Journal of Vocational Behavior 74 (2009) 128133

Contents lists available at ScienceDirect

Journal of Vocational Behavior

journal homepage: www.elsevier.com/locate/jvb

Predictive validity of the medical specialty preference inventory

Kevin W. Glavin a,*, George V. Richard b, Erik J. Porfeli a
a
b

Northeastern Ohio Universities College of Medicine and Pharmacy, Department of Behavioral Sciences, 4209 St. Rt. 44, P.O. Box 95, Rootstown, OH 44272, USA
Association of American Medical Colleges, 2450 N Street, N.W., Washington, DC 20037, USA

a r t i c l e

i n f o

Article history:
Received 9 November 2008
Available online 6 December 2008

Keywords:
Predictive validity
Medical specialty preference inventory
Medical interests
MSPI
Medical specialty choice
Medical students

a b s t r a c t
Medical schools can assist students by providing them with quality career counseling to
help them choose a medical specialty. Many schools use interest inventories to help identify students specialty interests. This study examined the predictive validity of one such
inventory, the Medical Specialty Preference Inventory (MSPI). In a longitudinal design,
we used discriminant function analysis to examine how well students scores on the MSPI
t their chosen medical specialty one year later. The MSPI correctly predicted students
future medical specialty choice 58.1% of the time. These results can help career advisors
interpret MSPI scores, and identify students most likely medical specialty choice, as well
as their second most likely choice.
2008 Elsevier Inc. All rights reserved.

1. Introduction
Medical students specialty choice constitutes an important personal decision with far reaching consequences for individuals, and their families. Obtaining a medical degree requires the student to invest signicant personal and economic resourses, and to delay transitions to work and family roles. State governments invest enormous resources to subsidize medical
education at state institutions, and students mortgage their future to pay for the high costs of a medical education. Consequently, medical care providers, medical school faculty and staff, and students families encourage medical students to think
proactively and carefully about their future medical specialty choice.
Medical students choose a specialty for a variety of reasons, including experiences and exposure in medical school, academic performance in relevant clinical clerkships, personality attributes, and ratings of the content of medical practice (Reed,
Jernstedt, & Reber, 2001). Unfortunately, students must often make this life-changing choice with inadequate information.
Medical specialty counseling has developed over time to assist students in making specialty choices. Some career advisors
use the Medical Specialty Preference Inventory (MSPI: Zimny, 1979) to facilitate the decision-making process.
The MSPI is an assessment of students specialty interests and is used to inform their choice of medical specialty. In addition, the instrument can be used to assist physicians who are considering a medical specialty change at some point during
residency. The original MSPI consisted of 199 items pertaining to medical activities and settings. Zimny (1979) asked a sample of physicians in each of six major medical specialties (Family Medicine, Internal Medicine, Obstetrics and Gynecology,
Pediatrics, Psychiatry, and Surgery) to rate the extent to which each item from the MSPI was characteristic of general practice
in their specialties. This method avoided the inherent problem in using generic interests, which may co-vary across specialties, and therefore, fail to distinguish clearly between them. The authors considered this approach to be favorable because it
relates the characteristics of the practice of the specialties themselves rather than the characteristics of the physicians who

* Corresponding author.
E-mail address: kevinglavin@gmail.com (K.W. Glavin).
0001-8791/$ - see front matter 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jvb.2008.11.004

