GERIATRICS/ORIGINAL RESEARCH
Impact of a New Senior Emergency Department on Emergency Department Recidivism, Rate of Hospital Admission, and Hospital Length of Stay Daniel C. Keyes, MD, MPH*; Bonita Singal, MD, PhD; Charles W. Kropf, BA, BS; Andrea Fisk, MD *Corresponding Author. E-mail: [email protected].
Study objective: Senior (geriatric) emergency departments (EDs) are an emerging phenomenon across the United States, designed to provide greater comfort for elders, screening for common morbidities, and selective contact with social workers. We hypothesize that the senior ED will reduce recidivism, rate of admission, and hospital length of stay. Methods: This was a pre/postintervention observational study of seniors (65 years) before and after opening of a new senior ED in a large community hospital. Older nonseniors treated during the same periods were included to detect temporal trend bias. Outcomes included admission to the hospital, hospital length of stay, and ED return visits. Cox proportional hazards models, controlling for patient age, sex, triage level, insurance type, admission on the index visit, and hospital length of stay, were used to test association with time to return within 30 and 180 days. Multivariable regression modeling was used to determine whether the intervention was associated with admission on the index visit, and hospital length of stay. Results: There was no significant difference in time to return within 30 days (HR¼1.09; 95% confidence interval [CI] 0.95 to 1.23), 180 days (HR¼0.99; 95% CI 0.91 to 1.08), or average hospital length of stay. Risk of being admitted on the index visit was lower for seniors treated in the senior ED compared with the regular ED (Relative Risk¼0.93; 95% CI 0.89 to 0.98). Conclusion: A new senior ED was not associated with reduced ED recidivism or hospital length of stay, but was associated with decreased rate of admission. [Ann Emerg Med. 2014;63:517-524.] Please see page 518 for the Editor’s Capsule Summary of this article. A feedback survey is available with each research article published on the Web at www.annemergmed.com. A podcast for this article is available at www.annemergmed.com. 0196-0644/$-see front matter Copyright © 2013 by the American College of Emergency Physicians. http://dx.doi.org/10.1016/j.annemergmed.2013.10.033
SEE EDITORIAL, P. 525. INTRODUCTION The US geriatric population is increasing rapidly, and geriatric patients are increasingly high users of emergency services.1,2 Older patients are more likely to be admitted to the hospital or experience adverse outcomes after visiting an emergency department (ED).3 Health service use and costs increase with patient age.4 Returns to the ED are frequent in this population, especially within the first month after index visit,5 and calls have been made for screening and other interventions that can reduce recidivism in this cohort.6-8 The ED has thus become an attractive target for modification.9 In some patient populations who are high users of emergency care, a case management approach focusing on the core needs of the individuals involved has resulted in improved care, decreased use of health care resources, and decreased cost.10,11 A specialized senior ED can be seen as an example of this: an acute care Volume 63, no. 5 : May 2014
interdisciplinary model to identify the distinct needs of the geriatric population.12 The model used here makes use of screening for highrisk conditions with geriatric-specific instruments, along with initiation of treatment, disposition, and follow-up planning in the ED.13 In one Australian large-population model that used screening, a decrease in return visits within 4 weeks was observed.14 Importance The senior ED in the current study implemented supplemental education for physicians and nurses, changes to the physical space to make the ED stay more comfortable for older patients, and universal screening of patients for common comorbidities in this age group (Figure 1). Senior EDs have garnered increasing interest as a means of directly managing the needs of older patients.15 The goal of a senior ED is to treat the patient’s chief complaint while identifying common comorbidities such as depression, delirium, dementia, medication interactions, fall risk, and substance abuse.16-18 Annals of Emergency Medicine 517
Senior Emergency Departments and Recidivism
Keyes et al
Editor’s Capsule Summary
What is already known on this topic Some hospitals have opened specialized geriatric “senior” emergency departments (EDs). What question this study addressed Can a senior ED reduce return ED visits, admissions, or hospital length of stay as studied in a time-series pre-post design with 2 6-month time blocks and roughly 12,000 patients aged 55 years and older? What this study adds to our knowledge There was no change in hospital length of stay or time until a return ED visit occurred, but hospital admissions decreased by 3% after the introduction of the senior ED. How this is relevant to clinical practice These findings tentatively suggest that the version of a senior ED examined in this study may slightly reduce hospital admissions for older adults.
