student selection
Predicting medical student performance from
attributes at entry: a latent class analysis
Paul Lambe & David Bristow
OBJECTIVES The objectives of this study
were two-fold. Firstly, we aimed to model
typologies of student examination perfor-
mance by grouping students into like
categories based on measures of prior
academic achievement (particularly in the
science subjects) and interview rating at time
of entry to a medical degree course, and out-
come measures of subsequent performance
across the course. Secondly, we aimed to
illustrate and evidence the utility of the latent
class analysis (LCA) clustering technique to
provide meaningful information on the
effectiveness of a student selection process with
respect to the likelihood of poor examination
performance.
METHODS For this retrospective study,
anonymised data on two sequential cohorts of
students who graduated from a 5-year Bachelor
of Medicine, Bachelor of Surgery degree course
were analysed using LCA. In order to triangu-
late the findings, the same data were analysed
Medical Education 2011: 45: 308316
doi:10.1111/j.1365-2923.2010.03897.x
Institute of Clinical Education, Peninsula Medical School, University
of Plymouth, Plymouth, Devon, UK
308
using the more conventional approach of
logistic regression.
RESULTS The LCA identified three distinct
classes or typologies of student examination
performance using measures of prior academic
achievement and interview rating at time of
course entry. Measures of prior academic
achievement and score on a structured admis-
sions interview made significant contributions
to the models ability to discriminate between
typologies. Strong prior academic achievement,
especially in chemistry, and high interview score
were positively related to the likelihood of suc-
cessful test performance. These findings were
supported by the logistic regression analysis.
CONCLUSIONS The LCA clustering tech-
nique provided meaningful information on the
performance of a selection process. As a com-
plementary tool to existing methods used in
this area of research, LCA has the potential to
empirically inform the selection process.
Correspondence: Paul Lambe, C409 Portland Square, University of
Plymouth, Drake Circus, Plymouth, Devon PL4 8AA, UK.
Tel: 00 44 1752 588950; Fax: 00 44 1752 586788;
E-mails: paul.lambe@pms.ac.uk, david.bristow@pms.ac.uk
Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
 Predicting student performance from entry attributes
INTRODUCTION
The General Medical Council1 requires that UK
medical schools provide evidence that their stu-
dent selection procedures are based on relevant
guidance, recognised best practice, and research
into effective, reliable and valid selection processes
which can have the confidence of applicants and the
public. In response to this requirement, this study
examines the relationships between prior academic
achievement, interview score and examination
performance on assessments of applied medical
knowledge (AMK), clinical skills and professionalism.
Prior academic achievement and interview rating are
the most widely used determinants in the medical
school selection and admission process.2 Applicants to
UK medical schools are mainly offered a place based
on a combination of academic ability and an inter-
view3 and prior academic achievement has generally
been found to have better ability to predict study
success than interview rating.4,5 Given the demanding
science content of medical courses, applicants are
generally required to have studied two or more
sciences to A-level and, in particular, to have achieved
the top grade in A-level chemistry or biology. Indeed,
good science grades, particularly in chemistry and
biology, have been shown to be associated with
success on the undergraduate medical course,
especially in terms of knowledge-based performance,
6,7
and a lower mean examination grade at A-level has
been reported to be associated with greater risk for
poor performance.8,9 Further, there is evidence that
A-level results predict performance in postgraduate
medical examinations and postgraduate careers,10,11
and that better performance in postgraduate exam-
inations is significantly correlated with graduation
from UK medical schools with student intakes that
have a higher mean of qualifications at A-level.12
It is generally accepted that there is much more to
being a doctor than academic success and that
selection should discriminate not only on cognitive
ability, but also on a range of non-academic
qualities.1,3,13 Performance at interview is a major
determinant in the selection and admission decision-
making process, but there is debate about its ability to
reliably predict non-clinical and clinical academic
success in medical training.14 However, there is much
evidence that structure is a major moderator of
interview validity and that highly structured interviews
conducted by trained interviewers increase the pre-
dictive ability of interview ratings with respect to
medical student performance.1416
Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
Nevertheless, much of the variance in medical school
performance remains unexplained by prior academic
achievement and, indeed, by interview rating.4
Correlation is the most widely used measure in
studies of the predictive validity of selection proto-
cols. However, the strength of correlation coeffi-
cients between predictors and outcomes are
weakened by data range restriction caused by a lack of
variability in scores on measures of prior academic
achievement and interview ratings of highly qualified
