The relatioship betwee primary medical qualificatio regio ad atioality at the time of registratio Workig paper 2 March 2017
Cotets Executive summary... 3 Itroductio... 4 Statemet of problem... 5 PMQ regio... 5 Natioality group at the time of registratio... 5 Method... 7 Developig a estimatio model... 7 Iformatio used i the modellig... 7 Testig the estimatio model... 8 Results... 9 Couts of doctors... 9 Proportios of the workforce... 10 Discussio... 12 Limitatios of the aalysis... 12 Coclusios... 12 Further iformatio... 12 Aex A - estimatio model desig ad testig... 13 Model desig... 13 Model testig... 21 2
Executive summary This workig paper reports the results of a aalysis of how the regios i which doctors gaied their primary medical qualificatios (PMQs) relate to their atioalities at the time that they joied the register. By regios, we mea here UK, Europea Ecoomic Area (EEA), ad iteratioal medical graduate (IMG). Of about 232,000 licesed doctors, almost 67,000 doctors who joied the register before 2004 have o atioality iformatio recorded. However, these missig data about atioality at the time of registratio may be imputed from other iformatio o the register with a reasoable degree of accuracy. Havig imputed the missig atioality data, the couts of EEA atioal doctors are compared with those for EEA qualified doctors practisig i each coutry of the UK. The couts of o-eea atioal doctors are compared with those for IMGs. The paper cocludes that: EEA regio PMQ is ot a reliable idicator for coutig doctors who were of EEA atioality at the time of their registratio - therefore i cosiderig the workforce implicatios of Brexit, it will be importat to have regard to the fact that EEA regio PMQ data uder-estimates the umber of EEA atioals; Coutig IMG doctors may produce a reasoably accurate surrogate measure for the umbers of doctors who were of o-eea atioality at the time of their registratio. 3
Itroductio The referedum vote to leave the Europea Uio (EU) has geerated iterest i the GMC's data about the atioalities of the doctors practisig i the UK. This is our secod workig paper o the topic. The first describes, i detail, the characteristics of doctors who are licesed to practise i the UK ad have a primary medical qualificatio (PMQ) from the Europea Ecoomic Area (EEA). 1 We refer to these as EEA graduate doctors. We have produced this secod workig paper to make our data as useful as possible to those who wat to uderstad more about the compositio of the medical workforce. It reports the results of a aalysis of how the regios i which doctors gaied their PMQs relate to their atioalities at the time that they joied the register. The relatioship betwee PMQ regio - where we have precise iformatio for each ad every licesed doctor - ad atioality at time of registratio - where sigificat amouts of missig data have to be imputed or estimated - is ot a simple oe. The estimates i this secod workig paper are, we believe, robust at a high level of aggregatio (by UK coutry), but are ot suitable for more detailed aalyses, such as by medical specialty. For these more detailed breakdows, we refer readers to our first workig paper o EEA graduate doctors. 1 Our data about doctors with a Europea primary medical qualificatio; Workig paper 1 - February 2017, available at www.gmc-uk.org/publicatios/30409.asp 4
Statemet of problem O 30 Jue 2016, 2 there were 232,192 licesed doctors o the register. For aalysis purposes, they may be classified accordig to either their PMQ regio or their atioality group at time of registratio. PMQ regio PMQ coutries are, by covetio, grouped by the GMC ito three regios: UK EEA Iteratioal medical graduate (IMG). The umber of licesed doctors with a PMQ from each of these regios is show i Table 1 (below). Table 1: Licesed doctors o 30 Jue 2016, by PMQ regio PMQ regio Cout Per cet UK 155,031 66.8 EEA 21,539 9.3 IMG 55,622 24.0 Total 232,192 Natioality group at the time of registratio I the same way, atioalities at time of registratio may also be grouped ito three categories: British EEA atioality No-EEA atioality. At the time a doctor applies for registratio, up to two atioalities may be recorded. 2 The aalysis date of 30 Jue was chose as the moth ed closest to the date o which a available data extract used for establishig doctor practice locatios was dowloaded. 5
