Allocation of residency training position in Spain: contextual effects on speciality preferences

ALLOCATION OF RESIDENCY TRAINING POSITIONS IN SPAIN: CONTEXTUAL EFFECTS ON SPECIALTY PREFERENCES JEFFREY E. HARRIS!a,* BEATRIZ G. LOPEZ-VALCARCEL!b PATRICIA BARBER!b VICENTE ORTÚN!c 21-Dec-2014 a. Department of Economics, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. Fax: +1 617 253 6915. b. Department of Quantitative Methods in Economics and Management, University of Las Palmas de Gran Canaria, Las Palmas, Spain. c. Faculty of Economic and Business Sciences, Universitat Pompeu Fabra, Barcelona, Spain. *Correspondence to: Department of Economics, Massachusetts Institute of Technology, Cambridge MA 02139, USA. E-mail: jef[email protected]. KEY WORDS: medical residency; mixed logit; physician income; professional prestige; MIR system; independence of irrelevant alternatives RUNNING HEAD: Medical Specialty Preferences in Spain WORD COUNT: 4,998!TABLE COUNT: 3!FIGURE COUNT: 2 APPENDIX TABLE COUNT: 2 FUNDING SOURCE: This project was funded by the Spanish Ministry of Science and Innovation through Grant ECO2010–21558 to BGLV as the principal investigator. !! !! ABSTRACT In Spain’s “MIR” system, medical school graduates are ranked by their performance on a national exam and then sequentially choose from the available residency training positions. We took advantage of a unique survey of participants in the 2012 annual MIR cycle to analyze preferences under two different choice scenarios: the residency program actually chosen by each participant when it came her turn (the “real”); and the program that she would have chosen if all residency training programs had been available (the “counterfactual”). Utilizing conditional logit models with random coefficients, we found significant differences in medical graduates’ preferences between the two scenarios, particularly with respect to three specialty attributes: work hours/lifestyle, prestige among colleagues, and annual remuneration. In the counterfactual world, these attributes were valued preferentially by those nearer to the top, while in the real world, they were valued preferentially by graduates nearer to the bottom of the national ranking. Medical graduates’ specialty preferences, we conclude, are not intrinsically stable, but depend critically on the “rules of the game.” The MIR assignment system, by restricting choice, effectively creates an externality in which those at the bottom, who have fewer choices, want what those at the top already have. (199 words) JEL Classification: I11, I18, C25 MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 2 1. INTRODUCTION Studies of physician specialty choice have generally pursued a common research strategy. Enumerate the specific attributes to be studied and then use data on prospective or recent medical graduates’ survey responses, experimental decisions or choices of residency programs to determine quantitatively which attributes are most important. The list of attributes considered by researchers is extensive: the length of the residency program, the anticipated debt upon completion of training, hospital versus ambulatory orientation, the expected financial remuneration, life style and work hours, prestige among colleagues or the general public, employability, malpractice litigation risk, direct patient interaction and continuity of care, research and teaching opportunities, and potential for career advancement (Dorsey et al., 2003; Gagne and Leger, 2005; Goldacre et al., 2010; Harris et al., 2013; Harris et al., 2005; Heikkila et al., 2011; Hurley, 1991; Nicholson, 2002; Rosenthal et al., 1994; Sivey et al., 2012; Thornton, 2000; Thornton and Esposto, 2003). In this article, we suggest that this research paradigm is inadequate. We posit that the institutional rules for allocating medical school graduates to different specialties – what we call “contextual effects” – can themselves alter individuals’ preferences. We focus on the current national system for allocating residency training positions in Spain, widely known as “MIR,” in which medical school graduates are ranked nationally and then sequentially choose from the available training positions. We take advantage of a unique survey of participants in the 2012 annual MIR cycle to analyze preferences under two different choice scenarios: the residency program actually chosen by each participant when it came her turn (the “real world”); and the program that she would have chosen if all residency training programs had been available (the “counterfactual world”). Utilizing conditional logit models with random coefficients, we find