Anda di halaman 1dari 10

clinician training

Patients of doctors further from medical school


graduation have poorer outcomes
John J Norcini, John R Boulet, Amy Opalek & W Dale Dauphinee

CONTEXT There is an apparent contradic- family physicians. Patient severity of illness at


tion between the findings of studies indicating admission and in-hospital mortality, hospital
that patient outcomes are better when physi- location and volume, and the physicians recent
cians have a greater volume of practice and practice volume, time since medical school
those that find outcomes to be worse with graduation, board certification, and citizenship
increased time since training, which implies or medical school location were analysed.
greater volume.
RESULTS After adjustment, recent practice
OBJECTIVES This study was designed to esti- volume did not have a statistically significant
mate the adjusted relationships between physi- association with in-hospital mortality for all of
cians characteristics, including recent practice the conditions combined. By contrast, each
volume and time since medical school gradua- decade since graduation from medical school
tion, and patient outcomes. was associated with a 4.5% increase in relative
risk for patient mortality.
METHODS This is a retrospective observa-
tional study based on all Pennsylvania hospitali- CONCLUSIONS Recent practice volume does
sations over 7 years for acute myocardial not mitigate the increase in patient mortality
infarction, congestive heart failure, gastroin- associated with physicians time since medical
testinal haemorrhage, hip fracture and pneu- school graduation. These findings underscore
monia. It refers to 694 020 hospitalisations in the need to finds ways to support and encour-
184 hospitals attended by 5280 internists and age learning.

Medical Education 2017: 51: 480489


doi: 10.1111/medu.13276

Foundation for Advancement of International Medical Correspondence: John J Norcini, Foundation for Advancement of
Education and Research (FAIMER), Philadelphia, Pennsylvania, International Medical Education and Research (FAIMER), 3624
USA Market Street, 4th Floor, Philadelphia, Pennsylvania 19104, USA.
Tel: 00 1 215 823 2170; E-mail: jnorcini@faimer.org

480 2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education;
MEDICAL EDUCATION 2017 51: 480489
Doctors further from medical school graduation

open vascular procedures. It is not known whether a


INTRODUCTION similar skill overlap exists for medical conditions.
Consequently, a second purpose of the current
A 2005 review of the literature suggested that older study is to determine how composite volume relates
physicians and those who had been in practice to patient outcomes for medical conditions.
longer had less factual knowledge, were less likely to
adhere to standards for diagnosis, screening, pre- We analysed all hospitalisations for acute myocardial
vention and therapy, and may have had poorer infarction (AMI), congestive heart failure (CHF),
patient outcomes.1 More recent studies have added gastrointestinal haemorrhage (GIH), hip fracture
further evidence of the trends in patient outcomes and pneumonia in the state of Pennsylvania in the
noted in this systematic review.25 According to the USA over a 7-year period. We limited our analyses
authors of the review, these results are consistent to the patients of self-identified family physicians
with the view that doctors have not kept up with the and internists. Included in the analyses was informa-
continual advances in medicine that should other- tion on the number of conditions these physicians
wise accrue to the quality of patient care.1 had treated during the period of study, as well as
the number of years since their graduation from
By contrast, and as noted in a different systematic medical school. The outcome measure studied was
review of the literature, other research indicates that in-hospital mortality. We controlled for a variety of
increased individual practice volume is associated hospital, physician and patient characteristics.
with better patient outcomes across a number of dif-
ferent conditions and procedures.6 However, the
magnitude of the effects varies considerably, and METHODS
their estimation is made difficult by a series of
methodological challenges. Nonetheless, these Sources of data
results are also consistent with findings in the psy-
chology literature indicating that better perfor- In Pennsylvania, facilities offering in-patient services
mance comes with learning by watching, doing and are required to send Uniform Billing data (as set by
reflecting on performance, especially if it is under- the National Uniform Billing Committee) to the
taken in association with near-peer feedback.710 Pennsylvania Health Care Cost Containment Coun-
cil (PHC4) each time a patient is discharged.12 They
On the surface, these findings appear to be contra- must also send patient information, including
dictory because performance improves with practice demographic details, and information on co-morbid
volume but declines with time since training, conditions and important clinical findings (e.g. lab-
although the latter implies additional volume. This oratory values, clinical signs and symptoms, and
apparent conflict is compounded by the fact that pathophysiological data) from the beginning of the
some of the studies conducted on this topic look in-patient stay to MediQualTM, which applies a statis-
either at time since training or at practice volume, tical model to produce the AtlasTM index for severity
but not at both together. One purpose of the pre- of illness (SoI) at admission.13 This measure reflects
sent study is to estimate the contributions of recent the severity of a patients illness at admission using
volume and time since training to a physicians a 5-point scale on which a score of 0 indicates no
patient outcomes. clinical instability (probability of death: < 0.001), a
score of 1 indicates minimal instability (probability
Much of the work performed on practice volume of death: 0.0020.011), a score of 2 indicates moder-
has focused on its relationship with outcomes for a ate instability (probability of death: 0.0120.057), a
specific condition or procedure. A recent study score of 3 indicates severe instability (probability of
explored whether the overall volume of a surgeons death: 0.0580.499), and a score of 4 indicates maxi-
open vascular procedures was an independent pre- mal instability (probability of death: > 0.50). The
dictor of his or her patient outcomes for a specific PHC4 combines the severity of illness data with the
vascular procedure (e.g. open abdominal aortic Uniform Billing data for each patient.12
aneurysm repair).11 Although this investigation was
based on a relatively small number of surgeons, the For this study, the patient data included the
results indicated that overall volume was a more 1 404 894 hospitalisations that occurred from 1 Jan-
important determinant of outcomes than condition- uary 2003 to 31 December 2009 in which the princi-
specific volume. This finding speaks of the overlap pal diagnosis was AMI, CHF, GIH, hip fracture or
in the technical skills required to perform various pneumonia. These conditions were chosen because

