Accepted for publication in The Journal of the American Geriatrics Society
Outcomes of infection in nursing home residents with and without early hospital transfer
Kenneth S. Boockvar, MD, MS1, Ann L. Gruber-Baldini, PhD2, Lynda Burton, ScD3, Sheryl Zimmerman, PhD4, Conrad May, MD2, Jay Magaziner, PhD, MSHyg2
1Bronx Veterans Affairs Medical Center, Bronx, NY and Mount Sinai School of Medicine, New York, NY; 2University of Maryland, Baltimore; 3The Johns Hopkins University; 4University of North Carolina at Chapel Hill
Corresponding author: Kenneth Boockvar, MD, MS, Bronx Veterans Affairs Medical Center, 130 W. Kingsbridge Rd, Rm 4A-17, Bronx, NY 10468
Tel: 718-584-9000 x3807; Fax: 718-741-4211; Email:
Alternate correspondent: Ann Gruber-Baldini, PhD, Department of Epidemiology and Preventive Medicine, University of Maryland School of Medicine, 660 Redwood St, Suite 200, Baltimore, MD 21201-1596 Tel: 410-706-2444; Fax: 410-706-4433; Email:
This research was supported by grants from the National Institute on Aging (R01 AG08211 and R29 AG11407) and a Pfizer/Foundation for Health in Aging Junior Faculty Scholarship for Research on Health Outcomes in Geriatrics to Dr. Boockvar. Presented in part at the 2003 meeting of the American Geriatrics Society, Baltimore, MD.
Running head: Outcomes of infection in nursing home
Word counts: Abstract 238; Text 4083 Graphics: 2 tables, 1 figure References: 43
Abstract
Objective: To compare outcomes of infection in nursing home residents with and without early hospital transfer.
Design: Observational cohort study
Setting and Participants: 2153 individuals admitted to 59 nursing homes in Maryland from 1992-5.
Measurements: Incident infection was recorded when a new infectious diagnosis was documented in the medical record or non-prophylactic antibiotic therapy was prescribed. Early hospital transfer was defined as transfer to the emergency department and/or admission to the hospital within 3 days of infection onset. Infection, resident, and facility characteristics were entered into a multivariate model to create a propensity score for early hospital transfer. Association between early hospital transfer and outcomes of infection, namely pressure ulcers and death between days 4 and 34 after infection onset, were examined controlling for propensity score.
Results: 4990 infections occurred in 1301 residents. Genitourinary (28%), skin (19%), upper respiratory (13%), and lower respiratory (12%) were the most common types. 375 episodes in which residents survived 3 days (7.6%) resulted in early hospital transfer. In multivariate regression, individuals with early hospital transfer had higher mortality (OR 1.44; 95% CI 1.04 – 1.99) and, among 1-month survivors, a greater occurrence of pressure ulcers (OR 1.61; 95% CI 1.17 – 2.20) than those without, after adjusting for propensity score.
Conclusions: Using observational data and propensity score methods, outcomes are worse in nursing home residents transferred to the hospital within 3 days of infection onset compared with those who remained in the nursing home.
Key words: infection, nursing homes, treatment outcome
INtroduction
Infections are common in nursing home residents and cause significant morbidity and mortality despite efforts at prevention and treatment. Infection incidence ranges between 4.7 and 12 episodes per 1000 resident-days (1.7-4.4 episodes per resident per year) 1-4, and 11-15% of residents have 3 or more episodes during their nursing home stay 2, 3. In nursing home residents with all types of infection, three-week mortality has been reported to be 10% 2, and mortality is even higher for residents with lower respiratory tract infection 5-7 and for residents with infection and coexisting dementia 8, 9. Residents with infection are also at high risk for physical function decline5.
Hospital transfer is a triage option for clinicians managing nursing home residents with infection, and is a decision that is influenced by illness, individual, provider, and facility factors. Many nursing home clinicians and researchers note that the benefits of hospital transfer (e.g., closer monitoring, greater access to diagnostic testing, more treatment options) are balanced by the risk of iatrogenic harm from inter-institutional transfer and inpatient hospital care. Origination from a nursing home, as well as older age, worse physical function, and worse cognitive function, is associated with adverse events in the hospital such as adverse drug events, falls, and delirium 10-16. In addition, adverse drug events occur because of changes in medications during transfer both ways between nursing home and hospital 17. It is unknown whether the potential benefits of hospital transfer outweigh the potential hazards, since few studies have compared outcomes of nursing home-acquired infection with and without hospital transfer 5-7, and no prospective, controlled trials have been published. Partly as a result of ongoing uncertainty, there is no standard approach to triaging nursing home residents with infection, and there exists great variability among nursing homes in the frequency of hospital transfer 18. There also is evidence that some transfers may be unnecessary and even inappropriate 19.