British This group comprises doctors where at least oe of their atioalities is British, or derives from a coutry that qualifies them for British citizeship. These coutries are set out withi the www.gov.uk website ad iclude territories such as Gibraltar ad Bermuda. EEA atioality This group cotais doctors who are ot British ad where at least oe of their atioalities is from a coutry withi the EEA. For the purposes of registratio, the EEA is the 28 coutries of the EU, together with Norway, Switzerlad, Icelad ad Liechtestei. I this paper, 'EEA atioality' therefore excludes British doctors ad those doctors from outside the EEA with eforceable Commuity rights. No-EEA atioality Those doctors with all of their atioalities from coutries outside the EEA are placed i this group. Applyig this categorisatio to the first ad secod atioalities recorded o the register, the resultig couts of licesed doctors are set out i Table 2 (below). Table 2: Licesed doctors o 30 Jue 2016, by recorded atioality group at time of registratio Natioality group recorded Cout Per cet British 103,524 44.6 EEA atioality 22,713 9.8 No-EEA atioality 39,327 16.9 Not recorded 66,628 28.7 Total 232,192 Note: The vast majority of the doctors with o atioality iformatio recorded joied the register before 2004. Each of these two sets of categories, PMQ regio ad atioality group at time of registratio, is exhaustive ad mutually exclusive - every licesed doctor belogs to oe, ad oly oe, category i each set. The aim of the remaider of the aalysis reported here is to allocate the 66,628 doctors i the "Not recorded" atioality group category to oe of the three recorded groups, based upo the other iformatio available about them o the register; ad the to compare the resultig distributio by estimated atioality at time of registratio with the distributio by PMQ regio. 6
Method Developig a estimatio model A database of licesed doctors was created ad split ito three parts. The first part cotaied the 66,628 doctors with o recorded atioality at time of registratio. The secod part cotaied a radom sample of three-quarters of the doctors with recorded atioality groups, to be used i buildig a estimatio model to explai their atioality groups based o the other iformatio kow about them (the "traiig" data set). The fial part cotaied the remaiig quarter of the doctors with recorded atioality groups, to be used i testig the estimatio model (the "test" data set). The techical details of the estimatio model desig ad testig are documeted i Aex A. I summary, the followig rules were derived to predict atioality group at time of registratio based o other registratio iformatio. Iformatio used i the modellig There is a strog degree of associatio betwee the PMQ regio of a doctor ad their atioality group at time of registratio, but this is ot determiative. May doctors move to aother coutry for their medical traiig. I geeral, therefore, the model starts by assumig that doctors with a EEA PMQ regio were likely to have a EEA atioality, ad doctors with a IMG PMQ regio were likely to have a o-eea atioality. It supplemets this with further iformatio, as set out uder the followig three atioality group headigs. 3 British If a doctor's ethicity was British or their qualificatio was described as "ChB", the doctor was almost certai 4 to be of British atioality at time of registratio. EEA atioality If a doctor's PMQ regio was "EEA - Souther" or their qualificatio was described as ay of "Artsexame", "LMS", "Ptychio Iatrikes", "State Exam" or "State Exam Med", the doctor was almost certai to be of EEA atioality at time of registratio. 3 The factors set out here are ot ecessarily the strogest idicators of the atioality group of a doctor; they are the strogest idicators that allow o-trivial umbers of doctors with o recorded atioality group to be assiged to a group. 4 "Almost certai" here implies that more tha 90 per cet of the doctors i the traiig ad test data sets coform to the patter described. 7
