Randomized Trial Of A Warfarin Communication Protocol For Nursing Homes: an SBAR-based Approach

Abstract

Background:

More than 1.6 million Americans currently reside in nursing homes. As many as 12% receive long-term anticoagulant therapy with warfarin. Prior research has demonstrated compelling evidence of safety problems with warfarin therapy in this setting, often associated with suboptimal communication between nursing home staff and prescribing physicians.

Methods:

We conducted a randomized trial of a warfarin management protocol using facilitated telephone communication between nurses and physicians in 26 nursing homes in Connecticut in 2007–2008. Intervention facilities received a warfarin management communication protocol using the approach “Situation, Background, Assessment, and Recommendation” (SBAR). The protocol included an SBAR template to standardize telephone communication about residents on warfarin by requiring information about the situation triggering the call, the background, the nurse’s assessment and recommendations.

Results:

435 residents received warfarin therapy during the study period for 55,167 resident days in the intervention homes and 53,601 in control homes. In intervention homes, residents’ international normalized ratio (INR) values were in the therapeutic range a statistically signiciant 4.50% more time than in control homes (95% CI, 0.31%, 8.69%). There was no difference in obtaining a follow-up INR within 3 days after an INR value ≥4.5 (OR 1.02 (95% CI, 0.44, 2.4). Rates of preventable adverse warfarin-related events were lower in intervention homes, although this result was not statistically significant: the incident rate ratio for any preventable adverse warfarin-related event was .87 (CI, .54, 1.4).

Conclusion:

Facilitated telephone communication between nurses and physicians using the SBAR approach modestly improves the quality of warfarin management for nursing home residents.

Introduction

As many as 12% of the 1.6 million American nursing homes residents receive long-term oral anticoagulant therapy with warfarin to prevent strokes and other thromboembolic events. Prior research has demonstrated compelling evidence of safety problems with warfarin therapy in nursing homes resulting from suboptimal warfarin management and errors in prescribing and monitoring.^1–3 In our previous studies, we estimated that there may be as many as 34,000 fatal, life-threatening, or serious adverse warfarin-related events per year in the nursing home setting, the majority of which may be preventable.^{1, 4–5} An important contributor to many of these events appeared to be the lack of key information regarding prior international normalized ratio (INR) values, warfarin dosing, and interacting medications.

Decisions about medical care in nursing homes, including decisions about prescribing, often occur in the context of brief telephone discussions between physicians and nursing staff, as physicians are frequently off-site when such decisions occur.^6–9 Thus, the system of medication management in this setting relies heavily on the quality of communication between nursing staff within the facility and the physicians who provide medical care to individual residents. Studies focused on communication between physicians and nursing staff in long-term care facilities have identified challenges in communicating all the relevant information required to make optimal clinical decisions.^{7, 9–12} Several recent studies have suggested that some of these challenges may be overcome through communication protocols and training. Rates of preventable hospitalizations for nursing home residents were lower for facilities with nurses trained to communicate effectively with physicians.¹² Implementation of a systematic approach to telephone communication of health information at a Veterans Administration-affiliated extended care center improved nurses’ perceptions of their ability to handle communications with physicians.¹³ Therefore, in an effort to improve the quality and safety of anticoagulation management in the nursing home setting, we conducted a randomized trial of a warfarin management communication protocol. The approach was standardized communication of Situation, Background, Assessment, and Recommendation (SBAR). SBAR was originally developed by the United States armed forces to structure situation briefings. It has been implemented in a number of healthcare settings where communications between nurses and physicians are critically important for patient safety.^14–17

Methods

The study was conducted in 26 nursing homes in the state of Connecticut in 2007–2008. Patients in short-term care areas of the facilities (e.g., subacute care, hospital-level care, or rehabilitation) were not included in the study. The study was approved by the institutional review board of the University of Massachusetts Medical School.

Nursing homes were randomized to intervention or control status by the central project office within blocks according to bed size (less than 150, 150 to 249, 250 or more) using a random number generator. Table 1 provides characteristics of the intervention and control homes in comparison to all facilities in Connecticut and across the United States.^18–19 Facilities randomized to the intervention group received a warfarin communication protocol developed around the SBAR approach. The protocol included customized methods of identifying and highlighting residents taking warfarin, systematic procedures for tracking and communicating international normalized ratio (INR) results, a targeted training program for nursing staff focused on using the SBAR approach for telephone-based communication, and message templates to standardize telephone communication about residents on warfarin between nursing staff and physicians. The message templates were provided on printed forms designed to facilitate use of the SBAR approach (figure 1). Control homes continued with usual care. A representative of each nursing home participated in initial training after which all materials were shipped to the home in preparation for the facility’s selected start date. Each nursing home participated in the study for one year. The assigned representative participated in monthly calls with the central project office and periodic cross-facility conference calls.

Table 1.