significant differences in medical graduates’ preferences between the two scenarios, particularly with respect to three specialty attributes: work hours/lifestyle, prestige among colleagues, and annual remuneration. In the counterfactual world, these attributes were valued preferentially by those nearer to the top, while in the real world, they were valued preferentially by graduates nearer to the bottom of the national ranking. Medical graduates’ specialty preferences, we conclude, are not intrinsically stable, but depend critically on the “rules of the game.” The MIR assignment system, by MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 3 restricting choice, effectively creates an externality in which those at the bottom, who have fewer choices, want what those at the top already have. 2. SPAIN’S MIR SYSTEM OF ALLOCATING RESIDENCY TRAINING POSITIONS The allocation of residency training positions in Spain is organized and regulated at the national level by a system widely known as MIR, which stands for “médico interno residente,” literally “resident medical intern.” On an annual basis, the central government’s Ministry of Health authorizes postgraduate training programs in 47 specialties. To be eligible for a residency training position, each “candidate” must have an approved diploma from a Spanish or foreign medical school and take a national examination. Candidates are then ranked on the basis of their MIR combined score, which is a weighted average of their national exam score (90%) and medical school grade point average (10%). Once all training positions are authorized and all candidates are nationally ranked, the final phase of the annual MIR cycle functions essentially as a one-sided sequential allocation mechanism or “serial dictatorship,” in which the training programs play only a passive role (Harris et al., 2014). The top-ranked candidate chooses her preferred residency training position from the entire set of nationally available training programs. Then the second-ranked candidate chooses from the remaining available residency positions, and the process continues iteratively until all training positions are exhausted or all candidates have elected positions. The national rank ordering is a critical element of the MIR allocation scheme. With each annual MIR cycle, the graduates of the nation’s top medical schools consistently attain the highest combined scores and thus get their first choices among the most highly valued residency training programs in such sought-after specialties as plastic surgery, dermatology and cardiology (Gonzalez Lopez-Valcarcel et al., 2013; Lopez-Valcarcel et al., 2013). At the bottom of the national ranking, the residual claimants are left with a Hobson’s choice between enrolling in a residency in family and community medicine or dropping out in order to retake the national exam the following year (Gonzalez Lopez-Valcarcel et al., 2011). In the 2002 MIR cycle, as shown in Figure 1, only 200 candidates within the top-ranked 3,000 chose a residency position in family and community medicine, and by the 2012 cycle, only 50 had done so. [FIGURE 1 ABOUT HERE.] MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 4 3. DATA Our principal database consisted of the individual assignments to residency training programs for all candidates participating in the 2012 MIR nationwide competition (for short, the “2012 MIR registry”). This database was provided by Spain’s Ministry of Health, Social Services and Equality. For each candidate, the registry contained: the candidate’s national ranking (an ordinal number ranging from 1 up to the total number of participants); the residency program chosen (including medical specialty and training center); the candidate’s residential postal code, sex, nationality, and medical school attended, including foreign medical schools. In the 2012 MIR competition, a total of 231 training centers offered residency positions in one or more of 44 specialties.1 For hospital-based specialties, such as cardiology, neurology and urology, these training centers were located in 181 different hospitals. For non-hospital based specialties, such as family medicine,2 occupational medicine, public health, and some psychiatry training programs, we grouped the training centers according to 50 provinces throughout the country. Together, these training centers offered a total of 2,527 distinct residency training programs, classified by specialty and center. With each residency training program offering multiple positions, there was a grand total of 6,555 available residency positions nationwide. Among these, 1,860 (28.4%) were residency