reasons. Descriptive characteristics of the Middle East 13 31.8 26.3–37.9
study populations are shown in Table S1. A full list North America 19 30.3 23.6–38.0
is available in Table S3. Meta-regression and sub- South America 13 26.8 21.6–32.7
group analyses are shown in Tables 1 and 2. 2 Course structure
Undergraduate 55 27.0 22.7–31.8
Prevalence of depression Postgraduate 22 30.8 24.6–37.8
3 Gender
The global prevalence of depression amongst medi- Female 23 31.5 26.1–37.5
cal students was 28.0% (95% CI 24.2–32.1%; Male 22 24.2 18.5–31.0
Z= 9.37, d.f. = 76, s2 = 0.76, I2 = 99.06) based
4 Year of study
on the random-effects model. Figure 2 demonstrates
Year 1 20 33.5 25.2–43.1
the results of 77 studies that reported prevalences
Year 2 19 32.4 24.9–41.0
of depression amongst medical students. In view of
the substantial heterogeneity in the papers analysed, Year 3 18 29.3 22.7–36.9
meta-regressions against mean age (p = 0.868), pro- Year 4 17 26.0 20.2–32.7
portion of female students (p = 0.478) and propor- Year 5 8 20.5 13.2–30.5
tion of students whose relationship status was single 5 Course of study
(p = 0.510) were conducted, but none of the mod- Medical 6 28.7 15.7–46.5
erators were found to have statistically significant Non-medical 6 30.6 19.3–44.9
effects (Table 1).
95% CI = 95% confidence interval.
Analyses conducted by geographic region (Fig. 3)
showed differences, with the highest prevalences of
depression in the Middle East at 31.8% (95% CI (27.0%, 95% CI 22.7–31.8%), but this difference
13.2–29.2%), followed by North America (30.3%, was not statistically significant (Fig. 4). When preva-
95% CI 23.6–38.0%), Asia (30.1%, 95% CI 21.3– lences were stratified by gender, female medical stu-
40.6%), South America (26.8%, 95% CI 21.6– dents were found to have a higher prevalence of
32.7%) and Europe (20.0%, 95% CI 13.2–29.2%). depression (31.5%, 95% CI 26.1–37.5%) than male
However, these differences were not statistically sig- students (24.2%, 95% CI 18.5–31.0%), but again
nificant. Postgraduate medical students faced higher this difference was not statistically significant
rates of depression (30.8%, 95% CI 24.6–37.8%) (Fig. 5). When prevalences were stratified by year of
compared with undergraduate medical students study in medical school, Year 1 students showed the

460 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468
 Depression and medical students

Figure 2 Data from studies (n = 77) assessing prevalences of depression amongst medical students (n = 62 728). 95%
CI = 95% confidence interval

ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468 461
R Puthran et al

Figure 3 Data from studies (n = 77) assessing prevalences of depression amongst medical students (n = 62 728) stratified by
region. 95% CI = 95% confidence interval

462 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468
 Depression and medical students

Figure 4 Data from studies assessing prevalences of depression in undergraduate and postgraduate medical students. 95%
CI = 95% confidence interval

ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468 463
R Puthran et al

Figure 5 Data from studies assessing prevalences of depression amongst medical students stratified by gender. 95%
CI = 95% confidence interval

highest rates of depression (33.5%, 95% CI 25.2– s2 = 0.26, I2 = 91.58) based on the random-effects
43.1%) and rates gradually decreased until Year 5 model, using eight studies. The proportion of
(20.5%, 95% CI 13.2–30.5%) (Fig. 6). This trend depressed medical students who sought treatment
represented a significant decline (B = 0.324, was 12.9% (95% CI 8.1–19.8%; Z = 7.30, d.f. = 5,
p = 0.005). s2 = 0.23, I2 = 72.44), using six studies.

Suicide ideation and treatment rate Comparison with non-medical students

The overall prevalence of suicide ideation amongst Of 13 studies that compared medical students with
all medical students, regardless of depression status, non-medical students, seven used comparison
was 5.8% (95% CI 4.0–8.3%; Z = 14.25, d.f. = 7, groups from only one specific faculty, and hence

464 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468
 Depression and medical students

Figure 6 Prevalences of depression amongst medical students (n = 62 728) stratified by year group