2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education; 481
MEDICAL EDUCATION 2017 51: 480489
J J Norcini et al

they are among the Agency for Healthcare Research of insurance, principal diagnosis, year of admission,
and Qualitys (AHRQ) in-patient quality indicators, attending physician, AtlasTM SoI index at admission,
they occur often, and they are frequently used to discharge status and hospital. From the AMA
judge quality of care.14 After applying the procedure Masterfile, we obtained data on the physicians year
codes and standard exclusions described in the of medical school graduation, self-reported practice
AHRQ technical specifications, we matched these specialty, and specialty board certification (specialty
data with the 2010 American Medical Association board certification is voluntary in the USA and is
(AMA) Physician Masterfile, which included data on not required to practise medicine). From the
specialty board certification. Information from the ECFMG records, we established which physicians
files of the Educational Commission for Foreign were international medical graduates (IMGs) and
Medical Graduates (ECFMG) was added to the data- obtained the names of their countries of citizenship
set for those physicians who had graduated from on entry to medical school.
medical schools outside the USA and Canada.
These data served as the basis for the creation of
Both the PHC4 data and the AMA Masterfile are pub- other variables. For hospitalisations, we combined
licly available datasets; the agencies responsible for those scoring 0 and 1 on the AtlasTM SoI index at
these resources have secured the appropriate releases. admission because there were relatively few hospital-
All patient records were anonymised and de-identified isations in the 0 category. We then created sepa-
prior to their release from PHC4 and therefore also for rate indicators for none/minimal, moderate, severe
our analysis. Although their data are not publicly avail- and maximal instability.
able, candidates for ECFMG certification agree to allow
their de-identified information to be used for research We developed an indicator to categorise physicians
purposes. Hence, institutional review board approval as US medical school graduates (USMGs), non-US
was not needed for this study. citizen international medical school graduates
(non-USIMGs), or US citizen international medical
Of the 1 404 894 hospitalisations for these principal school graduates (USIMGs). We also added indica-
diagnoses, we confined the analyses to the 709 207 tors for specialty board certification in family med-
hospitalisations for which the attending physician icine or internal medicine (indicating both
had graduated from medical school in 1958 or later completion of accredited training and successful
(when the ECFMG began operation), spent the examination performance). We calculated the
majority of time in direct patient care, and was a number of years since the physician had gradu-
self-identified family physician or internist. The ated from medical school for each hospitalisation
PHC4 defines the attending physician as: . . .the as an indicator of career length. As a marker of
physician who would normally be expected to certify recent experience, we counted the number of
and re-certify the medical necessity of the services times the doctor had been the attending physician
rendered and/or who has primary responsibility for for each of the conditions during the 7-year per-
the patients medical care and treatment.12 This iod of study; we also summed these to develop an
group of doctors accounts for the majority of hospi- overall total.
talisations and is relatively homogeneous in terms of
training compared with subspecialists. In the USA, We developed an indicator to show whether each
both family doctors and general internists are hospital in the study was in a rural location by refer-
required to have undergone 3 years of postgraduate ence to the county list maintained by the Pennsylva-
training and the majority of these doctors practise nia Office of Rural Health.15 As an indicator of
in out-patient settings. To give most physicians time institutional experience, we counted the number of
to complete their postgraduate training and to times, during the period of study, that each condi-
become established in practice, we further elimi- tion was treated by the study physicians at each hos-
nated the 15 187 hospitalisations (2%) in which the pital; we also developed an overall count for this.
attending physician was within 5 years of graduation
from medical school; this left 694 020 hospitalisa- Analyses
tions for analysis.
Descriptive statistics were calculated for the hospi-
Data elements talisations, doctors and institutions. For all five of
the conditions combined, we applied a multivari-
From the PHC4 records, we were able to obtain ate model to assess the effects of the independent
data on the patients age, sex, race, ethnicity, type measures on mortality. The model included