K.W. Glavin et al. / Journal of Vocational Behavior 74 (2009) 128133

129

practice in those specialties. Subsequently, it compares the medical activities that students prefer to the activities that physicians in that specialty perform.
This approach has received support from the literature. Savickas, Brizzi, Brisbin, and Pethtel (1988) compared the predictive validity of the MSPI with another medical specialty preference inventory, the Medical Specialty Preference Scales (MSPS:
Gough, 1979). Whereas the MSPI uses practitioner-based scales and contains items related to actual medical practice, the
MSPS uses student-based scales and contains generic interest items. The results from two consecutive medical student cohorts found that the MSPI successfully predicted a students specialty 59% of the time, whereas the MSPS successfully predicted a students specialty 19% of the time. Furthermore, the prediction rate for the MSPS only slightly exceeded what one
would expect to see by chance alone. These results suggest that practitioner-based scales containing items related to specic
medical practice predict students medical specialty choice better than student-based scales containing generic interest
items.
The current study employed the MSPI 2nd edition, which Zimny revised and updated in 2002 (Zimny, 2002). This revision
brought about a number of important changes including a reduction in the number of items and subscales. Although previous research demonstrated the strong predictive validity of the original version of the MSPI, the predictive power of the updated version has not been tested. Therefore, this study investigated the predictive validity of the MSPI 2nd edition.
2. Methodology
2.1. Participants
The participants were 506 medical students who completed the MSPI on the Association of American Medical Colleges
Careers in Medicine (CiM) website, which can be found at http://www.aamc.org/students/cim/. Careers in Medicine is a comprehensive career planning program available to all US and Canadian medical schools. To gain access to the site students
must obtain a free token from their school liaison, usually the schools associate or assistant dean of student affairs. At registration users are presented with an IRB-approved informed consent statement that indicates any data stored in the CiM site
may be used for research. While most schools actively provide access to the program, there are small number of schools that
do not participate as actively as others. Registered users of the site represent approximately 5565% of all enrolled medical
students in US and Canadian medical schools. Once registered, students are free to use all of the condential resources available on the site. Use of this site is voluntary as is the completion of the MSPI. The data were gathered from January 2005 to
December 2006, while the respondents were in their nal year of medical school, and contained 190 male students, and 316
female students. Respondents identied their race as follows: Caucasian (67%), Asian American/Pacic Islander (12%), African
American (9%), Indian American (7%), Native American (1%), Other (2%), and Unknown (2%). Upon completion of medical
school, and approximately one year later, students entered residency training for their chosen specialty. The time between
completing the MSPI and reporting residency choice was approximately one year. Information about specialty choice was
obtained from the AAMCs GME Track system containing resident census information for all training programs in the US.
Those who were specied as active residents in one of six medical specialties (Family Medicine, Internal Medicine, Obstetrics
and Gynecology, Pediatrics, Psychiatry, and Surgery) were selected for the purposes of this study.
2.2. Instrument
The MSPI measures medical interests, and is used to assist students in choosing a specialty. A 2002 revision of the MSPI
resulted in a reduction in the number of items to 150. Of the 150 items, 104 are used for scoring purposes. The remaining 46
provide additional data intended for use in constructing future scales. MSPI items reect job specic tasks that relate directly
to medical practice. Students rate each item on a seven-point scale which reects their degree of desirability for each item. A
score of 12 indicates low desirability, 35 indicates moderate desirability, and 67 indicates high desirability. Differences
between students subscale scores and specialists scores determine students preference for each medical specialty. Students
whose subscale scores are similar to specialist scores for the same subscales will receive a high preference score for that particular specialty. Scores are calculated for: Family Medicine (FAM), Obstetrics Gynecology (OBGYN), Surgery, Psychiatry
(PSY), Pediatrics (PED), and Internal Medicine (MED). Higher scores indicate a greater preference for a specialty, while lower
scores indicate less preference for a specialty. The instrument was self-administered on the internet, and preference scores
were reported to participants immediately upon completion.
Zimny (1979) reported on the reliability and validity indices used in the initial development of the MSPI. Reliability was
estimated in two separate analyses. In the rst analysis, using the Spearman Brown formula, reliability estimates ranged
from .74 (Pediatrics) to .93 (Family Medicine), and in the second analysis estimates ranged from .66 (Pediatrics) to .91 (Surgery). Most reliability estimates in both analyses fell in the .80s and .90s. Zimny also conducted a study on the predictive
validity of the MSPI using National Resident Matching Program (NRMP) data (1980). The NRMP contains data about the initial specialty choice of medical students as they enter into residency training. He found that the level of predictive accuracy
over all specialties ranged from about 50 to 55%. This range represents a level of prediction well above the conservative
chance expectancy level of 17% accuracy, and indicates that a substantial relationship exists between specialty preference
scores on the MSPI and subsequent specialty choice.