If a doctor's route to registratio was "EEA" or their ethicity was Irish, the doctor was likely 5 to be of EEA atioality. The older they were at the time of their first registratio, the more likely they were to be of EEA atioality. No-EEA atioality If a doctor's PMQ coutry was Myamar, the doctor was almost certai to be of o-eea atioality at time of registratio. If a doctor's route to registratio was "IMG" or their registratio type was "Full i Approved Practice Settigs", the doctor was likely to be of o-eea atioality. Testig the estimatio model Havig built the model usig the records i the traiig data set, its performace was assessed by usig it to explai the atioality groups at time of registratio of the doctors i the test data set. The estimatio model was the used as the basis for assigig doctors with o recorded atioality to their most likely atioality group at time of registratio, ad couts based o this assigmet were regarded as our "best estimates". 5 "Likely" is used to describe factors that have a statistically sigificat ad material effect o improvig the quality of the estimatio model. 8
Results Couts of doctors The best estimates described above are first preseted for doctors licesed to practise i the UK, ad the for those curretly practisig i each of Eglad, Scotlad, Wales ad Norther Irelad, i Table 3 (below). For compariso, the umber of licesed doctors with a EEA PMQ is show below each EEA atioality estimate, ad the umber with a IMG PMQ is show below each No-EEA atioality estimate. Table 3: Licesed doctors, o 30 Jue 2016, by atioality group / PMQ regio, by UK coutry Estimate EEA atioal / EEA No-EEA atioal / IMG UK - Best estimate 24,546 51,774 UK - By PMQ 21,539 55,622 Eglad - Best estimate 17,083 44,779 Eglad - By PMQ 15,859 47,968 Scotlad - Best estimate 1,478 2,444 Scotlad - By PMQ 1,155 2,295 Wales - Best estimate 694 2,406 Wales - By PMQ 629 2,586 Norther Irelad - Best estimate /a 337 Norther Irelad - By PMQ 549 363 Note: The UK figures iclude all doctors licesed to practise i the UK; the coutry-level figures iclude those licesed doctors curretly i those coutries; figures for doctors curretly outside the UK are ot preseted, so the Eglad, Scotlad, Wales ad Norther Irelad couts do ot total to those for the UK. For example, at UK level, whereas there are 21,539 licesed doctors holdig a EEA PMQ, our best estimate of the umber of EEA atioals licesed to practise i the UK is higher 24,546. No reliable best estimate for Norther Irelad is available from the modellig. The recorded registratio data implyig EEA atioality for doctors practisig there may be ureliable, because uder the Belfast Agreemet people bor i Norther Irelad ca choose to be British citizes, Irish citizes or both. They may choose to idetify themselves at registratio as either havig bee bor i part of the UK or beig Irish, for a variety of persoal reasos. 9
Of the 549 doctors with a EEA regio PMQ practisig i Norther Irelad, 398 doctors obtaied their qualificatio i the Republic of Irelad. Proportios of the workforce EEA atioality I order to allow the couts reported i Table 3 to be compared, they are preseted as percetages of the workforce i Chart 1 (below). Chart 1: Proportios of licesed doctors who were EEA atioals at time of registratio / have a EEA PMQ, 30 Jue 2016, by UK coutry 30 25 20 15 10 9.0 8.4 7.5 5.8 7.2 6.5 8.9 5 0 Eglad - Eglad - EEA EEA PMQ atioality Scotlad - Scotlad - EEA EEA PMQ atioality Wales - EEA atioality Wales - EEA PMQ /a Norther Irelad - EEA atioality Norther Irelad - EEA PMQ Note: This is based o 189,519 doctors i Eglad, 19,803 i Scotlad, 9,695 i Wales ad 6,197 i Norther Irelad. Figures for doctors curretly practisig outside the UK are ot show. Whe comparig best estimates, the proportio of the workforce that was of EEA atioality at time of registratio decreases o movig from Eglad (9.0 per cet), to Scotlad (7.5 per cet), to Wales (7.2 per cet). It is otable that, i every coutry where a estimate is available, the best estimate of the proportio of doctors of EEA atioality at time of registratio is higher tha the proportio with a EEA PMQ. A more detailed compariso of the accuracy of the estimatio method agaist simply usig PMQ regio as a surrogate for atioality group is documeted i Aex A. It cofirms that EEA regio PMQ is ot a reliable surrogate for coutig doctors who were of EEA 10
