CONCEPTS
emergency medicine residency, cost
C o s t of Care in the Emergency Department: Impact of an Emergency Medicine Residency Program From the Department of Emergency Medicine, The Medical College of Pennsylvania, Philadelphia. Received for publication October 16, 1991. Revision received February 5, 1992. Accepted for publication March 2, 1992. Presented in part at the Society for Academic Emergency Medicine Annual Meeting in Washington, DC, May 1991.
Robert M McNamara, MD, FACEP John J Kelly, DO, FACEP
Study objective: To evaluate the impact of an emergency medicine residency training program on the cost of care in the emergency department. Design: A retrospective chart review was conducted of all ED encounters for a three-month period, six months before and six months after the introduction of an emergency medicine residency program into an urban community hospital. Physician staffing of this ED before the residency period was by nonemergency medicine residency-trained emergency physicians. Setting:
A 27,O00-visit-per-year urban community hospital ED.
Type of patients: A consecutive sample of all patients discharged home from the emergency center with one of six diagnoses. The diagnoses studied were viral upper respiratory infection, pharyngitis, acute asthma, seizure, lumbosacral strain, and cervical strain. outcome measures: Frequency of laboratory test and radiograph ordering pertinent to the evaluation of each diagnostic category were used as a marker of cost of care. Main
Results: The presence of the residency training program did not increase the cost of care as measured by test use and, for three of the six diagnoses, actually lowered the cost of care, This effect was most prominent in the evaluation of lumbosacral and cervical strain when the residency physicians ordered radiographs at a rate five and 2.3 times lower, respectively, than the previous group and in the approach to pharyngitis when they ordered throat cultures 2.8 times less frequently. Conclusion: As measured by selected test use for six common discharge diagnoses, the introduction of an emergency medicine residency program did not increase the cost of care in this urban community hospital ED. [McNamara RM, Kelly J J: Cost of care in the emergency department: Impact of an emergency medicine residency program. Ann ErnergMed August 1992;21:956-962.]
8 2/ 9 5 6
ANNALS OF EMERGENCY MEDICINE 21:8 AUGUST 1992
RESIDENCY P R O G R A M McNamara & Kelly
INTRODUCTION
MATERIALS AND METHODS
The presence of medical teaching programs has generally been found to increase the costs of inpatient care. 1-8 The introduction of more extensively trained physicians to an inpatient area has been shown to increase costs. 9 In the outpatient setting, residency-trained physicians o r d e r more diagnostic tests than general practitioners with specialists having the highest rate of test use. lo-12 Training in academically oriented medical schools or residency programs has been found to increase test ordering for ambulatory patients. 13 The effect of teaching programs or more extensive training on the cost of care has not been studied in the emergency department setting. The emergency physician is playing an increasingly important role in the overall health care cost picture. In the decade from 1980 to 1989 yearly visits to the nation's EDs increased from 82 million to nearly 90 million. 14,15 Since 1967, expenditures for outpatient and ED care have been the most rapidly growing component of health care costs. 16 In 1989, these areas represented 19% of all hospital costs and 7% of the total health care expenditures for the nation. 16 The ED also has become a m a j o r source of p r i m a r y care,17,18 particularly in those areas in which access to other forms of p r i m a r y care is limited. 19 The impact of emergency physicians on cost extends to the inpatient arena, as they often determine the initial stages of evaluation and treatment of the patients admitted.t8 Residency training programs in emergency medicine have existed since the early 1970s. The effect of specialists in emergency medicine on the cost of care in the ED has not been studied, b u t some have suggested that their presence increases costs. 20,21 The purpose of this research was to study the impact of an emergency medicine residency program with primarily emergency medicine residency-trained faculty on the cost of care in the ED. This was made possible when such a p r o g r a m assumed responsibility for physician staffing of an u r b a n community hospital ED from a group of nonresidency-trained emergency physicians.
On July 1, 1989, the emergency medicine residency program of the Medical College of Pennsylvania began providing physician coverage of a 27,000-visit u r b a n community hospital ED. This facility had been previously staffed by emergency physicians supplied to the hospital by a regional contract group. This n a t u r a l experiment afforded an excellent opportunity to study the impact of a teaching p r o g r a m on the cost of care in the ED. A retrospective chart review was conducted on a consecutive sample of patients t r e a t e d in the ED from J a n u a r y through March of 1989 and 1990 who were discharged home with one of six common diagnoses. The cost of care was assessed by the performance frequency of selected tests pertinent to the evaluation of each diagnosis. Overall charges for each diagnostic category were also measured and compared. An almost complete change in physician staffing allowed for direct comparison of two groups of emergency physicians in relation to the cost of care. F r o m July 1, 1989, on, physician coverage of the ED consisted of continuous presence of one faculty member and one resident from the residency program 24 hours a day. During the residency study period ( J a n u a r y to March 1990), there was an additional ten hours of resident coverage on 62 (69%) of the 90 days. The residents were generally in their second or third y e a r of training in emergency medicine. Thirteen of the 16 emergency medicine faculty who provided coverage during the study period had completed a residency in this specialty. Of the three remaining faculty, all were board-certified in internal medicine, with one also board-certified in emergency medicine and another also board-certified in pediatrics. The large number of faculty is because most rotate frequently between the three EDs within the residency program. One of the previous group's physicians was retained as clinical faculty during the changeover and all patients managed or supervised by this physician were excluded from analysis in each of the two study periods.