130

K.W. Glavin et al. / Journal of Vocational Behavior 74 (2009) 128133

To test the predictive validity of the MSPI, scores for each of the six medical specialties were compared to students residency choices. Six v2-tests were conducted to test the predictive validity of the six MSPI scale scores. The general hypothesis
was that students pattern of scores across the six MSPI scores would be predictive of their medical specialty choice one year
later. Students were classied into three MSPI groups for each of the six specialties. Group one included those students
whose score for a particular specialty was their highest score relative to the other ve specialties, and this score was at least
3.5 points greater than their next highest scale score. The researchers used 3.5 points because this represented approximately half of the average difference between students highest and second highest scores. Group two included those students whose score for a particular specialty was their highest score relative to the other ve specialties, but this score
was not at least 3.5 points greater than their next highest scale score. Group three included the remaining students. The
aim was to discern if students choose, for example, family medicine for their residency with greater frequency if they scored
highest on the family medicine subscale of the MSPI, and if the frequency of this specialty choice increased as this score became more pronounced (by more than 3.5 points) relative to the other ve scale scores.
3. Results
Results showed that the overall hit rate (i.e., rate of correctly classied students based upon their interests and specialty
choice), for groups one and two combined, was 46%. The overall hit rates by medical specialty were as follows; Family Medicine 41%, OBGYN 53%, Pediatrics 46%, Psychiatry 56%, Internal Medicine 27%, Surgery-General 88%. Of the participants who
chose Family Medicine as their medical specialty, 27.7% of them met both criteria (i.e., group 1), 13.3% met the second criteria only (i.e., group 2), and the remainder, 59%, met neither criteria (i.e., group 3), v2 (2, N = 506) = 81.09 p < 0.001. Of the
participants who chose OBGYN as their medical specialty, 44.2% of them met both criteria, 9.3% met the second criteria only,
and the remainder, 46.5%, met neither criteria, v2 (2, N = 506) = 82.92 p < 0.001. Of the participants who chose Pediatrics as
their medical specialty, 23.2% of them met both criteria, 23.2% met the second criteria only, and the remainder, 53.7%, met
neither criteria, v2 (2, N = 506) = 60.9 p < 0.001. Of the participants who chose Psychiatry as their medical specialty, 48.8% of
them met both criteria, 7% met the second criteria only, and the remainder, 44.2%, met neither criteria, v2 (2,
N = 506) = 172.94 p < 0.001. Of the participants who chose Internal Medicine as their medical specialty, 11.2% of them
met both criteria, 15.4% met the second criteria only, and the remainder, 73.4%, met neither criteria, v2 (2,
N = 506) = 59.36 p < 0.001. Of the participants who chose Surgery-General as their medical specialty, 78.1% of them met both
criteria, 9.6% met the second criteria only, and the remainder, 12.3%, met neither criteria, v2 (2, N = 506) = 117.9 p < 0.001.
Although these v2 results are promising, they mainly speak to bivariate associations between categories of MSPI subscale
scores and specialty choice. They do not permit a simultaneous comparison of the discriminatory power of the six subscales
in combination. Discriminant function analysis was employed to test if, and to what extent, the entire MSPI predicts medical
specialty choice and to what extent each of the subscales contributes to its discriminatory power while controlling for the
others.
Fig. 1 shows the mean scores for each of the subscales for each of the medical specialty choices. Each line represents residents who chose that specialty, and their mean scores on each of the six preference scales. For example, students who chose

75.00

Preference Score

70.00

Medical Specialty
Choice

65.00

FAM
MED

60.00

OBGYN
PED

55.00

PSY
SUR

50.00
45.00
FAM

MED

OBGYN

PED

PSY

SUR

Medical Specialty Preference Scales

Fig. 1. Mean scores for medical specialty preference scales by specialty choice. *FAM, Family Medicine; MED, Internal Medicine; Pediatrics, Pediatrics,
Psychiatry, Psychiatry, SUR, Surgery-General.