The noninferiority of a senior ED using a staff of geriatricians was reported in Italy, caring for nontraumatic, nondemented patients.19 In the current study, demented patients were included in the senior ED, and they completed assessment tools to the extent that their cognitive abilities allowed. To our knowledge, there are no published studies investigating the effect of using a senior ED, staffed by emergency physicians and nurses, with support from social workers and pharmacists on recidivism, hospital admission, or hospital length of stay. Goals of This Investigation The primary purpose of this investigation was to determine whether a new senior ED, caring for patients 65 years and older through case management with specific attention to medication use, activities of daily living, depression, delirium, and alcohol abuse, will result in decreased recidivism, defined as a longer time to return to the ED. Secondary outcomes included evaluation of the number of hospital admissions and hospital length of stay after the new program was established. We hypothesized that the senior ED is associated with decreased frequency of return visits, decreased admissions, and decreased length of stay.
MATERIALS AND METHODS The Saint Joseph Mercy Health System institutional review board approved this study. Study Design We conducted a retrospective pre/postintervention comparison study of 2 cohorts of patients—those aged 65 years and older and 518 Annals of Emergency Medicine
Figure 1. Characteristics of the senior ED.
those aged 55 to 64 years—presenting to an ED during 2 periods. We evaluated the time to return visit for patients during a 6-month period (October 2009 to April 2010) before the opening of the new senior ED and during a 6-month period after the development of the new senior ED (October 2010 to April 2011). This resulted in 4 groups: seniors in the ED before the senior ED opened, those in the new senior ED, younger patients treated before the senior ED opened, and younger patients treated after it opened. Data were drawn from the study site’s Quality Institute for administrative purposes and are not representative of a patient’s full chart. No data abstraction was performed because patient information was already deidentified and summarized by the administrative system. Volume 63, no. 5 : May 2014
Keyes et al Setting The setting was a single-site, large community hospital ED with 85,000 annual visits in southeastern Michigan. In 2011, patients in this ED were 69% white, 26% black, 1% Asian, and 4% all other races. A senior ED opened at this site in October 2010. In 2011, this senior ED accommodated 12% of patients treated in the ED, or approximately 10,000 patients. Senior ED patients were 59% women compared with 55% in the entire ED. Two hospitals are located within 10 miles of this site: a university hospital providing standard ED services but with a lower annual volume and a Veterans Administration hospital. There are no other hospitals within 20 miles of the research site. Changes to the senior ED included staff training, facility redesign, and new processes. Nurses working in the senior ED received a geriatric emergency nursing education course (Emergency Nurses Association, Des Plaines, IL), and physicians completed a web-based curriculum (Geriatric Online Education). In the senior ED, physical conditions were modified to include nonskid, nonglare floors, side rails, pressure-reducing mattresses, softer lighting, and larger clocks and televisions. For patients treated in the senior ED and all patients older than 64 years who meet the same acuity inclusion criteria used in the general ED, nurses administer a senior packet that includes the following: a Triage Risk Screening Tool, an evaluation of polypharmacy/fall risks, the Katz Activities of Daily Living Index, the Geriatric Depression Scale, the Confusion Assessment Method, the MiniCog Mental Status Evaluation, and the Short Michigan Alcoholism Screening Tool–Geriatric Version20-25 (Figure 1). The case management approach used at this senior ED involves nurses, social workers, pharmacists, and physicians. Patients are considered to be at increased risk if they present with a fall or altered mental status while receiving any of a list of selected Beers criteria medications,26 they score 5 of 6 or below on the Katz Index or 5 of 15 or above on the Geriatric Depression Scale, score positively for any item on the Confusion Assessment Method, score lower than a 2 of 3 with a good clock drawing on the Mini-Cog, or score 2 of 10 or higher on the Short Michigan Alcoholism Screening Tool–Geriatric Version. When a patient has screened positively, social workers interview the patient and the family. If the patient cannot consult with the social worker during the index visit, the social worker conducts a telephone