Table 2. Pharyngitis: Aged 5 to 50 years, discharged home*
Table 1. Viral upper respiratory infection: Aged 15 to 50 years, discharged home*
Preresidency (94)
Residency (61)
P
28 ± 9 45 {48%) 38 (40%) 25 (27%) 27 (29%)
30 _+8 26 (43%) 25 (41%) 15 (25%) 12 (20%)
.11 .63 .92 .92 .28
15 (16%) 32 (34%)
3 (5%) 14 (23%)
.029 .19
(94)
P
Mean age (yr) + SD Male sex History of odynophagia Temperature > 38.3 C Tons±liar exudates
27 _+8 30 (40%) 18 (24%) 21 (28%) 34 (45%)
26 + 8 39 (41%) 29 (31%) 27 (29%) 61 (65%)
.36 .97 .41 .94 .016
5 (7%) 30 (40%)
2 (2%) 13 (14%)
.14 < .0011
Test ordering
Test ordering CBC Chest radiograph Mean charge + SD
per patient $50 ± 69 $33 ± 60 .11 *Exclusion criteria: Currentantibiotic therapy,significant comorbidity,vomiting,diarrhea, abdominal pain.
AUGUST1992
Residency
(75)
Comparison factors
Comparisonfactors Mean age (yr) ± SD Male sex Productive cough Chest pain Temperature > 3&3 C
Preresidency
21:8
ANNALS OF EMERGENCY MEDICINE
CBC Throat culture Mean charge _+SD
per patient $22 ± 26 $7.7 + 18 *Exclusion critieria: Currentantibiotic therapy,significant comorbidity. t Statistically significant after appficationof Bonferroni correction (seetext).
< .001+
957/83
RESIDENCY PROGRAM McNamara & Kelly
The remaining physician staff in the preresidency phase of the study (January to March 1989) was comprised of nine emergencyphysicians hired by a regional contract group a n d approved f o r clinical privileges in emergency medicine by the hospital. Three of these nine physicians were certified by the American B o a r d of Internal Medicine, and none was board-certified in any other specialty, including emergency medicine. The staffing p a t t e r n of this group consisted of an attending 24 hours p e r day with 12 hours of double coverage. During the periods reviewed, a limited n u m b e r of patients were examined and treated by house officers from the hospital's internal medicine residency p r o g r a m while u n d e r the supervision of the attending physician. These charts were excluded from analysis. There were no changes in the physical plant of the ED between th e study periods or in ancillary staffing patterns. Importantly, 22,23 both groups of physicians were salaried With no incentive programs for increased charges. In addition, the study was conceived and designed several months after the last period studied, so neither group of physicians was aware of a plan to study their test-ordering behavior. The availability of all diagnostic tests studied was identical in each study period. F o r the six diagnoses studied, no nursing protocols existed in either time period for the routine ordering of tests. The frequency of ordering l a b o r a t o r y and radiographic studies was chosen as a m a r k e r of cost of care, as this is readily measured and is a significant contributor to medical costs. 4 Test utilization has been previously used to study the costs of teaching programs 1-4 as well as to compare the behavior of practicing physicians.9-11,13,22-25 In comparing costs of care between separate groups of physicians, it is i m p o r t a n t to control for case mix, patient demography, and the environment of care. 6,7,13,26 Fortunately, in this study, the setting did not change, and significant demographic shifts in the ED population were not expected as the community served was a stable, u r b a n type. To control for case mix, we chose six common discharge diag-
noses including viral upper respiratory infection, pharyngitis, acute asthma, seizure, lumbosacra] strain, and cervical strain. We chose to study diagnostic categories instead of presenting complaints, as the former would likely represent a more homogeneous group. In deciding on which diagnoses to study, several variables were considered by the authors. First, the diagnosis needed to be common to ensure adequate numbers for study. Second, we sought conditions likely to be treated by the emergency physician without the aid of consultants who might o r d e r tests. Similarly, the diagnostic category h a d to be one in which most patients were discharged home, as it would have been difficult to discern whether the emergency physicians or the admitting team o r d e r e d a test on the patient while in the ED. To study differences in test-ordering behavior, we also avoided diagnoses in which tests are infrequently o r d e r e d (eg, headache) or those in which the obtaining of tests is fairly routine (eg, ankle sprain). A set of inclusion and exclusion criteria for each was developed in an attempt to obtain uniformity within each diagnostic category (Tables 1 through 6). The inclusion and exclusion criteria were specifically chosen in an attempt to narrow the spectrum of patients in each diagnostic category to those in whom the measured test ordering was more at the discretion of the physician r a t h e r than a commonly accepted practice pattern. To further study the issue of comparability of case mix we also conducted age, sex, and p a y o r status comparisons and sought from the written record symptoms or physical signs that might influence test ordering. The tests selected for evaluation under each diagnosis were specifically chosen because of their more discretionary nature. We did not examine test ordering that is routinely conducted (eg, anticonvulsant levels in seizures) to isolate discretionary testing. As physicians other than the emergency physician o r d e r tests on admitted patients, we chose to study only discharged patients. Table 4. Seizures: Aged 15 years or older, recent (< 12 hours) generalized seizure, discharged home
Table 3. Acute asthma: Aged 5 to 50 years, discharged home* Preresidenoy
Residency
(76)
(66)
P
31 + 9 23 (30%) 12 (16%) 4 (5%)
27 +6 31 (47%) 18 (27%) 7 (11%)
.003" .047 .14 ,19
17 (22%) 7 (9%) 11 (14%)
10 (15%) 1 (1.5%) 15 (23%)
.38 .048 .29
per patient $31 -+67 $35 _+65 .663 *Exclusioncriteria: Temperature > 38.3C, current antibiotic therapy,significant comorbidity,first attack of asthma. t Statistically significant after application of Bonferroni correction (seetext).