atioality at time of registratio. I particular, see Table A4 for a cross tabulatio of PMQ regio agaist estimated atioality ad Table A11 for the results of the compariso. No-EEA atioality Similar percetages for doctors i the o-eea atioality group at time of registratio are preseted i Chart 2 (below). Chart 2: Proportios of licesed doctors who were o-eea atioals at time of registratio / have a IMG PMQ, 30 Jue 2016, by coutry 30 25 23.6 25.3 24.8 26.7 20 15 10 12.3 11.6 5.4 5.9 5 0 Eglad - Eglad - No-EEA IMG PMQ atioality Scotlad - Scotlad - No-EEA IMG PMQ atioality Wales - Wales - No-EEA IMG PMQ atioality Norther Norther Irelad - Irelad - No-EEA IMG PMQ atioality Note: This is based o 189,519 doctors i Eglad, 19,803 i Scotlad, 9,695 i Wales ad 6,197 i Norther Irelad. Figures for doctors curretly practisig outside the UK are ot show. Whe comparig best estimates, the proportio of the workforce that was of o-eea atioality at time of registratio decreases o movig from Wales (24.8 per cet), to Eglad (23.6 per cet), to Scotlad (12.3 per cet), to Norther Irelad (5.4 per cet). The best estimates of the proportios of doctors of o-eea atioality at time of registratio are geerally somewhat lower tha the proportios with a IMG PMQ, except i Scotlad where it is somewhat higher. The differeces are, however, relatively small. See Aex A for a detailed compariso of the two differet measures, which suggests that IMG may be a reasoable surrogate for coutig doctors with a o-eea atioality group at time of registratio. 11
Discussio Limitatios of the aalysis It must be kept i mid that the atioality iformatio recorded by the GMC is maily 6 that at the time of registratio ad, for some doctors, this may have chaged sice. The aalysis therefore has little to say about the relatioship betwee PMQ regio ad curret atioality. The modellig approach assumes that the uderlyig relatioships betwee registratio iformatio ad atioality at time of registratio operate cosistetly over time, ad this is u-tested. Furthermore, the two stage approach idetifies about 900 doctors with o atioality recorded as likely to be i the o-eea group i the first stage, the as likely to be i the EEA group i the secod stage. The fial estimates of umbers thus deped somewhat o the orderig of the stages i the aalysis. The estimatio methods used here are ot reliable for use i the particular circumstaces of Norther Irelad. To ivestigate atioality i Norther Irelad further, it is likely that primary research (that is, cotactig doctors for further iformatio) would be eeded. Coclusios The aalysis described i this paper suggests that, i each of Eglad, Scotlad, Wales ad Norther Irelad: EEA regio PMQ is ot a reliable idicator for coutig doctors who were of EEA atioality at the time of their registratio - therefore i cosiderig the workforce implicatios of Brexit, it will be importat to have regard to the fact that EEA regio PMQ data uder-estimates the umber of EEA atioals Coutig IMG doctors may produce a reasoable accurate surrogate measure for the umbers of doctors who were of o-eea atioality at the time of their registratio. Further iformatio Further detail about this project is available from Tom Frakli at tom.frakli@gmcuk.org. 6 "Maily", because the GMC has from time to time udertake projects to update the iformatio held about doctors' atioality o the register. 12
Aex A - Estimatio model desig ad testig Model desig The work to derive a best estimate atioality group for each of the 66,628 licesed doctors with o recorded atioality proceeded i three stages: First, imputatios of a British, EEA or o-eea atioality group were made, based o register variables that have a very strog (90 per cet or more) associatio with oe group; The, a first logistic regressio model was developed to idetify those of the remaiig doctors who were more likely to be o-eea atioals tha ot; Fially, a secod logistic regressio model was developed to idetify those of the last remaiig doctors who were more likely to be EEA atioals tha ot. Source data The followig four sources of data were used for the aalysis. 13