students who are accepted to oversubscribed courses,
and the unreliability of outcome measures.4,11
Although there are corrective formulae to address the
problem of range restriction, they are only appro-
priate in particular circumstances and should not be
routinely applied when the selection process is
complex.17 Furthermore, selection measures are
likely to be inter-correlated, as are outcome mea-
sures,3 making multivariate analysis problematic.
The need for a clearer understanding of the
predictive ability of prior academic achievement and
interview score has greater urgency given the large
numbers of high academic achievers applying for
places at medical schools, grade inflation (grades
awarded increase faster than any real increase in
standards) and increased numbers of applicants with
non-traditional qualifications (which makes discrim-
inating fairly among candidates increasingly prob-
lematic).18,19 A primary aim of this study was to obtain
a more refined picture of the relationship between
prior academic performance, interview score and
performance across the medical degree course. To do
this, selection and outcome criteria are scored
categorically to enable the use of a non-parametric
statistical method. Although the rescaling of range-
restricted continuous data into categorical data
involves a further loss of information, the general
benefit of this strategy is that it allows for a more
multivariate approach. A particular benefit of the
method is that it mitigates the problem of inter-
correlation among the indicators of interest.
METHODS
In this paper we take a novel alternative approach and
model the quality of students by grouping students
into like categories based on measures of prior
academic achievement and interview rating at time of
entry, and outcome measures of performance across
the degree course, using the latent class analysis
(LCA) clustering technique.20,21 Latent class analysis
is a non-parametric statistical method (it does not
309
P Lambe & D Bristow
assume a linear relationship, normal distribution of
data or homogeneity of variance) for identifying
subtypes (classes) of related cases from multivariate
categorical data. The classes are defined by the
criterion of conditional independence: conditional
on latent class membership, the observed variables are
mutually independent of one another. Thus, the
variables within any one class are uncorrelated. Pos-
terior membership probabilities (maximum likeli-
hood estimates based on patterns of scores on the
observed variables) assign cases to homogeneous
latent classes. Cases in a particular latent class are
similar to one another because their response patterns
are generated by the same probability distribution.
Two useful parameters are produced: latent class
probabilities and conditional probabilities. Latent
class probabilities indicate the relative size of each
class and whether the sample population is relatively
evenly distributed among the classes. Within each
class there is a set of conditional probabilities relating
to each indicator. The conditional probabilities
represent the probabilities of an individual in a
particular class being at a particular response level for
a particular indicator and thereby enable characteri-
sation of the nature of the types defined by each of
the latent classes. Analogous to factor loadings in
factor analysis, the conditional probabilities provide
the measurement structure that defines the latent
classes.
Specification of the basic LCA model does not involve
the issue of causal ordering. In these analyses, our
interest is in the underlying association between
variables and we make no distinction between what is
dependent and what is explanatory. Used here as a
data reduction technique, the method empirically
establishes the dimensions of any underlying
associations.
Data
For this retrospective study, anonymised data were
collected on two sequential cohorts of students who
graduated from a 5-year Bachelor of Medicine,
Bachelor of Surgery degree course (n = 142; 73
females, 69 males). All students included in these
analyses were direct school-leavers at time of entry.
Prior academic achievement
The medical school employs academic criteria
(predicted or attained A-levels and AS-levels or
equivalent qualifications) and, since 2006, the UK
Clinical Aptitude Test (UKCAT) to select students
310 Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
who will be invited for a structured admissions
interview. A-level and AS-level qualifications are post-
compulsory education (1618 years) advanced-level
examinations taken by those who want to study at
university in the UK. All applications to UK univer-
sities are made through the Universities and Colleges
Admission Service (UCAS); the UCAS tariff system
allocates points for qualifications, equivalised
according to the level and type of qualification and
related to grades (AE) that the applicant is pre-
dicted to achieve by his or her institution. In these
analyses, the UCAS22 points system was used to
calculate a score for each student based on his or her
actual attained qualifications at entry to Year 1 of the
programme.
Selection interview
A structured interview informs the offer of places on
the undergraduate medical programme. A students
interview total score is based on his or her answers to
a series of questions constructed to elicit perfor-
mance in 10 areas (communication skills, empathy
and pro-social behaviour, flexibility, decision making,
teamwork, honesty, manner of reflection, insight
about self, insight about illness and medicine, insight
