Original Contribution

Predictors and outcomes of frequent chest pain unit users

Miquel Sanchez MD, PhD?, Beatriz Lopez MD, Ernest Bragulat MD, PhD,

Elisenda Gomez-Angelats MD, PhD, Sonia Jimenez MD, PhD, Mar Ortega MD, PhD, Blanca Coll-Vinent MD, PhD, Oscar Miro MD, PhD

Seccio d’Urgencies Medicina, Area d’Urgencies, Hospital Clinic, Universitat de Barcelona, Barcelona, 08036 Catalonia, Spain

Received 25 January 2008; revised 9 May 2008; accepted 13 May 2008

Abstract

Aim: To determine predictors of frequent chest pain unit (CPU) users and to identify characteristics and outcomes of their CPU visits.

Patients and Methods: Observational prospective case-control study. Frequent CPU user was defined by 3 or more CPU visits within the study year. A control patient and a control visit were randomly selected for each case patient and case visit. Demographic, clinical, and outcome variables were collected from medical record and phone interview performed in a 30-day interval. A multivariate logistic regression analysis was used to identify frequent CPU users? predictors.

Results: Of 1934 patients presenting during the year, 80 (4.1%) met the definition for case patient. They accounted for 352 (13%) of 2709 CPU visits. Sixty-seven (83.7%) case patients and 71 (88.7%) control patients were contacted. The final predictors were the following: Karnofsky Performance Scale of 70 or lesser (odds ratio [OR], 5.24 [95% confidence interval {CI}, 1.71-16.06]), previous hospitalization (OR, 3.76 [95% CI, 1.49-9.49]), previously known coronary artery disease (OR, 3.72 [95% CI, 1.32-10.52]), and symptoms of depression (OR, 2.98 [95% CI, 1.14-7.78]). Case visits were more likely at night (OR, 2.41 [95% CI, 1.64- 3.52]), generated more diagnostic uncertainty (OR, 2.39 [95% CI, 1.71-3.35]), but did not increase the need of hospital admission.

Conclusions: Frequent CPU user is associated with previously known coronary artery disease, previous hospitalization, impaired performance status, and presence of symptoms of depression. They are more likely to arrive on CPU at night and generate more diagnostic uncertainty.

(C) 2009

Introduction

In most industrialized countries, emergency departments (EDs) receive a large number of visits as they play a pivotal role as society’s health care safety net for people seeking urgent medical attention. Unfortunately, the high demand is often not adjusted to the supply, and overcrowding in EDs has become a worldwide problem [1-5]. This situation ultimately leads to

* Corresponding author. Urgencies. Hospital Clinic, 08036 Barcelona, Spain. Tel.: +34 93 227 9833; fax: +34 93 227 5693.

E-mail address: [email protected] (M. Sanchez).

deterioration in the quality of care with a huge number of patients waiting to be seen by a physician [6,7]. Patients with chest pain, who can reach more than 5% of ED visits in some areas [8], have been long considered in the group of “immediate care” because they are at risk for a life-threatening problem, and their outcome has directly been related to the prompt beginning of the treatment [9]. Chest pain units (CPUs) have been proven both to provide fast and efficient protocol- driven diagnostic testing for patients with chest pain and avert poor outcome when ED overcrowding is present [10,11].

Some authors have put the blame on a specific group of patients, the “frequent” or “heavy” users as among the factors

0735-6757/$ – see front matter (C) 2009 doi:10.1016/j.ajem.2008.05.007

contributing to ED overcrowding [12,13]. These patients have been defined as a group who accounts for a disproportionate number of ED visits [14]. They have been characterized by a special social and economic distress [12,13,15,16] and high rates of chronic physical illness as well as psychiatric distress [15,17]. Furthermore, frequent ED use has been correlated with adverse clinical outcomes such as increased mortality [15,17], admission rate [17], and use of other Health care resources [18]. If this phenomenon were also present in the group of chest pain patients attending either CPUs or EDs, their study could be an essential point to better know their profile, to find out their potential interference in the medical attention of other chest pain patients, and to help manage care organizations and administration control costs.