positions in the specialty of family medicine. Initially, a total of 11,713 medical graduates passed the MIR exam in order to be eligible to opt for one of the 6,555 training positions. Of these candidates, 5,158 (44.0%) withdrew from the competition without choosing a residency, in many cases because their test scores were so low that they had no chance of choosing their desired specialty or training center. That left 6,555 candidates for exactly as many training positions. A total of 4,839 (73.8%) of these participating candidates were Spanish nationals, while the remaining 1,716 (26.2%) were foreign nationals. MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 5 1 While the Ministry of Health authorizes a total of 47 specialties, we excluded the school-based specialties of forensic medicine, sports medicine, and medical hydrology, which together accounted for only 149 candidates in the 2012 MIR competition. Rather than receiving salaries as medical residents, trainees in these three specialties pay tuitions to attend a professional school. They ranked at the bottom in the MIR competition. 2 Formally, the specialty is known as “family and community medicine” (or “medicina familiar y comunitaria”), but for brevity we shall sometimes refer to it simply as “family medicine.” We supplemented the 2012 MIR registry with additional data on the characteristics of the 231 training centers. These included information on hospitals’ bed capacity and high-technology facilities, derived from the official National Catalog of Hospitals of Spain 2011. We also used Google Maps to create a 50 ×50 matrix of travel times between the capital of a candidate’s province of residence and the capital of the province in which each training center was located. We matched the 2012 MIR registry database with two cross-sectional surveys specifically designed for this research project. The first was a survey of students in their final semester of medical school in Spain, administered in April 2011 (Harris et al., 2013). This survey provided us with the perceived values of seven key attributes of each specialty, as described in Appendix Table A. These included: the probability of obtaining employment ( X1 ), favorable working hours and working conditions ( X2 ), recognition by patients ( X3 ), prestige and recognition by colleagues ( X4 ), possibilities for advancement and professional development ( X5 ), average annual remuneration ( X6 ), and proportion of income derived from private practice ( X7 ). In earlier work (Harris et al., 2013), we found that employability ( X1 ) had a significant impact on specialty choice in the context of Spain’s economic crisis. In a nationalized healthcare system such as Spain, where employed physicians receive salaries negotiated through collective bargaining, the extent to which a physician can engage in outside private practice ( X7 ) is a superior proxy for earnings ( X6 ). As a measure of earnings, outside income is more orthogonal to the other specialty characteristics than total earnings. There is good evidence that professional prestige ( X4 ) is an important driver of the decision not to pursue a career in primary care medicine, independent of income (Kolstad, 2013). The second cross-sectional survey (for short, the “2012 post-MIR survey”) was performed in May 2012 on those candidates who had just been assigned to residency positions in the 2012 MIR competition. These candidates belonged to the same cohort that answered the first survey as medical students in 2011. We asked them not only about their actual specialty/center selection, but also about their preferred choice if they had been ranked first in the competition. We refer to the latter as a candidate’s “counterfactual” choice. We used respondents’ e-mail addresses, voluntarily provided under our assurance of anonymity, to match respondents with the MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 6 principal database. We thus had additional data from this survey on 3,432 (or 52.4%) of the 6,555 candidates in the 2012 MIR registry. In our econometric models, to be described below, we dropped observations with missing values, leaving an estimation sample of 6,254 MIR candidates, of whom 3,117 (50%) were matched to the survey database and answered the counterfactual question on their preferred specialty and training center. Table I displays comparative statistics for the 2012 MIR registry and the 2012 post-MIR survey, including tests for comparison of group means. While there were no significant differences with respect to Spanish nationality or gender, those candidates responding to the post-MIR survey