into stress). A panel of at least three trained
interviewers, generally comprising a practising
doctor, an academic and a representative of the
community, rate each applicants performance in the
interview. The average over the three interviewers of
the sum of ratings for the 10 attributes was used in
these analyses.
Examination performance
Medical students AMK is tested regularly by progress
tests delivered four times per year.23 The test is
designed as a longitudinal assessment of the growth of
a students medical knowledge across the whole
programme. Scores on the fourth test in each of the
5 years of the degree course were used in the analyses.
An integrated structured clinical examination (ISCE)
is undertaken by students at the end of Years 2 and 4;
this is comprised of a number of timed stations that
aim to assess a students clinical competence and
professional ability.24 Students final scores were used
in the analyses.
In Year 5 of study, students undertake patient-based
presentations (PBPs) which assess ability to clerk,
present and lead a patient-based discussion. Clinical
and professional competence is graded on a scale
of 14 (1 = unsatisfactory, 2 = borderline,
 Predicting student performance from entry attributes
3 = satisfactory, 4 = excellent) in eight PBPs.
Students total scores were used in the analyses.
Data analysis
A retrospective study using LCA was conducted using
Latent GOLD Version 4.0 (Statistical Innovations,
Inc., Belmont, MA 02478, USA).25 The indicators
included in the model are the dependent variables
used to define the classes. The inactive covariates do
not influence the model parameter estimates and are
included in the model to enable further description
of the nature of the classes.20,26
The categorical variables included as indicators in the
model were:
1 bottom-quartile score on UCAS entry tariff
(1 = student scores in bottom quartile of score
distribution for respective cohort, 0 = student
scores outside bottom quartile);
2 bottom-quartile interview score (1 = student
scores in bottom quartile of score distribution for
respective cohort, 0 = student scores outside
bottom quartile);
3 lack of grade A in A-level chemistry (1 = Yes,
0 = No);
4 possession of one science A-level (1 = student has
chemistry or biology or physics at A-level,
0 = student has more than one of these science
subjects at A-level), and
5 bottom-quartile score on progress test
(1 = student scores in bottom quartile of score
distribution of his or her cohort on any one of
the end-of-year progress tests of AMK over the
5 years of the undergraduate course, 0 = student
scores outside bottom quartile).
The categorical variables included as inactive covari-
ates in the model were:
1 top-quartile score on Year 2 ISCE (1 = student
scores in top quartile of score distribution for
respective cohort, 0 = student scores outside top
quartile);
2 top-quartile score on Year 4 ISCE (1 = student
scores in top quartile of score distribution for
respective cohort, 0 = student scores outside top
quartile), and
3 top-quartile score on PBP (1 = student scores in
top quartile of score distribution for respective
cohort, 0 = student scores outside top quartile).
To triangulate the LCA findings, binary logistic
regression was employed to estimate the relative
Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
probability of the outcome scoring in the bottom
quartile or not in an end-of-year progress test, given
an individuals level on selected predictors. The
predictors used in the binary logistic regression
model were: grade at A-level chemistry; grade at
A-level biology; having a bottom-quartile interview
score or not, and having a bottom-quartile UCAS
entry tariff or not.
Ethical considerations
Ethical approval for this study was sought from the
medical schools ethics committee, which considered
that, as the research did not raise any human subject
protection issues, ethical approval was not required.
RESULTS
Latent class analysis
A sequence of models was fitted to the data with
different numbers of classes and a three-class model
was found to provide the best fit to our data
(Table 1). A good model fit was indicated by a low
likelihood ratio statistic (L2) not substantially
greater than the model degrees of freedom, a p-value
> 0.05, a low Bayesian information criterion (BIC)
statistic, a low Aikake information criterion (AIC)
statistic, and a dissimilarity index (D) of < 0.1.20
The three-class models indicators (bottom-quartile
UCAS entry tariff, bottom-quartile interview score,
grade A at A-level chemistry, one science A-level,
bottom-quartile progress test score) all had Wald
statistic p-values of < 0.05, indicating that all five
contributed significantly towards the ability to
discriminate between the typologies (Table 2). Thus,
each of the three classes contains a homogeneous
group of students who share common characteristics
and behaviour with respect to the observed
indicators.
The best performing students had the lowest like-
lihood of not having a grade A at A-level chemistry
(Table 2, conditional probabilities 0.26 versus 0.86
and 0.99), were highly unlikely to have just one
science subject at A-level, and were less likely to have a
low UCAS tariff or admission interview score. The
intermediate performing students had a much
higher likelihood of not having a grade A at A-level
chemistry than the best performing students and,
although they were highly unlikely to have only one
science A-level, they had the highest likelihood of
holding a bottom-quartile UCAS tariff score and
311
P Lambe & D Bristow