The objective of this study was to quantify the number of frequent CPU users and their visits, to determine their demographic and clinical characteristics, and to identify independent predictors as to be at risk for becoming a frequent CPU user. Demographic and clinical characteristics along with outcomes of their visits were also analyzed.

Patients and methods

Study design

This was an observational prospective case-control analysis of all patients attending our CPU from July 1, 2003, to June 30, 2004.

Frequent CPU user (case patient) was defined by 3 or more CPU visits within the study year. These patients and their number of visits were automatically identified by our CPU Tracking system. Each CPU visit constituted 1 independent “case visit.”

All the other patients seen in the CPU within the study period were eligible as control patients. All CPU visits other than those from case patients were eligible as control visits. Epidemiological, clinical, and outcome variables were collected from medical records. Some specific demographic variables not included in medical record were queried by means of a prescribed telephone questionnaire that was done after 30 days from the ED visit. There were at least 5 attempts to reach each patient included in the study. Each case visit was paired with a randomly selected control visit.

This simple random sampling of controls was chosen because the source population was explicitly identified in this study. This design was preferred to match controls to avoid the risk of overmatching cases and controls and the bias due to the loss of the effect of the matched variables. The study was approved by our institutional review board.

Study setting and population

Our hospital is an urban 700-bed tertiary care adult teaching hospital that treats approximately 125000 patients

per year in the ED. It is made up of 3 main areas: medicine, surgery, and orthopedics. Chest pain unit is located inside the medicine area along with the emergent, urgent, and nonurgent areas. Depending on the acuity of the complaint, all patients are triaged to the appropriate care provider area. Accordingly, any patient older than 18 years complaining of a nontraumatic chest pain is seen in our CPU, which is equipped with 7 stretchers and staffed with 2 registered nurses, an attending physician, and a first- or second-year medicine resident that does not take care of or make decisions on patients without direct attending supervision. All stretchers are prepared to be monitored. According to Spanish Cardiology Society Guidelines [19], once initial clinical evaluation and first electrocardiogram (ECG) is carried out, all patients are categorized into one of these initial protocols: (1) ST elevation myocardial infarction ; (2) definitive acute coronary syndrome (ACS) without ST elevation, where most patients are treated and admitted (if in-hospital bed is available); (3) possible ACS; and (4) noncoronary chest pain, where disposition is pronounced depending on the final diagnosis. All patients from group 3, and those from group 2 remaining in the CPU because of bed unavailability, meet protocol criteria for Continuous ECG monitoring and serial cardiac troponin I determination. Patients from group 3 (possible ACS) with symptom recurrence and/or ECG changes at anytime, and/or elevated Troponin I levels, are finally admitted as non-STEMI or unstable angina. All the remaining patients are considered low-risk patients for ACS and undergo an exercise Stress testing. Those with a negative result are finally discharged.

Measurements

For each case patient and control patient, the following variables were analyzed: age, sex, marital status, education, work status, living situation, Cardiovascular risk factors (hypertension, diabetes, hypercholesterolemia, and current smoking), previously known history of coronary artery disease (CAD), previous history of psychiatric disorder, previous admittance to the hospital for any reason other than CAD, functional or performance status by means of the Karnofsky Performance Scale [20], and potential depression mood by means of the Goldberg Depression Scale [21]. This scale is ideal for use by telephone because it is brief, well accepted, and of satisfactory reliability and validity [22]. All these variables were obtained either from the medical record or from a prescribed phone questionnaire (Table 1).

For each case visit and control visit, we collected the following from the medical record: the weekday, part of the day, referred to or not by other physician, initial protocol, final diagnosis, and disposition. For further analysis, the variable “initial protocol” was dichotomized according to the need or not, once initial clinical assessment and first ECG were performed, of further CPU observation for a definitive ACS rule-out or rule-in diagnosis (no observation needed– protocols 1, 2, and 4 rule out or in an ACS once initial