Table A1: Data sources File Source Cases Variables Licesed_PMQ-EEA.csv Dowloaded from the GMC maagemet iformatio systems as at 30 Jue 2016 232,192 licesed doctors GMC Referece Number; Date of Birth; Geder; Natioality; Secod Natioality; Coutry (Cotact); First Registratio Year; PMQ Qualificatio; PMQ Coutry; Ethic Origi - Level 1; Ethic Origi - Level 2. Registratio_PMQ- EEA.csv Dowloaded from the GMC maagemet iformatio systems as at 30 Jue 2016 604,842 doctors who have ever bee registered GMC Referece Number; Route (First Grated Applicatio); Type (First Grated Applicatio). 06_Prioritised_postcodes. csv Take from the State of Medical Educatio ad Practice 2016 245,803 registered doctors with a postcode GMC Referece Number; Postcode; Regio. EthicityWithNTS.xlsx Extracted from NTS, HESA, GAMSAT & UKCAT data files 34,536 UIDs with o ethicity data o the register GMC Referece Number; FialEthicity. The first data file, Licesed_PMQ-EEA.csv, was the master list of licesed doctors as at 30 Jue 2016 for the aalysis. The data i the other three files was added to it, to supplemet the registratio data cotaied. Specifically, Registratio_PMQ-EEA.csv added iformatio about the route to register ad type of registratio; 06_Prioritised_postcodes.csv added iformatio about the UK coutry i which each doctor was practisig i Jue 2016; ad ' EthicityWithNTS.xlsx ' supplemeted missig ethicity data from various educatio datasets. 14
Distributios of variables used The distributios of the variables used (durig ay of the three stages) i the aalysis are summarised below. Table A2: Licesed doctors with a PMQ coutry of Myamar, by recorded atioality group Recorded atioality group Total PMQ coutry British EEA No-EEA Not recorded Myamar Cout 32 0 404 168 604 Per cet 7.3 0.0 92.7 - Table A3: Licesed doctors with a PMQ regio (twelve areas) of EEA - Souther Europe, by recorded atioality group PMQ regio (12) EEA - S. Europe Recorded atioality group Total British EEA No-EEA Not recorded Cout 236 6,767 190 395 7,588 Per cet 3.3 94.1 2.6 - Table A4: Licesed doctors, by PMQ regio (three areas), by recorded atioality group Recorded atioality group Total PMQ regio (3) British EEA No-EEA Not recorded UK Cout 95,153 3,384 5,184 51,310 155,031 Per cet 91.7 3.3 5.0 - EEA Cout 1,541 16,996 766 2,236 21,539 Per cet 8.0 88.1 4.0 - IMG Cout 6,830 2,333 33,377 13,082 55,622 Per cet 16.1 5.5 78.5 - Total Cout 103,524 22,713 39,327 66,628 232,192 15
Table A5: Licesed doctors with selected PMQ qualificatios, by recorded atioality group Recorded atioality group Total PMQ Qualificatio British EEA No-EEA Not recorded ChB Cout 36 0 0 19 55 Per cet 100.0 0.0 0.0 - Artsexame Cout 19 536 0 146 701 Per cet 3.4 96.6 0.0 - LMS Cout 23 1,061 3 170 1,257 Per cet 2.1 97.6 0.3 - Ptychio Iatrikes Cout 46 2,260 108 65 2,479 Per cet 1.9 93.6 4.5 - State Exam Cout 29 506 2 78 615 Per cet 5.4 94.2 0.4 - State Exam Med Cout 85 1,900 39 258 2,282 Per cet 4.2 93.9 1.9 - Table A6: Licesed doctors with a recorded route to registratio, by route, by recorded atioality group Recorded atioality group Total Route to registratio British EEA No-EEA Not recorded UK Cout 66,875 3,712 10,281 3,886 84,754 Per cet 82.7 4.6 12.7 - EEA Cout 899 11,909 467 1 13,276 Per cet 6.8 89.7 3.5 - IMG Cout 2,070 157 13,911 39 16,177 Per cet 12.8 1.0 86.2 - Licece Cout 1,315 296 255 712 2,578 Per cet 70.5 15.9 13.7 - Restoratio Cout 497 584 230 3 1,314 Per cet 37.9 44.6 17.5 - Other route Cout 1,529 486 633 708 3,356 Per cet 57.7 18.4 23.9 - Total Cout 73,185 17,144 25,777 5,349 121,455 16