482 2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education;
MEDICAL EDUCATION 2017 51: 480489
Doctors further from medical school graduation

indicators for each of the conditions and was In total, 5280 physicians were represented in the
adjusted for: (i) severity of illness (coded as mod- analyses. Their average year of graduation from
erate, severe or maximal with none/minimal as medical school was 1987 (standard deviation [SD]
the reference); (ii) the condition (coded as AMI, 9.7), and 79% of them were specialty board-certi-
CHF, GIH or hip fracture with pneumonia as the fied. Of these physicians, 3156 (60%) were internists
reference); (iii) whether the physician was board- and 2124 (40%) practised family medicine. Of the
certified; (iv) whether the physician was a USIMG total, 3847 (73%) physicians were USMGs, 1067
or a non-USIMG (with USMGs as the reference); (20%) were non-USIMGs and 366 (7%) were
(v) the number of years the physician had been USIMGs. During the period of study, the average
in practice; (vi) the number of hospitalisations for physician managed a mean  SD of 131  159
which the doctor was the attending physician; (vii) hospitalisations. Of these, a mean  SD of 14  22
whether the hospital was in a rural location, and were for AMI, 49  65 were for CHF, 22  27 were
(viii) the number of hospitalisations in the for GIH, 3  6 were for hip fracture, and 43  54
institution for that condition. We also applied a were for pneumonia.
similar model to each of the individual conditions.
These analyses were undertaken to ensure that Of the 184 hospitals in the study, 74 were in rural
the effects observed in the combined data did locations. During the period of study, the average
not mask the influence of experience with a facility managed a mean  SD of 3722  2934
specific condition. To adjust the confidence hospitalisations. Of these, a mean  SD of
intervals for the clustering of patients within physi- 403  484 were for AMI, 1406  1115 were for
cians and physicians within hospital, we used gen- CHF, 618  511 were for GIH, 98  117 were for
eralised estimating equations (GENMOD procedure; hip fracture, and 1247  884 were for pneumonia.
SAS Version 9.3; SAS Institute, Inc., Cary, NC,
USA). For the total group, Table 2 presents the
mean  SD crude mortalities for hospitalisations
Potential biases broken into 10-year periods according to the num-
ber of years since the attending physician had grad-
Given the nature of these analyses, it was possible uated from medical school. Having an attending
that physician effects might be confounded with physician closer to medical school graduation was
hospital effects. To test this possibility, we con- associated with lower mortality.
ducted conditional logistic regression analysis using
the overall model specified above, eliminating hos- Multivariate analyses
pital-level variables and including hospital as a fixed
effect. It produced parameter estimates comparable Table 3 shows the results of the multivariate analy-
with those presented below; consequently, we do sis for the total group. Adjusting for characteristics
not report them. of the patients, physicians and facilities, the rela-
tive risk for mortality was lower in hospitalisations
Further, it is possible that differences in patient out- for which the attending physician had more
comes were related to whether the attending physi- recent experience and was greater with increasing
cian specialised in either family medicine or time since the completion of training. For each
internal medicine. To test this possibility, we con- hospitalisation managed by their physician,
ducted the analysis described above with this factor patients had a 0.007% (95% confidence interval
included. This did not make a statistically significant [CI] 0.016% to 0.001%) change in relative risk
contribution to the model; consequently, we do not for mortality. Although the direction of change
report it. was as expected, it did not reach traditional levels
of statistical significance (p = 0.08). By contrast,
each additional year since the physician had grad-
RESULTS uated from medical school was associated with a
0.455% (95% CI 0.2800.630%) increase in
Characteristics of physicians, patient hospitalisations relative risk for mortality.
and hospitals
Results for the total group presented in Table 3 also
Table 1 presents descriptive data for the hospitalisa- show that specialty board certification was associated
tions broken down for each of the conditions and with a 7.7% (95% CI 11.7% to 3.7%) change
all conditions combined. in relative risk for mortality. Likewise, being the