Variables	Obtained from	Case patients (n = 67)	Control patients (n = 71)	P
Age (y)	MR	69.9 +- 15.3	59.6 +- 20.4	.003
Sex, male (%)	MR	55.2	56.3	.735
Marital status (%)	PPQ			.881
Single		20.9	22.5
Married		47.8	50.7
Divorced/separated		4.5	5.6
Widowed		26.9	21.1
Education (%)	PPQ			.011
None		26.9	11.3
Primary school		55.2	49.3
High school		10.4	15.5
College or higher		7.5	23.9
Work status (%)	PPQ			.018
Student		—	7
Employee		16.4	31
Unemployed		3	2.8
Retired person		80.6	59.2
Living situation (%)	PPQ			.718
Lives alone		23.9	18.3
With children only		3	2.8
With other adults		73.1	78.9
Cardiovascular risk factors (%)	MR/PPQ
Hypertension		70.1	53.5	.046
Diabetes mellitus		23.9	16.9	.308
Hypercholesterolemia		32.7	33.8	.849
Current smoking		16.4	23.9	.272
Health status	MR/PPQ
Previously known CAD (%)		76.1	40.8	b.001
Previous psychiatric disorder (%)		28.8	23.9	.520
Karnofsky Performance Scale (mean +- SD)		67.4 +- 24.5	86.1 +- 16.2	b.001
Goldberg Depression Scale (mean +- SD)		31.7 +- 23.3	9.3 +- 16.3	b.001
Previous hospital admission (%)		82.1	35.2	b.001
MR indicates medical record; PPQ, prescribed phone questionnaire.

clinical assessment and first ECG are performed–and observation needed–protocol 3 or possible ACS).

Table 1 Demographic characteristics and clinical features of patients for the 2 groups: frequent CPU users (case patients) and controls (control patients)

Statistical analysis

Categorical variables were expressed as percentages, and continuous variables were expressed as the mean +- SD. The normality distribution of variables was assessed by means of Kolmogorov-Smirnov test. For comparisons, the ?2 test or Fisher exact test (categorical variables) and the Student t test or, alternatively, Mann-Whitney U test (continuous variables) were used as statistically appropriate. Some categorical and continuous variables having a P b.10 in the univariate analysis were dichotomized for multivariate analysis. The crude associations between independent variables and frequent CPU users and between independent variables and frequent CPU user visits were estimated by the odds ratios (ORs) and the 95% confidence interval (CI) limit. Finally, a multivariate

stepwise forward logistic regression analysis was used to identify frequent CPU user determinants or predictors. The adjusted ORs and their 95% CIs were obtained from the final model and used for substantive interpretation. The same logistic regression techniques were applied to find out frequent CPU user visit characteristics. Model calibration was assessed by the Hosmer-Lemeshow “goodness-of-fit statistic” and discrimination with receiver operating curve (ROC) analysis. All reported P values were 2 tailed, and P b.05 was considered to indicate statistical significance. The statistical analysis was performed using the SPSS software package (SPSS 10.0 version; SPSS Inc, Chicago, Ill).

Results

Of 1934 patients presenting during the study time, 80 (4.1%) met the definition for frequent CPU user or case

patient. They accounted for 352 (13%) of 2709 CPU visits registered within that period. Therefore, for analysis purpose, 2 random samples, 1 of 80 control patients and 1 of 352 control visits, were selected. Sixty-seven case patients (83.7%) and 71 control patients (88.7%) were contacted and accepted to participate. Their Demographic and clinical data along with the univariate analysis are listed in Table 1. The remaining patients either refused to respond the phone interview or were not contacted.

Continuous variables having P b .10 were dichotomized as follows: age according to the median of the pooled distribution of cases and controls, Karnofsky Performance Scale according to whether the patient was able to carry on normal activity (score, N70 points) and Goldberg Depression Scale according to whether the patient was suffering or not from any symptom of depression. The ORs of all these variables along with categorical variables with P b .10 are shown in Table 2.

The unadjusted multivariate analysis found 4 variables statistically significant. Once adjustment by age (the major potential confounder) was made, the final set of determinants that increased the odds of being a frequent CPU user was the

Table 2 Likelihood of being a frequent CPU user (case patient)

Variables Univariate analysis

? coefficient

OR (95% CI)

P

following (Table 3): a Karnofsky Performance Scale of 70 or lesser (OR, 5.24 [95% CI, 1.71-16.06]), having a previous hospitalization for any reason other than CAD (OR, 3.76 [95% CI, 1.49-9.49]), having previously known CAD (OR, 3.72 [95% CI, 1.31-10.52]), and experiencing any symptom

of depression (OR, 2.98 [95% CI, 1.14-7.78]). The Hosmer- Lemeshow statistic was 2.39 (P = .98). The area under the ROC was 0.84 (95% CI, 0.78-0.91).