Table A7: Licesed doctors with a specified registratio type, by type, by recorded atioality group Recorded atioality group Total Type of registratio British EEA No-EEA Not recorded Provisioal Cout 48,047 2,936 4,009 233 55,225 Per cet 87.4 5.3 7.3 - Full Cout 938 6,688 844 8 8,478 Per cet 11.1 79.0 10.0 - Full i APS Cout 6,776 615 12,805 13 20,209 Per cet 33.6 3.1 63.4 - Full with Specialist Cout 333 3,296 596 123 4,348 Per cet 7.9 78.0 14.1 - GP Cout 4,250 317 1,786 403 6,756 Per cet 66.9 5.0 28.1 - Specialist Cout 10,081 1,486 4,922 3,242 19,731 Per cet 61.1 9.0 29.9 - Other type Cout 2,760 1,806 815 1,327 6,708 Per cet 51.3 33.6 15.2 - Total Cout 73,185 17,144 25,777 5,349 121,455 17
Table A8: Licesed doctors with a recorded ethicity, by ethicity, by recorded atioality group Recorded atioality group Total Ethicity British EEA No-EEA Not recorded Idia Cout 10,217 289 13,333 7,500 31,339 Per cet 42.9 1.2 55.9 - Pakistai Cout 4,662 393 5,278 1,589 11,922 Per cet 45.1 3.8 51.1 - Bagladeshi Cout 990 13 484 243 1,730 Per cet 66.6 0.9 32.6 - Chiese Cout 1,822 62 1,952 891 4,727 Per cet 47.5 1.6 50.9 - Other Asia Cout 3,455 294 3,573 1,702 9,024 backgroud Per cet 47.2 4.0 48.8 - Africa Cout 2,215 385 3,542 1,174 7,316 Per cet 36.1 6.3 57.7 - Caribbea Cout 273 9 179 156 617 Per cet 59.2 2.0 38.8 - Other Black backgroud Cout 234 32 113 148 527 Per cet 61.7 8.4 29.8 - British Cout 64,491 448 468 35,230 100,637 Per cet 98.6 0.7 0.7 - Irish Cout 953 2,803 16 1,492 5,264 Other White backgroud Per cet 25.3 74.3 0.4 - Cout 2,955 14,213 2,704 3,199 23,071 Per cet 14.9 71.5 13.6 - Mixed White / Asia Cout 1,356 132 119 250 1,857 Per cet 84.4 8.2 7.4 - Mixed White / Black Africa Mixed White / Black Caribbea Other Mixed backgroud Cout 308 101 242 176 827 Per cet 47.3 15.5 37.2 - Cout 190 36 30 53 309 Per cet 74.2 14.1 11.7 - Cout 1,241 209 367 288 2,105 Per cet 68.3 11.5 20.2 - Arab Cout 586 128 1 027 171 1,912 Per cet 33.7 7.4 59.0 - Other ethic group Cout 1,746 340 2,112 843 5,041 Per cet 41.6 8.1 50.3 - Total Cout 97,694 19,887 35,539 55,105 208,225 18
Figure A1: Box plots for ages at first registratio for licesed doctors, by recorded atioality group 20 30 40 50 60 British EEA No-EEA Not recorded The shaded area of each box exteds from the lower quartile of the ages for that atioality group to the upper quartile. No outlyig or extreme values are show. The box plots i Figure A1 (above) suggest that doctors from EEA coutries ted to register first at a somewhat older age tha those from o-eea coutries, ad at a much older age tha British doctors. Stage 1: Imputig missig atioality groups The missig atioality groups were imputed usig five rules, applied i priority order: If PMQ coutry is Myamar, doctor is o-eea atioality; If PMQ regio is EEA - Souther, doctor is EEA atioality; If Level 1 ethicity is British, doctor is British atioality; If qualificatio is ChB, doctor is British atioality; If qualificatio is ay of Artsexame, LMS, Ptychio Iatrikes, State Exam or State Exam Med, doctor is EEA atioality. These imputatios reduced the umber of doctors with a ukow atioality group from 66,628 to 30,524. 19
Stage 2: Estimatig missig o-eea atioality groups Table A9: Logistic regressio results, o-eea atioality oto five selected doctor characteristics Number of observatios = 151,250 Likelihood ratio Chi 2 (5) = 78,466 Chi 2 sigificace = 0.0000 Pseudo R 2 = 0.52 Odds ratio Stadard Error Z-score Z-score sigificace Odds ratio 95% cofidece iterval No-EEA atioality UK PMQ 0.0151 0.0003-184.6 0.000 0.0144 0.0157 EEA PMQ 0.0159 0.0007-91.6 0.000 0.0145 0.0173 Irish ethicity 0.0504 0.0152-9.9 0.000 0.0280 0.0909 IMG route 1.5541 0.0736 9.3 0.000 1.4163 1.7053 Full i APS 1.9596 0.0872 15.1 0.000 1.7960 2.1381 registratio Costat 2.5389 0.0393 60.2 0.000 2.4630 2.6172 The model was estimated usig all of the 151,250 licesed doctor records i the traiig data set. The probability that each doctor with o recorded atioality was of o-eea atioality was predicted, ad those of them with a predicted probability of more tha 0.5 were added to the estimated o-eea atioality group. Stage 3: Estimatig missig EEA atioality groups Table A10: Logistic regressio results, EEA atioality oto selected doctor characteristics Number of observatios = 120,675 Likelihood ratio Chi 2 (5) = 74,375 Chi 2 sigificace = 0.0000 Pseudo R 2 = 0.74 Odds ratio Stadard Error Z-score Z-score sigificace Odds ratio 95% cofidece EEA atioality iterval EEA route 4.6692 0.2743 26.2 0.000 4.1615 5.2390 IMG PMQ 5.0630 0.2998 27.4 0.000 4.5083 5.6860 Age at first registratio 1.1030 0.0036 29.6 0.000 1.0959 1.1102 EEA PMQ 68.5816 3.4856 83.2 0.000 62.0791 75.7652 Irish ethicity 61.4632 2.5224 100.4 0.000 56.7131 66.6113 Costat 0.0008 0.0001-77.1 0.000 0.0007 0.0009 20