84/958
(57)
P
Mean age (yr) + SD Male sex History of alcohol abuse Multiple seizures Significant comorbidity Alert on arrival
41 + 14 44 (71%) 33 (53%) 15 (24%) 11 (18%) 57 (92%)
39-+ 14 40 (70%) 24 (42%) 17 (30%) 16(28%) 46 (81%)
.48 .01 .30 .63 .26 .13
30 (48%) 46 (74%) 2 (3.2%) 1 (1.6%)
19 (33%) 28 (49%) 14 (25%) 14 (25%)
.14 .008 < .001 < .001
Test ordering
Test ordering CBC Arterial blood gas Chest radiograph Mean charge + SD
Residency
(62)
Comparison factors
Comparison factors Mean age (yr) + SD Male sex Productive cough Chest pain
Preresidency
CBC Serum chemistry panel t Serum calcium Serum magnesium Mean charge _+SD
per patient $71 _+43 $59 -4-63 *Exclusion critieria: First-time seizure,temperature > 38.3C. Statistically significant after application of Bonferroni correction (seetext).
ANNALS OF EMERGENCY MEDICINE
21:8
.246
AUGUST1992
RESIDENCY
PROGRAM
McNamara & Kelly
The chart of every ED encounter for the six months of the study period was studied by one of the two physician authors. The discharge diagnosis was reviewed, and only those closely related to one of the six study categories were accepted for inclusion. F o r example, tonsillitis was acceptable under pharyngitis but bronchitis was not included under viral u p p e r r e s p i r a t o r y infection. The patient chart then was scrutinized for the other inclusion and exclusion criteria, and data were extracted onto a simple form when appropriate. Illegible charts were excluded. Statistical analysis included the use of Z2 analysis and Fisher's exact test for the frequency of test ordering and nonlinear results. An u n p a i r e d t-test was used for age and overall charge comparisons. Charge comparisons were conducted only for those tests evaluated u n d e r each diagnosis. It was thought that this would be a more accurate representation of test-ordering behavior versus total patient charges, which are influenced by other variables. The hospital charge for each test during the month of April 1990 was used to calculate test charges for both groups. Statistical analysis involving multiple comparisons between groups raises the possibility of committing a Type ] e r r o r and incorrectly claiming statistical significance. The main results of this study are the mean overall charge comparisons between groups that were analyzed using an u n p a i r e d t-test. To account for multiple comparisons regarding this, we used the deliberately conservative Bonferroni adjustment. In this method, the chosen level of significance (P < .05) is divided by the number of comparisons (six) between groups to arrive at a required level of significance of P < .008 for mean charges under each diagnostic category. Similarly, in the statistical analysis of the groups regarding comparison factors and individual tests ordered, an adjustment must be made for multiple comparisons. As these are not the main study measures, a p a r t i a l Bonferroni correction was used that required a level 6f significance of P < .005 for each item.