Regarding the CPU visits, demographic and clinical characteristics along with outcomes of both groups are listed in Table 4. More detailed information on initial and final diagnosis and disposition of case visits depending on the order of the visit is shown in Table 5. Apart from patients with STEMI at the First visit, 2 new patients were diagnosed with STEMI, one at the third and another at the fourth CPU attendance. In both cases, elapsed time since the previous visit was at least 6 weeks. At those visits, 1 patient had been diagnosed with unstable angina and admitted, and the other was a well-known coronary patient with nonrevascularizable CAD and then sent home. It is also remarkable that the percentage of case visits and control visits with ACS as final diagnosis (30.1% and 33.5%, respectively) was higher than

Age
<=65 y		1 (reference group)
N65 y	.899	2.456 (1.212-4.977)	.013
Education
None		1 (reference group)
Primary school	-.755	0.470 (0.181-1.218)	.120
High school	-.631	0.532 (0.283-0.999)	.050
College or higher	-.678	0.508 (0.329-0.783)	.002
Work status
Student		NA
Employee		1 (reference group)
Unemployed	.693	2.000 (0.248-16.159)	.516
Retired person	.472	1.604 (1.060-2.427)	.025
Karnofsky Performance Scale
N70		1 (reference group)
<=70	2.111	8.260 (3.734-18.272)	b.001
Symptoms of depression
No		1 (reference group)
Yes	1.624	5.075 (2.296-11.219)	b.001
Hypertension
No		1 (reference group)
Yes	.713	2.041 (1.012-4.113)	.046
Previously known CAD
No		1 (reference group)
Yes	1.530	4.616 (2.215-9.622)	b.001
Previous hospital admission
No		1 (reference group)
Yes	2.132	8.433 (3.820-18.618)	b.001
Continuous variables (age, Karnofsky Performance Scale, and Goldberg Depression Scale) were dichotomized.

Table 3 Multivariate logistic model as to be or become a frequent CPU user Included variables Crude analysis Analysis adjusted for age
		? coefficient	OR (95% CI)	P		? coefficient	OR (95% CI)	P
	Karnofsky Performance Scale, <=70	1.342	3.825 (1.543-9.479)	.004		1.657	5.244 (1.713-16.060)	.004
	Symptoms of depression	1.186	3.275 (1.268-8.457)	.013		1.092	2.981 (1.142-7.780)	.026
	Previously known CAD	1.083	2.954 (1.186-7.357)	.011		1.314	3.723 (1.317-10.520)	.013
	Previous hospital admission Intercept	1.272 -2.402	3.568 (1.432-8.890)	.005		1.326 -1.534	3.764 (1.493-9.491)	.005
	Results were adjusted for age.

the percentage of visits eventually admitted in cardiology (19% and 19.6%, respectively). This apparent discrepancy, which accounted for 88 patients with ACS (39.3%), was seen because 21 patients (9.4%) were admitted to other hospital services, and 67 (29.9%) were sent home and transferred to their primary care physician after stabilization and treatment optimization in the CPU. Among these discharged patients, 39 (17.4%) had been previously labeled as not amenable to revascularization, and 28 (12.5%) had poor quality of life (different levels of functional or cognitive impairment that discourage further studies).

Univariate analysis of CPU visits? variables along with their ORs are shown in Table 6. After multivariate analysis, 2 variables were found to be statistically significant and associated with frequent CPU user visits (Table 7): attending CPU at night (OR, 2.41 [95% CI, 1.64-3.52]) and requiring a CPU observation period to definitively rule out or in an ACS (OR, 2.39 [95% CI, 1.71-3.35]). That means that frequent CPU users were more frequently included in protocol 3 or possible ACS and, therefore, generated more initial diag- nostic uncertainty because coronary origin of their chest pain could not be safely ruled out or in after the initial clinical assessment and the first ECG. The Hosmer-Lemeshow statistic was 0.105 (P = .95). The area under the ROC was

0.62 (95% CI, 0.58-0.66).