interview on the next available shift. Policy dictates that all patients be contacted about any risk, but data were not available on the frequency of these telephone contacts. Social workers interface with the emergency physician and the primary care physician to provide the patients with appropriate resources. Selected patients receive evaluation by a pharmacist for inappropriate prescribing or pharmaceutical interactions, resulting in communication between the pharmacist and the primary care physician. Before the opening of the senior ED, the social workers dedicated to the ED consulted with approximately 5 patients per day. Because no systematic screening was in place, social workers performed their own “case-finding” by reviewing charts to identify potential patients. Most referrals were for medication Volume 63, no. 5 : May 2014
Senior Emergency Departments and Recidivism assistance, management of patients without insurance, and transportation. If the social worker was engaged with assisting a patient, no one was reviewing charts to find other patients, and in this way many patients likely were missed. Screening tools were used only on rare occasions. With the opening of the senior ED came widespread screening of geriatric patients and delivery of these results to social workers. This effectively increased the number of patients consulting with social workers to 40 or more per day. Discharge/ disposition planning by social work case management is now performed on about 8 times as many patients daily. Selection of Participants Sample size was fixed by the number of patients treated in the ED during the evaluation periods. Patients were included if they were treated in the ED at Saint Joseph Mercy Hospital during the study periods and were either aged 55 to 64 years or older than 64 years. Patients with an Emergency Severity Index27 level of 1, indicating severe illness or the need for acute resuscitation, or for whom the Emergency Severity Index was not recorded, were excluded from the study. In an effort to avoid contamination of the second sample with frequently returning visitors who were treated in the first sample, only new cases, beginning within each cohort, were included. Patients who were treated within 12 months before the start of that particular cohort were therefore excluded. Data Collection and Processing Patient information was obtained from an administrative database assembled by the Quality Institute at Saint Joseph Mercy Hospital, which included the following variables: date admitted to the ED, date admitted to the hospital, hospital length of stay, age, sex, insurance type, discharge diagnosis, Emergency Severity Index score, whether the patient was placed in an observation unit, and whether a flu test was administered during the first site visit. Information on patient deaths was obtained from the Social Security Death Index. Outcome Measures Recidivism was defined as a return visit within 30 and 180 days from index visit. The number of days to first return visit to the ED was calculated as the difference between the return visit date and the index visit date. If a patient did not make a revisit up to the end of the 30-day or 6-month evaluation period, the time was censored at the number of days from the initial visit to the end of the evaluation period, either 30 or 180 days. Admission to the hospital was recorded as positive or negative for each patient. Length of stay for patients admitted to the hospital was calculated from date of ED admission to date of discharge. Primary Data Analysis Statistical analyses were performed with SAS (version 9.3; SAS Institute, Inc., Cary, NC). Demographic and baseline variables Annals of Emergency Medicine 519
Keyes et al
Senior Emergency Departments and Recidivism were summarized with means, medians, or percentages, as appropriate. We determined whether time to return ED visit for seniors by 30 and 180 days was different before and after opening of the senior ED by fitting Cox proportional hazards models.28,29 Younger patients were included in the analysis to detect temporal trend bias. The main independent variable was categorical, representing 4 case types: 55- to 64-year-olds who were treated in the regular ED in the presenior ED period, patients aged 65 years or older who were treated in the regular ED in the presenior ED period, 55- to 64-year-olds who were treated in the regular ED after the senior ED was open, and patients aged 65 years and older who were treated in the senior ED. We attempted to control for chance differences between the groups that could contribute to differences in recidivism by including potentially confounding variables as covariates in the models. We chose variables that were deemed, on clinical grounds, to be possible predictors of recidivism and that were available to us