The 607 Ontario nursing homes are presented by location (urban or rural) and tertile of antibiotic use (high, medium, or low).
eTable. Administrative Database Codes Used for Detection of Antibiotic-Related Adverse Events
Daneman N, Bronskill SE, Gruneir A, Newman AM, Fischer HD, Rochon PA, Anderson GM, Bell CM. Variability in Antibiotic Use Across Nursing Homes and the Risk of Antibiotic-Related Adverse Outcomes for Individual Residents. JAMA Intern Med. 2015;175(8):1331-1339. doi:10.1001/jamainternmed.2015.2770
Copyright 2015 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Antibiotics are frequently and often inappropriately prescribed to patients in nursing homes. These antibiotics pose direct risks to recipients and indirect risks to others residing in the home.
To examine whether living in a nursing home with high antibiotic use is associated with an increased risk of antibiotic-related adverse outcomes for individual residents.
Design, Setting, and Participants
In this longitudinal open-cohort study performed from January 1, 2010, through December 31, 2011, we studied 110 656 older adults residing in 607 nursing homes in Ontario, Canada.
Nursing home–level antibiotic use was defined as use-days per 1000 resident-days, and facilities were classified as high, medium, and low use according to tertile of use. Multivariable logistic regression modeling was performed to assess the effect of nursing home–level antibiotic use on the individual risk of antibiotic-related adverse outcomes.
Main Outcomes and Measures
Antibiotic-related harms included Clostridium difficile, diarrhea or gastroenteritis, antibiotic-resistant organisms (which can directly affect recipients and indirectly affect nonrecipients), allergic reactions, and general medication adverse events (which can affect only recipients).
Antibiotics were provided on 2 783 000 of 50 953 000 resident-days in nursing homes (55 antibiotic-days per 1000 resident-days). Antibiotic use was highly variable across homes, ranging from 20.4 to 192.9 antibiotic-days per 1000 resident-days. Antibiotic-related adverse events were more common (13.3%) in residents of high-use homes than among residents of medium-use (12.4%) or low-use homes (11.4%) (P < .001); this trend persisted even among the residents who did not receive antibiotic treatments. The primary analysis indicated that residence in a high-use nursing home was associated with an increased risk of a resident experiencing an antibiotic-related adverse event (adjusted odds ratio, 1.24; 95% CI, 1.07-1.42; P = .003). A sensitivity analysis examining nursing home–level antibiotic use as a continuous variable confirmed an increased risk of resident-level antibiotic-related harms (adjusted odds ratio, 1.004 per additional day of nursing home antibiotic use; 95% CI, 1.001-1.006; P = .01).
Conclusions and Relevance
Antibiotic use is highly variable across nursing homes; residents of high-use homes are exposed to an increased risk of antibiotic-related harms even if they have not directly received these agents. Antibiotic stewardship is needed to improve the safety of all nursing home residents.
Antibiotics are the most commonly prescribed medications among nursing home residents, with approximately two-thirds of residents receiving antibiotic treatment each year.1- 3 Many of these prescriptions are inappropriate, unnecessary, or unnecessarily prolonged, thereby directly exposing individual recipients to Clostridium difficile infection and diarrhea, antibiotic-resistant organisms, antibiotic allergies, and other medication toxic effects.1,4- 6 By promoting selection of C difficile and antibiotic-resistant organisms, antibiotic overuse also poses a potential indirect threat beyond the recipients of these medications because these pathogens can lead to environmental contamination and cross-transmission among residents.
We have previously reported that the variation of antibiotic treatment duration in nursing homes is more dependent on prescriber than recipient characteristics.6 Because most nursing home prescribers in Ontario, Canada, are associated with a single nursing home,7 we hypothesized that antibiotic use would also vary across homes. The capacity of antibiotic treatments to harm individual recipients and neighboring nursing home residents implies that residents could be subject to a higher risk of antibiotic-related adverse outcomes purely on the basis of their place of residence.8 Measuring antibiotic use at the nursing home level is also useful because antibiotic stewardship interventions and solutions could potentially be introduced at that level.9- 11The objectives of this longitudinal open-cohort study were to measure the variability in antibiotic use across Ontario nursing homes and to determine whether residence within a nursing home with high antibiotic use is associated with a greater risk of antibiotic-related adverse outcomes for individual residents.