RESULTS The total n u m b e r of patients who presented to the El) and their disposition during the two study periods are shown (Table 7). The total number of patients entered in the study during the preresidency period was 435, whereas 368 were entered during the residency period. These numbers represent 8.8% and 7.1%, respectively, of the patients treated and discharged during the study time. The main results of the study are presented (Tables 1 through 6). P a y o r status and racial composition of all diagnostic categories were similar between groups. Of the history and physical comparison factors used, the only one achieving a statistically significant difference after application of the p a r t i a l Bonferroni correction was the higher mean age of asthmatics in the preresidency period (P = .003). As can be seen in the tables, the only tests o r d e r e d more frequently by the residency group were serum calcium and magnesium in seizure patients. The preresidency group, after applying the Bonferroni adjustment of P < . 005, however, ordered significantly more throat cultures in pharyngitis (P < .001) and radiographs in both lumbosacral (P < .001) and cervical strain (P < .001). Charge comparisons are reported as mean charge p e r patient in each group only for those tests looked at in the study. The residency group charges did not exceed those of the preresidency group for any diagnoses and were significantly lower for three of the six diagnoses. Although the main focus of this study was to look at the ordering of tests that are discretionary in nature, we did collect information on two routinely ordered tests, anticonvulsant levels in seizures and theophylline levels in asthma. The n u m b e r of anticonvulsant levels ordered in seizure patients was 52 (91%) for the residency group and 52 (84%) for the preresidency group. The residency group o r d e r e d theophylline levels in 12 (18%) of the asthmatics, and the preresidency groups obtained the test in 22 (29%) of these patients. The differences in the groups for these tests were not statistically significant and were consistent with the results obtained for discretionary testing.
Table 5. Lumbosacral strain: Aged 15 to 50 years, ambulatory, alert; mechanism of injury either motor vehicle accident, lift, pull, or twist; discharged home
Preresidency (46)
Residency (41)
P
31 + 9 39 (85%)
28 +_8 26 (63%)
.053 .041
38 (83%) 20 (43%)
31 (76%) 11 (27%)
.59 .16
35 (76%)
6 (15%)
< .001t
Table 6. Cervical strain: Aged 15 to 50 years, ambulatory, alert; mechanism of injury was motor vehicle accident; discharged home
Comparisonfactors Mean age (yr) _+SD Male sex Motor vehicle accident as mechanism Midline low back tenderness
Residency (49)
30 + 11 50 (61%) 15 (18%)
30 + 10 24 (49%) 13(26%)
.93 .25 .16
71 (87%)
18 (37%)
< .O01t
P
Comparison factors Mean age (yr) + SD Male sex Midline cervicaltenderness
Test ordering
Test ordering Lumbosacra[ radiograph Mean charge _+SD
per patient $264 + 150 $51 _+124 < .OOl* • Exclusion critieria: Significant underlying illness or associated injury, abnormal neurologic examination. t Statistically significant after application of Bonferroni correction (see text).
AUGUST1992
Preresidency (82)
21:8
ANNALS OF EMERGENCY MEDICINE
Cervical-spine radiograph
Mean charge + SD per patient
$219 +87 $93 + 123 < .001t *Exclusion cdtieria: Significant underlying illness or associated injury, abnormal neurologic examination. Statistically significant after application of Bonferroni correction (see text).
959/85
RESIDENCY PROGRAM McNamara & Kelly
During the residency period, 55% of the entered patients were evaluated primarily by emergency medicine residents. The number of patients excluded because they were initially examined by the hospital's internal medicine house officers was 16 and 14, respectively, in the preresidency and residency periods. No charts were excluded for illegibility.
DISCUSSION The results indicate that the introduction of an emergency medicine residency p r o g r a m into this community hospital did not increase the cost of care as measured by test use for six common diagnoses. Actual charges were significantly less in the residency period for three of the six categories, with a fivefold reduction in the number of himbosacral spine radiographs and more than halving the rate of cervical-spine radiography. Cost of care was assessed in this study by focusing on the "little ticket" items of l a b o r a t o r y testing and radiography. Although individually not costly, the use of such diagnostic tests has a m a j o r impact on health care costs,a, 28 comprising one fourth of all hospital costs.28, 29 Most of the previous research on test use has been conducted on inpatients, but outpatient care is accounting for a steadily increasing proportion of health care expenditures.X2,16 Similar to inhospital costs, l a b o r a t o r y tests and radiographs account for approximately 25% of ambulatory care expenses in the United States. 22 The results of our study conflict with most of the previous research on the effect of teaching programs by not finding an increase in the cost of care. 1-8 Hueston,30 however, found that the introduction of family practice residents did not increase inpatient charges. It has generally been found that physician test use is inversely related to e x p e r i e n c e j °-12,32 and the large percentage of patients p r i m a r i l y managed by residents in the second period of our study would have been expected to raise costs. However, the structure of resident experience in accredited emergency medicine programs with continuous physical presence of supervising faculty creates a unique environment in which test use by the house officer is directly monitored for each patient. As test-ordering behavior of housestaff is significantly influenced by their attending staff32 and the individual physician behaves in a manner similar to clinical leaders,12, 24 it is likely that our study was predominantly comparing attending test use.
Table 7. ED patients and disposition (January to March 1989 and 1990) Preresidency Admitted/transferred*
Residency
1,101 (18%) 1,212 (19%) Left without treatment 353 244 Leftwhile receiving treatment 42 17 Discharged home 4,898 5,144 Died 52 39 Total ED patients 6,446 6,656 * Percentageof all EDpatientsexcludingthosewho left withouttreatment,.