Comparison of Study Nursing Homes with Facilities in Connecticut and the United States

Characteristic	Intervention Homes (N=13)	Control Homes (N=13)	CT Nursing Homes (N=245)	US Nursing Homes (N=16,1001)
No. Beds
Mean ± SD	137.5 ± 56.5	135.5 ± 63.6	113.7	107.6
% Female residents
Mean ± SD	75.0 ± 5.6	70.6 ± 11.3	67.3	71.2
Ownership
For profit (%)	46.2	84.6	77.6	61.5
Nonprofit	46.2	7.7	21.6	30.8
Government	7.6	7.7	0.8	7.7
Staffing Licensed nursing staff hours per resident per day
Mean ± SD	1 h 25 m ± 12 m	1 h 27 m ± 24 m	1 h 24 m	1 h 18 m
Nursing assistant staff hours per resident per day
Mean ± SD	2 h 31 m ± 31 m	2 h 26 m ± 22 m	2 h 18 m	2 h 18 m

Figure 1.

Template for SBAR communication

Case-finding

Information on INR values, warfarin dosing, and warfarin-related incidents was identified through quarterly reviews of nursing home records by trained nurse abstractors for each eligible resident of a participating facility who was receiving warfarin at any time during the one-year period of the home’s participation in the study. Abstractors identified possible warfarin-related incidents based on specified triggers: INR values of 4.5 or greater, evidence of overt bleeding, hematocrit decreases of 3% or greater, orders for vitamin K, a stroke or thromboembolic event, and emergency department visits or hospitalizations, following the approach used in our previous study.¹ For each trigger identified, the abstractors completed an abstraction form describing the incident and all related INR values, medications and medication changes.

Outcome measures

Two measurements were used to assess quality of anticoagulation management: (1) the proportion of time that nursing home residents receiving warfarin had INR values within the target therapeutic range (defined as 2.0 to 3.0); and (2) the rate at which the response to an INR value 4.5 or greater was a follow-up INR value within 3 days. We identified preventable adverse events associated with warfarin therapy, defined as injuries resulting from the use of warfarin, including hemorrhagic events as well as thromboembolic events that were associated with subtherapeutic INR values. Adverse warfarin-related events were considered to be preventable if they were judged to be due to an error and were preventable by any means available. Errors could occur during the ordering, dispensing, administration, and monitoring stages of warfarin management. Error types were further categorized as wrong dose, wrong drug, missed dose, wrong frequency, extra dose, wrong resident, known drug interaction, inadequate laboratory monitoring, or delayed response to laboratory evidence of an out-of-range INR value. We also identified potential adverse warfarin-related events, defined as instances in which the INR value was 4.5 or greater with an associated error in warfarin management, but without patient injury. Warfarin-related incidents were presented to pairs of physician-reviewers (JHG, JT, LRH) blinded to the intervention status of the resident’s nursing home. Reviewers independently classified incidents including whether an adverse warfarin-related event or a potential adverse warfarin-related event was present, whether the event was preventable, and the severity of the event. The severity of adverse warfarin-related events was classified according to modification of the criteria of Landefeld and colleagues,²⁰ as used by White and colleagues.²¹ The severity of adverse warfarin-related events was categorized as minor, serious, life-threatening, or fatal. Minor events were those with no medical consequence (e.g., bruising). Serious events were those that required specific treatment or a medical evaluation. Life-threatening events included the need for a surgical intervention to stop the bleeding, irreversible sequelae (e.g., myocardial infarction or stroke), or any 2 of the following: transfusion of 3 or more units of blood, hypotension, critical anemia, or acute bleeding (<3 days).

When the two reviewers disagreed on the classification of an incident, its preventability or its severity, they met and reached consensus. Consensus was reached in all instances. We compared the initial assessments of the physician reviewers and calculated interrater reliability using the kappa statistic with kappa values of 0.93 for judgments regarding the presence of an adverse warfarin-related event, 0.61 for preventability, and 0.74 for severity. A kappa score of 0.4 to 0.6 reflects “moderate agreement,” 0.6 to 0.8 reflects “substantial agreement,” and 0.8 to 1.0 is considered “almost perfect.”²²

Statistical analysis

We hypothesized that warfarin management among residents in the intervention nursing homes would be superior to that in the control nursing homes, leading to higher percent time in the therapeutic range and higher rates of follow-up to INR values ≥4.5 within three days. Conservative sample size estimates suggested that the study would be able to find differences in these rates of 20% or greater statistically significant at the 0.05 level. We also tracked rates of preventable warfarin-related events (AWEs) and potential adverse warfarin-related events, although we recognized that our sample size was insufficient for identifying moderate differences in these rates. Based on data from our previous study of warfarin management in Connecticut nursing homes,¹ we estimate that a sample size of 40 nursing homes per arm would be required to find a 20% reduction in preventable AWEs statistically significant.²²

The primary analyses are intention-to-treat analyses comparing the intervention and control nursing homes with regard to: the percentage of time within the therapeutic range for residents taking warfarin; the rate of follow-up to INR values ≥4.5 within three days; and the rates of preventable adverse warfarin-related events and potential adverse warfarin-related events. All analyses accounted for the clustering of individuals within the randomized nursing homes.