The study was approved by the research ethics board of Sunnybrook Health Sciences Centre. Informed consent was not required. The study was conducted in the most populous province of Ontario, Canada, using population-based administrative databases at the Institute for Clinical Evaluative Sciences. These well-validated databases are linked through encoded health care numbers and have been used extensively in prior research of medication use in general12- 14 and antibiotic prescribing in particular.6,15 These data sets include the following: (1) the Registered Persons database, which contains demographic data for all users of Ontario’s universal single-payer health care system; (2) the Continuing Care Reporting System Long-Term Care database generated from the Resident Assessment Instrument Minimum Data Set 2.0, a mandatory resident assessment completed at quarterly intervals16; (3) the Ontario Drug Benefit Program database, which contains detailed drug information for Ontario’s more than 1.5 million older adults; (4) the Ontario Health Insurance Plan database, which includes physician billing claims; (5) the Canadian Institute for Health Information Discharge abstract database, which details all hospitalization events; (6) the Ontario Mental Health Reporting System database, which documents admissions to designated psychiatric beds; (7) the National Ambulatory Care Reporting System database, which describes all emergency department visits; and (8) the National Rehabilitation Reporting System.
The cohort consisted of all older adults (aged ≥66 years) residing in an Ontario nursing home from January 1, 2010, through December 31, 2011. We excluded the residents who were younger than 66 years, stayed only in homes with fewer than 25 beds, or switched homes more than once during the study.
Residents of nursing homes are often transferred to the emergency department or hospital. This time spent out of the nursing home poses a methodologic challenge for studies of exposures introduced within nursing homes. To address this issue, we created an open cohort, which allowed us to calculate all resident-days spent in Ontario nursing homes. All residents were followed up from January 1, 2010, or their new arrival into a nursing home (whichever came last) until December 31, 2011, or death (whichever came first). Time intervals spent in nursing homes were considered at risk and contributed to the denominator of resident-days in which an individual was eligible to receive an antibiotic. Time intervals spent outside the nursing home were considered off risk, including time spent in the emergency department, acute care facility, psychiatric facility, chronic care facility, and/or rehabilitation hospitals. If, during the study period, a resident returned to the nursing home from one of these alternate locations, the resident could continue to contribute resident-days to the cohort. A small proportion of residents (7.0%) were transferred between different homes during the study period, and the corresponding resident time was apportioned separately to these facilities. To cross-validate the accuracy of our at-risk intervals, we reviewed the proportion of intervals that included at least one drug claim identified as being dispensed to a nursing home resident. Our cross-validation confirmed that 99.5% of these at-risk intervals were associated with at least one such claim.
Consistent with previous work, dispensed antibiotics were ascertained from the Ontario Drug Benefit Program database, which includes information on drug name, drug dose, drug route, date of drug claim, days supplied, and prescription location.6,15 This program provides universal coverage of publicly funded medications for all Ontario adults older than 65 years; the database is greater than 99% concordant with pharmacy medical record review.17 We included all systemic antibacterials administered via the enteral or parenteral route; we excluded antifungals, antivirals, and topical agents.15
Antibiotic use was calculated as the number of days a resident received antibiotics per 1000 at-risk resident-days spent in each home. The primary predictor of interest was nursing home–level antibiotic use during the study interval, categorized in tertiles as low, medium, or high use.
We also measured nursing home size (number of beds),18,19 ownership (for profit vs not for profit),20 and location (urban vs rural)21,22 because these facility structural characteristics are known to be associated with antibiotic use and/or medication-related adverse events.
Detailed resident-level factors were collected from the Continuing Care Reporting System Long-term Care database, as well as hospital, emergency department, and physician claims databases.6 The Continuing Care Reporting System assessments are performed using the Resident Assessment Instrument Minimum Data Set 2.0, which provides a comprehensive, well-validated assessment of the functional status and care needs of nursing home residents.16 Resident factors that could potentially affect the need for antibiotics or vulnerability to antibiotic-related complications included extensive demographic data, nursing home experience, health care use, diagnoses, functional dependence, continence, and device use variables.1,2,23,24
The primary outcome was a composite of potential antibiotic-related adverse outcomes, including C difficile, diarrhea or gastroenteritis, antibiotic-resistant organisms, allergic reactions to antibiotics, or general medication adverse events. Each component outcome was detected through International Statistical Classification of Diseases,10th Revision (ICD-10), codes from hospital admission and emergency department databases (eTable in the Supplement). The available ICD-10 codes for antibiotic resistance include methicillin-resistant Staphylococcus aureus, vancomycin-resistant Enterococcus, extended-spectrum β-lactamase–producing Gram-negative bacteria, and more broad codes for resistance to β-lactams, fluoroquinolones, or multiple antibiotic classes (eTable in the Supplement). The hospital and emergency department databases have been validated through extensive medical record reabstraction studies, but the adverse drug reaction codes were not specifically examined. The accuracy of C difficile coding has been validated separately, with excellent test characteristics.25- 27 Diarrhea or gastroenteritis was also measured through outpatient physician billing claims to capture antibiotic-associated diarrhea and C difficile infections that did not require hospitalization because there are no specific physician claims for these entities. Residents with a do-not-hospitalize order were excluded from all outcome analyses; by definition, they are not at risk of hospitalization outcome events. A nested secondary outcome measure included only those events that have the potential to cause direct harm to recipients and indirect harm through transmission to other residents (C difficile, diarrhea or gastroenteritis, and antibiotic-resistant organisms). All outcomes were measured at the level of individual residents.