in the administrative data set we were using for the analysis. Age, sex, and insurance status might account for some variation in the risk of returning after an ED visit. Triage level, one measure of the severity of the initial complaint, might also influence the risk to return. Whether the patient was admitted or sent to the observation unit might decrease early recidivism. Length of stay is another measure of severity of illness in which patients who stay longer are presumably sicker and have more complicated medical problems. We considered using an indicator variable for influenza testing to try to account for any temporal differences that may have occurred because of an outbreak of influenza in the 2009 to 2010 winter season.30,31 However, this test was conducted on only 4 patients in the entire cohort, and it was not included in the models. Age was coded as a continuous variable. Sex, admission, and observation unit status were coded as dichotomous variables. Triage level was in 3 categories. Insurance status was coded as commercial, Medicare, Medicaid, county plan, and uninsured. Length of stay was divided into 4 categories (1 day and 2 to 3, 4 to 10, and >10 days). The first category for length of stay included patients treated in the ED and discharged directly or sent to the observation unit and discharged from there. For the primary outcome variable, days to first return visit to the ED, we calculated the number of days between the index visit and the first return visit. If a given patient did not make a revisit up to the end of the 1- or 6-month evaluation interval, the result was censored at 30 or 180 days, respectively. If a patient died before the end of the evaluation period, the result was censored at death. Patients censored for death were counted as not returning up to the time of death. The models were tested for the proportional hazard assumptions with graphic techniques, and the Cox models appeared adequate. We compared the frequency of hospitalization on the index visit between the 4 groups, using a generalized estimating equation with a Poisson distribution, a log link, and a sandwich estimator, controlling for triage level, age, sex, and insurance type.32 This approach was used instead of logistic regression 520 Annals of Emergency Medicine
analysis so that we could derive an estimate of relative risk instead of an odds ratio. The odds ratio does not approximate the risk ratio when the risk of the outcome is not small, as in the case of admission to the hospital. Hospital length of stay (in days) for the first visit was compared between groups by fitting a general linear model. The logarithm of length of stay was used as the dependent variable to better comply with linear model assumptions. Age, sex, insurance status, and triage level were included in these models to control for confounding. Age squared was tested in all models to account for nonlinearity in the relationship between age and the outcome. Statistical significance was set at P
Impact of a New Senior Emergency Department on Emergency Department Recidivism, Rate of Hospital Admission, and Hospital Length of Stay Daniel C. Keyes, MD, MPH*; Bonita Singal, MD, PhD; Charles W. Kropf, BA, BS; Andrea Fisk, MD *Corresponding Author. E-mail: [email protected].
Study objective: Senior (geriatric) emergency departments (EDs) are an emerging phenomenon across the United States, designed to provide greater comfort for elders, screening for common morbidities, and selective contact with social workers. We hypothesize that the senior ED will reduce recidivism, rate of admission, and hospital length of stay. Methods: This was a pre/postintervention observational study of seniors (65 years) before and after opening of a new senior ED in a large community hospital. Older nonseniors treated during the same periods were included to detect temporal trend bias. Outcomes included admission to the hospital, hospital length of stay, and ED return visits. Cox proportional hazards models, controlling for patient age, sex, triage level, insurance type, admission on the index visit, and hospital length of stay, were used to test association with time to return within 30 and 180 days. Multivariable regression modeling was used to determine whether the intervention was associated with admission on the index visit, and hospital length of stay. Results: There was no significant difference in time to return within 30 days (HR¼1.09; 95% confidence interval [CI] 0.95 to 1.23), 180 days (HR¼0.99; 95% CI 0.91 to 1.08), or average hospital length of stay. Risk of being admitted on the index visit was lower for seniors treated in the senior ED compared with the regular ED (Relative Risk¼0.93; 95% CI 0.89 to 0.98). Conclusion: A new senior ED was not associated with reduced ED recidivism or hospital length of stay, but was associated with decreased rate of admission. [Ann Emerg Med. 2014;63:517-524.] Please see page 518 for the Editor’s Capsule Summary of this article. A feedback survey is available with each research article published on the Web at www.annemergmed.com. A podcast for this article is available at www.annemergmed.com. 0196-0644/$-see front matter Copyright © 2013 by the American College of Emergency Physicians. http://dx.doi.org/10.1016/j.annemergmed.2013.10.033