Figure 7 Data from studies (n = 6) assessing prevalences of depression in medical and non-medical students. 95% CI = 95%
confidence interval

were deemed unsuitable for the comparative analy- Publication bias and quality of studies
sis as comparisons against specific faculties would
not be useful because each study used a different There was no significant publication bias for the
faculty. Odds ratios from the six remaining studies prevalence of depression amongst medical students
are represented in Fig. 7. There was a difference in the 77 studies (intercept = 2.11, 95% CI 6.26
in prevalence between the two groups, with non- to 2.03; t = 1.02; d.f. = 75, p = 0.31). There was no
medical students facing a prevalence of depression significant difference in prevalences of depression
of 30.6% and medical students subject to a preva- between studies that reported the provision of ethi-
lence of 28.7%. However, the pooled OR between cal approval (28.5%, 95% CI 23.9–33.5%) and those
the two groups was found to be insignificant for that did not (27.6%, 95% CI 22.9–32.8%). When
this result (OR 0.965, 95% CI 0.651–1.429; the Newcastle–Ottawa cohort scale for cross-sec-
p = 0.857). tional studies22 was used to critically appraise the

ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468 465
R Puthran et al

quality of studies, all of the studies included scored and future career, denial of illness, poor insight,
≥ 7 out of a possible 10 points. lack of time, the stigma associated with using psy-
chiatric health services and fear of unwanted inter-
vention.12,29 Chew-Graham et al.29 also found that
DISCUSSION student knowledge of the existing support services
available to medical students was poor and that
To our knowledge, this is the first meta-analysis to medical students were reluctant to consult psychi-
investigate the prevalence of depression amongst atric health services.
medical students globally. Our findings suggest that
medical students have a high overall prevalence of In a comparison of medical students with non-medi-
depression of 28.0%. In a study conducted amongst cal students, we found no significant difference in
university students from 23 countries of varying the prevalence of depression between the two
income levels, Steptoe et al.11 found that preva- groups. This finding adds statistical value to the
lences of depressive symptoms were 19% in male qualitative review carried out by Bacchi and Lici-
students and 22% in female students, which are nio,4 who suggested that the similar rates in the two
much lower than the current findings. Chew-Gra- groups may reflect a problem of depression of
ham et al.29 found that the transition of moving broader scope amongst university students, and that
away from home and academic stressors were two further longitudinal studies comparing medical with
main themes that emerged from their interviews non-medical students are required to determine the
with medical students. One factor unique to medi- underlying issues that affect university students in
cal students, which was commonly cited in that general and medical students specifically. Regard-
study,29 was the need to be seen as a competent less, prevalences of depression amongst both medi-
clinician. This extra pressure may come from profes- cal and non-medical students are high and neither
sors, mentors, peers or patients, as well as from group should be discounted.
medical students themselves.
These findings confer significant implications as
Our analysis showed that the prevalence of depres- they aid senior leaders of medical schools, medical
sion is significantly highest in Year 1 medical stu- educators and mental health professionals in shap-
dents, and that this decreases in subsequent years ing the medical curriculum and taking care of stu-
(Fig. 6). This may be attributable to issues of dents’ psychiatric health as it will impact on patient
adjustment when prospective medical students first care in the long run.1 Given the high global preva-
enter medical school, which later improve as stu- lence of depression amongst medical students and
dents learn to cope better, even with the increased the aforementioned reasons for their avoidance of
workload and the transition to the clinical phase. treatment, medical schools should aim to reduce
Longitudinal studies suggest that adjustment issues the associated stigma and should encourage
may be responsible for the high prevalence of depressed students to seek treatment from profes-
depression in Year 1 students by showing that levels sionals outside medical school in order to ensure
of depressive symptoms in batches of medical stu- that students are provided with independent assess-
dents increase significantly from their pre-medical ments. We suggest that such strategies be imple-
school baseline.30,31 This supports the suggestion mented at a Year 1 level, rather than in the final
that medical school itself is a significant stressor for year, because our findings show that depression is
students, especially in the first year of study. significantly highest in the first year of medical
school.
We found the prevalence of suicide ideation in the
overall sample to be high at 5.8%. Although our As medical students have a high prevalence of
finding is much lower than that of Dyrbye et al.,14 depression, medical schools should look into the
it is still considered a high prevalence, and it is provision of structured programmes that are evi-
important because physicians have been found to dence-based. Examples include life skills training
have much higher suicide rates than the general and mindfulness therapy, both of which have been
population.15 Undertreatment is also evident in shown to significantly reduce levels of depression in
our study findings, with only 12.9% of depressed medical student populations.32,33 Social support
medical students seeking treatment. Common rea- should also be given to female students, as well as
sons for not seeking treatment include fear of a graduate students, who face higher levels of stress
negative impact on the student’s academic record and depression in medical school.