Baseline nursing home and resident characteristics were compared across tertiles of nursing home antibiotic use. Crude primary and secondary outcome rates were calculated across these same tertiles. The primary outcome rate was also examined among the subset of nursing home residents who never directly received an antibiotic.
The primary analysis involved a multivariable marginal (nonconditional) logistic regression analysis, assessing the affect of nursing home–level antibiotic use tertile on the resident-level risk of an antibiotic adverse event. These models accounted for all the other measured nursing home–level and resident-level characteristics, as well as for clustering within homes. Similar modeling was performed for the secondary outcome measure, which was limited to complications that could affect antibiotic recipients and nonrecipients. We compared high-use and low-use nursing homes because recent literature suggests that interfacility comparisons should focus on extreme tiers rather than intermediate tiers28; this approach also protects against multiple hypothesis testing. In a sensitivity analysis, we examined nursing home–level antibiotic use as a continuous variable. All analyses were performed using SAS statistical software, version 9.3 (SAS Institute Inc).
From January 1, 2010, through December 31, 2011, the open cohort included 110 656 unique nursing home residents residing for 118 394 at-risk intervals in 607 nursing homes for a total of 50 953 000 resident-days. Antibiotic treatment was provided on 2 783 000 of the 50 953 000 resident-days (55 antibiotic-days per 1000 resident-days). Antibiotic use varied from as low as 20.4 to as high as 192.9 days per 1000 resident-days (Figure). Facilities were separated into tertiles of antibiotic use: low use, 20.4 to 45.7 antibiotic-days per 1000 resident-days; medium use, 45.8 to 62.2 antibiotic-days per 1000 resident-days; and high use, 62.3 to 192.9 antibiotic-days per 1000 resident-days. Antibiotic use was stable across the 2 study years, with good agreement between tertile assignment of facilities in 2010 and 2011 (68% agreement; weighted κ = 0.62).
The classes of antibiotic agents were similar across the facilities, with penicillins being the most commonly prescribed agents in low-, medium-, and high-use facilities (32.6%, 33.5%, and 33.7% of antibiotic prescriptions, respectively), followed by second-generation fluoroquinolones (31.7%, 29.4%, and 27.1% of antibiotic prescriptions, respectively). Sulfonamides, first-generation cephalosporins, nitrofurantoin, and macrolides were the next most common classes of antibiotics, each accounting for 4% to 9% of antimicrobial prescriptions in each tertile. However, there were more antibiotic prescriptions in the high- vs medium- vs low-use facilities (191 809 vs 162 337 vs 116 532 prescriptions, P < .001), and treatment durations were significantly longer (mean [SD] of 15.5 [53.1] days vs 12.2 [36.6] days vs 10.4 [24.9] days, P < .001).
The median number of beds across nursing homes was 120 (interquartile range, 75-160 beds). Most nursing homes were in urban centers (473 [77.9%]) and had for-profit ownership (378 [62.3%]). In general, high-use facilities were smaller, more likely to have for-profit ownership, and more likely to be rural (Table 1).
The 110 656 residents had a median age of 85 years, were predominantly female (77 337 [69.9%]), had a high prevalence of dementia (61 733 [55.8%]), and usually required at least moderate assistance with activities of daily living (93 096 [84.1%]). Resident characteristics were stable across tertiles with the exception of do-not-resuscitate orders, which were more prevalent in high-use facilities (Table 2).