Discussion

This study showed that less than 5% of patients admitted to our CPU with nontraumatic chest pain were or became a frequent user. They actually accounted for 13% of CPU visits. We also identified determinants strongly associated with frequent CPU users and their visits. Specifically, those patients with a previously known CAD, with a previous hospitalization for any reason other than CAD, with a Karnofsky Performance Scale of 70 or lesser, and with a symptom of depression were more likely to become a frequent CPU user. They usually attended CPU at night and generated more diagnostic uncertainty than did those not being frequent users.

Overcrowding in EDs has become a worldwide problem. Increases in patient volume, in illness severity, and the lack

of Hospital beds are factors contributing to this crisis [1,5,23]. Prior studies suggest that heavy ED users would also play an important role in explaining overcrowding [12-18]. In fact, 3% to 4% of patients attending EDs can account for 12% to 20% of total ED visits [12,13].

The first remarkable finding of this study is that these figures mirror those obtained from our CPU. Like heavy ED users, frequent CPU users do exist and are responsible for a disproportionately large number of visits. Chest pain units

Table 4 Characteristics and clinical outcomes of CPU visits for the 2 groups: frequent CPU users (case visits) and controls (control visits)

Variables	Case visits (n = 352)	Control visits (n = 352)	P
Weekday (%) Monday Tuesday Wednesday Thursday Friday Saturday Sunday Part of the day (%) Morning (8-16 h) Evening (16-24 h) Night (0-8 h) Referred by a physician (%) Initial protocol (%) Protocol 1 or STEMI Protocol 2 or definitive ACS Protocol 3 or possible ACS Protocol 4 or noncoronary CP Final diagnosis (%) ACS Unspecific CP/anxiety Other diagnosis Disposition (%) Admitted in cardiology Admitted in other services Discharged home			.143
	11.9	13.4
	15.1	17.9
	17.9	13.9
	14.5	16.8
	11.6	15.9
	13.9	11.9
	15.1	10.2
			b.001
	44	46.3
	27.6	38.4
	28.4	15.3
	13.3	9.7	.128
			b.001
	2.6	4.3
	9.4	9.7
	77.3	60.5
	10.8	25.6
			.624
	30.1	33.5
	42.6	40.6
	27.3	25.9
			.001
	19	19.6
	12	4.3
	69	76.1
CP indicates chest pain.

First visit Second visit Third visit Fourth visit Fifth visit Sixth visit or Total (n = 80) (n = 80) (n = 80) (n = 36) (n = 21) more (n = 55) (n = 352)
Initial protocol, n (%)
Protocol 1 or STEMI	7 (8.7)	0	(0)	1 (1.2)	1 (2.8)	0 (0)	0	(0)	9	(2.6)
Protocol 2 or definitive ACS	5 (6.3)	3	(3.7)	10 (12.5)	5 (13.9)	3 (14.3)	7	(9.1)	33	(9.4)
Protocol 3 or possible ACS	61 (76.3)	72	(90)	57 (71.3)	27 (75)	15 (71.4)	40	(56.4)	272	(77.3)
Protocol 4 or non-coronary CP	7 (8.7)	5	(6.3)	12 (15)	3 (8.3)	3 (14.3)	8	(34.5)	38	(10.8)
Final diagnosis, n (%)
ACS	28 (35)	19	(23.7)	26 (32.4)	14 (38.9)	6 (28.6)	13	(23.6)	106	(30.1)
Unspecific CP/anxiety	32 (40)	36	(45)	27 (33.8)	12 (33.3)	11 (52.4)	32	(58.2)	150	(42.6)
Other diagnosis	20 (25)	25	(31.3)	27 (33.8)	10 (27.8)	4 (19)	10	(18.2)	96	(27.3)
Disposition, n (%)
Admitted in cardiology	18 (22.5)	14	(17.5)	19 (23.7)	10 (27.8)	3 (14.3)	3	(5.5)	67	(19)
Admitted in other services	10 (12.5)	10	(12.5)	12 (15)	6 (16.7)	1 (4.8)	3	(5.5)	42	(12)
Discharged home	52 (65)	56	(79)	49 (61.3)	20 (55.5)	17 (80.9)	49	(89)	243	(69)
CP indicates chest pain.