Table 1 Goodness-of-fit statistics of latent class analysis models containing different numbers of classes or typologies

Latent class models v2 L2 d.f. Npar BIC AIC D

I 295.73 314.54 26 5 143.84 262.54 0.29

II 48.55 44.62 20 11 ) 66.69 4.62 0.07
III 19.00 18.99 14 17 ) 72.92 ) 9.00 0.04
IV 9.13 9.96 8 23 ) 42.52 ) 6.04 0.029

BIC = Bayesian information criterion statistic; AIC = Aikake information criterion statistic

Table 2 Outline of conditional and latent class probabilities for the three-class model

Conditional probabilities

Class I Class II Class III

Best Intermediate Poorest
Three-class model performers performers performers Wald p-value

Indicators
Bottom-quartile score on AMK progress test 0.25 0.32 0.52 12.77 0.00
Bottom-quartile score on UCAS entry tariff 0.01 0.97 0.57 6.52 0.04
Bottom-quartile interview score 0.23 0.34 0.33 9.68 0.01
No grade A at A-level chemistry 0.26 0.86 0.99 103.97 0.00
Only one science A-level 0.02 0.00 0.67 11.23 0.00
Covariates
Top-quartile score on Year 2 ISCE 0.36 0.32 0.19
Top-quartile score on Year 4 ISCE 0.30 0.25 0.15
Top-quartile score on PBP 0.27 0.22 0.15
Latent class probability 0.52 0.42 0.06

AMK = applied medical knowledge; UCAS = Universities and Colleges Admission Service; ISCE = integrated structured clinical examination;
PBP = patient-based presentation

therefore lower aggregate A-level grades than their
counterparts in the best performing class. They were
also more likely than the best performing students
to have achieved a bottom-quartile interview score. By
stark contrast, the poorest performing students had
the highest probability of gaining a bottom-quartile
score on an end-of-year progress test, the highest
probability of not having a grade A at A-level
chemistry and the highest probability of having only
one science subject at A-level. Indeed, the poorest
performing students were over three times as likely
(odds ratio [OR] = 3.28) to achieve a bottom-quartile
score on an end-of-year progress test over the 5 years
of the degree course than the best performing
students and twice as likely (OR = 2.23) as the
intermediate performing students.
Furthermore, the inactive covariates included in the
model indicate that those in the poorest performing
typology were half as likely to achieve a top-quartile
score on Year 2 and Year 4 tests of clinical and
professional competence, and Year 5 assessments of
professionalism, than their counterparts in the best
performing typology (Table 2).

312 Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
 Predicting student performance from entry attributes

Table 3 Logistic regression modelling the probability of gaining a bottom-quartile score on an end-of-year progress test over the 5 years of
the degree course

Predictor* b-coefficient SE Z p > |z| 95% confidence interval

Grade at A-level chemistry 0.7319 0.1426 5.13 0.000 0.45231.0115

Grade at A-level biology 0.7743 0.1375 5.63 0.000 0.50471.0439

Bottom-quartile interview score ) 0.7008 0.2110 ) 3.32 0.001 ) 1.114458 to ) 0.2873

Bottom-quartile UCAS tariff score ) 0.4222 0.2125 ) 1.99 0.047 ) 0.8387558 to ) 0.0058
Constant 0.5351 0.6188 0.86 0.387 ) 0.6779 to 1.7481

* Chemistry and biology grades: 1 = A, 2 = B, 3 = other grades, 4 = did not take subject at A-level; bottom-quartile interview score:
1 = Yes, 2 = No; bottom-quartile UCAS tariff score: 1 = Yes, 2 = No
Wald tests: grade at A-level chemistry, v2 = 26.33, d.f. = 1, p < 0.01; grade at A-level biology, v2 = 31.69, d.f. = 1, p < 0.01; bottom-
quartile interview score, v2 = 11.033, d.f. = 1, p < 0.01; bottom-quartile UCAS tariff score, v2 = 3.95, d.f. = 1, p < 0.05
UCAS = Universities and Colleges Admission Service; SE = standard error