The risk of an antibiotic-related adverse event was higher in residents of high-use (3311 [13.3%] of 24 943) vs medium-use (3890 [12.4%] of 31 425) vs low-use facilities (3869 [11.4%] of 33 822) (P < .001) (Table 3). The risk difference in adverse events between the high- and low-use antibiotic tertiles translates to a number needed to harm of 53. Among the subset of residents who directly received an antibiotic, there was a higher rate of antibiotic-related adverse events in residents of high-use (2724 [14.3%] of 18 999) vs medium-use (3024 [13.5%] of 22 436) vs low-use facilities (2831 [12.9%] of 21 942) (P ≤ .001). Among the subset of residents who did not directly receive any antibiotics during the study period, there was also a higher rate of antibiotic-related adverse events in residents of high-use (587 [9.9%] of 5944) vs medium-use (866 [9.6%] of 8989) vs low-use facilities (1038 [8.7%] of 11 880) (P = .02). The number needed to harm was 71 for direct antibiotic recipients and 83 for nonrecipients. The secondary outcome measure restricted to the subset of complications that can potentially affect recipients and nonrecipients of antibiotics also occurred at a higher rate among residents of the highest-use tertile. For some of the less common individual adverse events, statistically significant differences were not detected across the tertiles (Table 3).
Multivariable logistic regression modeling of the primary outcome indicated that residence in a high-use (compared with low-use) home was associated with an increased risk of experiencing any antibiotic-related adverse event (adjusted odds ratio [aOR], 1.24; 95% CI, 1.07-1.42; P = .003) (Table 4). Other important predictors of antibiotic-related complications included previous hospitalization and emergency department visits, diabetes mellitus, peripheral vascular disease, cancer, gastrointestinal disease, liver disease, renal failure, functional dependence, incontinence, bladder catheterization, and feeding tube use. The secondary analysis evaluated only those outcomes with the potential to cause indirect harm through transmission to other nursing home residents and confirmed an increased risk for residents of high-use homes (aOR, 1.23; 95% CI, 1.07-1.42; P = .005). A sensitivity analysis examining nursing home–level antibiotic use as a continuous variable confirmed an increased risk of resident-level, antibiotic-related harms (aOR, 1.004 per additional day of nursing home antibiotic use; 95% CI, 1.001-1.006; P = .01).
This study of more than 100 000 older residents of more than 600 nursing homes confirmed a 10-fold variation in nursing home antibiotic use. Residents of high-use nursing homes were exposed to a 24% greater risk of antibiotic-related adverse events, including emergency department visits and hospitalizations related to C difficile infections, diarrhea or gastroenteritis, antibiotic-resistant organisms, allergies, and general medication adverse events. Each additional day of antibiotic use in a facility is associated with a 0.4% increased risk of an antibiotic-related harm for a resident. High antibiotic use in a facility is associated with increased individual-level risk, even among nonrecipients of antibiotics.
The high variability of antibiotic use across nursing homes corroborates our previous finding of high variability in the use of prolonged antibiotic treatment durations in nursing homes, likely driven by culture and prescriber preference rather than resident characteristics.6 Our results are in line with work by the European Surveillance of Antimicrobial Consumption, which revealed that the point prevalence of antibiotic use in nursing homes varies within and between European countries from as low to 1% to as high as 35% of residents,29 as well as a multijurisdiction sampling of 73 US nursing homes that indicated a variation in the number of antibiotic courses prescribed from 0.44 to 23 per 1000 resident-days.30 Our study extends these findings by generating an open cohort of person time spent in nursing homes to calculate antibiotic use per resident time in a manner similar to the way density of antibiotic use is measured in acute care facilities.18,31 More importantly, our study extends prior research by linking antibiotic variability to an increased risk of harms for residents.
Our group and others12,13 have previously found that use of antipsychotics and benzodiazepines in older nursing home residents is associated with serious harms to the recipients of these unsafe medications. The harms of antibiotics, however, threaten not only individual recipients but also neighboring and future residents given the exertion of a selective pressure on the local microbial ecology, leading to higher rates of antibiotic-resistant pathogens and C difficile.32,33 These harms may extend beyond the walls of the nursing home because of transfer of residents in and out of hospitals. Thus, although ongoing, rational use of all medication classes is important in nursing homes,34 our findings indicate a particular urgency for improving antibiotic use. Our results support the Infectious Diseases Society of America call for antibiotic stewardship within nursing homes and the European surveillance work to standardize and reduce nursing home antimicrobial consumption.1,29
Our multivariable analysis of nursing home– and resident-level characteristics confirmed a number of other expected predictors of increased vulnerability to antibiotic-related harms, including advanced age, previous health care use, diabetes mellitus, peripheral vascular disease, liver and renal disease, and use of indwelling devices. Resident-level factors associated with a reduced risk of antibiotic-related harms included a do-not-resuscitate order and dementia, perhaps because of a lower propensity to subject these patients to aggressive diagnostics and treatment.