2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education; 483
MEDICAL EDUCATION 2017 51: 480489
J J Norcini et al

Table 1 Descriptive information for hospitalisations for each of the studied conditions

Patients, n (%)

AMI CHF GIH Hip fracture Pneumonia Total

All patients 74 132 258 697 113 744 17 946 229 501 694 020
Patient age
1849 years 4597 (6%) 10 917 (4%) 10 587 (9%) * 29 608 (13%) 55 709 (8%)
5064 years 12 717 (17%) 34 908 (14%) 17 632 (16%) * 39 400 (17%) 104 657 (15%)
6574 years 12 538 (17%) 46 886 (18%) 20 903 (18%) 1838 (10%) 40 813 (18%) 122 978 (18%)
7584 years 22 478 (30%) 89 018 (34%) 37 885 (33%) 6998 (39%) 67 996 (30%) 224 375 (32%)
85 years 21 802 (29%) 76 968 (30%) 26 737 (24%) 9110 (51%) 51 684 (23%) 186 301 (27%)
Patient sex
Female 38 245 (52%) 145 698 (56%) 62 079 (55%) 12 818 (71%) 123 059 (54%) 381 899 (55%)
Male 35 885 (48%) 112 999 (44%) 51 665 (45%) 5128 (29%) 106 440 (46%) 312 117 (45%)
Unknown 2 (< 1%) 2 (< 1%) 4 (< 1%)
Patient race
White 66 677 (90%) 215 812 (83%) 97 206 (85%) 17 050 (95%) 202 076 (88%) 598 821 (86%)
Black 4915 (7%) 35 023 (14%) 12 522 (11%) 515 (3%) 19 560 (9%) 72 535 (10%)
Asian or Pacific Islander 183 (< 1%) 573 (< 1%) 469 (< 1%) 39 (< 1%) 758 (< 1%) 2022 (< 1%)
Native American 22 (< 1%) 87 (< 1%) 32 (< 1%) 5 (< 1%) 65 (< 1%) 211 (< 1%)
or Alaskan
Other/Mixed 950 (1%) 3409 (1%) 1586 (1%) 92 (< 1%) 3330 (1%) 9367 (1%)
Unknown 1385 (2%) 3793 (2%) 1929 (2%) 245 (1%) 3712 (2%) 11 064 (2%)
Patient ethnicity
Hispanic 1208 (2%) 4365 (2%) 1917 (2%) 151 (1%) 4379 (2%) 12 020 (2%)
Non-Hispanic 72 924 (98%) 254 326 (98%) 111 823 (98%) 17 795 (99%) 225 108 (98%) 681 976 (98%)
Unknown 6 (< 1%) 4 (< 1%) 14 (< 1%) 24 (< 1%)
Type of insurance
Uninsured 837 (1%) 1964 (1%) 1468 (1%) 49 (< 1%) 3039 (1%) 7357 (1%)
Medicare 56 141 (76%) 216 206 (84%) 86 109 (76%) 16 950 (94%) 165 749 (72%) 541 155 (78%)
Medicaid 3681 (5%) 15 818 (6%) 7372 (6%) 49 (<1%) 18 564 (8%) 45 484 (7%)
Blue Cross 8046 (11%) 13 788 (5%) 11 127 (10%) 391 (2%) 25 272 (11%) 58 624 (8%)
Commercial 4759 (6%) 8875 (3%) 6713 (6%) 391 (2%) 14 794 (6%) 35 532 (5%)
Government 550 (1%) 1439 (1%) 669 (1%) 83 (1%) 1492 (1%) 4233 (1%)
Unknown 96 (< 1%) 421 (< 1%) 214 (< 1%) 29 (< 1%) 426 (< 1%) 1186 (< 1%)
Missing 22 (< 1%) 186 (< 1%) 72 (< 1%) 4 (< 1%) 165 (< 1%) 449 (< 1%)
Severity of illness
None/minimal 9434 (13%) 52 326 (20%) 56 693 (50%) 2851 (16%) 52 525 (23%) 173 829 (25%)
Moderate 29 879 (40%) 155 352 (60%) 45 658 (40%) 11 229 (63%) 125 601 (55%) 367 719 (53%)
Severe 32 255 (44%) 49 984 (19%) 10 788 (9%) 3817 (21%) 49 788 (22%) 146 632 (21%)
Maximal 2564 (3%) 1035 (< 1%) 605 (1%) 49 (< 1%) 1587 (1%) 5840 (1%)
Mortality 6977 (9%) 8867 (3%) 2732 (2%) 850 (5%) 9312 (4%) 28 738 (4%)