Specific person-time within the therapeutic range in each nursing home was computed according to the method of Rosendaal which assigns an INR value to each person-day based on interpolated values between measurements.²⁴ We excluded periods from analysis when a resident was not taking warfarin or was absent from the home. We stopped interpolating INR values at the last INR before an excluded period, and did not restart until the first INR value after that period. Analyses comparing the percentage of time spent in the therapeutic range between intervention and control homes were modeled using a generalized estimating equation model to account for residents clustered within nursing homes and to control for resident characteristics including age, gender, race, and comorbidity, as well as abstractor. To control for the potential impact of the length of time residents were on warfarin during the study, we developed two models which provided consistent estimates of the intervention effect: a mixed model with individual resident INR values weighted by total days taking warfarin during the study and a mixed model adjusted for total days. Comparisons of the rates of follow-up INRs occurring within 3 days after an INR value of 4.5 or greater used logistic regression analyses modeled with generalized estimating equations clustered within nursing homes and adjusted for the resident characteristics and abstractor.^25–26

To determine crude rates of events, the outcome measures (i.e., numbers of preventable adverse warfarin-related events, serious or more severe preventable adverse warfarin-related events, and potential adverse warfarin-related events) were divided by the total number of nursing home resident days on warfarin during the study period. These were estimated by obtaining census data for all eligible residents on warfarin throughout the course of the study, accounting for absences from the nursing homes and breaks in warfarin use. Comparisons of the rates between intervention and control homes were modeled using Poisson regression and generalized estimating equations to account for clustering within nursing homes.^25–27 All models controlled for the resident characteristics and abstractor.

Results

The 26 nursing homes continued to participate for the full one-year follow-up period and all are included in analyses of outcomes. Across the 26 homes, 435 residents received warfarin therapy during the study period (Table 2), contributing 55,167 days of observation in the intervention homes and 53,601 days in the control homes. Approximately 71% of residents in the intervention homes were female compared to 65% in the control homes, with an average age of 83 years in intervention and 82 years in control homes. The most common indication for warfarin therapy in both groups of homes was prevention of stroke due to atrial fibrillation (66%).

Table 2.

Characteristics of Residents in Intervention and Control Nursing Homes

	Intervention nursing homes	Control nursing homes
Total # residents on warfarin	211	224
Average % residents on warfarin	11.79	12.70
Total resident days on warfarin during the study period	55167	53601
				p-value
Age Mean, SD	82.8, 9.1	82.2, 11.8		0.554
% Female	70.6	65.2		0.225
Race
% white	90.5	78.6		0.001
% black	3.3	11.6
% other or unknown	6.2	9.8
Comorbidity burden (Charlson Index) Mean ± SD	2.15 ±1.59	2.44 ±1.66		0.064
Indication for warfarin
% Atrial fibrillation	66.3	65.7		0.585
% Venous thrombosis/pulmonary embolism	22.9	20.7
% Stroke prevention without atrial fibrillation	8.8	12.5
% Valvular heart disease	0.5	0.0
% Other	4.4	5.1

Among the intervention homes, residents had INR values within the therapeutic range 53.1% of their time on warfarin compared to 50.0% among residents in the control homes for an absolute difference of 3.1%. With adjustment for clustering within homes, length of time on warfarin, characteristics of the residents, and abstractor, the difference in the percent time in therapeutic range was 4.5% (95% confidence interval (CI), 0.3%, 8.7%). For a patient taking warfarin for a full year, this represents 16 additional days in the therapeutic range (95% CI, 1, 32 days) The adjusted intracluster correlation for nursing homes was 0.01.

In the intervention homes, 64.6% of INRs ≥4.5 were followed by another INR measure within 3 days while in the control homes the result was 71.7%. Adjusting for clustering within homes, length of time on warfarin, characteristics of the residents, and abstractor, the odds ratio for follow-up in more than three days comparing residents in intervention nursing home to those in control homes was 1.02 (CI, 0.44, 2.4). The adjusted intracluster correlation for nursing homes was 0.10.