Some other nursing home–level characteristics were associated with an increased risk of antibiotic-related harms. Larger facilities were associated with more adverse events, potentially related to increased crowding and interpatient transmission of pathogens.1 Rural facilities were overrepresented in the highest tertile of antibiotic use, but after accounting for other nursing home– and patient-level characteristics, rurality was found to be protective against antibiotic-related harms. The mechanism for this protective association of rurality may be an artifact related to hospital access but merits further study.1
Our study is subject to limitations from the use of administrative databases, but misclassification of antibiotic use is highly unlikely given the accuracy of the Ontario Drug Benefit Program database.17 In addition, misclassification of resident characteristics is unlikely in the well-validated Resident Assessment Instrument Minimum Data Set 2.0 database,16 and accurate assignment of resident time in nursing homes was confirmed by 99.5% cross-agreement between the independent nursing home and drug data sets. Moreover, any misclassification in hospital and emergency department outcome databases should be nondifferential for residents of low- and high-use nursing homes, and so we may have underestimated the magnitude of risk associated with high antibiotic use. Given the lack of inpatient medication data in the hospital data sets, our study cannot account for potential harms that originate from antibiotics received during off-risk exposure times. Our outcome definition likely undercounts antibiotic-resistant organisms and will miss mild-moderate antibiotic harms that do not result in emergency department visits or hospitalization, but this fact only further supports the clinical relevance of the number needed to harm estimate of 53 residents because these events were serious enough to prompt transfer to an emergency department or hospital. We cannot assess the appropriateness of antibiotic use given that medical record review is beyond the scope of a study of this size, but the consistent finding of inappropriate antibiotic use in previous related nursing home research suggests that the variability of antibiotic use across institutions in a large part relates to variability in unnecessary days of treatment.5,32 Our causal inference that nursing home–level antibiotic use is responsible for individual-level adverse events could be influenced by unmeasured confounding related to characteristics associated with likelihood of antibiotic receipt and vulnerability to antibiotic-related harm, but this scenario is unlikely given the magnitude of the association, the robustness across multiple outcome measures, and the adjustment for a rich array of resident-level and nursing home–level factors.
Our findings indicate extensive variability in antibiotic use across nursing homes and link this variability to an increased risk of harm for individuals living in facilities with high antibiotic use. Future prospective research is needed to evaluate potential antibiotic stewardship interventions to reduce nursing home antibiotic use, and these interventions may be worth targeting broadly at the nursing home level. Reducing overall antibiotic use at the nursing home level has the potential to reduce harms to direct recipients and nonrecipients of antibiotics and will provide particular protection to residents who are more vulnerable to antibiotic-related harms. Antibiotic stewardship in the nursing home can be challenging because of atypical infection presentations in elderly patients, challenging communication with impaired residents, lack of access to diagnostic testing, and inconsistent availability of onsite pharmacists and physicians.1 However, there have been some encouraging recent successes in reducing inappropriate antibiotic use in the nursing home.32,35 Our study emphasizes the urgency of bolstering these antibiotic stewardship efforts to improve the safety of all nursing home residents.
Accepted for Publication: April 10, 2015.
Corresponding Author: Nick Daneman, MD, MSc, Division of Infectious Diseases, Sunnybrook Health Sciences Centre, University of Toronto, 2075 Bayview Ave, Toronto, ON M4N 2M5, Canada (firstname.lastname@example.org).
Published Online: June 29, 2015. doi:10.1001/jamainternmed.2015.2770.
Author Contributions: Dr Daneman had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Daneman, Bronskill, Gruneir, Fischer, Rochon, Anderson, Bell.
Acquisition, analysis, or interpretation of data: Daneman, Bronskill, Gruneir, Newman, Fischer, Bell.
Drafting of the manuscript: Daneman, Bell.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Daneman, Gruneir, Newman.
Obtained funding: Rochon, Anderson.
Study supervision: Anderson, Bell.
Conflict of Interest Disclosures: Dr Daneman reported receiving support from a Canadian Institutes of Health Research (CIHR) Clinician Scientist Award. Dr Bell reported receiving support from a CIHR and Canadian Patient Safety Institute Chair in Patient Safety and Continuity of Care. Dr Gruneir reported receiving salary support from a CIHR New Investigator Award. Dr Bronskill reported receiving support from a CIHR New Investigator Award in the Area of Aging.
Funding/Support: This study was supported by team grant OTG-88591 from the CIHR Institute of Nutrition, Metabolism and Diabetes and by the Institute for Clinical Evaluative Sciences, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Disclaimer: The opinions, results, and conclusions reported in this article are those of the authors and are independent of the funding sources. No endorsement by the Institute for Clinical Evaluative Sciences or the Ontario Ministry of Health and Long-Term Care is intended or should be inferred.