86/960
P .24 < .001
emergency medicine residency, cost
C o s t of Care in the Emergency Department: Impact of an Emergency Medicine Residency Program From the Department of Emergency Medicine, The Medical College of Pennsylvania, Philadelphia. Received for publication October 16, 1991. Revision received February 5, 1992. Accepted for publication March 2, 1992. Presented in part at the Society for Academic Emergency Medicine Annual Meeting in Washington, DC, May 1991.
Robert M McNamara, MD, FACEP John J Kelly, DO, FACEP
Study objective: To evaluate the impact of an emergency medicine residency training program on the cost of care in the emergency department. Design: A retrospective chart review was conducted of all ED encounters for a three-month period, six months before and six months after the introduction of an emergency medicine residency program into an urban community hospital. Physician staffing of this ED before the residency period was by nonemergency medicine residency-trained emergency physicians. Setting:
A 27,O00-visit-per-year urban community hospital ED.
Type of patients: A consecutive sample of all patients discharged home from the emergency center with one of six diagnoses. The diagnoses studied were viral upper respiratory infection, pharyngitis, acute asthma, seizure, lumbosacral strain, and cervical strain. outcome measures: Frequency of laboratory test and radiograph ordering pertinent to the evaluation of each diagnostic category were used as a marker of cost of care. Main
Results: The presence of the residency training program did not increase the cost of care as measured by test use and, for three of the six diagnoses, actually lowered the cost of care, This effect was most prominent in the evaluation of lumbosacral and cervical strain when the residency physicians ordered radiographs at a rate five and 2.3 times lower, respectively, than the previous group and in the approach to pharyngitis when they ordered throat cultures 2.8 times less frequently. Conclusion: As measured by selected test use for six common discharge diagnoses, the introduction of an emergency medicine residency program did not increase the cost of care in this urban community hospital ED. [McNamara RM, Kelly J J: Cost of care in the emergency department: Impact of an emergency medicine residency program. Ann ErnergMed August 1992;21:956-962.]
8 2/ 9 5 6
ANNALS OF EMERGENCY MEDICINE 21:8 AUGUST 1992
RESIDENCY P R O G R A M McNamara & Kelly
INTRODUCTION
MATERIALS AND METHODS
The presence of medical teaching programs has generally been found to increase the costs of inpatient care. 1-8 The introduction of more extensively trained physicians to an inpatient area has been shown to increase costs. 9 In the outpatient setting, residency-trained physicians o r d e r more diagnostic tests than general practitioners with specialists having the highest rate of test use. lo-12 Training in academically oriented medical schools or residency programs has been found to increase test ordering for ambulatory patients. 13 The effect of teaching programs or more extensive training on the cost of care has not been studied in the emergency department setting. The emergency physician is playing an increasingly important role in the overall health care cost picture. In the decade from 1980 to 1989 yearly visits to the nation's EDs increased from 82 million to nearly 90 million. 14,15 Since 1967, expenditures for outpatient and ED care have been the most rapidly growing component of health care costs. 16 In 1989, these areas represented 19% of all hospital costs and 7% of the total health care expenditures for the nation. 16 The ED also has become a m a j o r source of p r i m a r y care,17,18 particularly in those areas in which access to other forms of p r i m a r y care is limited. 19 The impact of emergency physicians on cost extends to the inpatient arena, as they often determine the initial stages of evaluation and treatment of the patients admitted.t8 Residency training programs in emergency medicine have existed since the early 1970s. The effect of specialists in emergency medicine on the cost of care in the ED has not been studied, b u t some have suggested that their presence increases costs. 20,21 The purpose of this research was to study the impact of an emergency medicine residency program with primarily emergency medicine residency-trained faculty on the cost of care in the ED. This was made possible when such a p r o g r a m assumed responsibility for physician staffing of an u r b a n community hospital ED from a group of nonresidency-trained emergency physicians.
On July 1, 1989, the emergency medicine residency program of the Medical College of Pennsylvania began providing physician coverage of a 27,000-visit u r b a n community hospital ED. This facility had been previously staffed by emergency physicians supplied to the hospital by a regional contract group. This n a t u r a l experiment afforded an excellent opportunity to study the impact of a teaching p r o g r a m on the cost of care in the ED. A retrospective chart review was conducted on a consecutive sample of patients t r e a t e d in the ED from J a n u a r y through March of 1989 and 1990 who were discharged home with one of six common diagnoses. The cost of care was assessed by the performance frequency of selected tests pertinent to the evaluation of each diagnosis. Overall charges for each diagnostic category were also measured and compared. An almost complete change in physician staffing allowed for direct comparison of two groups of emergency physicians in relation to the cost of care. F r o m July 1, 1989, on, physician coverage of the ED consisted of continuous presence of one faculty member and one resident from the residency program 24 hours a day. During the residency study period ( J a n u a r y to March 1990), there was an additional ten hours of resident coverage on 62 (69%) of the 90 days. The residents were generally in their second or third y e a r of training in emergency medicine. Thirteen of the 16 emergency medicine faculty who provided coverage during the study period had completed a residency in this specialty. Of the three remaining faculty, all were board-certified in internal medicine, with one also board-certified in emergency medicine and another also board-certified in pediatrics. The large number of faculty is because most rotate frequently between the three EDs within the residency program. One of the previous group's physicians was retained as clinical faculty during the changeover and all patients managed or supervised by this physician were excluded from analysis in each of the two study periods.