were introduced to provide fast and efficient protocol-driven diagnostic testing for patients with chest pain [10,11]. However, frequent CPU users? arrivals can lead CPU to become a busy area where potential coronary patients could be put at risk for adverse outcome if they had to wait too long to be seen by a physician. This CPU overuse is to be considered inappropriate when such frequent CPU users are not actually at risk for or having an ACS. Needless to say that this reasoning is to be applied not only to CPU but also to EDs without structural CPU in which chest pain patients are seen together with all other patients. Therefore, under- standing the characteristics of these patients could help improve patient flow, efficiency, and eventually, ease potential ED/CPU overcrowding. Moreover, it could be

Table 5 Initial diagnosis and outcome (final diagnosis and disposition) of case visits according to the order of the visit

Table 6 Likelihood of characteristics and outcomes of frequent CPU user visits

Variables Univariate analysis

? coefficient

OR (95% CI)

P

essential to also improve frequent CPU users? medical care in both the ED and the Primary care setting.

The second remarkable conclusion of this study is that, to our knowledge, this is the first time that determinants or predictors that better define a frequent CPU user have been analyzed and summarized. Several studies demonstrated that demographic characteristics, health status, and health care access correlated with high ED use [12,13,16,17]. Factors such as poverty, advanced age, and homelessness suggested that frequent ED visits were embedded in social distress and disadvantages [12,24]. However, none of these demographic characteristics have been identified in our group of frequent CPU users. We think that a part of the explanation would have to be sought in the different national health care system.

Weekday
Weekday		1 (reference group)
Weekend	.360	1.433 (1.019-2.016)	.039
Arriving on CPU…
Not at night		1 (reference group)
At night	.784	2.190 (1.511-3.174)	b.001
Initial protocol
No observation needed (protocol 1, 2 or 4)		1 (reference group)
Observation needed (protocol 3)	.797	2.219 (1.598-3.081)	b.001
Final diagnosis
ACS		1 (reference group)
Non ACS	.115	1.121 (0.816-1.541)	.479
Disposition
Discharged		1 (reference group)
Admitted	.362	1.436 (1.028-2.005)	.034
Variables were dichotomized. CP indicates chest pain.

Table 7 Multivariate logistic model for frequent CPU user visits

Included Multivariate analysis variables

Arriving on CPU at night

CPU observation needed

Intercept

? coefficient

.877

.872

OR (95% CI)

2.405

(1.641-3.524)

2.391

(1.707-3.349)

P

b.001

b.001

-.771

Heavy ED user studies were mostly carried out in the United States, whereas our study has been performed in Spain. The United States has a majority-private health care system wherein EDs play a pivotal role as a health safety net for those without economic resources. Under these premises, when medical attention is needed, such patients can only afford to attend EDs. Instead, Spain has socialized medical services that allow people feeling ill to seek attention not only at the ED but also at other settings. However, these conclusions are difficult to generalize because most previous observations have used patient data from 1 ED. In fact, a recent population-based study has failed to recognize poor uninsured people as potential Frequent ED users [25]. This finding is markedly more in agreement with our results. Therefore, other reasons regarding education, usual habits, and medical accessibility may play also a role as potential ED user predictors.

Health status has also been largely proposed as a predictor of frequent ED use. Studies suggested a high chronic physical and psychologic Disease burden in this population [12,15,17,18,25]. These findings would provide a context for understanding the higher rates of hospitalization and illness severity [15,26]. Our results agree with previous conclusions and describe the frequent CPU user as an individual with higher prevalence of CAD, worse performance status, and clearer trend toward depression disorder when compared with other chest pain patients. This background finally justifies a higher percentage of previous hospitalization for such population. However, when visits generated by frequent CPU users were analyzed, it is remarkable that they created more diagnostic uncertainty but not more hospital admis- sions when compared with other chest pain patients. This apparent contradiction can be explained by the fact that admissions from a CPU are mainly due to ACS, whereas reasons for admission of a heavy ED user can be multiple. After 1 or more hospitalizations, some frequent CPU users could be already labeled either as not amenable to revascularization or as not suitable for further workup because of their level of functional impairment. At this point, new CPU visits will not lead to hospital admission; they will rather remain in the CPU and finally discharged after stabilization and treatment optimization. This cost-effective strategy justifies differences noted between the percentage of patients finally admitted in cardiology and those actually