Binary logistic regression

Table 4 Comparisons between the predicted probabilities of a
Wald tests indicated that each b-coefficient in the student to gain a bottom-quartile score on an end-of-year
binary logistic model had a significant effect on the progress test over the 5 years of the degree course according
to grades at A-level chemistry and biology and whether the
probability of a student gaining a bottom-quartile student gained a bottom-quartile interview score
score on an end-of-year progress test (Table 3). Thus,
holding all other variables in the model constant, a
unit decrease in grade in A-level chemistry is 95%
associated with an increase in the odds of gaining a confidence
bottom-quartile score on an end-of-year progress test Type of student Probability interval
over the 5 years of the degree course. The same is
true for a unit decrease in grade at A-level biology. Grade A chemistry, grade 0.54 0.48510.6090
The odds of gaining a bottom-quartile score on an A biology (other variables
end-of-year progress test are increased for students held at mean)
with bottom-quartile interview scores, as well as for Grade A chemistry, grade 0.67 0.47220.8408
those with bottom-quartile UCAS tariff scores.27 A biology, bottom-quartile
interview score (other
Predicted values of the probability of the outcome
variables held at mean)
computed for specified values of the independent
Grade B chemistry, grade 0.84 0.80340.8863
variables in the logistic model provide further
B biology (other variables
congruent evidence to support the findings of the
LCA (Table 4).27 Predicted probabilities are com- held at mean)
puted from the b-coefficients and have a value of 01. Grade B chemistry, grade 0.90 0.85830.9419
The predicted probability of gaining a bottom- B biology, bottom-quartile
quartile score on an end-of-year progress test for interview score (other
students with A grades in both chemistry and biology variables held at mean)
(holding other variables at mean) is 0.55 and is used
as a comparative baseline. However, this baseline
increases almost 12% to 0.67 for students who were to 0.90 for those with B grades in both chemistry and
equally qualified but scored in the bottom quartile biology and with bottom-quartile interview scores
at interview. From our baseline the predicted prob- (holding other variables at mean).
ability increases by 29% to 0.85 for students with B
grades in both chemistry and biology (holding Although there is some overlap in the confidence
other variables at mean), and increases around 35% intervals for these predicted probabilities, it is still

Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316 313
P Lambe & D Bristow
true that the probability of gaining a bottom-quartile
score on an end-of-year progress test increases as
grades on chemistry and biology A-levels decrease.
Further, having a bottom-quartile interview score
appears to have an effect independent of our
measure of prior academic achievement upon the
likelihood of gaining a bottom-quartile score on an
end-of-year progress test.
DISCUSSION
This study illustrates a method that has the potential
to inform the selection process and identify the
profiles of students who are most likely to need
learning support. It adds to understanding of the
importance of A-level science subjects and grades,
and interview scores, in the prediction of subsequent
performance across the undergraduate medical
degree course.
The LCA shows that the difference in performance
among our sample of students on tests of AMK can
in part be explained by differences in measures of
prior academic achievement and interview score at
time of entry. It identified distinct subgroups which
differ in their probability for poor examination
performance according to scores on attributes mea-
sured at time of entry to the course.
Good A-level grades in more than one science
subject, particularly in chemistry, a high interview
score and a high UCAS tariff score were shown to be
positively associated with the likelihood of better
performance. Having only one science A-level, a low
UCAS tariff score and a low interview score were
shown to significantly increase the likelihood of poor
performance. The results of this study support the
predictive value of A-level chemistry and biology for
performance on tests of AMK and the contention that
good science grades, particularly in chemistry and
biology, are associated with success in knowledge-
based examinations on the medical degree course.
The findings also support the view that a lower
mean examination grade at A-level is associated
with greater risk for poor performance. Thus, with
respect to prior academic achievement, our findings
concur with those of many previous studies511 and
are not unexpected. What the study adds to this
literature is evidence indicating that, among equally
qualified students, those with lower interview scores
are more likely to perform less well.
Thus, the results of the study add to our under-
standing of the predictive ability of the structured
314 Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
interview. The LCA and the binary regression analysis
show that a structured interview can give valuable
added information about the quality of those selected
with respect to performance on tests of AMK across
the medical degree course. This finding concurs with
those of previous studies.15,2830
Additionally, the inclusion in the LCA model of
inactive covariates measuring clinical competence
and professionalism indicates that students who are
least likely to rank among the top performers in tests
of medical knowledge are also least likely to rank
among the top performers in assessments of clinical
competence and professionalism. This finding
concurs with Wilkinson and Framptons contention
that a students theoretical knowledge leads to good
performance in tests of practical clinical knowledge
and skills, and that good students tend to perform
well across a range of assessments because they have
ability in all components of clinical practice.31
Although the study indicates that results on science
A-levels may be more important than previously
thought for indicating likely academic progress, this
finding needs to be assessed in the context of the
general acceptance that considering a wide range
of personal qualities and skills may be just as
important as investigating academic achievement
when assessing which candidates are most likely to
become good doctors.1 Consideration of prior
academic achievement is just one element of a fair
selection procedure. As this study shows, non-cogni-
tive attributes measured at interview also have some
predictive validity.
All studies have limitations; the LCA cluster model
approach involves a trade-off between the loss of
information entailed by the rescaling of continuous
data into categorical data and the advantages facili-
tated by a more multivariate approach using a non-
parametric method, which enables a more nuanced
insight into the individual and combined effects of
selection instruments on outcomes of interest.
In general, the application of latent class models
requires comparatively large samples. However, mod-
erate sample size is unproblematic if the number of
polytomous variables included in the model does not
cause sparseness of data. This occurs when the number
of variables or the number of categories of these
variables are large and the number of cells in the
resultant multi-way frequency table exceeds the
sample size. In this study, five dichotomous variables
were included in the LCA model, producing a 32-fold
response pattern, a number well below the sample size
 Predicting student performance from entry attributes
of n = 142. If the inactive covariates had been included
as indicators, the resulting response pattern of 256
would have exceeded the sample size and cast doubt
on the precision of the parameter estimates. Clearly,
the number of categories of the continuous variables
rescaled into categorical variables and the number of
categorical variables included in the latent class model
for this study was constrained by sample size.26
Unfortunately, the latent class cluster model is also
limited in that it is unable to add to our knowledge
about the incremental validity of interview score and
can only provide evidence as to whether interview
score contributed significantly towards the ability to
discriminate between the typologies of student