performed better on the national exam and tended to choose residencies closer to their residence. [TABLE I ABOUT HERE.] For the 3,117 candidates with complete data in both databases, Table II compares their actual specialty choices in the MIR competition with their counterfactual ones. For a highly preferred specialty with a limited number of training positions, such as plastic surgery, only a small fraction of candidates were assigned to their preferred choices. By contrast, virtually every one of the candidates who preferred family medicine was assigned to his top choice. Those candidates who preferred cardiology, dermatology and plastic surgery tended to have high MIR rankings, so that few ended up assigned to family medicine. [TABLE II ABOUT HERE.] In our econometric analysis of the 2012 MIR registry, the dependent variable was the training program actually chosen by each candidate (the real choice). In the 2012 post-MIR survey, the dependent variable was the training program that the candidate would have made if he had been top ranked (the counterfactual choice). In both analyses, the explanatory variables included the seven key attributes of the specialty; the characteristics of the training center (number of beds, and availability of positron emission tomography (PET) in the affiliated hospital); the characteristics of the candidate, including gender, nationality and MIR ranking, interacted with specialty attributes; the distance from the candidate’s residence to the training center, measured in minutes of travel time between provincial capitals; and an indicator variable MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 7 equal to 1 if the candidate’s residence and the training center were located in the same province. The complete list of explanatory variables is shown in Appendix Table B. 4. DISCRETE CHOICE MODELING Our estimation strategy was based upon a conditional logit model with random parameters, also called the “mixed logit model” (McFadden and Train, 2000; Train, 2009). To analyze the data from the 2012 MIR registry, we accounted for the endogeneity of the choice set available to each candidate as her turn came up in the MIR sequence. To analyze the data from the 2012 post-MIR survey, where respondents were asked to designate their preferred specialty under the counterfactual assumption that they had been ranked first, we assumed all candidates had the same, complete choice set. We let P=pjj=1,…,M { } denote the set of distinct residency training programs, each identified by a particular specialty and location, and let mj denote the total number of training positions available in program pj . Let C=cii=1,…,N { } denote the set of candidates participating in the MIR sequential assignment process. We first consider estimation of the mixed logit model with an endogenous choice set. We assume that the candidates are already ordered so that candidate c1 is first to elect a training program, while candidate cN is last. Let P i denote the set of distinct residency training programs available to candidate ci when it is her turn to elect a program, and let mij denote the number of remaining unassigned positions in training program pj available to candidate ci when her turn comes up. Since the top-ranked candidate c1 can choose any program, we have P 1=P and m1j=mj . If candidate ci chooses a position in training program pj , the number of positions available in that training program is decremented by 1, that is, mi+1, j=mij −1 , while the number of positions available in all other training programs remains unchanged, that is, mi+1,k=mik for all k≠j . If candidate ci chooses the last available slot in training program pj , that is, mij =1 , MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 8 then P i+1=P i−pj { } and that training program is dropped from the choice sets P i+1,P i+2,… of all remaining candidates. Let yi denote the training program chosen by candidate ci . Given the unobserved parameter vector β , the probability that candidate ci choses training program yi=pj∈P i is given by the conditional logit model (1)! Pr yi=pjXij , β { } =eXij β eXik β pk∈P i ∑=Lij β ( ) where Xij is a vector of observed characteristics that candidate ci associates with program pj . Some elements of Xij may depend only on the specific training program and not on the candidate, such as the particular specialty or the facilities of the training center. Other elements may depend on both the program and the candidate, such as the distance of the candidate’s home province from the training center. Included in the latter category are interactions between a candidate’s characteristics (e.g., gender, nationality, ranking) and a program’s characteristics. We further assume that the vector of unobserved parameters β has a multivariate normal distribution βµ ,Σ~N µ ,Σ ( ) with density function φ β µ ,Σ ( ) . To simplify the notation, we let θ = µ ,Σ ( ) . Conditional on θ , the probability that candidate ci choses training program yi=pj∈P i is therefore given by3 (2)! Pr yi=pjXij , θ { } =Lij β ( ) φ β θ ( ) d β ∫=Lij θ ( ) As noted by Train (Train, 2009) and others, the parameters θ = µ ,Σ ( ) can be estimated by maximum likelihood, where the integral in (2) can be computed by simulation. Let ˆ θ denote MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 9 3 In (1) and (2), we have intentionally overloaded the notation Lij ⋅ ( ) , where Lij β ( ) refers to the logit probability conditional on β , while Lij θ ( ) refers to the probability conditional on θ . of IIA based upon the lack of independence of the error terms eij across candidates, namely, Cov eik ,ejk ( ) ≠0 for i≠j . 6.1. Limitations of this Study This study has a number of limitations. First, we relied upon a one-time survey of students in their final year of medical school (Harris et al., 2013) to ascertain the values of the seven specialty attributes ( X1 through X7 ). While the survey had a wide coverage of all 27 of Spain’s medical schools, we do not know whether candidates’ valuations remained stable during the nearly one-year period between the date of the survey (April 2011) and the moment of decision when they had to make their commitments (March 2012). The intervening time typically spent in post-graduate courses preparing for the national exam may have altered candidates’ perceptions of their specialty choices. Second, only half of the participants in the 2012 MIR cycle responded to our post-2012 MIR survey, and those who did respond had a somewhat higher MIR ranking (Table 2). Still, the concordance of parameter estimates between Models I and II points to the absence of selection bias among the 2012 post-MIR survey respondents. Third, we cannot draw strong conclusions about the generalizability of our results. During the past decade, the number of candidates in the first half of the MIR rankings who chose family medicine declined by over 75 percent (Figure 1). While the recent economic crisis and the resulting concerns about employability may have helped to stabilize preferences (Harris et al., 2013), we cannot state with precision that our results will be applicable in the future. Fourth, we do not have adequate explanations for some of our counterintuitive results. In all three models of specialty choice, recognition by patients ( X3 ) had a significant negative coefficient, while prestige among colleagues ( X4 ) had a significantly positive coefficient. While our models may have correctly identified a genuine distinction that prospective physicians made between the two attributes, there remains the concern that recognition by patients was correlated by some unobserved characteristic that deterred candidates from choosing a training program. Similarly, the coefficients of annual remuneration after 10–15 years of experience ( X6 ) were negative with marginal significance in Model I and insignificant in Models II and III. The MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 16 significant positive interactions between X6 and Ranking suggest that those candidates at the bottom of the MIR selection queue do, in fact, value their annual income. However, in Spain’s nationalized system where most physicians are salaried, the proportion of income from private practice ( X7 ) is likely to be a superior indicator of differential compensation among specialties (Harris et al., 2013). Even in the U.S., where private fee-for-service medical practice remains highly prevalent, many medical students do not appear to have accurate knowledge of differential compensation (Nicholson, 2005). Fifth, in our graphical candidate-level comparisons of preferences in the real and counterfactual worlds, we relied upon the posterior mean values of the coefficients ˆ β i , based on the posterior density function given in equation (3) above. While this practice appears to be commonplace in the application of mixed logit models (Hastings et al., 2010), it understates the degree of uncertainty in our results. An alternative would be to simulate repeated draws from the posterior density of ˆ β i (Revelt and Train, 2000). 