The objective of this study was to compare outcomes of infection with and without hospital transfer in an observational cohort of 2153 nursing home residents from 59 facilities. We hypothesized that, even after controlling for potentially confounding differences in clinical status between hospital and nursing home triage groups, and after controlling for differences in administration and staffing in the nursing homes in which residents resided, outcomes would be worse for residents transferred to the hospital. We used detailed data about infection episodes, individuals, and facilities, and two-stage propensity modeling to adjust for these potential confounding factors.
METHODS
Setting and participants
All individuals admitted to one of 59 nursing homes in Maryland between September 1992 and March 1995 were eligible to be enrolled in a prospective observational cohort study on the prevalence and cost of dementia in nursing homes. The study design and cohort characteristics have been described previously 20, 21. The 221 licensed long-term care facilities in Maryland in 1992 were grouped into 15 strata by geographic region (5 strata) and bed-size (3 strata: <50 beds, 50-150 beds, >150 beds). A sample of facilities within each stratum was recruited, selected by random sampling weighted by the statewide proportion of beds represented by that stratum. Of 64 facilities that were contacted, 60 agreed to participate. One facility that agreed to participate reported no new admissions during the study period, leaving 59 facilities in the study.
A total of 3283 individuals age 65 years and older who had not resided in a nursing home or chronic care facility for 8 or more days in the previous year and for whom baseline data could be collected within 2 months of admission were eligible to be enrolled. Of these, 2285 individuals (70%) or their legal surrogates gave written informed consent to participate, and follow-up data were available for 2153. The human subjects institutional review board at the University of Maryland, Baltimore approved the study protocol.
Measures
Baseline
Information about individual residents was gathered from interviews with residents, nursing home staff, and family members, and from medical record review, including review of the nursing home Minimum Data Set (MDS) and hospital discharge summaries. Data included socio-demographic characteristics (age, gender, race, education), function in activities of daily living (ADLs), chronic medical conditions, body mass index, and presence or absence of dementia. A score for function in 6 ADLs (bathing, dressing, toileting, transferring, feeding, and continence) was calculated using a modified version of the Katz index 22. A count of 14 possible chronic medical conditions (coronary artery disease, congestive heart failure, cerebrovascular disease, chronic obstructive pulmonary disease, arthritis, cancer, diabetes mellitus, malnutrition, uncontrolled hypertension, liver disease, peripheral vascular disease, seizure disorder, pressure ulcer, and dementia) was calculated as a measure of chronic illness burden. A geriatric psychiatrist and neurologist examined the medical record and all other data collected to determine dementia status (yes or no), using Diagnostic and Statistical Manual III-R criteria 21, 23.
Facility information was obtained from interviews with facility administrators and directors of nursing, and from abstraction of administrative forms. Facility data included bedsize, proprietary status (for profit versus not-for-profit), rural versus urban location (according to Department of Agriculture criteria24), distance to hospital (0-5 miles versus 6 or more), nurse staffing (number of full-time equivalent RN and LPN nurses per 100 beds), and a count of 4 medical services available (intravenous fluid administration, respiratory therapy, tracheostomy care, and tube feeding).
Follow-up: Incident infection
Individuals were followed until nursing home discharge or until 2 years after nursing home admission. Incident infection was defined as 1) documentation of an infectious diagnosis in any part of the medical record, including nursing notes, physician notes, orders, consults, and outside medical reports, or 2) in the absence of documentation of an infectious diagnosis, receipt of antibiotics for any length of time except if prescribed the day preceding, day of, or day following a surgical procedure. Infection onset was defined as the date the infection was first documented in the chart or the date antibiotics were first ordered, whichever occurred earlier. Infection end was defined as the date of documentation of infection resolution or the last date of receipt of antibiotics, whichever occurred later. For residents who experienced more than one infection, the first episode was considered the index infection.