466 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468
 Depression and medical students

This study is strengthened by its finding of minimal which is consistent with the findings of other reviews
publication bias and also by the fact that it is the on medical student depression,4,19,34 as well as the
first meta-analysis to estimate the global prevalence high prevalence of suicide ideation and low treat-
of depression amongst medical students. Nonethe- ment rate, it is clear that the time has come for con-
less, this meta-analysis is not without limitations. certed efforts by senior administrators of medical
Firstly, we were unable to assess many other factors schools, medical educators and mental health profes-
that might predispose a medical student to depres- sionals to develop new systems for the prevention,
sion, such as family history, life stressors and poor screening and treatment of psychiatric health disor-
coping strategies as a result of the variability in fac- ders in our future physicians.
tors reported in multiple studies. Secondly, all of
the studies included in the analysis used standard-
ised screening questionnaires to estimate the preva- Contributors: RP contributed to the acquisition, analysis
lence of depression; however, although screening and interpretation of data, and the drafting of the manu-
questionnaires provide a practical means to assess script. MWBZ contributed to the conception and design
of the work, and to the acquisition, analysis and interpre-
depression in large groups of students, full diag-
tation of data. WWT contributed to the critical revision of
nostic interviews would be a more accurate way to
the analysis and interpretation of data. RCH contributed
assess for the presence of depression. Furthermore, to the conception and design of the work. All authors
different screening instruments were adopted in contributed to the critical revision of the paper and
different studies, and researchers differed in their approved the final manuscript for publication.
choice of cut-off points for depression; however, Acknowledgements: none.
these thresholds were largely in agreement with Funding: none.
internationally accepted norms used to accurately Conflicts of interest: none.
diagnose depression and to screen for major Ethical approval: not applicable.
depressive disorder. Thirdly, the heterogeneity of
the study populations, wide variety of course struc-
tures, varied locations and cultural influences intro- REFERENCES
duce numerous confounding variables that may
influence the self-reporting of depression, and such 1 Rosenthal JM, Okie S. White coat, mood indigo –
heterogeneity is thus a limitation of this meta-analy- depression in medical school. N Engl J Med 2005;353
sis. The high heterogeneity in the results may (11):1085–8.
2 Adams J. Straining to describe and tackle stress in
reflect poor and varied methodologies in the
medical students. Med Educ 2004;38 (5):463–4.
papers analysed, which makes it difficult to draw 3 Dahlin M, Joneborg N, Runeson B. Stress and
definitive conclusions, and may also explain differ- depression among medical students: a cross-sectional
ences between groups. We were unable to assess study. Med Educ 2005;39 (6):594–604.
for some potential moderators, in view of the high 4 Bacchi S, Licinio J. Qualitative literature review of the
heterogeneity, such as the proportion of students prevalence of depression in medical students
with a personal or family history of a psychiatric compared to students in non-medical degrees. Acad
health disorder, as well as additional stressors such Psychiatry 2015;39:293–9.
as parenthood and financial constraints, amongst 5 Peterlini M, Tiberio IFLC, Saadeh A, Pereira JCR,
others. Martins MA. Anxiety and depression in the first year of
medical residency training. Med Educ 2002;36 (1):66–
72.
6 Schenk TL, Davis L, Wimsatt LA. Depression, stigma,
CONCLUSIONS
and suicidal ideation in medical students. JAMA
2010;304 (11):1181–90.
In conclusion, the present meta-analysis estimated 7 Bramness JG, Fixdal TC, Vaglum P. Effect of medical
the prevalence of depression to be 28.0% amongst school stress on the mental health of medical
medical students globally. Female students, graduate students in early and late clinical curriculum. Acta
students and Middle Eastern students trended Psychiatr Scand 1991;84 (4):340–5.
towards higher prevalences of depression, although 8 Levine RE, Litwins SD, Frye AW. An evaluation of
these were not statistically significant. Year 1 students depressed mood in two classes of medical students.
had the highest prevalence of depression, reflecting a Acad Psychiatry 2006;30 (3):235–7.
difference that was statistically significant, and subse- 9 Haldorsen H, Bak NH, Dissing A, Petersson B. Stress
and symptoms of depression among medical students
quent year groups showed a gradual decrease in
at the University of Copenhagen. Scand J Public Health
prevalences of depression to the final year. In view of 2014;42 (1):89–95.
the high prevalence of depression identified here,

ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468 467
R Puthran et al

10 James D, Ferguson E, Powis D, Bore M, Munro D, 25 Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal
Symonds I, Yates J. Graduate entry to medicine: Analysis. Hoboken, NJ: Wiley-Interscience 2004.
widening psychological diversity. BMC Med Educ 26 Laird NM, Ware JH. Random-effects models for
2009;9:67. longitudinal data. Biometrics 1982;38 (4):963–74.
11 Steptoe A, Tsuda A, Tanaka Y, Wardle J. Depressive 27 Ho RC, Zhang MW, Tsang TY et al. The association
symptoms, socio-economic background, sense of between internet addiction and psychiatric
control, and cultural factors in university students from co-morbidity: a meta-analysis. BMC Psychiatry
23 countries. Int J Behav Med 2007;14 (2):97–107. 2014;14:183.
12 Tjia J, Givens JL, Shea JA. Factors associated with 28 United Nations Statistics Division. Standard Country
undertreatment of medical student depression. J Am and Area Codes Classification (M49), revised October
Coll Health 2005;53 (5):219–24. 2013. http://unstats.un.org/unsd/methods/m49/
13 Givens JL, Tjia J. Depressed medical students’ use of m49regin.htm. [Accessed 1 December 2015]
mental health services and barriers to use. Acad Med 29 Chew-Graham CA, Rogers A, Yassin N. ‘I wouldn’t
2002;77 (9):918–21. want it on my CV or their records’: medical students’
14 Dyrbye LN, Thomas MR, Massie FS et al. Burnout and experiences of help-seeking for mental health
suicidal ideation among US medical students. Ann problems. Med Educ 2003;37 (10):873–80.
Intern Med 2008;149 (5):334–41. 30 Rosal MC, Ockene IS, Ockene JK, Barrett SV, Ma Y,
15 Schernhammer ES, Colditz GA. Suicide rates among Hebert JR. A longitudinal study of students’
physicians: a quantitative and gender assessment (meta- depression at one medical school. Acad Med 1997;72
analysis). Am J Psychiatry 2004;161 (12):2295–302. (6):542–6.
16 Schernhammer E. Taking their own lives – the high 31 Yusoff MS, Abdul Rahim AF, Baba AA, Ismail SB, Mat
rate of physician suicide. N Engl J Med 2005;352 Pa MN, Esa AR. The impact of medical education on
(24):2473–6. psychological health of students: a cohort study.
17 Sun L, Sun LN, Sun YH, Yang LS, Wu HY, Zhang Psychol Health Med 2013;18 (4):420–30.
DD, Cao HY, Sun Y. Correlations between 32 Li C, Chu F, Wang H, Wang XP. Efficacy of Williams
psychological symptoms and social relationships LifeSkills training for improving psychological health:
among medical undergraduates in Anhui Province of a pilot comparison study of Chinese medical students.
China. Int J Psychiatry Med 2011;42 (1):29–47. Asia Pac Psychiatry 2014;6 (2):161–9.
18 Klonsky ED, May AM. The three-step theory (3ST): a 33 Warnecke E, Quinn S, Ogden K, Towle N, Nelson
new theory of suicide rooted in the ‘ideation-to- MR. A randomised controlled trial of the effects of
action’ framework. Int J Cogn Therapy 2015;8 mindfulness practice on medical student stress levels.
(2):114–29. Med Educ 2011;45 (4):381–8.
19 Dyrbye LN, Thomas MR, Shanafelt TD. Systematic 34 Hope V, Henderson M. Medical student depression,
review of depression, anxiety, and other indicators of anxiety and distress outside North America: a
psychological distress among US and Canadian systematic review. Med Educ 2014;48 (10):963–79.
medical students. Acad Med 2006;81 (4):354–73.
20 Dyrbye LN, Thomas MR, Power DV et al. Burnout SUPPORTING INFORMATION
and serious thoughts of dropping out of medical
school: a multi-institutional study. Acad Med 2010;85 Additional Supporting Information may be found
(1):94–102.
in the online version of this article:
21 Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson
GD, Rennie D, Moher D, Becker BJ, Sipe TA,
Thacker SB. Meta-analysis of observational studies in Table S1. Descriptive characteristics of studies
epidemiology: a proposal for reporting. Meta-analysis reporting prevalences of depression amongst medi-
of Observational Studies in Epidemiology (MOOSE) cal students.
group. JAMA 2000;283 (15):2008–12.
22 Herzog R, Alvarez-Pasquin MJ, Diaz C, Barrio JLD, Table S2. Appraisal of studies for participant selec-
Estrada JM, Gil A. Are healthcare workers’ intentions tion, comparability of outcome groups and assess-
to vaccinate related to their knowledge, beliefs and ment of outcome measures.
attitudes? A systematic review BMC Public Health
2013;13:154. Table S3. Full descriptive characteristics of the study
23 Han B, Eskin E. Random-effects model aimed at
populations.
discovering associations in meta-analysis of genome-
wide association studies. Am J Hum Genet 2011;88
(5):586–98. Received 7 June 2015; editorial comments to author 30 July
2015, 7 October 2015, accepted for publication 3 November
24 Diggle P. Analysis of Longitudinal Data, 2nd edn.
2015
Oxford; New York, NY: Oxford University Press 2002.

468 ª 2016 John Wiley & Sons Ltd. MEDICAL EDUCATION 2016; 50: 456–468