Table 2. Pharyngitis: Aged 5 to 50 years, discharged home*
Table 1. Viral upper respiratory infection: Aged 15 to 50 years, discharged home*
Preresidency (94)
Residency (61)
P
28 ± 9 45 {48%) 38 (40%) 25 (27%) 27 (29%)
30 _+8 26 (43%) 25 (41%) 15 (25%) 12 (20%)
.11 .63 .92 .92 .28
15 (16%) 32 (34%)
3 (5%) 14 (23%)
.029 .19
(94)
P
Mean age (yr) + SD Male sex History of odynophagia Temperature > 38.3 C Tons±liar exudates
27 _+8 30 (40%) 18 (24%) 21 (28%) 34 (45%)
26 + 8 39 (41%) 29 (31%) 27 (29%) 61 (65%)
.36 .97 .41 .94 .016
5 (7%) 30 (40%)
2 (2%) 13 (14%)
.14 < .0011
Test ordering
Test ordering CBC Chest radiograph Mean charge + SD
per patient $50 ± 69 $33 ± 60 .11 *Exclusion criteria: Currentantibiotic therapy,significant comorbidity,vomiting,diarrhea, abdominal pain.
AUGUST1992
Residency
(75)
Comparison factors
Comparisonfactors Mean age (yr) ± SD Male sex Productive cough Chest pain Temperature > 3&3 C
Preresidency
21:8
ANNALS OF EMERGENCY MEDICINE
CBC Throat culture Mean charge _+SD
per patient $22 ± 26 $7.7 + 18 *Exclusion critieria: Currentantibiotic therapy,significant comorbidity. t Statistically significant after appficationof Bonferroni correction (seetext).
< .001+
957/83
RESIDENCY PROGRAM McNamara & Kelly
The remaining physician staff in the preresidency phase of the study (January to March 1989) was comprised of nine emergencyphysicians hired by a regional contract group a n d approved f o r clinical privileges in emergency medicine by the hospital. Three of these nine physicians were certified by the American B o a r d of Internal Medicine, and none was board-certified in any other specialty, including emergency medicine. The staffing p a t t e r n of this group consisted of an attending 24 hours p e r day with 12 hours of double coverage. During the periods reviewed, a limited n u m b e r of patients were examined and treated by house officers from the hospital's internal medicine residency p r o g r a m while u n d e r the supervision of the attending physician. These charts were excluded from analysis. There were no changes in the physical plant of the ED between th e study periods or in ancillary staffing patterns. Importantly, 22,23 both groups of physicians were salaried With no incentive programs for increased charges. In addition, the study was conceived and designed several months after the last period studied, so neither group of physicians was aware of a plan to study their test-ordering behavior. The availability of all diagnostic tests studied was identical in each study period. F o r the six diagnoses studied, no nursing protocols existed in either time period for the routine ordering of tests. The frequency of ordering l a b o r a t o r y and radiographic studies was chosen as a m a r k e r of cost of care, as this is readily measured and is a significant contributor to medical costs. 4 Test utilization has been previously used to study the costs of teaching programs 1-4 as well as to compare the behavior of practicing physicians.9-11,13,22-25 In comparing costs of care between separate groups of physicians, it is i m p o r t a n t to control for case mix, patient demography, and the environment of care. 6,7,13,26 Fortunately, in this study, the setting did not change, and significant demographic shifts in the ED population were not expected as the community served was a stable, u r b a n type. To control for case mix, we chose six common discharge diag-
noses including viral upper respiratory infection, pharyngitis, acute asthma, seizure, lumbosacra] strain, and cervical strain. We chose to study diagnostic categories instead of presenting complaints, as the former would likely represent a more homogeneous group. In deciding on which diagnoses to study, several variables were considered by the authors. First, the diagnosis needed to be common to ensure adequate numbers for study. Second, we sought conditions likely to be treated by the emergency physician without the aid of consultants who might o r d e r tests. Similarly, the diagnostic category h a d to be one in which most patients were discharged home, as it would have been difficult to discern whether the emergency physicians or the admitting team o r d e r e d a test on the patient while in the ED. To study differences in test-ordering behavior, we also avoided diagnoses in which tests are infrequently o r d e r e d (eg, headache) or those in which the obtaining of tests is fairly routine (eg, ankle sprain). A set of inclusion and exclusion criteria for each was developed in an attempt to obtain uniformity within each diagnostic category (Tables 1 through 6). The inclusion and exclusion criteria were specifically chosen in an attempt to narrow the spectrum of patients in each diagnostic category to those in whom the measured test ordering was more at the discretion of the physician r a t h e r than a commonly accepted practice pattern. To further study the issue of comparability of case mix we also conducted age, sex, and p a y o r status comparisons and sought from the written record symptoms or physical signs that might influence test ordering. The tests selected for evaluation under each diagnosis were specifically chosen because of their more discretionary nature. We did not examine test ordering that is routinely conducted (eg, anticonvulsant levels in seizures) to isolate discretionary testing. As physicians other than the emergency physician o r d e r tests on admitted patients, we chose to study only discharged patients. Table 4. Seizures: Aged 15 years or older, recent (< 12 hours) generalized seizure, discharged home
Table 3. Acute asthma: Aged 5 to 50 years, discharged home* Preresidenoy
Residency
(76)
(66)
P
31 + 9 23 (30%) 12 (16%) 4 (5%)
27 +6 31 (47%) 18 (27%) 7 (11%)
.003" .047 .14 ,19
17 (22%) 7 (9%) 11 (14%)
10 (15%) 1 (1.5%) 15 (23%)
.38 .048 .29
per patient $31 -+67 $35 _+65 .663 *Exclusioncriteria: Temperature > 38.3C, current antibiotic therapy,significant comorbidity,first attack of asthma. t Statistically significant after application of Bonferroni correction (seetext).