* These cases were specifically excluded by the Agency for Healthcare Research and Quality (AHRQ) guidelines

patient of a non-USIMG was associated with a Table 4 presents the results of the multivariate analy-
10.7% (95% CI 14.8% to 6.6%) change in sis for each of the conditions. In four of the five analy-
relative risk for mortality. ses, time since graduation from medical school had a

484 2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education;
MEDICAL EDUCATION 2017 51: 480489
Doctors further from medical school graduation

Table 2 Crude mortality for hospitalisations by years since DISCUSSION


the attending physicians graduation from medical school*
The aim of this study was to estimate the contribu-
Years since Hospitalisations, Mortality, tions of various physician characteristics to the out-
graduation n proportion (SE)
comes of their patients. Of the educational and
experiential characteristics examined, we focused
514 222 466 0.0381 (0.0004)
particularly on recent experience and time since
medical school graduation. For all five conditions
1524 235 094 0.0424 (0.0004)
combined, a physicians recent experience did not
2534 187 759 0.0436 (0.0005)
have a statistically significant relationship with
35 48 701 0.0429 (0.0009)
adjusted mortality. When the conditions were anal-
ysed separately, recent experience had a statistically
* F = 30.63; p < 0.0001
SE = standard error
significant relationship with adjusted mortality only
for CHF. By contrast, time since graduation from
medical school had a statistically significant relation-
ship with adjusted mortality for all conditions com-
statistically significant association with mortality after bined. Each decade since graduation was associated
adjustment for all other factors. Hip fracture was the with a 4.5% increased relative risk for patient mor-
only exception, but the number of such hospitalisa- tality. In addition, time since graduation was a statis-
tions was small relative to those for the other condi- tically significant predictor of outcomes in four of
tions. In only one of the conditions, CHF, did recent the five conditions.
experience have a statistically significant association
with adjusted mortality. In this instance the reduction For all five conditions combined, mortality was
in relative risk was 0.06% (95% CI 0.08% to lower in patients whose doctors were specialty
0.03%) for each hospitalisation.

Table 3 Parameter estimates and adjusted odds ratios for all conditions together*

Parameter Adjusted
estimate 95% CI OR

Admission severity Moderate 1.79998 1.71509 to 1.88488 6.04956


Severe 3.52249 3.43871 to 3.60627 33.8686
Maximal 5.47459 5.37645 to 5.57273 238.5529
Condition AMI 0.39645 0.36014 to 0.43275 1.48654
CHF 0.08852 0.05626 to 0.12078 1.09256
GIH 0.10922 0.06254 to 0.15591 1.11541
Hip fracture 0.30707 0.22839 to 0.38575 1.35944
Physician Board certification 0.07744 0.11743 to 0.03745 0.92548
USIMG 0.04841 0.00959 to 0.10642 1.04961
Non-USIMG 0.10699 0.14785 to 0.06613 0.89853
Years since graduation 0.00460 0.00285 to 0.00636 1.00462
Case volume 0.00007 0.00016 to 0.00001 0.99993
Facility Case volume 0.00001 0.000003 to 0.00002 0.99999
Urban location 0.12171 0.084250.15918 1.12943

* Reference groups: Admission severity, none/minimal; Condition, pneumonia; medical school, USMGs

This represents the odds of mortality given a particular exposure that has been adjusted for all other variables in the model
95% CI = 95% confidence interval; AMI = acute myocardial infarction; CHF = congestive heart failure; GIH = gastrointestinal haemor-
rhage; OR = odds ratio; USIMG = US citizen international medical school graduate; non-USIMG = non-US citizen international medical
school graduate; USMG = US medical school graduate

2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education; 485
MEDICAL EDUCATION 2017 51: 480489
J J Norcini et al