diagnosed with ACS. Conversely, it could be argued that this approach had adverse outcomes. In fact, 2 patients were readmitted because of STEMI. Adverse outcomes, including ED revisits and hospital admission, have also been reported [15,17,25,26]. However, the long elapsed time from the time when patients were previously seen makes unlikely that they were inappropriately discharged.

Overall, although the group of frequent CPU users is more likely to have real CAD, they infrequently have actual ACS and so unnecessarily and inefficiently consume CPU stretchers and require stress testing. If they could be labeled as such, they would release resources for other cases, relieving the burden and poor outcomes that come with overcrowding.

All determinants found in the present study should help explain frequent CPU users? behavior, address their needs, and solve this issue in a more efficient fashion. The combination of fragile physical condition and depressive features, added to a previous history of CAD, could make them highly hesitant about any chest discomfort, with the consequence of new CPU attendance. Once this patient is detected, efforts should be made to socially support their impaired physical condition and to correct the underlying depression disorder. If patients are able to gain in confidence and improve their self-esteem, it is likely that they leave behind the label of frequent CPU user and use resources in a more cost-effective fashion.

Limitations

Despite these results, this study has limitations that deserve commentary. The predictors for frequent CPU users were obtained from a population of patients with chest pain in a single center. Therefore, concerns about its external validity may arise. So it cannot be assumed to be representative of all patients with chest pain not only here but also abroad. However, CAD incidence and distribution among white people in industrialized countries with socialized medicine are similar, and so are issues related to opened accessibility to tertiary care teaching hospitals? ED. In fact, similar predictors have been obtained from previous studies performed in other populations and countries alike [12,15,17,18,25].

In addition, in the absence of a deliberate definition of frequent CPU user from available literature, the authors relied upon a couple of articles for guidance in reaching consensus. In general terms, more than 4 ED attendances per year are needed to consider any patient as a frequent ED user [27], but only 3 visits per year are enough when it comes to the same main complaint [28], as it was the case in the present study. Limitations regarding recall bias and lack of response were also present because, similar to other survey data, it was

impossible to contact with all selected patients.

Finally, the selection of controls was randomly made instead of matching them with cases by age and/or sex. This allowed us to find that frequent CPU users were actually

older than controls. To control this confounding variable, results presented were adjusted for age.

Conclusions

We demonstrated that frequent CPU user is associated with previously known CAD, previous hospitalization for any reason other than CAD, impaired performance status, and presence of symptoms of depression. Their CPU attendance generates more diagnostic uncertainty (possible ACS), consumes more time and resources, but does not increase the need for hospital admission. Special attention to fragile physical condition and depressive features may explain the underlying reasons that justify the behavior of these patients. Further studies are warranted to know if addressing these issues may lead frequent users to gain confidence and use health care resources in a more efficient fashion.