84/958
(57)
P
Mean age (yr) + SD Male sex History of alcohol abuse Multiple seizures Significant comorbidity Alert on arrival
41 + 14 44 (71%) 33 (53%) 15 (24%) 11 (18%) 57 (92%)
39-+ 14 40 (70%) 24 (42%) 17 (30%) 16(28%) 46 (81%)
.48 .01 .30 .63 .26 .13
30 (48%) 46 (74%) 2 (3.2%) 1 (1.6%)
19 (33%) 28 (49%) 14 (25%) 14 (25%)
.14 .008 < .001 < .001
Test ordering
Test ordering CBC Arterial blood gas Chest radiograph Mean charge + SD
Residency
(62)
Comparison factors
Comparison factors Mean age (yr) + SD Male sex Productive cough Chest pain
Preresidency
CBC Serum chemistry panel t Serum calcium Serum magnesium Mean charge _+SD
per patient $71 _+43 $59 -4-63 *Exclusion critieria: First-time seizure,temperature > 38.3C. Statistically significant after application of Bonferroni correction (seetext).
ANNALS OF EMERGENCY MEDICINE
21:8
.246
AUGUST1992
RESIDENCY
PROGRAM
McNamara & Kelly
The chart of every ED encounter for the six months of the study period was studied by one of the two physician authors. The discharge diagnosis was reviewed, and only those closely related to one of the six study categories were accepted for inclusion. F o r example, tonsillitis was acceptable under pharyngitis but bronchitis was not included under viral u p p e r r e s p i r a t o r y infection. The patient chart then was scrutinized for the other inclusion and exclusion criteria, and data were extracted onto a simple form when appropriate. Illegible charts were excluded. Statistical analysis included the use of Z2 analysis and Fisher's exact test for the frequency of test ordering and nonlinear results. An u n p a i r e d t-test was used for age and overall charge comparisons. Charge comparisons were conducted only for those tests evaluated u n d e r each diagnosis. It was thought that this would be a more accurate representation of test-ordering behavior versus total patient charges, which are influenced by other variables. The hospital charge for each test during the month of April 1990 was used to calculate test charges for both groups. Statistical analysis involving multiple comparisons between groups raises the possibility of committing a Type ] e r r o r and incorrectly claiming statistical significance. The main results of this study are the mean overall charge comparisons between groups that were analyzed using an u n p a i r e d t-test. To account for multiple comparisons regarding this, we used the deliberately conservative Bonferroni adjustment. In this method, the chosen level of significance (P < .05) is divided by the number of comparisons (six) between groups to arrive at a required level of significance of P < .008 for mean charges under each diagnostic category. Similarly, in the statistical analysis of the groups regarding comparison factors and individual tests ordered, an adjustment must be made for multiple comparisons. As these are not the main study measures, a p a r t i a l Bonferroni correction was used that required a level 6f significance of P < .005 for each item.