Table 4 Parameter estimates and adjusted odds ratios for each condition*

Parameter Adjusted
Condition estimate 95% CI OR

AMI Admission severity Moderate 2.26726 1.907965 to 2.626548 9.65288


Severe 3.89474 3.538143 to 4.251338 49.14328
Maximal 5.88584 5.523936 to 6.247752 359.9065
Physician Board certification 0.02830 0.09972 to 0.043112 0.972092
USIMG 0.05668 0.04473 to 0.158089 1.058317
Non-USIMG 0.07282 0.14385 to 0.0018 0.929767
Years since graduation 0.00423 0.001074 to 0.007391 1.004242
Case volume 0.00044 0.00128 to 0.000386 0.999555
Facility Case volume 0.00015 0.0002 to 0.00011 0.999846
Urban location 0.18248 0.121312 to 0.243657 1.200196
CHF Admission severity Moderate 1.531017 1.398165 to 1.663868 4.622875
Severe 3.213425 3.081567 to 3.345283 24.8641
Maximal 5.199926 5.023395 to 5.376457 181.2589
Physician Board certification 0.10855 0.16783 to 0.04927 0.897133
USIMG 0.024006 0.06382 to 0.111833 1.024297
Non-USIMG 0.13118 0.19301 to 0.06935 0.877062
Years since graduation 0.005893 0.0032120.008573 1.00591
Case volume 0.00055 0.00082 to 0.00029 0.999445
Facility Case volume 2.2968E-05 4.5E-05 to 6.2E-07 0.999977
Urban location 0.142853826 0.084792 to 0.200915 1.153561
GIH Admission severity Moderate 1.94 1.773889 to 2.106111 6.958751
Severe 3.774752 3.615237 to 3.934267 43.58669
Maximal 5.655279 5.433372 to 5.877186 285.7961
Physician Board certification 0.22398 0.3209 to 0.12707 0.799328
USIMG 0.009536 0.13792 to 0.156997 1.009582
Non-USIMG 0.04107 0.14611 to 0.063959 0.959757
Years since graduation 0.006033 0.001659 to 0.010406 1.006051
Case volume 0.000284 0.00096 to 0.00153 1.000284
Facility Case volume 5.7E-05 0.00014 to 2.28E-05 0.999943
Urban location 0.087478 0.00806 to 0.183012 1.091419
Hip fracture Admission severity Moderate 0.799109 0.504713 to 1.093506 2.22356
Severe 1.642047 1.340204 to 1.94389 5.165734
Maximal 2.833524 2.134415 to 3.532632 17.00527
Physician Board certification 0.04721 0.21656 to 0.122144 0.953887
USIMG 0.08308 0.35105 to 0.184887 0.920278
Non-USIMG 0.29853 0.48746 to 0.1096 0.741908
Years since graduation 0.004522 0.00324 to 0.012288 1.004532
Case volume 0.00352 0.01049 to 0.00345 0.996488
Facility Case volume 0.00141 0.00192 to 0.00089 0.998594
Urban location 0.191715 0.032475 to 0.350956 1.211325
Pneumonia Admission severity Moderate 1.978883 1.817357 to 2.140409 7.234659
Severe 3.821336 3.662555 to 3.980116 45.66516
Maximal 5.715121 5.528173 to 5.90207 303.421

486 2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education;
MEDICAL EDUCATION 2017 51: 480489
Doctors further from medical school graduation

Table 4 (Continued)

Parameter Adjusted
Condition estimate 95% CI OR

Physician Board certification 0.05663 0.11694 to 0.003687 0.944946


USIMG 0.061921 0.02901 to 0.152856 1.063879
Non-USIMG 0.11805 0.18227 to 0.05383 0.888654
Years since graduation 0.004484 0.001743 to 0.007225 1.004494
Case volume 0.000165 0.00011 to 0.000443 1.000165
Facility Case volume 9.26E-06 1.8E-05 to 3.68E-05 1.000009
Urban location 0.117056 0.061211 to 0.172901 1.124182

* Reference groups: Admission severity, none/minimal; Condition, pneumonia; medical school, USMGs

This represents the odds of mortality given a particular exposure that has been adjusted for all other variables in the model
95% CI = 95% confidence interval; AMI = acute myocardial infarction; CHF = congestive heart failure; GIH = gastrointestinal haemor-
rhage; OR = odds ratio; USIMG = US citizen international medical school graduate; non-USIMG = non-US citizen international medical
school graduate; USMG = US medical school graduate