6.2. Policy Implications Numerous authors in many countries have lamented the shortage of primary care physicians (Barber and Lopez-Valcarcel, 2010; Bodenheimer, 2006; Colwill et al., 2008; Huibers et al., 2009; Mariolis et al., 2007; Rosenblatt et al., 2006; Scott et al., 2006; Steinbrook, 2009; Thistlethwaite et al., 2008). Numerous corrective measures have been proposed, including changes in physician compensation, improvements in working conditions, policies to counter the low prestige of primary care medicine, and the training of non-physician practitioners (Dorsey et al., 2003; Gagne and Leger, 2005; Gonzalez Lopez-Valcarcel and Barber Perez, 2012; Goroll et al., 2007; Krueger and Halperin, 2010; Ortun et al., 2008; Sivey et al., 2010; Thornton and Esposto, 2003). Our study has focused on the critical juncture where medical school graduates choose residency training positions. So long as the MIR selection system remains essentially intact, our empirical results for real-world Models I and II capture the local effects of changes in the key specialty attributes X1 through X7 . Our results imply that increased professional prestige and MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 17 financial remuneration could increase the likelihood that a qualified medical school graduate will elect a career in family medicine. Our finding of contextual effects suggests that the institutions designed to allocate residencies are not neutral with respect to preferences. In the MIR system, the choices made by those candidates at the top may not simply remove options from the remaining candidates’ choice sets. They may change the remaining candidates’ preferences as well. This raises the possibility that the MIR system itself has exacerbated the shortage of qualified primary care physicians in Spain. Our results suggest a strategy for studying the effects of alternative public policies, based upon simulation of the choice models estimated here. In the evaluation of alternative policies that depart from the current MIR allocation, it will be important to consider not only the potential efficiency gains from assigning more qualified candidates to primary care, but also the potential equity losses from preventing those who scored highest to choose first (Harris et al., 2014). MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 18 FUNDING This project was funded by the Spanish Ministry of Science and Innovation through Grant ECO2010–21558 to BGLV as the principal investigator. The funder had no influence in the conduct of this study or the drafting of this manuscript. ACKNOWLEDGMENTS We gratefully acknowledge the Ministry of Health, Social Services and Equality, General Sub-Directorate on Professional Planning, Spain, for providing us the 2012 MIR registry data. We acknowledge the many medical students and MIR candidates who responded to our 2011 and 2012 surveys. We thank Jaime Pinilla Domínguez for technical assistance. Finally, we thank the participants in workshops and seminars that we have given concerning this research, including: the Pontificia Universidad Católica, Chile (JEH), Universidad Pública de Navarra (BGLV), Universidad de Vigo (BGLV), and the Grupo Evaluación de Políticas y Servicios de Salud, Asociación de Economía de la Salud (EvaluAES), Valencia (VO). An earlier version of this article appeared as National Bureau of Economic Research Working Paper No. 19896 (February 2014). The opinions expressed in this paper are ours and ours alone. CONFLICTS OF INTEREST We have none to report. See enclosed conflict of interest forms. AUTHOR CONTRIBUTIONS All authors (JEH, BGLV, PB, VO) participated in the conceptualization and the design of the study, as well as the analysis and interpretation of the data. JEH and BGLV were principally responsible for writing the manuscript. All authors have reviewed and approved the final draft of this article. STATEMENT CONCERNING ETHICAL APPROVAL No ethical approval was required for this research. MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 19 REFERENCES Barber P, Gonzalez Lopez-Valcarcel B. Forecasting the need for medical specialists in Spain: application of a system dynamics model. Human Resources for Health 2010;8. Barber P, Lopez-Valcarcel BG. Forecasting the need for medical specialists in Spain: application of a system dynamics model. Hum Resour Health 2010;8; 24. Berry S, Levinsohn J, Pakes A. Differentiated Products Demand Systems from a Combination of Micro and Macro Data: The New Car Market. Journal of Political Economy 2004;112; 68-105. Bikhchandani S, Hirschleifer D, Welch I. A Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades. Journal of Political Economy 1992;100; 992-1026. Bodenheimer T. Primary care--will it survive? N Engl J Med 2006;355; 861-864. Budish E, Che Y-K, Kojima F, Milgrom P. Designing