131

K.W. Glavin et al. / Journal of Vocational Behavior 74 (2009) 128133

Family Medicine, on average, scored 67.73 on Family Medicine, 65.1 on Internal Medicine, 67.13 on OBGYN, 67.43 on Pediatrics, 55.68 on Psychiatry, and 60.35 on Surgery-General. The most differentiated group proles occur for those students
who selected Psychiatry and Surgery-General specialties. Students who chose Psychiatry tended to score highest on the Psychiatry subscale and lowest on the Surgery-General subscale, relative to the other four subscales, and relative to all the other
students in the other medical specialties. The least differentiated prole occurred for Family Medicine. Students choosing
OBGYN and Pediatrics had the most elevated scores, while students choosing Psychiatry had the least elevated scores. Medical residents in Family Medicine and Internal Medicine demonstrated mid-range elevation scores, when compared to the
residents in other medical specialties.
A direct discriminant analysis was performed using six interest variables as predictors of membership in six medical specialty groups. Predictors were scores on Family Medicine, Internal Medicine, OBGYN, Pediatrics, Psychiatry, and SurgeryGeneral MSPI scales. The criterion variable was the medical specialty the participant entered. The researchers used 70% of
the original dataset to run the discriminant analysis, and reserved the remaining 30% for cross-validation purposes. No cases
were found to be missing data, n = 365. The results yielded ve signicant discriminant functions, with a combined v2
(30) = 613.29, p < 0.001. Functions two through ve yielded a v2 (20) = 364.293, p < 0.001. Functions three through ve
yielded a v2 (12) = 227.175, p < 0.001. Functions four through ve yielded a v2 (6) = 114.957, p < 0.001. Function ve yielded
a v2 (2) = 34.107, p < 0.001.
Table 1 displays bivariate correlations, means, and standard deviations for each of the predictor variables. Strong intercorrelations exist between Family Medicine and Internal Medicine (.81), and OBGYN and Pediatrics (.75). These correlations
give cause for concern because they violate one of the assumptions of discriminant function analysis. The effect sizes for the
discriminant functions were as follows; .5, .32, .27, .2, and .09, for functions 1, 2, 3, 4, and 5, respectively. Therefore, the ve
functions accounted for 50%, 32%, 27%, 20%, and 9% of the total relationship between predictors and groups. The amount of
between-group variance accounted for by each function was as follows; 45.8%, 21.3%, 16.8%, 11.6%, and 4.6% for functions 1,
2, 3, 4, and 5, respectively.
The classication results displayed in Table 2 demonstrate that the MSPI may be used to correctly classify students specialty choice 58.1% of the time, which exceeds what we would expect to see by chance alone (i.e., 20%). The results show that
the MSPI best predicts participants who chose Surgery-General, with a 74% correct classication rate, followed by psychiatry,
with a hit rate of 71%. Although the MSPIs predictive power is weakest for family medicine, the hit rate (47%) remains well
above chance. The pattern of misclassications for Family Medicine and Pediatrics specialties suggest that students who select these specialties may exhibit similar patterns of interests. The same can be said for Surgery-General and Internal Medicine. The patterns of misclassications were inconsistent for Psychiatry relative to any other discipline, which suggests that
participants who chose this specialty appear to have a relatively unique interest prole. The stability of the classication procedure was checked by a cross-validation run. Approximately 30% of the cases were withheld from calculation of the classication functions in this run. For the 70% of the cases from which the functions were derived, there was a 58% correct
classication rate. For the cross-validation cases, classication was 57%. This result suggests there exists a high degree of
consistency in the classication scheme, and suggests that the results are not dependent on the sample data used to test
the discriminatory power of the MSPI.
4. Discussion
This study examined the predictive validity of the MSPI, which demonstrated an overall hit rate of approximately 46%
when using a v2 analysis. The highest hit rate occurred for Surgery-General, and the lowest hit rate occurred for Psychiatry.
These ndings are signicantly better than what we would expect to see by chance alone, which was determined to be 20%.
Interest inventories based on six categories generally report hit rates between 35 and 45% (Holland, Magoon, & Spokane,
1981).
The researchers used discriminant function analysis to move beyond simple bivariate associations to test the simultaneous discriminatory power of the six MSPI scales to predict medical specialty choice. Fig. 1 demonstrates that residents
within the six medical specialties typically exhibited unique patterns of preference scores based on the elevation and variability of those scores. Table 2 suggests that in most specialties there exists a secondary, or perhaps tertiary, specialty that
is typically chosen by participants who do not choose the predicted specialty on the basis of their interests. One can think of