The model was estimated usig the 120,675 licesed doctor records i the traiig data set that had a imputed atioality group of British or EEA ad data preset for their age at first registratio. 954 doctors were excluded from the model buildig because they had o age at first registratio recorded. The probability that each doctor with o recorded atioality was of EEA atioality was predicted 7, ad those of them with a predicted probability of more tha 0.5 were added to the estimated EEA atioality group. The remaiig doctors with o recorded atioality were estimated to be British. Model testig Usig the test data set, the results of applyig the estimatios described above were assessed, by testig them agaist the imputed atioality groups - that is, the groups for those doctors that had bee either recorded o the register or imputed i Stage 1. The results of this testig are tabulated i the colums headed "Estimated atioality" i Table A11 (below). For compariso, the results of simply usig the PMQ regio to estimate atioality group are show alogside, i the colums headed "PMQ". The aalysis has bee repeated for the UK ad each of the four UK coutries. 7 For those doctors with o age at first registratio recorded, the sample mea value of 27 years was used to support a probability predictio. 21
Table A11: Percetages of doctors i test data set correctly or icorrectly assiged to their imputed atioality groups by PMQ ad by the estimatio process, by practice coutry Row per cet UK / British EEA IMG / No-EEA Imputed PMQ Estimated PMQ Estimated PMQ Estimated Practice atioality atioality atioality atioality regio group UK British 93.4 93.5 1.3 1.4 5.4 5.0 EEA 14.4 16.1 75.6 78.7 9.9 5.3 No-EEA 13.1 13.5 1.9 7.0 84.9 79.5 Eglad British 92.9 93.1 1.4 1.5 5.8 5.4 EEA 12.2 13.5 78.7 81.3 9.1 5.2 No-EEA 12.7 13.0 1.9 7.1 85.4 79.9 Scotlad British 97.4 97.2 0.4 0.7 2.2 2.1 EEA 27.1 29.8 66.7 67.8 6.2 2.4 No-EEA 26.9 27.5 1.4 6.6 71.7 65.9 Wales British 94.6 94.8 0.9 0.9 4.5 4.2 EEA 11.0 15.3 77.9 79.1 11.0 5.5 No-EEA 7.6 8.1 1.4 4.0 91.0 87.9 Norther British 96.8 /a 2.0 /a 1.2 /a Irelad EEA 74.6 /a 20.7 /a 4.6 /a No-EEA 17.6 /a 0.0 /a 82.4 /a Note: The UK percetages are based o 50,418 licesed doctors i the test data set. The bases for the four UK coutries were: Eglad 41,103, Scotlad 4,382, Wales 1,981 ad Norther Irelad 1,263. Lookig at the first row of Table A11 (above) for example, of those doctors i the test data set with a recorded or imputed atioality group of British, 93.4 per cet of them were correctly estimated by lookig at PMQ ad 93.5 per cet by the estimatio model. Ispectio of the results i Table A11 suggests that both methods are very accurate i idetifyig British doctors (better tha 90 per cet success i every UK coutry), but less accurate i idetifyig EEA ad o-eea atioality group doctors. The estimatio model cosistetly out-performs usig PMQ for EEA atioals, but the reverse is true for o-eea atioals. Both methods are at their least accurate i Scotlad. 22
Email: gmc@gmc-uk.org Website: www.gmc-uk.org Telephoe: 0161 923 6602 Geeral Medical Coucil, 3 Hardma Street, Machester M3 3AW Textphoe: please dial the prefix 18001 the 0161 923 6602 to use the Text Relay service Joi the coversatio @gmcuk facebook.com/gmcuk likd.i/gmcuk youtube.com/gmcuktv This iformatio ca be made available i alterative formats or laguages. To request a alterative format, please call us o 0161 923 6602 or email us at publicatios@gmc-uk.org. Published March 2017 2017 Geeral Medical Coucil The text of this documet may be reproduced free of charge i ay format or medium providig it is reproduced accurately ad ot i a misleadig cotext. The material must be ackowledged as GMC copyright ad the documet title specified. The GMC is a charity registered i Eglad ad Wales (1089278) ad Scotlad (SC037750) Code: GMC/PMQ&N/0317