SEE EDITORIAL, P. 525. INTRODUCTION The US geriatric population is increasing rapidly, and geriatric patients are increasingly high users of emergency services.1,2 Older patients are more likely to be admitted to the hospital or experience adverse outcomes after visiting an emergency department (ED).3 Health service use and costs increase with patient age.4 Returns to the ED are frequent in this population, especially within the first month after index visit,5 and calls have been made for screening and other interventions that can reduce recidivism in this cohort.6-8 The ED has thus become an attractive target for modification.9 In some patient populations who are high users of emergency care, a case management approach focusing on the core needs of the individuals involved has resulted in improved care, decreased use of health care resources, and decreased cost.10,11 A specialized senior ED can be seen as an example of this: an acute care Volume 63, no. 5 : May 2014
interdisciplinary model to identify the distinct needs of the geriatric population.12 The model used here makes use of screening for highrisk conditions with geriatric-specific instruments, along with initiation of treatment, disposition, and follow-up planning in the ED.13 In one Australian large-population model that used screening, a decrease in return visits within 4 weeks was observed.14 Importance The senior ED in the current study implemented supplemental education for physicians and nurses, changes to the physical space to make the ED stay more comfortable for older patients, and universal screening of patients for common comorbidities in this age group (Figure 1). Senior EDs have garnered increasing interest as a means of directly managing the needs of older patients.15 The goal of a senior ED is to treat the patient’s chief complaint while identifying common comorbidities such as depression, delirium, dementia, medication interactions, fall risk, and substance abuse.16-18 Annals of Emergency Medicine 517
Senior Emergency Departments and Recidivism
Keyes et al
Editor’s Capsule Summary
What is already known on this topic Some hospitals have opened specialized geriatric “senior” emergency departments (EDs). What question this study addressed Can a senior ED reduce return ED visits, admissions, or hospital length of stay as studied in a time-series pre-post design with 2 6-month time blocks and roughly 12,000 patients aged 55 years and older? What this study adds to our knowledge There was no change in hospital length of stay or time until a return ED visit occurred, but hospital admissions decreased by 3% after the introduction of the senior ED. How this is relevant to clinical practice These findings tentatively suggest that the version of a senior ED examined in this study may slightly reduce hospital admissions for older adults.
The noninferiority of a senior ED using a staff of geriatricians was reported in Italy, caring for nontraumatic, nondemented patients.19 In the current study, demented patients were included in the senior ED, and they completed assessment tools to the extent that their cognitive abilities allowed. To our knowledge, there are no published studies investigating the effect of using a senior ED, staffed by emergency physicians and nurses, with support from social workers and pharmacists on recidivism, hospital admission, or hospital length of stay. Goals of This Investigation The primary purpose of this investigation was to determine whether a new senior ED, caring for patients 65 years and older through case management with specific attention to medication use, activities of daily living, depression, delirium, and alcohol abuse, will result in decreased recidivism, defined as a longer time to return to the ED. Secondary outcomes included evaluation of the number of hospital admissions and hospital length of stay after the new program was established. We hypothesized that the senior ED is associated with decreased frequency of return visits, decreased admissions, and decreased length of stay.
MATERIALS AND METHODS The Saint Joseph Mercy Health System institutional review board approved this study. Study Design We conducted a retrospective pre/postintervention comparison study of 2 cohorts of patients—those aged 65 years and older and 518 Annals of Emergency Medicine
Figure 1. Characteristics of the senior ED.