board-certified. Likewise, the patients of non-US broadly on all accounts will be important to estab-
citizen IMGs had lower adjusted mortality than lish the generalisability of our results.
those of either US medical graduates or US citizen
IMGs. Both of these results are consistent with It is noteworthy that recent practice volume does
previous findings and highlight the relationship not mitigate the increase in patient mortality associ-
between physician characteristics, their training ated with physicians time since medical school
and examination performance, and patient graduation. This increase existed even after adjust-
outcomes.4,5 ing for hospital, patient and physician characteris-
tics, including initial specialty board certification. In
As with any observational study, our work has a addition, the finding was similar for individual con-
number of limitations. Although we have included a ditions and for the conditions in composite.
variety of patient, physician and hospital characteris- Although the causes are likely to be multi-factorial,
tics, there may be other important factors that we a plausible explanation for a sizeable portion of this
have not captured but that may have influenced the decline continues to be failure of physicians to keep
results. For example, although all of the physicians up with changes in the treatment of patients with
in the study were required to obtain continuing particular conditions, such as clinical therapeutics
education credits to retain their licences, we do not or imaging options.1
know whether they had enrolled in a specialty
boards maintenance of competence programme or It is also of interest that recent experience does not
made other educational efforts relevant to these overcome the increase in patient mortality that
specific conditions. In addition, although care was accompanies a physicians time since formal train-
exercised in collecting data from the hospitals, ing. Clearly, it is not sufficient to repeatedly treat
there may be differences across institutions in how patients with the same, or similar, conditions. This
the attending physician was identified or in the implies that some form of structured educational
treatment guidelines applied. Further, the growing intervention is required, along with appropriate
use of hospitalists in the USA may ultimately mean assessments. This will help to ensure that physicians
that fewer of these conditions are managed by inter- maintain their skills and provide optimal patient
nists and family physicians. Our analyses attempted care.1618
to rule out institutional confounding but may not
have completely eliminated it. Finally, these analyses Although the increased relative risk is of notable
are based on data for hospitalisations for only five magnitude for the population of doctors, it is less
conditions in one state in the USA. Sampling more pronounced when considered in the context of

2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education; 487
MEDICAL EDUCATION 2017 51: 480489
J J Norcini et al

individual physicians. Over the 7-year period under


study, the average physician managed 131 hospitali- Contributors: JJN contributed to the conception and design
sations for the five conditions. Given that the aver- of the study, and to the acquisition, analysis and interpre-
age absolute risk for mortality was 4.1%, the tation of data, and drafted the paper. JRB contributed to
number of an individual doctors patients actually the study design, and to the acquisition, analysis and
dying was small. Consequently, to have a meaningful interpretation of data. AO contributed to the acquisition,
analysis and interpretation of data. WDD contributed to
effect, systematic interventions must be aimed at
the conception and design of the study and to the inter-
populations of physicians rather than at individuals.
pretation of data. All authors contributed to the critical
Unfortunately, the physicians in this study were sub- revision of the paper and approved the final manuscript
ject to a wide range of requirements for continuing for publication. All authors have agreed to be accountable
education credits and mandatory or voluntary main- for the study.
tenance of competence programmes, many of which Acknowledgements: none.
are likely to have been insufficiently intense or Funding: none.
focused to be effective. Conflicts of interest: none.
Ethical approval: both the Pennsylvania Health Care Cost
The American Board of Medical Specialties Containment Council (PHC4) data and the American
(ABMS) maintenance of competence programme Medical Association (AMA) Masterfile are publicly avail-
offers an example of a broader approach. It able datasets that have secured the appropriate releases
from participants. All patient records were anonymised
involves ongoing measurement of six core compe-
and de-identified prior to their release from PHC4 and
tencies within the context of a four-part framework: therefore our analysis. Although their data are not pub-
(i) professionalism and professional standing; (ii) licly available, candidates for Educational Commission for
lifelong learning and self-assessment; (iii) assess- Foreign Medical Graduates (ECFMG) certification agree
ment of knowledge, judgement and skills, and (iv) to allow their de-identified information to be used for
improvement in medical practice. The programme research purposes. Hence, institutional review board
adapts itself to differences within and across the approval was not needed for this study.
specialties of medicine, and is focused on the iden-
tified needs of individual physicians. However, the
programme has been criticised as being too time- REFERENCES
consuming and intrusive.1922 Other, perhaps less
controversial, models exist to accomplish the same 1 Choudhry NK, Fletcher RH, Soumerai SB. Systematic
ends.23 review: the relationship between clinical experience
and quality of health care. Ann Intern Med 2005;142
Regardless of the exact characteristics of such pro- (4):26073.
2 Southern WN, Bellin EY, Arnsten JH. Longer lengths
grammes, our findings offer evidence of the need to
of stay and higher risk of mortality among inpatients
find ways to prevent the increase in patient mortal-
of physicians with more years in practice. Am J Med
ity as physicians mature in their practices. It is clear 2011;124 (9):86874.
that neither volume alone nor years of experience 3 Duclos A, Peix J-L, Colin C et al. Influence of
are sufficient to ensure that a physician keeps up experience on performance of individual surgeons in
with rapid changes in the practice environment. thyroid surgery: prospective cross sectional
This problem is exacerbated in the USA by the fact multicentre study. BMJ 2012;344:d8041.
that a sizeable group of practising physicians has 4 Norcini JJ, Boulet JR, Dauphinee WD, Opalek A,
never achieved specialty certification and has poorer Krantz ID, Anderson ST. Evaluating the quality of
patient outcomes, even after adjusting for a variety care provided by graduates of international medical
of patient and hospital characteristics. These indi- schools. Health Aff (Millwood) 2010;29 (8):14618.
5 Norcini JJ, Boulet JR, Opalek A, Dauphinee WD.
viduals cannot participate in the maintenance of
Outcomes of cardiac surgery: associations with
competence programmes of the specialty boards.
physician characteristics, institutional characteristics,
and transfers of care. Med Care 2013;51 (12):10349.
Finally, as unpopular as these programmes may be 6 Halm EA, Lee C, Chassin MR. Is volume related to
with physicians, some form of assessment process, outcome in health care? A systematic review and
perhaps centred on the outcomes of care, is methodologic critique of the literature. Ann Intern
needed to protect patients and provide them with Med 2002;137 (6):51120.
a way to choose among doctors. At the same time, 7 Ericsson KA. Deliberate practice and the acquisition
health care systems must provide doctors with the and maintenance of expert performance in medicine
time and resources they need to maintain their and related domains. Acad Med 2004;79 (10
competence. Suppl):7081.