References

1. Derlet RW, Richards JR. Overcrowding in the nation?s emergency departments: complex causes and disturbing effects. Ann Emerg Med 2000;35:63-8.
2. Ospina MB, Bond K, Schull M, Innes G, Blitz S, Rowe BH. Key indicators of overcrowding in emergency departments in Canada: a national survey. Healthc Q 2007;10:32-40.
3. Braitberg G. Emergency department overcrowding: dying to get in. Med J Aust 2007;187:624-5.
4. Jairod M, Carretero J, Closa R, Allue X. La densidad horaria de pacientes acumulados como indicador de saturacion de urgencias. Emergencias 2006;18:215-8.
5. Sanchez M, Miro O, Coll-Vinent B, et al. Saturacion del servicio de urgencias: factores asociados y cuantificacion. Med Clin (Barc) 2003;121:167-72.
6. Carbonell MA, Girbes J, Calduch JV. Determinantes en el tiempo de espera en urgencias hospitalarias y su relacion con la satisfaccion del usuario. Emergencias 2006;18:30-5.
7. Miro O, Sanchez M, Milla J. Hospital mortality and staff workload. Lancet 2000;356:1356-7.
8. Bragulat E, Lopez B, Miro O, et al. Performance assessment of an emergency department chest pain unit. Rev Esp Cardiol 2007;60: 276-84.
9. Braunwald E, Antman EM, Beasley JW, et al. ACC/AHA 2002 guidelines update for the management of patients with unstable angina and Non-ST-segment elevation myocardial infarction–summary article: a report of the American College of Cardiology/American Heart Association task force on practice guidelines (Committee on the Management of Patients with Unstable Angina). J Am Coll Cardiol 2002;40:1366-74.
10. Zalensky RJ, Rydman RJ, Ting S, et al. A national survey of emergency department chest pain centers in the United States. Am J Cardiol 1998;81:1305-9.
11. Roberts RR, Zalensky RJ, Mensah EK, et al. Costs of an emergency department-based accelerated Diagnostic protocol vs hospitalization in patients with chest pain. JAMA 1997;278:1670-6.
12. Mandelberg JH, Kuhn RE, Kohn MA. Epidemiologic analysis of an urban, public emergency department?s frequent users. Acad Emerg Med 2000;7:637-46.
13. Murphy AW, Leonard C, Plunkett PK, et al. Characteristics of attendees and their attendances at an urban accident and emergency department over a one year period. J Accid Emerg Med 1999;16: 425-7.
14. Purdie FR, Honigman B, Rosen P. The chronic emergency department patient. Ann Emerg Med 1981;10:298-301.
15. Hansagi H, Allebeck P, Edhag O, Magnusson G. Frequency of emergency department attendances as a predictor of mortality: nine- year follow-up of a population-based cohort. J Public Health Med 1990;12:39-44.
16. Spillane LL, Lumb EW, Cobaugh DJ, Wilcox SR, Clark JS, Schneider SM. Frequent users of the emergency department: can we intervene. Acad Emerg Med 1997;4:574-80.
17. Sun BC, Burstin HR, Brennan TA. Predictors and outcomes of frequent emergency department users. Acad Emerg Med 2003;10: 320-8.
18. Hansagi H, Olsson M, Sjoberg S, Tomson Y, Goransson S. Frequent use of the hospital emergency department is indicative of high use of other health care services. Ann Emerg Med 2001;37:561-7.
19. Bayon J, Alegria E, Bosch X, et al. Chest pain units. Organization and protocol for the diagnosis of acute coronary syndromes. Rev Esp Cardiol 2002;55:143-54.
20. Karnofsky DA, Burchenal J, editors. The clinical evaluation of chemotherapeutic agents in cancer. New York: Columbia University Press; 1949. p. 191-205.
21. Goldberg DP, Bridges K, Duncan-Jones P, Grayson D. Detecting anxiety and depression in general medical settings. Psychol Med 1988;17:461-70.
22. Caldwell TM, Rodgers B, Jorm AF, Christensen H, Jacomb PA, Korten AE, et al. Patterns of association between alcohol consumption and symptoms of Depression and anxiety in young adults. Addiction 2002;97:583-94.
23. Moreno E. Crisis en los servicios hospitalarios de urgencias norteamericanos: ?social, clinica o economica. Emergencias 2006;6: 336-7.
24. Malone RE. Heavy users of emergency services: social construction of a policy problem. Soc Sci Med 1995;40:469-77.
25. Hunt KA, Weber EJ, Showstack JA, Colby DC, Callaham ML. Characteristics of frequent users of emergency departments. Ann Emerg Med 2006;48:1-8.
26. Lucas RH, Sanford SM. An analysis of frequent users of emergency care at an urban university hospital. Ann Emerg Med 1998;32:563-8.
27. Locker TE, Baston S, Mason SM, Nicholl J. Defining frequent use of an urban emergency department. Emerg Med J 2007;24:398-401.
28. Pines JM, Buford K. Predictors of frequent emergency department utilization in Southeastern Pennsylvania. J Asthma 2006;43:219-23.