Random Allocation Mechanisms: Theory and Applications. American Economic Review 2013;103; 585–623. Cheng TC, Scott A, Jeon SH, Kalb G, Humphreys J, Joyce C. What Factors Influence the Earnings of General Practitioners and Medical Specialists? Evidence from the Medicine in Australia: Balancing Employment and Life Survey. Health Econ 2011. Colwill JM, Cultice JM, Kruse RL. Will generalist physician supply meet demands of an increasing and aging population? Health Aff (Millwood) 2008;27; w232-241. de Bakker DH, Struijs JN, Baan CB, Raams J, de Wildt JE, Vrijhoef HJ, Schut FT. Early results from adoption of bundled payment for diabetes care in the Netherlands show improvement in care coordination. Health Aff (Millwood) 2012;31; 426-433. Dorsey ER, Jarjoura D, Rutecki GW. Influence of controllable lifestyle on recent trends in specialty choice by US medical students. JAMA 2003;290; 1173-1178. Gagne R, Leger PT. Determinants of physicians' decisions to specialize. Health Econ 2005;14; 721-735. Goldacre MJ, Laxton L, Lambert TW. Medical graduates' early career choices of specialty and their eventual specialty destinations: UK prospective cohort studies. BMJ 2010;341; c3199. Gonzalez Lopez-Valcarcel B, Barber Perez P. [Health workforce planning and training, with emphasis on primary care. SESPAS Report 2012]. Gac Sanit 2012;26 Suppl 1; 46-51. Gonzalez Lopez-Valcarcel B, Barber Perez P, Ortun Rubio V. 2011. ¿Por qué los médicos huyen y rehúyen la Medicina de Familia? Datos y claves sobre el problema en busca de !! !! soluciones. In: Ortún V (Ed)^(Eds), La refundación de la Atención Primaria. Springer Healthcare: Madrid; 2011. Gonzalez Lopez-Valcarcel B, Ortun V, Barber P, Harris JE. [Important differences between faculties of medicine. Implications for family and community medicine.]. Aten Primaria 2013. Goroll AH, Berenson RA, Schoenbaum SC, Gardner LB. Fundamental reform of payment for adult primary care: comprehensive payment for comprehensive care. Journal of General Internal Medicine 2007;22; 410-415. Harris JE, Gonzalez Lopez-Valcarcel B, Ortun V, Barber P. Specialty choice in times of economic crisis: a cross-sectional survey of Spanish medical students. BMJ open 2013;3. Harris JE, Lopez-Valcarcel BG, Barber P, Ortún V. Efficiency versus Equity in the Allocation of Medical Specialty Training Positions in Spain: A Health Policy Simulation Based on a Discrete Choice Models. National Bureau of Economic Research Working Paper No. 19896, Feb.: Cambridge MA; 2014. Harris MG, Gavel PH, Young JR. Factors influencing the choice of specialty of Australian medical graduates. Med J Aust 2005;183; 295-300. Hastings JS, Kane TJ, Staiger DO. Heterogeneous Preferences and the Efficacy of Public School Choices. New Haven CT, Yale University Department of Economics, Working Paper; 2010. Heikkila T, Hyppola H, Kumpusalo E, Halila H, Vanska J, Kujala S, Virjo I, Mattila K. Choosing a medical specialty--study of Finnish doctors graduating in 1977-2006. Med Teach 2011;33; e440-445. Hole A. Fitting mixed logit models by using maximum simulated likelihood. The Stata Journal 2007;7; 388-401. Huibers L, Giesen P, Wensing M, Grol R. Out-of-hours care in western countries: assessment of different organizational models. BMC health services research 2009;9; 105. Hurley JE. Physicians' Choices of Specialty, Location, and Mode: A Reexamination within an Interdependent Decision Framework. Journal of Human Resources 1991;26; 47-71. Kolstad JT. Information and Quality when Motivation is Intrinsic: Evidence from Surgeon Report Cardss. National Bureau of Economic Research, Working Paper No. 18804, February; 2013. MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 21 Krueger KJ, Halperin EC. Perspective: Paying physicians to be on call: a challenge for academic medicine. Academic medicine : journal of the Association of American Medical Colleges 2010;85; 1840-1844. Lopez-Valcarcel BG, Ortun V, Barber P, Harris JE, Garcia B. Ranking Spain's Medical Schools by their performance in the national residency examination. Revista clinica espanola 2013;213; 428-434. Machado MP, Mora R, Romero-Medina A. Can We Infer Hospital Quality from Medical Graduates’ Residency Choices? Journal of the European Economic Association 2012;10; 1400–1424. Mariolis A, Mihas C, Alevizos A, Gizlis V, Mariolis T, Marayiannis K, Tountas Y, Stefanadis C, Philalithis A, Creatsas G. General Practice as a career choice among undergraduate medical students in Greece. BMC Med Educ 2007;7; 15. McFadden D. 1974. Conditional Logit Analysis of Qualitative Choice Behavior. In: Zarembka P (Ed)^(Eds), Frontiers in Econometrics. Academic Press: NY; 