Nurse abstractors identified 782 possible warfarin-related incidents during the study. There were 83 classified as preventable adverse warfarin-related events, and 183 as potential adverse warfarin-related events (Table 3). Rates of preventable adverse warfarin-related events were lower in intervention homes, although this result was not statistically significant. Expressed as rates per 100 resident months, the rate of preventable adverse events related to warfarin in the intervention homes was 2.26 while the rate in the control homes was 2.38. The rate of serious preventable events was 0.39 in intervention homes and 0.62 in control homes. The rate of potential adverse warfarin-related events (situations in which the INR was 4.5 or greater and no injury occurred, but an error in warfarin management was identified) was 4.52 in the intervention homes and 5.74 in the control homes. The adjusted incident rate ratio (IRR) for any preventable adverse warfarin-related event was 0.87 (CI, 0.54, 1.4); for a serious preventable adverse warfarin-related event the IRR was 0.50 (CI, 0.17, 1.5) (Table 3). For potential adverse warfarin-related events, the IRR was 0.77 (CI, 0.45, 1.3).

Table 3.

Adverse Warfarin-related Events in Intervention and Control Nursing Homes

	Intervention nursing homes	Control nursing homes	Rate Ratio	Adjusted Rate Ratio (95% CI)
Peventable adverse warfarin related event	41	42	0.95	0.87 (0.54, 1.4)
Preventable serious, life-threatening or fatal warfarin-related event	7	11	0.62	0.50 (0.17, 1.5)
Potential adverse warfarin-related event	82	101	0.79	0.77 (0.45, 1.3)

^*Adjusted for nursing home, abstractor, and resident’s age, sex, race, and Charlson comorbidity score

Comments

In this randomized trial, a warfarin management communication protocol based on SBAR modestly improved the time that residents were maintained in the therapeutic range and appeared to lower the rates of preventable adverse warfarin-related events, although the latter findings are not statistically significant. However, the intervention had no impact on the rate of 3-day follow-up after INR values ≥4.5, with a subsequent INR.

Several interventions have been developed to improve warfarin management, including use of dosing algorithms,^28–30 anticoagulation clinics,³¹ patient home monitoring,³² and pharmacist based management.³³ However, there are logistical, financial, informational, and even regulatory challenges associated with providing specialized anticoagulation services to multiple non-affiliated nursing homes across a geographic region. In a previous study surveying physicians caring for residents of nursing homes, we actually found some resistance to external management of warfarin for their patients who reside in nursing homes.³⁴ Further, prior warfarin management interventions have been tested in the hospital and ambulatory setting, but have not been evaluated in the nursing home setting. While highly technological interventions to improve warfarin prescribing and monitoring are options for select nursing homes,³⁵ they are not options for many nursing homes because financial, structural and cultural barriers to full adoption of computerized order entry and electronic medical records remain in these facilities.^36–37

For these reasons, and because quality of communication has been shown to affect quality of medication prescribing in nursing homes,^{11, 38} our study evaluated an innovative, low-technology intervention to improve warfarin management in nursing homes by implementing structured communication. Based upon an established approach for situation briefing drawn from the U.S. armed forces, the SBAR approach to nurse-physician communication has been refined and tested in various clinical settings.^{14–17, 39} SBAR establishes a standardized structure for medical communication, ensuring that critical information that directly affects physician decision-making is transmitted in a predictable manner. Its use is particularly well suited to telephone communication in the nursing home, where medication management involves the critical flow of communication between nursing staff and off-site physicians. Frequently, decisions about warfarin dosing and INR measurement are made during brief telephone conversations with inadequate information about prior INR values, warfarin doses, and interacting medications available to the decision-maker.¹ In this trial, we operationalized SBAR in a template that requires nurses to collect information and consider their own assessment of the resident’s situation before placing calls to physicians about residents on warfarin to report INR values, indicators of possible adverse events, infections such as urinary tract infections, or medication issues. The template guides the user through communication in a consistent fashion, using standard language.

Our study has several limitations including inadequate power to identify the intervention’s impact on preventable adverse events. However, the intervention was associated with a statistically significant and clinically meaningful higher percent of time in the therapeutic range indicating better warfarin management with the potential for reduction in risk of both thromboembolic and hemorrhagic events. Our study relied entirely on information abstracted from existing medical records within each study nursing home and, thus, is subject to those limitations. This may have caused us to miss some information relevant to warfarin-treated residents. To overcome this potential problem, we trained abstractors to search for and report on a variety of indicators of possible events from multiple sources in the nursing homes. We used the subsequent physician review process to cull incidents that were not found to identify a warfarin-related event.

The study also has several important strengths. It is set among 26 community based nursing homes that are similar to nursing homes across the United States so that our results should be broadly generalizable. The study design is a randomized controlled trial, providing assurance that the impact is not confounded by underlying differences in the facilities’ staff, infrastructure, management or overall patient safety culture.

Conclusions

We conclude that use of a communication protocol based on SBAR can modestly improve the quality of warfarin management in nursing homes, as reflected by increased time in therapeutic range. This low technology approach may also serve as a model for improving the safety of other medications associated with high rates of preventable adverse drug events and for improving safety for vulnerable nursing home residents at special risk for medication-related problems.