Infections were classified by anatomical location as recorded in the medical record. Location categories were: eye, gastrointestinal, genitourinary, gynecologic, lower respiratory, skin, systemic, toenail, and upper respiratory. Infections of other anatomical locations were classified as “other,” and infections for which location could not be determined were classified as “undocumented.” Additional recorded characteristics of the infection episode included occurrence of fever 100.0F (yes or no), highest measured fever (4 categories: 100.0-100.9, 101.0-101.9, 102.0-102.9, and 103), and physician visit within 3 days of infection onset.
Follow-up: Triage group
Residents transferred to the hospital within 3 days of infection onset were classified as the hospital triage group. Residents with infection who were not transferred to the hospital and those who were transferred to the hospital more than 3 days after infection onset were classified as the nursing home triage group. The interval of 3 days from infection onset to potential hospital transfer was considered the maximum amount of time required for the provider to gather information for the triage decision. Residents who died within 3 days of infection onset were not assigned to either group and were excluded from the analysis.
Since the nursing home provider may not know at the time of resident transfer to a hospital emergency department whether the resident actually will be admitted to the hospital, residents who were transferred to the emergency department within 3 days of infection onset and discharged back to the nursing home (i.e., those who were not admitted) were included in the hospital triage group. This classification is analogous to intention-to-treat grouping in prospective controlled clinical trials, and addresses the question of whether triage to a hospital with intention to treat there benefits a nursing home resident with infection compared to remaining in the nursing home.
Follow-up: Outcomes
Outcomes of infection were examined from day 4 to 34 after infection onset. Outcomes examined included second infection, pressure ulcer, and death, selected because they are possible complications of institutional care of older adults with disability, and because they were available in the existing dataset. Second infection was defined as 1) an infection in an anatomical location different from the index infection location, the onset of which followed the index infection, or 2) an infection in the same anatomical location as the index infection, the onset of which occurred more than 7 days after the end date of the index infection. Pressure ulcer was defined as documentation in any part of the medical record of a pressure ulcer stage 2 or greater. Death was ascertained by examination of the nursing home medical record and hospital discharge summaries.
Analyses
We compared outcomes of infection between the hospital and nursing home triage groups. Episodes of eye, gynecologic, and toenail infection were excluded because of the absence of hospital transfers for these infections in our sample and because hospital transfer would not likely be considered necessary for them. We used propensity score methods to create resident groups in which the likelihood of being transferred to the hospital within 3 days of infection onset was similar, regardless of triage group membership, 25, 26 and to aggregate the potential confounding effects of infection, resident, and facility factors.
In the first step of propensity analysis we used Generalized Estimating Equations (GEE)27 to predict which residents might be transferred to the hospital on the basis of infection, resident, and facility characteristics. Only episodes with complete data were included. Infection characteristics in the model were: type of infection, occurrence of fever 100.0F, highest measured fever, and physician visit within 3 days of infection onset. Resident characteristics included in the model were: demographics, ADL function, chronic medical conditions, body mass index, and dementia status. Facility characteristics included in the model were: bedsize, proprietary status, rural versus urban location, distance to hospital, nurse staffing, and medical services availability. A propensity score for hospital transfer was generated for each resident using the coefficients from this model. Residents were stratified according to propensity score into quintiles or tertiles, as indicated. Model discrimination (c) was estimated using logistic regression. The c statistic ranges from 0.5 to 1 where 1 represents perfect discrimination and 0.5 represents no better than random discrimination28. Of note, a model that has exact discrimination (c = 1) would produce propensity scores that do not overlap between hospital and nursing home triage groups, prohibiting causal conclusions about the differential effect of hospital triage 25.
In the second step we used GEE to examine the relationship between hospital transfer and each outcome, controlling for the resident’s propensity score and length of follow-up. We created two sets of models to examine 1) the relationship between hospital transfer and outcomes of all infection episodes, and 2) the relationship between hospital transfer and outcomes of the index infection. Results were similar whether we treated the propensity score as a continuous variable or as an integer for each propensity score stratum. Only results of the former are presented.
All GEE analyses were estimated in STATA using a generalized linear model (xtgee) with a binomial distribution, and a robust variance estimate that adjusts for within- facility correlation29-32. This technique, described by Huber and White29, 30 and expanded on by others to include correlated cases31, 32, makes use of a grouping variable, in this case nursing facility, without requiring its inclusion in the regression model. It uses a theoretical bootstrap method for correcting the standard errors of the regression coefficients and can be applied in regression analyses using many distributions (including binomial). The first stage propensity models utilized a logit-link function, providing odds ratios. Second stage models used a log-link function, providing relative risks of outcomes.