those aged 55 to 64 years—presenting to an ED during 2 periods. We evaluated the time to return visit for patients during a 6-month period (October 2009 to April 2010) before the opening of the new senior ED and during a 6-month period after the development of the new senior ED (October 2010 to April 2011). This resulted in 4 groups: seniors in the ED before the senior ED opened, those in the new senior ED, younger patients treated before the senior ED opened, and younger patients treated after it opened. Data were drawn from the study site’s Quality Institute for administrative purposes and are not representative of a patient’s full chart. No data abstraction was performed because patient information was already deidentified and summarized by the administrative system. Volume 63, no. 5 : May 2014
Keyes et al Setting The setting was a single-site, large community hospital ED with 85,000 annual visits in southeastern Michigan. In 2011, patients in this ED were 69% white, 26% black, 1% Asian, and 4% all other races. A senior ED opened at this site in October 2010. In 2011, this senior ED accommodated 12% of patients treated in the ED, or approximately 10,000 patients. Senior ED patients were 59% women compared with 55% in the entire ED. Two hospitals are located within 10 miles of this site: a university hospital providing standard ED services but with a lower annual volume and a Veterans Administration hospital. There are no other hospitals within 20 miles of the research site. Changes to the senior ED included staff training, facility redesign, and new processes. Nurses working in the senior ED received a geriatric emergency nursing education course (Emergency Nurses Association, Des Plaines, IL), and physicians completed a web-based curriculum (Geriatric Online Education). In the senior ED, physical conditions were modified to include nonskid, nonglare floors, side rails, pressure-reducing mattresses, softer lighting, and larger clocks and televisions. For patients treated in the senior ED and all patients older than 64 years who meet the same acuity inclusion criteria used in the general ED, nurses administer a senior packet that includes the following: a Triage Risk Screening Tool, an evaluation of polypharmacy/fall risks, the Katz Activities of Daily Living Index, the Geriatric Depression Scale, the Confusion Assessment Method, the MiniCog Mental Status Evaluation, and the Short Michigan Alcoholism Screening Tool–Geriatric Version20-25 (Figure 1). The case management approach used at this senior ED involves nurses, social workers, pharmacists, and physicians. Patients are considered to be at increased risk if they present with a fall or altered mental status while receiving any of a list of selected Beers criteria medications,26 they score 5 of 6 or below on the Katz Index or 5 of 15 or above on the Geriatric Depression Scale, score positively for any item on the Confusion Assessment Method, score lower than a 2 of 3 with a good clock drawing on the Mini-Cog, or score 2 of 10 or higher on the Short Michigan Alcoholism Screening Tool–Geriatric Version. When a patient has screened positively, social workers interview the patient and the family. If the patient cannot consult with the social worker during the index visit, the social worker conducts a telephone interview on the next available shift. Policy dictates that all patients be contacted about any risk, but data were not available on the frequency of these telephone contacts. Social workers interface with the emergency physician and the primary care physician to provide the patients with appropriate resources. Selected patients receive evaluation by a pharmacist for inappropriate prescribing or pharmaceutical interactions, resulting in communication between the pharmacist and the primary care physician. Before the opening of the senior ED, the social workers dedicated to the ED consulted with approximately 5 patients per day. Because no systematic screening was in place, social workers performed their own “case-finding” by reviewing charts to identify potential patients. Most referrals were for medication Volume 63, no. 5 : May 2014
Senior Emergency Departments and Recidivism assistance, management of patients without insurance, and transportation. If the social worker was engaged with assisting a patient, no one was reviewing charts to find other patients, and in this way many patients likely were missed. Screening tools were used only on rare occasions. With the opening of the senior ED came widespread screening of geriatric patients and delivery of these results to social workers. This effectively increased the number of patients consulting with social workers to 40 or more per day. Discharge/ disposition planning by social work case management is now performed on about 8 times as many patients daily. Selection of Participants Sample size was fixed by the number of patients treated in the ED during the evaluation periods. Patients were included if they were treated in the ED at Saint Joseph Mercy Hospital during the study periods and were either aged 55 to 64 years or older than 64 years. Patients with an Emergency Severity Index27 level of 1, indicating severe illness or the need for acute resuscitation, or for whom the Emergency Severity Index was not recorded, were excluded from the study. In an effort to avoid contamination of the second sample with frequently returning visitors who were treated in the first sample, only new cases, beginning within each cohort, were included. Patients who were treated within 12 months before the start of that particular cohort were