Table 1
Inter-correlations, means, and standard deviations for medical specialty preference scores.

1.
2.
3.
4.
5.
6.
*

Family
Internal
OBGYN
Pediatrics
Psychiatry
Surgery
p < .001.

Mean

SD

60.37
63.38
65.97
65.65
55.11
64.33

11.20
8.81
7.63
7.20
10.34
10.77

1.00

0.81*
1.00

0.48*
0.36*
1.00

0.48*
0.45*
0.75*
1.00

0.35*
0.24*
0.31*
0.47*
1.00

0.00*
0.37*
0.06
0.00
-0.55*
1.00

132

K.W. Glavin et al. / Journal of Vocational Behavior 74 (2009) 128133

Table 2
Classications by medical specialty.
Predicted Medical specialty

Chosen medical specialty

Family Medicine
Internal Medicine
OBGYN
Pediatrics
Psychiatry
Surgery-General

Family Medicine

Internal Medicine

OBGYN

Pediatrics

Psychiatry

Surgery-General

46.7
8.3
8.6
15.2
17.6
2

10
55.8
8.6
12.1
0
14

8.3
7.5
60
6.1
2.9
6

18.3
12.5
5.7
53
8.8
4

6.7
2.5
0
4.5
70.6
0

10
13.3
17.1
9.1
0
74

Note: Numbers are row percentages and bolded numbers reect the percent of correct classications for each medical specialty.