488 2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education;
MEDICAL EDUCATION 2017 51: 480489
Doctors further from medical school graduation

8 Norman GR, Eva KW. Does clinical experience make evidence. J Contin Educ Health Prof 2013;33 (Suppl
up for failure to keep up to date? Equine Vet J 2007;39 1):2035.
(4):31921. 17 Hess BJ, Weng W, Holmboe ES, Lipner RS. The
9 Eva KW. The aging physician: changes in cognitive association between physicians cognitive skills and
processing and their impact on medical practice. Acad quality of diabetes care. Acad Med 2012;87 (2):
Med 2002;77 (10 Suppl):16. 15763.
10 Sargeant JM, Mann KV, van der Vleuten CP, 18 Norcini J. Understanding learning in the workplace
Metsemakers JF. Reflection: a link between receiving for those who practise: we cant wait another 50 years.
and using assessment feedback. Adv Health Sci Educ Med Educ 2016;50 (1):1820.
Theory Pract 2009;14 (3):399410. 19 Iglehart JK, Baron RB. Ensuring physicians
11 Modrall JG, Rosero EB, Chung J, Arko FR III, competence is maintenance of certification the
Valentine RJ, Clagett GP, Timaran CH. Defining the answer? N Engl J Med 2012;367 (26):25439.
type of surgeon volume that influences the outcomes 20 Drazen JM, Weinstein DF. Considering recertification.
for open abdominal aortic aneurysm repair. J Vasc N Engl J Med 2010;362 (10):9467.
Surg 2011;54 (6):1599604. 21 Teirstein PS, Topol EJ. The role of maintenance of
12 Pennsylvania Health Care Cost Containment Council certification programmes in governance and
(PHC4). Home page. http://www.phc4.org/. professionalism. JAMA 2015;313 (18):180910.
[Accessed 12 August 2016.] 22 Cook DA, Blachman MJ, West CP, Wittich CM.
13 Quantros, Inc., MediQualTM services. http://www.med Physician attitudes about maintenance of competence
iqual.com/. [Accessed 12 August 2016.] programmes: a cross-specialty national survey. Mayo
14 Agency for Healthcare Research and Quality Clin Proc 2016;91 (10):133645.
(AHRQ). Inpatient quality indicators overview. 23 Williams S, Holmes S, Laugharne K. Impact of
http://www.qualityindicators.ahrq.gov/Modules/iqi_ revalidation on appraisal in primary care: an initial
resources.aspx. [Accessed 12 August 2016.] evaluation of the experience of Welsh GPs. Educ Prim
15 PennState College of Health and Human Care 2016;27 (2):1218.
Development. Pennsylvania Office of Rural Health.
http://www.porh.psu.edu/porh/. [Accessed 12 Received 17 August 2016; editorial comments to author 7
August 2016.] November 2016, 21 December 2016; accepted for publication 5
16 Lipner RS, Hess BJ, Phillips RL Jr. Specialty board January 2017
certification in the United States: issues and

2017 John Wiley & Sons Ltd and The Association for the Study of Medical Education; 489
MEDICAL EDUCATION 2017 51: 480489

Anda mungkin juga menyukai