1974. McFadden D, Train K. Mixed MNL Models for Discrete Response. Journal of Applied Econometrics 2000;15; 447-470. Nicholson S. Physician Specialty Choice under Uncertainty. Journal of Labor Economics 2002;20; 816-847. Nicholson S. How Much Do Medical Students Know about Physician Income? Journal of Human Resources 2005;40; 100-114. Ortun V, Gonzalez Lopez-Valcarcel B, Barber Perez P. [Determinants of physician compensation]. Medicina clinica 2008;131; 180-183. Revelt D, Train K. Customer-specific taste parameters and mixed logit: Households' choice of electricity suppliers. Univeristy of California at Berkeley, Department of Economics, Working Paper No. E00-274; 2000. Rosenblatt RA, Andrilla CH, Curtin T, Hart LG. Shortages of medical personnel at community health centers: implications for planned expansion. JAMA 2006;295; 1042-1049. Rosenthal MP, Diamond JJ, Rabinowitz HK, Bauer LC, Jones RL, Kearl GW, Kelly RB, Sheets KJ, Jaffe A, Jonas AP, et al. Influence of income, hours worked, and loan repayment on medical students' decision to pursue a primary care career. JAMA 1994;271; 914-917. Scott J, Larson A, Jefferies F, Veenendaal B. Small-area estimates of general practice workforce shortage in rural and remote Western Australia. The Australian journal of rural health 2006;14; 209-213. MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 22 Servicio Público de Empleo Estatal (SEPE). Información mensual de mercado de trabajo de titulados universitarioss. Monthly Reports: Madrid; 2013. Sivey P, Scott A, Witt J, Joyce C, Humphreys J. Junior doctors' preferences for specialty choice. J Health Econ 2012;31; 813-823. Sivey PM, Scott A, Witt J, Humphreys J, Joyce C. Why Junior Doctors Don t Want to Become General Practitioners: A Discrete Choice Experiment from the MABEL Longitudinal Study of Doctors. SSRN eLibrary 2010. StataCorp. Stata 13 Base Reference Manuals. Stata Press: College Station TX; 2013. Steinbrook R. Easing the shortage in adult primary care—is it all about money? New England Journal of Medicine 2009;360; 2696-2699. Thistlethwaite JE, Leeder SR, Kidd MR, Shaw T. Addressing general practice workforce shortages: policy options. The Medical journal of Australia 2008;189; 118-121. Thornton J. Physician choice of medical specialty: do economic incentives matter? Applied Economics 2000;32; 1419-1428. Thornton J, Esposto F. How important are economic factors in choice of medical specialty? Health Econ 2003;12; 67-73. Timmins C, Murdock J. A revealed preference approach to the measurement of congestion in travel cost models. Journal of Environmental Economics and Management 2007;53; 230– 249. Train K. Discrete Choice Methods With Simulation, 2nd Editions. Cambridge University Press: New York; 2009. Watts JJ, Segal L. Market failure, policy failure and other distortions in chronic disease markets. BMC health services research 2009;9; 102. MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 23 Table I. Descriptive Statistics for the 2012 MIR Registry and the 2012 Post-MIR Survey 2012 MIR Registry 2012 Post-MIR Survey P-value!a Sample size 6,254 3,177 Spanish nationality!b 77.8% 77.3% 0.58 Women!b 66.0% 67.1% 0.29 MIR Ranking!c 3,307 (1,929) 3,049 (1,899) 0.00 Distance from home to the training center (minutes)!c 153.3 (334.5) 138.9 (321.5) 0.05 Home and training center in the same province!b 52.6% 56.4% 0.00 a. For binary variables (Spanish nationality, women, and home/training center in same province), P-value based two-group two tailed comparison test of proportions. For other variables (MIR ranking, distance from home to training center), P-value based on one-way ANOVA mean comparison test. b. Mean values of binary variables. c. Mean values with standard deviations in parentheses. The highest ranked candidate had a MIR ranking of 1. MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 24 Table II. Preferred Specialties Reported in the 2012 Post-MIR Survey!a Preferred Specialty Number of Respondents Assigned to Preferred Specialty (%) Assigned to Preferred Specialty (%) Assigned to Family Medicine (%)!c Assigned to Family Medicine (%)!c Cardiology 160 76 47.5% 18 11.2% Plastic Surgery 110 13 11.8% 15 13.6% Dermatology 193 50 25.9% 33 17.5% Family Medicine 234 229 97.9% 229 97.9% Internal Medicine 147 105 71.4% 37 25.2% Obstetrics & Gynecology 226 124 54.9% 60 26.6% Pediatrics 396 231 58.3% 106 26.8% Other Specialties 1,651 1,042 63.1% 278 16.8% Total(b) 3,117 1,870 60.0% 776 24.9% a. Based upon the question: “If you could choose specialty without regard to your score on the exam, what specialty would you have chosen?” MEDICAL SPECIALTY PREFERENCES IN SPAIN !21-Dec-2014 25