those secondary and tertiary specialties as a typical alternative. For example, while participants who scored highest in Family Medicine chose to enter Family Medicine 46.7% of the time, another 18.3% actually entered Pediatrics. This makes intuitive sense given that both Family Medicine and Pediatrics share some commonalities, such as the breadth of problems they
deal with, and in the role they play as primary-care providers. Participants with high OBGYN scores chose OBGYN 60% of the
time, but another 17.1% chose to enter General-Surgery. Both of these specialties share commonality in their surgical applications, and so these results make sense also. While most of the specialties seem to have interest companions, psychiatry
seems to be fairly distinct with respect to interests. On a larger level, these results hint at an underlying conceptual structure
of medical specialties on the basis of interest that could be explored in future research.
The two discriminant functions accounted for over 67% of the between-group variance for the six medical specialties, and
further help to show how the specialties are similar, or dissimilar, to each other. The rst function best differentiates between Surgery-General and Psychiatry. We would expect to see a statistically derived difference in the interest proles of
students choosing these two specialties because of the different activities each entail. Whereas Psychiatry involves a heavy
emphasis on interpersonal issues, and relies on talking cures (i.e., counseling practices) to address patients health disturbances, Surgery-General generally focuses less on interpersonal issues, and relies more heavily on physical interventions to
address health disturbances. As mentioned earlier, Psychiatry appears to share the least amount of interests with the other
ve specialties. Psychiatrists concentrate on mental health issues, whereas the other specialties attend to physical ailments.
Family Medicine and Pediatrics appear to be the least differentiated. Both specialties address a wide range of health problems, and only differ in the population of patients they see. Family physicians attend to all types of patients, while Pediatricians primarily treat children, but both ultimately interact with all family members. OBGYN appears to be different from the
other specialties, possibly because this specialty attends to a specic set of health problems and mainly treats females. The
other specialties may see a greater variety of patients and health problems.
The results of this study have implications for career practitioners who advise medical students. Practitioners should rst
examine students MSPI scores for each of the medical specialties and identify their highest score. Next, they should determine the distinctiveness of the highest score, by subtracting the second highest score from the highest score. If the highest
score is at least 3.5 points greater than the second highest score, students should be advised to consider the medical specialty
associated with their highest score. If the difference between the highest score and the second highest score is less than 3.5
points, students should be advised to consider the next most likely alternative. Practitioners can use the results displayed in
Table 2 to identify which medical specialty students will choose if they do not choose the specialty they were predicted to
enter. For example, students who do not enter Family Medicine are most likely to enter Pediatrics.
A case study using a students scores, and chosen medical specialty, helps to illustrate how practitioners can interpret
MSPI proles. The MPSI scores for a female student were as follows: Family Medicine = 83, Internal Medicine = 70, OBGYN = 63, Pediatrics = 62, Psychiatry = 57, and Surgery-General = 52. This student chose to enter Family Medicine. The MSPI
results show that the student scored highest on Family Medicine, and that this score is at least 3.5 points greater than the
second highest score. Therefore, this student would be best advised to consider choosing Family Medicine. Table 2 suggests
that individuals who score high on Family Medicine, but do not choose it, tend to choose Pediatrics.
Another case study illustrates a student whose highest MSPI score did not match his chosen medical specialty. The MPSI
scores for a male student were as follows: Family Medicine = 71, Internal Medicine = 74, OBGYN = 55, Pediatrics = 53, Psychiatry = 42, and Surgery-General = 65. This student chose to enter Surgery-General, despite the fact that this is his third highest
score. The MSPI results show that the student scored highest on Internal Medicine. However, the difference between the
highest and second highest scores is less than 3.5 points. This student did not enter Internal Medicine. Table 2 suggests that
individuals who score highest on Internal Medicine, but do not choose it, tend to choose Surgery-General.
4.1. Limitations
The results presented herein are limited to medical students who have made a specialty choice based on one of the six
major specialties, which include; Family Medicine, Internal Medicine, Pediatrics, OBGYN, Surgery-General, and Psychiatry.
These specialties only account for about 60% of all possible medical specialty choices. Future research should examine the
predictive validity of the MSPI for a larger number of medical specialty choices. The results are also limited to the degree

K.W. Glavin et al. / Journal of Vocational Behavior 74 (2009) 128133

133

that we only studied students in their rst year of residency. This presents a problem because some students, especially
those in Internal Medicine, use a specialty as an entry point for another specialty. Therefore, follow-up studies should compare MSPI scores to students medical specialty choice in their second and third years of residency. The self-selection process
by which students participated in this research may have impacted the results. Although the MSPI is available to all medical
students, on average, only 52% of students in each cohort completed the instrument. The researchers are aware that this process of self-selection may lead to a sample that is not representative of the overall population. There may be factors that differentiate those students who participated from those who did not, and these factors may have inuenced the results.
References
Gough, H. (1979). Medical specialty preference scales: A report for counselors. Palo Alto, CA: Consulting Psychologists Press.
Holland, J. L., Magoon, T. M., & Spokane, A. R. (1981). Counseling psychology: Career interventions, research, and theory. Annual Review of Psychology, 32,
270305.
Reed, V. A., Jernstedt, G. C., & Reber, E. S. (2001). Understanding and improving medical student specialty choice: A synthesis of the literature using decision
theory as a referent. Teaching and Learning in Medicine, 13(2), 117129.
Savickas, M. L., Brizzi, J. S., Brisbin, L. A., & Pethtel, L. L. (1988). Predictive validity of two medical specialty preferences inventories. Measurement and
Evaluation in Counseling and Development, 21, 106112.
Zimny, G. H. (1979). Manual for the medical specialty preference inventory. St. Louis, MO: St. Louis University School of Medicine.
Zimny, G. H. (1980). Predictive validity of the medical specialty preference inventory. Medical Education, 14, 414418.
Zimny, G. H. (2002). Updating the medical specialty preference inventory. St. Louis University School of Medicine. Unpublished manuscript.