RESULTS The total n u m b e r of patients who presented to the El) and their disposition during the two study periods are shown (Table 7). The total number of patients entered in the study during the preresidency period was 435, whereas 368 were entered during the residency period. These numbers represent 8.8% and 7.1%, respectively, of the patients treated and discharged during the study time. The main results of the study are presented (Tables 1 through 6). P a y o r status and racial composition of all diagnostic categories were similar between groups. Of the history and physical comparison factors used, the only one achieving a statistically significant difference after application of the p a r t i a l Bonferroni correction was the higher mean age of asthmatics in the preresidency period (P = .003). As can be seen in the tables, the only tests o r d e r e d more frequently by the residency group were serum calcium and magnesium in seizure patients. The preresidency group, after applying the Bonferroni adjustment of P < . 005, however, ordered significantly more throat cultures in pharyngitis (P < .001) and radiographs in both lumbosacral (P < .001) and cervical strain (P < .001). Charge comparisons are reported as mean charge p e r patient in each group only for those tests looked at in the study. The residency group charges did not exceed those of the preresidency group for any diagnoses and were significantly lower for three of the six diagnoses. Although the main focus of this study was to look at the ordering of tests that are discretionary in nature, we did collect information on two routinely ordered tests, anticonvulsant levels in seizures and theophylline levels in asthma. The n u m b e r of anticonvulsant levels ordered in seizure patients was 52 (91%) for the residency group and 52 (84%) for the preresidency group. The residency group o r d e r e d theophylline levels in 12 (18%) of the asthmatics, and the preresidency groups obtained the test in 22 (29%) of these patients. The differences in the groups for these tests were not statistically significant and were consistent with the results obtained for discretionary testing.
Table 5. Lumbosacral strain: Aged 15 to 50 years, ambulatory, alert; mechanism of injury either motor vehicle accident, lift, pull, or twist; discharged home
Preresidency (46)
Residency (41)
P
31 + 9 39 (85%)
28 +_8 26 (63%)
.053 .041
38 (83%) 20 (43%)
31 (76%) 11 (27%)
.59 .16
35 (76%)
6 (15%)
< .001t
Table 6. Cervical strain: Aged 15 to 50 years, ambulatory, alert; mechanism of injury was motor vehicle accident; discharged home
Comparisonfactors Mean age (yr) _+SD Male sex Motor vehicle accident as mechanism Midline low back tenderness
Residency (49)
30 + 11 50 (61%) 15 (18%)
30 + 10 24 (49%) 13(26%)
.93 .25 .16
71 (87%)
18 (37%)
< .O01t
P
Comparison factors Mean age (yr) + SD Male sex Midline cervicaltenderness
Test ordering
Test ordering Lumbosacra[ radiograph Mean charge _+SD
per patient $264 + 150 $51 _+124 < .OOl* • Exclusion critieria: Significant underlying illness or associated injury, abnormal neurologic examination. t Statistically significant after application of Bonferroni correction (see text).
AUGUST1992
Preresidency (82)
21:8
ANNALS OF EMERGENCY MEDICINE
Cervical-spine radiograph
Mean charge + SD per patient
$219 +87 $93 + 123 < .001t *Exclusion cdtieria: Significant underlying illness or associated injury, abnormal neurologic examination. Statistically significant after application of Bonferroni correction (see text).
959/85
RESIDENCY PROGRAM McNamara & Kelly
During the residency period, 55% of the entered patients were evaluated primarily by emergency medicine residents. The number of patients excluded because they were initially examined by the hospital's internal medicine house officers was 16 and 14, respectively, in the preresidency and residency periods. No charts were excluded for illegibility.
DISCUSSION The results indicate that the introduction of an emergency medicine residency p r o g r a m into this community hospital did not increase the cost of care as measured by test use for six common diagnoses. Actual charges were significantly less in the residency period for three of the six categories, with a fivefold reduction in the number of himbosacral spine radiographs and more than halving the rate of cervical-spine radiography. Cost of care was assessed in this study by focusing on the "little ticket" items of l a b o r a t o r y testing and radiography. Although individually not costly, the use of such diagnostic tests has a m a j o r impact on health care costs,a, 28 comprising one fourth of all hospital costs.28, 29 Most of the previous research on test use has been conducted on inpatients, but outpatient care is accounting for a steadily increasing proportion of health care expenditures.X2,16 Similar to inhospital costs, l a b o r a t o r y tests and radiographs account for approximately 25% of ambulatory care expenses in the United States. 22 The results of our study conflict with most of the previous research on the effect of teaching programs by not finding an increase in the cost of care. 1-8 Hueston,30 however, found that the introduction of family practice residents did not increase inpatient charges. It has generally been found that physician test use is inversely related to e x p e r i e n c e j °-12,32 and the large percentage of patients p r i m a r i l y managed by residents in the second period of our study would have been expected to raise costs. However, the structure of resident experience in accredited emergency medicine programs with continuous physical presence of supervising faculty creates a unique environment in which test use by the house officer is directly monitored for each patient. As test-ordering behavior of housestaff is significantly influenced by their attending staff32 and the individual physician behaves in a manner similar to clinical leaders,12, 24 it is likely that our study was predominantly comparing attending test use.
Table 7. ED patients and disposition (January to March 1989 and 1990) Preresidency Admitted/transferred*
Residency
1,101 (18%) 1,212 (19%) Left without treatment 353 244 Leftwhile receiving treatment 42 17 Discharged home 4,898 5,144 Died 52 39 Total ED patients 6,446 6,656 * Percentageof all EDpatientsexcludingthosewho left withouttreatment,.
86/960
P .24 < .001