examination performance.
Additionally, the analysis was carried out on just two
adjacent cohorts at a single institution. The study
group comprised only those who had successfully
completed the course and only those who were direct
school-leavers at time of entry. However, the results
reflect the study of a homogeneous group which
represents the vast majority of UK medical school
entrants.
Contributors: both authors contributed to the study
conception and design, data analysis and interpretation,
and the drafting and revision of the paper. Both authors
approved the final manuscript for publication.
Acknowledgements: none.
Funding: Peninsula Medical School funded this research.
Conflicts of interest: none.
Ethical approval: not required.
REFERENCES
1 General Medical Council. Tomorrows Doctors:
Outcomes and Standards for Undergraduate Medical
Education. 2009. http://www.gmc-uk.org/education/
undergraduate/tomorrows_doctors_2009.asp.
[Accessed 25 May 2010.]
2 Roberts C, Walton M, Rothie I, Crossley J, Lyon K,
Kumar K, Tillar D. Factors affecting the utility of the
multiple mini-interview in selecting candidates for
graduate-entry medical school. Med Educ 2008;42:396
404.
3 Patterson F, Ferguson E. Selection for medical educa-
tion and training. Understanding Medical Education Series.
Edinburgh: Association for the Study of Medical
Education 2007.
4 Ferguson E, James D, Madeley L. Factors associated
with success in medical school: systematic review of the
literature. BMJ 2002;331:5559.
Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
5 Yates J, Smith J, James D, Ferguson E. Should appli-
cants to Nottingham University Medical School study a
non-science A-level? A cohort study. BMC Med Educ
2009;9:5. http://www.biomedcentral.com/1472-6920/
9/5. [Accessed 15 April 2010].
6 James D, Chilvers C. Academic and non-academic
predictors of success on the Nottingham undergradu-
ate medical course 19701995. Med Educ 2001;35:1056
64.
7 Montague W, Odds F. Academic selection criteria and
subsequent performance. Med Educ 1990;24:1517.
8 Yates J, James D. Predicting the strugglers: a case-
control study of students at Nottingham University
Medical School. BMJ 2006;332:100913.
9 Yates J, James D. Risk factors for poor performance on
the undergraduate medical course: cohort study at
Nottingham University. Med Educ 2007;41:6573.
10 McManus C, Smithers E, Partridge P, Keeling A,
Fleming P. A-levels and intelligence as predictors of
medical careers in UK doctors: 20-year prospective
study. BMJ 2003;327:13942.
11 McManus C, Powis D, Wakeford R, Ferguson E, James
D, Richards P. Intellectual aptitude tests and A-levels
for selecting UK school-leaver entrants to medical
school. BMJ 2005;331:5559.
12 McManus C, Elder A, Champlain A, Dacre J, Mollon J,
Chis L. Graduates of different schools show substantial
differences in performance on MRCP(UK) Part 1 and
Part 2 and PACES examinations. BMC Med 2008.
http://www.biomedcentral.com/1741-7015/6/5.
[Accessed 23 April 2009.]
13 Greengross S. What patients want from their doctors.
In: Allen I, Brown P, Hughes P, eds. Choosing Tomorrows
Doctors. London: Policy Studies Institute 1997;129.
14 Morris J. The value and role of the interview in the
student admissions process: a review. Med Teach
1999;21:47381.
15 Powis D, Rolfe I. Selection and performance of medical
students at Newcastle, New South Wales. Educ Health
1998;11:1523.
16 Lemay J, Lockyer J, Collin T, Brownell K. Assessment of
non-cognitive traits through the admissions multiple
mini-interview. Med Educ 2007;41:5739.
17 Emery J.A report on the predictive validity of the BMAT
(2005) for first year examination performance on the
medicine and veterinary medicine courses at the
University of Cambridge. University of Cambridge Lo-
cal Examinations Syndicate. 2005.
18 McManus C, Woolf K, Dacre J. Even one star at A-level
could be too little too late for medical student selec-
tion. BMC Med Educ 2008;8:16. http://www.biomed
central.com/1472-6920-8-16. [Accessed 12 May 2010].
19 Department of Health. Medical Schools: Delivering the
Doctors of the Future. London: DoH 2004.
20 Hagenaars J, McCutcheon A, eds. Applied Latent Class
Analysis. Cambridge: Cambridge University Press 2002.
21 Rabe-Hesketh S, Skrondal A. Classical latent variable
models for medical research. Stat Methods Med Res
2008;17:532.
315
P Lambe & D Bristow
22 Universities and Colleges Admission Service (UCAS).
http://www.ucas.ac.uk/students/. [Accessed 12 May
2009.]
23 McHarg J, Bradley P, Chamberlain S, Ricketts C, Searle
J, McLachlan J. Assessment of progress tests. Med Educ
2005;39:2217.
24 Mattick K, Dennis I, Bradley P, Bligh J. Content speci-
ficity is it the full story? Statistical modelling of
a clinical skills examination. Med Educ 2008;42:58999.
25 Vermunt J, Magidson J. Technical Guide for Latent GOLD
4.0: Basic and Advanced. Belmont, MA: Statistical Inno-
vations, Inc. 2005.
26 Formann A, Kohlmann T. Latent class analysis in
medical research. Stat Methods Med Res 1996;5:179211.
27 Long J, Freese J. Regression Models for Categorical Depen-
dent Variables Using Stata. College Station, TX: Stata
Press 2003.
316 Blackwell Publishing Ltd 2011. MEDICAL EDUCATION 2011; 45: 308316
28 Powis D, Neame R, Bristow T, Murphy L. The objective
structured interview for medical student selection.
Br Med J (Clin Res Ed) 1988;296:7658.
29 Powis D, Waring T, Bristow T, OConnell D. The
structured interview as a tool for predicting premature
withdrawal from medical school. Aust N Z J Med
1992;22:6928.
30 Glick S. Selection for entry to medicine and specialist
training. Med Teach 2000;22:4437.
31 Wilkinson T, Frampton C. Comprehensive undergrad-
uate medical assessments improve prediction of clinical
performance. Med Educ 2004;38:11116.
Received 11 June 2010; editorial comments to authors 9 August
2010; accepted for publication 8 October2010