therefore excluded. Data Collection and Processing Patient information was obtained from an administrative database assembled by the Quality Institute at Saint Joseph Mercy Hospital, which included the following variables: date admitted to the ED, date admitted to the hospital, hospital length of stay, age, sex, insurance type, discharge diagnosis, Emergency Severity Index score, whether the patient was placed in an observation unit, and whether a flu test was administered during the first site visit. Information on patient deaths was obtained from the Social Security Death Index. Outcome Measures Recidivism was defined as a return visit within 30 and 180 days from index visit. The number of days to first return visit to the ED was calculated as the difference between the return visit date and the index visit date. If a patient did not make a revisit up to the end of the 30-day or 6-month evaluation period, the time was censored at the number of days from the initial visit to the end of the evaluation period, either 30 or 180 days. Admission to the hospital was recorded as positive or negative for each patient. Length of stay for patients admitted to the hospital was calculated from date of ED admission to date of discharge. Primary Data Analysis Statistical analyses were performed with SAS (version 9.3; SAS Institute, Inc., Cary, NC). Demographic and baseline variables Annals of Emergency Medicine 519
Keyes et al
Senior Emergency Departments and Recidivism were summarized with means, medians, or percentages, as appropriate. We determined whether time to return ED visit for seniors by 30 and 180 days was different before and after opening of the senior ED by fitting Cox proportional hazards models.28,29 Younger patients were included in the analysis to detect temporal trend bias. The main independent variable was categorical, representing 4 case types: 55- to 64-year-olds who were treated in the regular ED in the presenior ED period, patients aged 65 years or older who were treated in the regular ED in the presenior ED period, 55- to 64-year-olds who were treated in the regular ED after the senior ED was open, and patients aged 65 years and older who were treated in the senior ED. We attempted to control for chance differences between the groups that could contribute to differences in recidivism by including potentially confounding variables as covariates in the models. We chose variables that were deemed, on clinical grounds, to be possible predictors of recidivism and that were available to us in the administrative data set we were using for the analysis. Age, sex, and insurance status might account for some variation in the risk of returning after an ED visit. Triage level, one measure of the severity of the initial complaint, might also influence the risk to return. Whether the patient was admitted or sent to the observation unit might decrease early recidivism. Length of stay is another measure of severity of illness in which patients who stay longer are presumably sicker and have more complicated medical problems. We considered using an indicator variable for influenza testing to try to account for any temporal differences that may have occurred because of an outbreak of influenza in the 2009 to 2010 winter season.30,31 However, this test was conducted on only 4 patients in the entire cohort, and it was not included in the models. Age was coded as a continuous variable. Sex, admission, and observation unit status were coded as dichotomous variables. Triage level was in 3 categories. Insurance status was coded as commercial, Medicare, Medicaid, county plan, and uninsured. Length of stay was divided into 4 categories (1 day and 2 to 3, 4 to 10, and >10 days). The first category for length of stay included patients treated in the ED and discharged directly or sent to the observation unit and discharged from there. For the primary outcome variable, days to first return visit to the ED, we calculated the number of days between the index visit and the first return visit. If a given patient did not make a revisit up to the end of the 1- or 6-month evaluation interval, the result was censored at 30 or 180 days, respectively. If a patient died before the end of the evaluation period, the result was censored at death. Patients censored for death were counted as not returning up to the time of death. The models were tested for the proportional hazard assumptions with graphic techniques, and the Cox models appeared adequate. We compared the frequency of hospitalization on the index visit between the 4 groups, using a generalized estimating equation with a Poisson distribution, a log link, and a sandwich estimator, controlling for triage level, age, sex, and insurance type.32 This approach was used instead of logistic regression 520 Annals of Emergency Medicine
analysis so that we could derive an estimate of relative risk instead of an odds ratio. The odds ratio does not approximate the risk ratio when the risk of the outcome is not small, as in the case of admission to the hospital. Hospital length of stay (in days) for the first visit was compared between groups by fitting a general linear model. The logarithm of length of stay was used as the dependent variable to better comply with linear model assumptions. Age, sex, insurance status, and triage level were included in these models to control for confounding. Age squared was tested in all models to account for nonlinearity in the relationship between age and the outcome. Statistical significance was set at P
![[Length of stay in the hospital emergency department].](/assets/img/hold.png)
